|Home | About | Journals | Submit | Contact Us | Français|
Toll-like receptors (TLRs) detect conserved molecular patterns that are unique to microbes, enabling tailored responses to invading pathogens and modulating a multitude of immunopathological conditions. We investigated the ability of a naturally-occurring stearoyl-arachidonoyl form of phosphatidylserine (SAPS), to inhibit the proinflammatory effects of TLR agonists in models of inflammation investigating the interaction of leukocytes with epithelial and endothelial cells. The responses to LPS of both epithelial and endothelial cells were highly amplified in the presence of peripheral blood mononuclear cells (PBMCs). Coincubation with SAPS markedly inhibited activation of cocultures by LPS, principally through inhibition of the TLR4 signalling pathway in PBMCs, however this was not through downmodulation of TLR4 or coreceptor expression, nor was IL-1β-induced cytokine release affected. SAPS also impaired Pam3CSK4 (TLR2/1), Gardiquimod (TLR7/8), and Streptococcus pneumoniae-induced cytokine release, but had only modest effects on Poly(I:C) (TLR3) induced responses. FRET analysis of molecular associations revealed that SAPS disrupted the association of both TLR4 and TLR2 with their respective membrane partners that are required for signalling. Thus our data reinforces the existence and importance of cooperative networks of TLRs, tissue cells, and leukocytes in mediating innate immunity, and identifies a novel disrupter of membrane microdomains, revealing the dependence of TLR signalling on localisation within these domains.
It is now well-established that Toll-like receptors (TLRs) detect conserved molecular patterns unique to microbes, facilitating host defence against pathogens and enabling the construction of tailored responses that are essential for their clearance (1, 2). TLRs are key in the detection of many structurally unrelated proinflammatory stimuli, including a diverse range of microbial products and also endogenous ligands, generated from damaged or dying host cells (reviewed in (3, 4)).
TLR4 is the most widely studied member of the family and recognises an extensive range of agonists, but primarily LPS (5, 6). LPS-induced cellular activation is thought to occur when receptors are activated within membrane microdomains such as lipid rafts, which may help to confer ligand specificity through the association of diverse proteins within the raft (7, 8). CD14 is a key protein within rafts, and is involved in transfer of LPS to the TLR4/MD-2 complex (9-11). CD14 also binds to, and mediates the transport of, phospholipids such as phosphatidylserine (PS) (12-15). Whilst phosphatidylinositol has been reported to be an LPS antagonist (13, 16), PS has been shown to have only modest effects on LPS signalling, potentially as a result of interference with the LPS/CD14 interaction (16), although the exact mechanism remains unclear and PS may also have other effects on TLR signalling. It is ubiquitously present in the mammalian cells, where it is normally located on the cytosolic surface of the plasma membrane (17). During apoptosis, PS is externalised to the outer membrane targeting the cell for recognition and clearance by macrophages (18, 19). This process of apoptotic clearance leads to the production of anti-inflammatory cytokines and the active suppression of inflammatory mediator production (20, 21). Whilst PS is important in recognition of apoptotic cells, it is less clear whether PS directly signals to drive the phenotype of engulfing macrophages to a more anti-inflammatory state.
We were therefore interested to explore the consequences of PS exposure on subsequent TLR responses in a range of cell types and in vitro coculture models. We have previously demonstrated that responses to TLR agonists are often most efficaciously induced when leukocytes and tissue cells are allowed to interact, and proposed that tissue cell signalling is an important amplification mechanism for signals derived from leukocytes interacting with microbial agonists (22-25). In previous studies we have shown that the neutralisation of IL-1 can inhibit these networks (23). In this study we observed that a naturally-occurring species of PS, 1-stearoyl-2-arachidonoyl-sn-glycero-3-[Phospho-L-Serine] (abbreviated to SAPS), was an effective inhibitor of TLR4, but not IL-1β, signalling. Its non-toxic mode of action was not likely to be due to competition for LPS or activation of anti-inflammatory pathways, but was consistent with its ability to disrupt membrane microdomains, and revealed a substantial dependence of many TLRs on such domains for effective signalling.
Cell culture reagents were purchased from Invitrogen (Paisley, UK), and general laboratory reagents were purchased from Sigma-Aldrich (Poole, UK). FCS (endotoxin levels <0.5 EU/ml) was purchased from BioWhittaker (Verviers, Belgium). Purified LPS from Escherichia coli serotype R515 was from Axxora (Nottingham, UK). The synthetic lipopeptide Pam3CysSerLys4 (Pam3CSK4) was from EMC Microcollections (Tübingen, Germany). Poly(I:C) and gardiquimod were from InvivoGen (San Diego, USA). Recombinant human IL-1β and TNFα were from PeproTech EC (London, UK). Inert (heat-killed) type 2 Streptococcus pneumoniae (strain D39) was a kind gift from Dr. D. H. Dockrell (26). Highly purified SAPS (1-stearoyl-2-arachidonoyl-sn-glycero-3-[phospho-L-serine]; >99% pure as judged by HPLC) was supplied as a gift by Vaccine Technology Ltd., UK (PCT publication number WO 2008/068621) having been synthesised on their behalf by Avanti Polar Lipids Inc. (Alabaster, USA). alamarBlue™ was obtained from BioSource International (Camarillo, CA, USA). Cytochalasin D and staurosporine were from Sigma-Aldrich (Poole, UK) and Calbiochem (Merck Chemicals Ltd. Nottingham, UK) respectively. Nitrocellulose membrane and ECL reagent for western blotting were from GE Healthcare (Chalfont St. Giles, UK). Anti-phosho-ERK1/2, anti-phospho-JNK, anti-phospho-IκBα, and anti-rabbit secondary antibodies were from Cell Signaling Technology (Danvers, MA, USA). Anti-phospho-p38 was from Promega (Southampton, UK). Anti-actin was from Sigma-Aldrich (Poole, UK). For flow cytometry, PE-conjugated anti-TLR4 mAb (clone HTA125, isotype IgG2a) and PE-conjugated CD14 mAb (clone 61D3, isotype IgG1) and isotype controls were from eBioscience (San Diego, CA, USA). For FRET experiments: repurified LPS from Salmonella minnesota was purchased from List Biological Laboratories, Inc. (Surrey, UK), Lipoteichoic Acid (LTA) from Staphylococcus aureus was a generous gift from Professor Thomas Hartung, University of Konstanz. TLR4 and TLR2 specific mAbs, HTA125 and TL2.3, were purchased from Hycult Biotechnology (Uden, The Netherlands), whilst the TLR6 specific polyclonal antibody was from Autogen Bioclear (Wiltshire, UK). The CD14 specific mAb MY4 was from BioGenex (San Ramon, CA, USA). Hybridoma cells secreting 26ic (anti-CD14), and W6/32 secreting MHC class I specific mAbs were from ATCC (Manassas, VA, USA). Cholera toxin was purchased from List Biological Laboratories, Inc. (Surrey, UK). The antibodies used for FRET studies were conjugated to either Cy3 or Cy5 using labelling kits from Amersham Biosciences (Buckinghamshire, UK).
Peripheral blood mononuclear cells (PBMCs) were prepared from the venous blood of healthy volunteers taken with informed consent in accordance with a protocol approved by South Sheffield Local Research Ethics Committee. PBMCs were enriched by centrifugation over density gradients as described (27, 28). Monocytes were enriched further by negative magnetic selection using Monocyte Isolation Kit II (Miltenyi Biotech, Bergisch Gladbach, Germany) to a typical mean (±SEM) purity of 83.75±4% (n=4). The immortalised bronchial epithelial cell line BEAS-2B was maintained in RPMI 1640 containing 2 mM L-glutamine, supplemented with 10% FCS, 100 U.ml−1 penicillin, and 100 ug/ml−1 streptomycin. The monocytic cell-line, THP-1, was obtained from the American Type Culture Collection (Manassas, VA) and maintained in RPMI (supplemented as BEAS-2B). Cells were differentiated in 24 well plates with 0.5 μM PMA (Sigma-Aldrich, Poole, UK) for 3 hours, non-adherent cells were discarded and the adherent cells cultured in normal media for 24 hours prior to use. HUVECs were isolated from umbilical cords, donated with informed consent following a protocol approved by North Sheffield Local Research Ethics Committee. Cells were maintained in RPMI media supplemented with 2 mM L-glutamine, 20 μg/ml endothelial cell growth supplement (Harbour Bio-products, Norwood, MA, USA), 95 μg/ml heparin, 0.225% sodium bicarbonate, 100 U/ml penicillin, 100 μg/ml streptomycin, 10% FCS and 10% newborn calf serum.
Cocultures of BEAS-2B epithelial cells and PBMCs were created (in 24 well plates) with the addition of 30,000 PBMCs/well to 80-90% confluent BEAS-2B cells (seeded 24-48 hours prior to use to attain this confluency), giving a ratio of approximately 1 PBMC to 3 BEAS-2B cells. Monoculture controls were included in all experiments. HUVECs were used at passage 2-3 for all experiments. Cells were seeded in 12 well tissue culture plates and grown to 70-90% confluence, then washed and media replaced with low serum media (2%) for 24 hours. HUVEC/monocyte cocultures were created through the addition of 20,000 enriched monocytes/well to the HUVEC monolayers, giving a ratio of approximately 1 monocyte to 5 HUVECs. Monoculture controls were included in all experiments.
Cell treatment occurred in the same media as maintenance (see above); in brief, cells were pretreated with SAPS for 1 hour prior to the addition of the TLR or cytokine agonists for 24 hours (unless otherwise stated), with SAPS remaining present throughout. Each experiment was carried out multiple times using separate PBMC donors (19 donors used in total) and BEAS-2B or HUVEC cell culture passages.
Cell free supernatants were collected and stored at −80°C until use. Immunoreactive CXCL8, IL-1β, IL-6, and IL-10 were quantified by ELISA using matched Ab pairs from R&D systems (Abingdon, UK). The detection limits were 62.5, 19.5, 78, and 32.5 pg/ml respectively. CXCL10 was quantified using the BD™ Cytometric Bead Array (CBA) Human Soluble Protein Flex Set (BD Biosciences, Oxford, UK), and BD FACSArray bioanalyzer (BD Biosciences, San Jose, USA), in accordance with the manufacturer's protocols. The limit of detection was 10 pg/ml. Samples whose cytokine levels were undetectable were assigned the detection limit values for graphing and analysis.
Western blot analysis was carried out as described (29). The antibodies used were anti-phospho-ERK1/2 (1:500), anti-phospho-JNK (1:500), anti-phospho p38 (1:1000), anti-phospho-IκBα (1:1000), or anti-actin (1:10,000), all detected using a horseradish peroxidase-coupled anti-rabbit secondary antibody (1:2000) followed by enhanced chemilluminescense. Films were densitometrically analysed using NIH Image (version 1.62f).
Staining and flow cytometry were carried out as previously described (30). Briefly, cells were washed in ice-cold FACS buffer (10mM PBS without Ca2+ and Mg2+, containing 10mM HEPES and 0.25% BSA) before non-specific binding was blocked with mouse IgG (50 μg/ml) and cells were stained with anti-TLR4 or anti-CD14 (both a 1:25 dilution), or isotype control for 45 minutes at 4°C. Excess antibody was removed by washing, cells fixed in 1× CellFix™ (BD Biosciences, San Jose, USA) and cytometry performed on a dual laser FACSCalibur using CellQuest software (BD Biosciences, San Jose, USA). Data were analysed using FlowJo software v8.5.3 (Tree Star Inc., Oregon, USA).
BEAS-2B cells or 300,000 PBMCs were seeded in 96 well plates as described above. Triplicate wells were pretreated with SAPS for 1 hour prior to the addition of 1 ng/ml LPS for 24 hours, with SAPS remaining present throughout. Positive controls to induce cell death in BEAS-2B and PBMCs were cytochalasin D (5 μg/ml) and staurosporine (1 μM) respectively. At 20 hours 10% alamarBlue™ was added and the cells incubated for a further 4 hours. Fluorescence was measured on an fMax™ fluorimeter and Softmax Pro software (Molecular Devices, USA) with excitation at 544 nm and emission at 590 nm. The mean of the absolute fluorescence units for the triplicate wells was calculated, minus the average fluorescence units of media alone (no cells), and data expressed as percentage change from control cells.
FRET was carried out as described previously (31, 32). Briefly, monocytes were isolated from human A+ buffy coats and cultured on microchamber culture slides (Lab-tek, Gibco) in serum-free media supplemented with 0.01% L-glutamine and 40 μg/ml gentamicin. Cells were labelled with 100 μl of a 1:1 mixture of donor-conjugated antibody (Cy3) and acceptor-conjugated antibody (Cy5). The cells were pretreated with buffer or SAPS (100 μg/ml) for 1 hour prior to stimulation with LPS from Salmonella minnesota (100 ng/ml), or LTA from Staphylococcus aureus (10 μg/ml), for 10 minutes. The cells were washed twice in PBS/0.02% BSA prior to fixation with 4% formaldehyde for 15 minutes to prevent potential reorganisation of the proteins during the course of the experiment. Cells were imaged on a Carl Zeiss, Inc. LSM510 META confocal microscope (with an Axiovert 200 fluorescent microscope) using a 1.4 NA 63X Zeiss objective and images analysed using LSM image analysis software (Carl Zeiss, Inc.). The different fluorophores, Cy3 and Cy5, were detected using the appropriate filter sets. Using typical exposure times for image acquisition (<5 seconds), no fluorescence was observed from a Cy3-labelled specimen using the Cy5 filters, nor was Cy5 fluorescence detected using the Cy3 filter sets. The rate of energy transfer is inversely proportional to the sixth power of the distance between donor and acceptor. The efficiency of energy transfer (E) is defined with respect to r and R0, the characteristic Forster distance by: E = 1/[1+(r/R0)6]. Energy transfer was detected as an increase in donor fluorescence (dequenching) after complete photobleaching of the acceptor molecule by: E(%) × 100 = 10,000 × [(Cy3 postbleach − Cy3 prebleach)/Cy3 postbleach]. The scaling factor of 10,000 was used in order to expand E to the scale of the 12-bit images.
All data are presented as mean ± SEM (where appropriate) of at least three independent experiments on separate donors (19 donors used in total) and THP-1, BEAS-2B, or HUVEC cell culture passages. Data were analysed using the statistical tests stated, with ANOVA and the indicated post test being used for multiple comparisons. Data were analysed using Prism (version 4.03, GraphPad, San Diego, CA).
We have previously reported that coculture of immortalised airway epithelial cells with PBMCs results in a marked potentiation of LPS-induced cytokine release when compared to that of either cell type alone (22). Here, we show that release of the proinflammatory cytokines IL-1β and CXCL8 from epithelial cell/PBMC cocultures in response to LPS was substantially greater than that seen from either cell type alone, and that this was dose-dependently and potently inhibited by SAPS (Fig. 1A,D). Effective activation of cocultures by LPS is dependent upon activation of monocytes and their initial production of IL-1β (23). We therefore determined whether the modified PS acted on PBMCs to reduce their ability to activate epithelial cells, or upon the epithelial cells to reduce their sensitivity to activation by PBMCs. SAPS inhibited LPS-induced cytokine release from PBMCs alone (Fig. 1B,E) whilst the BEAS-2B cells alone did not release either IL-1β or CXCL8 at detectable levels when stimulating with the concentrations of LPS used here (Fig. 1C,F). A small increase in CXCL8 release was observed in BEAS-2B cells in response to the higher doses of SAPS (Fig. 1F), indicating that SAPS may have some modest proinflammatory effects, although these were restricted to tissue cells.
These data suggest that the main mechanism by which our modified PS inhibited inflammation in our model was through downregulation of proinflammatory effects during the initial PBMC response. However the anti-inflammatory cytokine IL-10 is produced by monocytes and is a potent suppressor of proinflammatory cytokine production, including IL-1β and CXCL8 (33). Consequently we investigated whether SAPS could directly increase IL-10 production from PBMCs and potentially provide a mechanism for suppression of LPS-induced pro-inflammatory cytokine release. Our data revealed that SAPS dose-dependently inhibited LPS-induced IL-10 release from PBMCs (Fig. 2) and thus our modified PS appeared to have a global inhibitory effect on TLR4 function.
TLR4 function is, in part, regulated by control of its expression and shuttling of the receptor from endosomal to plasma membrane compartments (34, 35). We therefore determined if exposure to PS, a membrane lipid, exerted effects on TLR surface expression that might underpin inhibition of TLR4 responses. We found that incubation of PBMCs with SAPS for 1 hour had little effect on TLR4 or CD14 expression as measured by flow cytometry (Fig. 3A,B). To rule out the possibility that SAPS was mediating its inhibitory effects by having detrimental actions on the cells, thus diminishing their ability to produce cytokines, we assessed the ability of PBMCs and epithelial cells to reduce the dye alamarBlue™, as a measure of their metabolic capacity and hence viability (36, 37). The metabolic capacity of PBMCs and epithelial cells was unaffected by SAPS (with or without LPS) (Fig. 3C,D). These data suggested SAPS was not acting to globally disturb membrane functions such as uptake of nutrients, was non-toxic, and did not interfere with cellular metabolism.
LPS activates the TLR4 receptor complex triggering two downstream signalling pathways termed MyD88-dependent and -independent, governed by the adaptor proteins that are recruited (38).
Production of proinflammatory cytokines, including IL-1β and CXCL8, is heavily dependent upon the MyD88-dependent arm that subsequently leads to activation of early-phase NF-κB and the MAP kinases, which diverge at the level of TRAF6/TAB2/TAK1/TAB1 complex (38). To gain further insights into the level at which SAPS exerted its actions, we investigated the actions of SAPS on LPS-mediated NF-κB and MAP kinase signalling. Treatment of PBMCs with LPS for 30 minutes activated NF-κB signaling as shown by IκBα phosphorylation, and this was inhibited by a 1 hour pretreatment with SAPS (Fig. 4A). SAPS also inhibited phosphorylation of the MAP kinase family members JNK (Fig. 4B), p38 (Fig. 4C), and ERK (Fig. 4D).
The MyD88-independent pathway involves recruitment of the adapters TRIF and TRAM, leading to activation of late-phase NF-κB and a second pathway involving activation of the transcription factor interferon (IFN) response factor 3 (IRF-3) and production of IFN-inducible cytokines such as interferon–inducible protein 10 (IP-10/CXCL10) (39). We investigated activation of this pathway in a macrophage-like cell line, differentiated THP-1 cells. Treatment of THP-1 cells with a low (10 ng/ml) and high (1 μg/ml) dose of LPS for 24 hours resulted in release of CXCL10 (Fig. 5A) and IL-1β (Fig. 5B; used to confirm SAPS was an effective inhibitor of LPS responses in THP-1 cells), which was significantly inhibited in the presence of SAPS. These data show SAPS potently inhibits both arms of the TLR4 signalling pathway and indicates it is most likely mediating its effects at the level of the receptor and/or adaptor complex.
The proinflammatory cytokine IL-1β also activates intracellular signalling via MyD88. Consequently, were SAPS to be acting at the level of the adapter complex it would likely prevent cellular responses to IL-1β. However, SAPS had no effect on IL-1β-induced CXCL8 release from cocultures (Fig. 6A), or the control monocultures (Fig. 6B,C). These data demonstrated that inhibition of LPS signalling was likely to be specific, was not mediated through non-selective disruption of cytokine processing or release, and was occurring at the level of the TLR4 complex.
We examined whether addition of exogenous IL-1β could restore LPS-induced coculture responses in the presence of SAPS. We found that dual stimulation of our coculture model with LPS and IL-1β produced an additive effect on CXCL8 release (Fig. 6D). Dual stimulation with exogenous IL-1β rendered SAPS a less effective inhibitor of coculture activation, suggesting that IL-1β could partially restore CXCL8 release. However at 50 μg/ml, SAPS completely inhibited LPS responses and CXCL8 release was reduced to the level produced by IL-1β alone (Fig. 6D).
LPS stimulation results in segregation of TLR4 into membrane microdomains (7, 8). We hypothesised that exposure of cells to SAPS would result in altered membrane biology and disturb microdomain function. We investigated this possibility using a FRET-based approach to determine consequences of SAPS on TLR4/CD14 association. FRET was measured in terms of dequenching of donor fluorescence after complete photobleaching of the acceptor fluorophore. Increased donor fluorescence, after complete destruction of the acceptor, indicated that the donor fluorescence was quenched in the presence of the acceptor because of energy transfer. The energy transfer efficiency of the system was tested using, as a positive control, the energy transfer from Cy3-26ic and Cy5-MY4 (mAbs to two different epitopes on CD14), which showed that the maximum energy transfer efficiency (E) was 37±1.5% in unstimulated monocytes, which was not affected by treatment with SAPS or LPS, alone or in combination (Table I). A negative control comprising Cy3-W6/32 (the mAb specific for MHC class I) and Cy5-cholera toxin (which recognises GM-1 ganglioside; a raft-associated lipid) was also used, which revealed no significant energy transfer in any of our 4 experimental groups (Table I).
Previously studies have shown that CD14 is situated within membrane microdomains in a resting state, whilst TLR4 moves into these domains only upon stimulation with LPS (32), and that LPS triggers a physical association between CD14 and TLR4 (40). In this study we confirmed these findings and reveal that SAPS disrupts this association. The results revealed large dequenching between CD14/GM-1 ganglioside in control cells confirming CD14 resides in membrane microdomains and treatment with LPS has no effect whilst SAPS disrupt this association both in the presence and absence of LPS (Table I). As expected there is no association between TLR4 and either GM-1 ganglioside or CD14 in the resting cell, but upon LPS stimulation a large dequenching occurs between TLR4/GM-1 and TLR4/CD14 revealing that TLR4 concentrates in microdomains after LPS stimulation and associates with CD14 within the membrane. Finally, monocytes stimulated with LPS in the presence of SAPS show reduced energy transfer in both cases, indicating that SAPS causes disassembly of these microdomains and a consequent inability of TLR4 and CD14 to associate (Table I).
A likely role for membrane microdomains in TLR4 signalling is established, but the importance of these domains in signalling of other TLRs is less well explored, and the potential for membrane disruption to affect highly geared inflammatory models such as our coculture systems is equally uncertain. We therefore explored the consequences of SAPS on inflammatory responses induced by a range of TLR and non-TLR agonists. We selected TLR agonists signalling from the cell surface and from endosomes, in order to gain insights into the dependence of TLR signalling on membrane microdomains.
We first determined the extent to which maximal epithelial cell responses to TLR and non-TLR agonists depended on cooperative signalling. We stimulated cocultures of BEAS-2B and PBMCs, and their monoculture controls, with Pam3CSK4 (TLR2/1), gardiquimod (which signals from acidified endosomes, activating TLR7>TLR8), and the proinflammatory cytokines TNFα and IL-1β (Fig. 7). Epithelial cells alone did not respond to Pam3CSK4 or gardiquimod, although a response to the cytokines TNFα and particularly IL-1β was observed. In contrast, PBMCs showed a modest response to the TLR agonists but not the cytokines. However, when the same numbers of cells were cultured together a significantly enhanced response to all four agonists was observed, demonstrating that tissue cells and infiltrating leukocytes will interact and cooperate in response to both exogenous pathogens and endogenous proinflammatory cytokines.
We then investigated the potential for such systems to be disrupted by SAPS. There is some evidence for a role for CD14 in TLR2 signalling (41-43) and we therefore explored whether SAPS would disrupt TLR2 responses. We found that SAPS dose-dependently and significantly inhibited IL-1β release from epithelial cell/PBMC cocultures stimulated with Pam3CSK4 (Fig. 8A). As with responses to TLR4, the ability of SAPS to inhibit activation of cocultures was again likely due to actions on the PBMCs, since SAPS also inhibited Pam3CSK4-induced IL-1β release from PBMCs monoculture controls, whilst the BEAS-2B cells alone showed no response to Pam3CSK4 (data not shown). Similar results were obtained for Pam3CSK4-induced CXCL8 release (data not shown).
We then determined whether SAPS disrupted the ability of TLR2 to associate with its signaling partners in membrane microdomains. A large dequenching was observed between TLR4/CD14 after LPS (Table (TableII and andII),II), but not after LTA (Table II), stimulation, and SAPS disrupted this association (Table II). In contrast, TLR2/CD14 associated after LTA, but not LPS, stimulation, and treatment with SAPS prevented this association (Table II). Treatment of monocytes with LTA also resulted in association of TLR2/GM-1 ganglioside and TLR6/GM-1 ganglioside, revealing that these receptors both concentrated within the microdomains of activated cells. These associations were disrupted by treatment of cells with SAPS (Table II).
Finally, S.pneumoniae is a leading cause of invasive bacterial disease capable of activating many TLRs and other pattern recognition systems. Notably, SAPS is also capable of significantly inhibiting heat-killed S.pneumoniae-induced IL-1β release from PBMCs (Fig. 8B). Similar results were obtained for S.pneumoniae-induced CXCL8 release (data not shown).
We subsequently determined if SAPS also disrupted signalling from intracellular compartments in tissue cells. In contrast to TLR2, TLR7/8 is exclusively expressed intracellularly and signals from acidified endosomes (44, 45). Epithelial cell/PBMC coculture responses to a TLR7/8 agonist, gardiquimod, were also dose-dependently and significantly inhibited by SAPS (Fig. 9A), which again was due to actions on the PBMCs (data not shown). As TLR3 is expressed by tissue cells but not PBMCs (22), we studied the effects of SAPS on polyI:C (a synthetic dsRNA analogue) induced CXCL8 release from epithelial cell monocultures. Only the highest dose of SAPS (50 μg/ml) reduced CXCL8 release (Fig. 9B). The control compound polydI:dC did not cause cytokine release, except in combination with the highest dose of SAPS, which is in keeping with the modest proinflammatory effects of SAPS alone that we observed previously (Fig. 1F).
We subsequently sought to compare the effects of SAPS as a potent disruptor of membrane microdomains in a further model system, using our established model of vascular inflammation in which human umbilical vein endothelial cells (HUVECs) are cultured with purified monocytes (22). In this study we showed that IL-6 release from LPS activated HUVEC/monocyte cocultures was dose-dependently inhibited by SAPS (Fig. 10A). The inhibitory effect of SAPS in cocultures appears to be through actions on both cell types since SAPS also inhibited LPS-induced IL-6 release from monocytes (Fig. 10B) and HUVECs (Fig. 10C) alone. Similar results were obtained for inhibition of LPS-induced IL-1β and CXCL8 release by SAPS (data not shown).
The work of this manuscript describes a number of important observations. We provide further evidence for the crucial role of cooperative networks in the mounting of effective inflammatory responses to TLR agonists. We identify a novel, non-toxic inhibitor of TLR responses that exerts broad-ranging inhibition of TLR, but not IL-1β, signalling, and show that this is mediated via disruption of membrane microdomains, extending our knowledge of the potential role of these domains in TLR signalling, including in TLRs located and signalling from the endosome. Disruption of membrane function nonetheless does not induce cell death or impair cytokine release, demonstrating that microdomain disruption has the potential to be a non-toxic strategy to intervene in inflammatory signalling.
The majority of in vitro studies of TLR responses consider actions of TLRs on single cell types. In recent publications we have shown that signalling to several TLR agonists is most effectively mediated by cooperative responses (22, 23). In particular, we have defined cooperative networks in which leukocytes, principally the monocyte, can use tissue cells to amplify inflammatory responses, and in which engagement of the tissue cell in the innate immune response is dependent upon such networks. In these systems, early production of IL-1 from the monocyte appears essential for effective tissue cell activation (23). Such networks enable effective tissue responses to agonists to which the tissue cells may be markedly less responsive.
Illustrating these concepts, here we show that epithelial cells are unresponsive to a TLR7/8 agonist, consistent with the lack of expression of this receptor seen in other epithelial cell lines (46). However, in the presence of PBMCs a profound response to the TLR7/8 agonist is observed as defined by production of proinflammatory cytokines. These data are similar to responses we have previously observed in airway smooth muscle (22), and reinforce the generic dependence of innate immunity on such networks. Even where tissue cells are responsive to TLR agonists such as LPS, we show again here (as previously, (23)), that signalling is profoundly greater when cooperative networks are established, whether the tissue cells are epithelial or endothelial in origin. Evidence for the in vivo relevance of such systems is accumulating (47).
Despite the high gearing of these networks, they are nonetheless amenable to therapeutic targeting. We have shown previously that IL-1 appears to play a crucial role in initiating inflammatory responses (23). Here, we identify another way to manipulate these systems yielding important insights into TLR function and the role of membrane microdomains in cell signalling.
We hypothesised initially that PS species would inhibit LPS-mediated activation of cocultures through antagonism of LPS binding to CD14. We explored this hypothesis using a specific PS species, SAPS, which is found in plasma in vivo (48). In keeping with our first hypothesis, we found that SAPS inhibited TLR4-mediated coculture activation and responses of individual cells, including freshly isolated human PBMCs, differentiated THP-1 cells (a macrophage model), and endothelial cells. In contrast, SAPS did not inhibit the actions of IL-1β in cocultures of epithelial cells and PBMCs, or in either cell type alone. Furthermore, we found that the addition of exogenous IL-1β could partially restore the inflammatory response induced by LPS even in the presence of SAPS, although this was evident only at lower SAPS concentrations (1 and 10 μg/ml). This suggests that whilst IL-1β is a crucial communicator between leukocytes and tissue cells other factors are also required to convey full activation of the LPS-induced inflammatory response, and at higher doses of SAPS (50 μg/ml) production of these additional factors are abolished.
We then sought to identify the mechanism of action whereby SAPS mediated its inhibitory effects on proinflammatory cytokine release. We found that this was not through production of the anti-inflammatory cytokine, IL-10. Nor did SAPS cause downregulation of expression of the TLR4 signalling complex, or death of either tissue cells or PBMCs. Although TLR2 signalling may also show some CD14 dependence (41-43), we were surprised by the degree to which SAPS abolished this. Our finding that SAPS inhibited both the MyD88-dependent and -independent signalling pathways equally, strongly suggested SAPS was acting at the level of the receptor or its immediate signalling complex. However, the lack of actions on IL-1β induced responses, which also utilise the MyD88 adaptor, suggested actions at the adaptor level were unlikely. We therefore hypothesised that SAPS might be directly influencing membrane dynamics. We found that SAPS disrupted the association of both TLR4 and TLR2 with their respective membrane partners that are required for signalling. These data are in keeping with observations showing that TLR2 and 4, but not IL-1β, signalling is dependent upon membrane microdomains, and that after stimulation, TLRs 2 and 4 associate with many proteins within these domains (8, 32, 40, 49-53). Importantly, the cytokine secretion observed in response to IL-1β revealed that membrane microdomain disruption did not result in complete cellular paralysis, and the lack of toxicity of SAPS with no evidence of impairment of cellular metabolic function suggests that normal uptake mechanisms for environmental sampling and nutrient uptake are unimpaired.
Some oxidised phosphocholine (PC) phospholipids have shown similar properties, with disruption of TLR2 and TLR4 signalling and impairment of TLR4 translocation to lipid rafts/caveolar fractions of endothelial cells (ECs) (54). However the inhibitory effects appear to require oxidation of the molecule as the unoxidised form had no effect. It is well documented that oxidised phospholipids induce ECs to interact with monocytes via the production of monocyte-specific chemoattractants (55, 56), adhesion molecules (57), and colony stimulating factors (58). In contrast, SAPS did not require oxidation to function as a potent inhibitor and we observed only minimal activation of epithelial cells and none of ECs by SAPS, nor did SAPS cause activation of monocyte/EC cocultures. Interestingly, however, the inhibitory oxPC phospholipids described by Walton et al (54) also shared an arachidonoyl group at a similar position on the molecule to SAPS (54). Recent evidence that the nature of the acyl chain structure of membrane lipids is important in the regulation of lipid raft stability (59), suggests that further exploration of the biology of such molecules is warranted. Of note, PS derivatives including SAPS have also been shown to affect secondary antibody production in brown Norway rats previously sensitized to ovalbumin, and though the mechanism remains to be explained, variations of the fatty acid groups altered the subsequent immune response (60).
In addition, our work revealed a much more profound inhibition of TLR-dependent signalling than just effects on TLR2 and TLR4. Whilst inhibition of individual TLR responses is of interest, it is striking that disruption of membrane microdomains also inhibited responses to heat-killed whole bacteria, which have the potential to engage multiple pattern recognition systems (61). Of particular note, we observed that signalling of a TLR7/8 agonist was markedly impaired in PBMCs and cocultures. This receptor signals solely from acidified endosomes, as does TLR3. In contrast to TLR7/8 signalling, however, SAPS had little impact on TLR3 signalling in epithelial cells. SAPS was active at inhibiting LPS responses in endothelial cell monocultures, and therefore this difference between inhibition of TLR3 and TLR7/8 is unlikely to be a result of global differences between effects of SAPS on leukocytes and tissue cells. The imidazoquinolines such as gardiquimod are small molecules for which no specific uptake mechanism has been described, thus it is unlikely that SAPS interfered with gardiquimod uptake, and global impairment of cellular environmental sampling is unlikely since (a) responses to the much larger, but also intracellular-acting molecule, poly(I:C) were preserved, and (b) the drug had no effect on cellular metabolic capacity. These data reveal that PS species may act as effective membrane disruptors in multiple cellular compartments including the plasma membrane and endosome. Though the role of lipid rafts/membrane microdomains in the signalling of TLRs recognising viral RNAs has received relatively little attention to date, our results are in keeping with evidence that TLR7/8 shows a greater dependence on membrane microdomains for its signalling than does TLR3 (62, 63). We have not determined the contribution of RIG-I/mda5 (64) to the observed responses to poly(I:C), and if these receptors were still responsive in the face of TLR3 antagonism by SAPS, it is possible that TLR3 signalling has been impaired to a greater extent than our data suggest. Interestingly, oxidised PCs inhibit TLR2 and 4 responses but have no effect on responses to TNFα (54), whose receptor can also reside in lipid rafts (65, 66). To what extent such responses are underpinned by cell-specific signalling systems or interactions of varying phospholipids with specific membrane microdomains showing some potential selectivity for TLRs requires further investigation.
In conclusion, we show here that effective signalling to a broad range of TLR agonists is mediated by cooperative networks that can be inhibited by a non-toxic phospholipid acting to disturb membrane microdomain formation, and which therefore has the potential to be useful in therapeutic interventions aimed at controlling acute inflammatory responses. The observation that SAPS is one of a variety of forms of PS found in membranes and tissues raises the possibility that its presence in plasma acts as a previously unrecognised autoregulatory anti-inflammatory activity.
We would like to thank Miss Kathryn Vaughan and Miss Emily Dick for preparation of the primary human PBMCs, Sue Newton for help with the CBA assay, and Professors Steven K. Dower and Moira K. Whyte for helpful discussions.
This work was funded in part by a grant from the Sheffield Hospitals Charitable Trust. I.S. is supported by a MRC Senior Clinical Fellowship (G116/170). K.T. is supported by the Wellcome Trust. J.R.W. is supported by a British Heart Foundation project grant (FS/06/004). This work was also supported by Allergy Therapeutics plc. and Vaccine Technology Ltd. who provided the SAPS (PCT publication number WO 2008/068621).