Fluorescence recovery after photobleaching reveals that LPS rapidly transfers from CD14 to hsp70 and hsp90 on the cell membrane

The recognition of bacterial lipopolysaccharide (LPS) or endotoxin by the innate immune system provokes a strong inflammatory response, triggering the production and release of potent inflammatory mediators. Although Gram-positive and Gram-negative bacteria differ in the composition of their outer cell wall, the host reactions to them are similar. In a severe bacterial infection caused by either form of bacterium, inflammatory mediators are produced and released in excess and can lead to haemodynamic shock, accompanied by symptoms such as fever, tachycardia, tachypnoea and possibly ending in multiple organ failure and sometimes death (Bone, 1991).

The outer cell wall of Gram-negative bacteria, which is composed of LPS, appears to be the cause of the immune activation. Our current understanding of the innate immune recognition of Gram-negative bacteria is based on the seminal discovery of LPS-binding protein (LBP) (Tobias et al., 1986) and the elucidation of its structure and function (Wright et al., 1989; Schumann et al., 1990), followed by the identification of CD14 as an LPS receptor (Wright et al., 1990). CD14 exists as a membrane bound protein (mCD14) and also as a soluble form (sCD14) (Durieux et al., 1994) that binds LPS. Complexes of soluble CD14 (sCD14) and LPS bind to a yet unknown receptor and activate cells of non-myeloid lineage (Frey et al., 1992; Labeta et al., 1993; Pugin et al., 1993; Lee et al., 1993; Demters et al., 1994). It has been suggested that the same signalling receptor might be involved in both mCD14-LPS and sCD14-LPS complexes. CD14 has also been implicated in mediating inflammatory responses to bacterial peptidoglycan (PGN) and lipoteichoic acid (LTA) (Weidemann et al., 1994; Gupta et al., 1996), which are the major components of the cell wall of Gram-positive bacteria.

However, because CD14 is a glycosylphosphatidylinositol (GPI)-linked protein and does not transverse the cell membrane, it has been suggested that it cannot deliver a signal for activation against LPS. It has been suggested that CD14 mediates LPS responses by interacting with other signal transducing molecules (Ingalls and Golenbock, 1995; Ingalls et al., 1997; Pugin et al., 1994; Gegner et al., 1995). The theory of the existence of multiple LPS receptors has been further strengthened by several lines of evidence, which have demonstrated that CD14 blocking monoclonal antibodies (mAbs) only partially inhibit LPS binding (Lynn et al., 1993; Shapira et al., 1994; Blondin et al., 1997; Troelstra et al., 1997; Triantafilou et al., 2000c). Furthermore, several reports have suggested the existence of LPS transducers, such as CD11c/CD18 (Ingalls and Golenbock, 1995), the human homologues of Toll (which are known as Toll-like receptor 2 (TLR2) (Kirschning et al., 1998) and TLR4 (Poltorac et al., 1998)), moesin (Tohme et al., 1999) and, most recently, heat shock proteins (Byrd et al., 1999), suggesting that the innate immune recognition of bacteria is highly complex.

Even though the exact mechanism whereby the innate immune system recognizes bacteria remains to be elucidated, the acquired immune response is less of a mystery. Monks et al. and Grakoui et al. have provided the first dynamic picture of the multimolecular choreography of receptors that takes place as the T cell engages an antigen-presenting cell (APC) (Monks et al., 1998; Grakoui et al., 1999). This revealed the molecular interactions and microdomains involved in APC-T cell engagement. It was shown that the APC-T cell contact zone, the immunological synapse, consisted of multiple receptors, which moved to different domains throughout the synapse while delivering different signals.

The innate immune recognition, being the more archaic form of host defence, lacks the complex interaction between two different cells but still possesses the multimolecular involvement of a variety of proteins and receptors. In this study, we set out to investigate whether the receptors involved in the innate bacterial recognition are also part of a complex immunological choreography involving multiple receptors and microdomains. It is possible that the central molecule involved in recognition of bacteria, CD14, being a GPI-linked protein, exists in detergent-insoluble glycolipid-enriched domains (DIGs) or GPI-microdomains, in which signalling molecules might accumulate following CD14 ligation by LPS.

In this study, fluorescence recovery after photobleaching (FRAP) was used in order to analyse the mobility of CD14 on the cell membrane before and after LPS stimulation. We report for the first time that the diffusion coefficient of CD14 remains virtually unchanged before and after LPS stimulation. Furthermore, we find that FITC-LPS and FITC-LTA, when bound to cells, have similar diffusion coefficients (which differ from that of CD14) in a fraction of cells and are completely immobile in the remainder. The results suggest that FITC-LPS/LTA must initially bind to CD14 but are subsequently rapidly transferred to a transducing molecule or complex of molecules that is immobile. Here, we present evidence that suggests that these molecules are heat shock proteins (hsps) 70 and 90. We find that these proteins are completely immobile on the plasma membrane and antibodies against them, when added prior to LPS stimulation, block the progressive immobilization of LPS and inhibit synthesis of interleukin 6 (IL-6) and subsequent cell activation. We thus propose that hsp70 and hsp90 are implicated in cell activation, probably as part of a multimeric transducing complex.

Rough LPS from Salmonella minessota Re595 was purchased from List Labs (Campbell, CA). All fine chemicals and human pooled serum were purchased from Sigma (St Louis, MO). Fluorescein isothiocyanate (FITC) and OG were purchased from Molecular Probes Europe (Leiden, The Netherlands). Hybridoma cells secreting 26ic (anti-CD14) was obtained from the American Type Culture Collection (ATCC, Rockville, MD). Monoclonal antibody, MY4, directed against the functional domain of CD14 was obtained from Coulter (San Francisco, CA). Recombinant human LBP (rLBP) (Theofan et al., 1994) and rsCD14 were provided by XOMA (Berkeley, CA) LLC. Hsp70 rabbit polyclonal serum was obtained from Dako (Cambridge, UK). Hsp90-specific rabbit polyclonal serum was obtained from Santa Cruz Biotechnology (Santa Cruz, CA). IL-6 enzyme-linked immunosorbent assay (ELISA) kit was obtained from Diaclone (Besancon, France).

Chinese hamster ovary (CHO) cells transfected with hCD14 cDNA in the expression vector pRc/RSV (CHO-CD14) were kindly provided by S. Viriyakosol and T. Kirkland (University of California) (Viriyakosol and Kirkland, 1995). CHO cells were maintained in DMEM/F12 (Dulbecco’s Modified Eagle’s Medium (DMEM)/Ham’s F12 1:1 mix) from Gibco BRL with 2 mM L-glutamine, 7.5% foetal calf serum (FCS), 500 μg ml⁻¹ gentamycin sulfate (G418, Sigma). Cells were grown in 80 cm³ tissue culture flasks (Nunc). Trypsin/EDTA (0.05% trypsin/0.53 mM EDTA) was used for passaging the cells.

The MonoMac 6 cell line was obtained from the Institute of Immunology, University of Munich, Germany. MonoMac 6 cells were cultured in 5% CO₂ at 37°C in Iscove modified Dulbecco medium (IMDM; Gibco BRL) containing 10% FCS.

The human umbilical vein endothelial cell line (ECV-304) was obtained from the ECACC (Cambridge, UK). The cells were maintained in Medium 199 supplemented with Glutamax^® (Gibco BRL) and 10% FCS. Cells were grown in 80 cm³ flasks (Nunc). Cultures of high viability were obtained by trypsinizing the cells in 0.25% trypsin/EDTA, followed by seeding the cells at a density of 1×10⁵ cells per 25 cm³ flask.

For FRAP experiments, the cells were seeded in Lab-tek slides (Nunc) at a density of 10,000 cells per well and cultured for a further 72 hours prior to the experiments.

R-phycoerythrin (PE)-Fab probe was prepared as previously described (Smith et al., 1998; Triantafilou et al., 2000b). Briefly IgG was purified and digested with papain. Aliquots of purified Fab fragments were labelled with PE, pyridyl disulphide derivative (Molecular Probes, Eugene, OR). Briefly, 1.25 mg of Fab was concentrated to 5 mg ml⁻¹ in a Centristart-1 microconcentrator (Sartorius). Thiol-reactive maleimide residues were introduced into Fab by incubating with 5 mM succinimidyl trans-4-(N-maleimidylmethyl) cyclohexane-1-carboxylate (SMCC) in dimethyl sulphoxide for 2 hours at room temperature. Excess SMCC was removed by extensive dialysis against 100 mM phosphate buffer, 150 mM NaCl pH 7.5.

In parallel, 2.0 mg of R-phycoerythrin, pyridyl disulphide derivative (average 2.2 pyridyl residues) was incubated with 50 mM dithiothreitol (DTT) for 15 minutes at room temperature. Excess DTT was eliminated by dialysis against 100 mM phosphate buffer, 150 mM NaCl pH 7.5. The PE was then incubated with the Fab for 20 hours at 4°C in the dark. After this time, in order to prevent Fab disulphide reduction or aggregation, a 20 M excess of N-ethyl maleimide was added.

Preparative size exclusion chromatography was performed on a Bio-Rad 5000T HRLC controlled by a 486 PC. 300 μl labelled Fab were loaded onto a Bio-Select SEC 250-5 column, eluted with 50 mM phosphate buffer, 150 mM NaCl pH 6.8 at 0.1 ml minute⁻¹ and 200 μl fractions collected. Integration was performed using ValueChrom^® integration analysis software (Bio-Rad). PE-Fab concentration was determined using a molar extinction coefficient of 1.96×10⁻⁶ M⁻¹cm⁻¹ for PE.

Antibodies against hsp70 and hsp90 were digested with papain overnight and purified in order to yield Fab fragments. Aliquots of purified Fab, 400 μg (1 mg ml⁻¹) were mixed with 20% by volume of sodium bicarbonate buffer (1 M, pH 8.0) and 40 μl of freshly prepared OG 488 succinimide dissolved in DMSO (10 mg ml⁻¹). The mixture was placed in the dark for 1 hour at room temperature. The reaction was terminated by adding ∼10% by volume of hydroxylamine (1.5 M, pH 8.5) and incubated for a further hour in the dark. The conjugate was subsequently separated from the unlabelled Fab and dye by passing the mixture through a PD-10 column that had previously been equilibrated with PBS.

LPS or LTA were labelled with FITC according to a previously described protocol (Troelstra et al., 1997). Briefly, LPS/LTA (4 mg) was made monomeric by treatment with 2 ml of 0.5% triethylamine (Sigma) and sonication for 15 minutes on ice. After the sonication, 200 μl of 100 mM EDTA (BDH,) were added. The pH was then adjusted to 5 by adding 10 μl of 1 N HCl. 800 μl of 0.25 M borate buffer (pH 10.5) containing 20 mg FITC (Molecular Probes) were added to the LPS solution. The mixture was sonicated again for 1 minute and, after the addition of 1 ml 1.6% sodium deoxycholate (Sigma), was incubated for 18 hours at 37°C while rotating. Aggregates were pelleted by centrifuging the mixture at 10,000 g and the supernatant was concentrated in a dialysis bag with polyethylene glycol 6000 (BDH). Afterwards, the conjugate was dialysed against PBS for 1 hour and passed through a PD-10 column (Pierce) in order to separate the FITC-LPS/LTA from the free FITC molecules. The fractions containing FITC-LPS/LTA were pooled, concentrated and dialysed against PBS. The concentration of FITC in the final preparation was determined spectrophotometrically at 492 nm, with an extinction coefficient for FITC (E₄₉₂) of 8.5×10⁴ M⁻¹ cm⁻¹. To determine the amount of LPS, the level of 2-keto-3-deoxyoctulosonic acid (KDO) was measured by the thiobarbiturate assay as described previously (Brade and Galanos, 1983). The FITC-LPS probe was assayed for its KDO content in order to estimate the LPS concentration and the absorbance at 492 nm was used to determine the FITC content. KDO and FITC data were used to calculate the labelling ratio. This revealed that a high degree of efficiency had been achieved, yielding a 1:1 labelling ratio for LPS.

The binding of FITC-LPS/LTA to CHO cells transfected with wild-type CD14 was measured by flow cytometry. CHO cells, 3×10⁵ per sample were incubated with increasing concentrations of FITC-LPS or FITC-LTA (1–250 ng) in the presence and absence of 5% human pooled serum (HPS). Samples were incubated for 30 minutes at 22°C, gently mixed. For blocking experiments, cells were first incubated with 1 μg MY4 (Coulter) mAb directed against a functional domain of CD14. Subsequently, the cells were washed three times in PBS before cytometric analysis. The mean fluorescence of 10,000 cells was measured in a flow cytometer (FACalibur^®, Becton Dickinson). The cells were not gated.

500,000 cells were harvested and washed twice with PBS, followed by incubation with PBS, 0.02% bovine serum albumin (BSA), 0.02% NaN₃ for 30 minutes. The probe was added for 1 hour at room temperature. To remove the unbound probe, the cells were washed twice with PBS, 0.02% BSA, 0.02% NaN₃. Finally the cells were washed twice with PBS, 0.02% donor calf serum (DCS), 0.02% NaN₃. The final cell pellet was resuspended in 500 μl of PBS prior to flow cytometric analysis.

CHO-CD14, MonoMac 6 and ECV-304 cells were allowed to adhere on eight-well microchamber slides for 72 hours at 37°C. Afterwards, cells were washed twice in PBS prior to labelling with PE-Fab or OG-Fab for 30 minutes at 22°C or FITC-LPS/LTA for 30 minutes at 22°C in serum-free medium supplemented with 5% HPS or 1 μg ml⁻¹ of LBP and 1 μg ml⁻¹ of rsCD14. To avoid aggregation, FITC-LPS/LTA probes were sonicated prior to cell labelling. After washing with PBS and serum-free medium, the cells were imaged at the desired temperature using a temperature-controlled microscope stage.

FRAP measurements were performed as previously described (Ladha et al., 1994; Ladha et al., 1996). Briefly, slides containing labelled cells were placed onto a temperature controlled microscope stage (Physitemp, Clifton, NJ; Model TS-4) and allowed to equilibrate to the desired temperature for FRAP measurements. After equilibration, the beam of an argon ion laser (Innova, Palo Alto, CA; 100-10) was focused onto the desired area on the cell. The laser beam was of Gaussian cross-sectional intensity, with a half-width at 1/e² height of the laser beam at its point of focus equal to 2.15 μm spot radius. FRAP measurements were recorded and analysed as previously described (Ladha et al., 1996).

MonoMac 6 cells (5×10⁵) were mixed with 50 μl of serial dilutions of LPS in the presence of 1% HPS in serum-free medium. After 2.5 hours, the supernatant was collected and analysed for IL-6 using an ELISA (Diaclone, France).

In order to study the mobility of CD14 on the cell surface before and after LPS stimulation, a fully characterized probe of small size and high fluorescent yield was prepared. Phycobiliprotein-Fab conjugates were prepared from PE and Fab fragments of the mAb 26ic. This antibody, which is specific for CD14, was chosen because it binds to a non-functional domain of CD14 that is distinct from the LPS binding domain (Wright et al., 1990). Therefore the binding of the LPS to CD14 would not interfere with the binding of the probe.

Fab fragments of the mAb 26ic were conjugated in a 1:1 molar ratio to PE. The PE-Fab conjugate, which had a molecular mass higher than either of the two starting materials, was purified by HPLC size exclusion chromatography. Using HPLC the PE-Fab conjugate was eluted first, followed by the free PE and Fab. Fig. 1A shows a typical HPLC profile obtained with a PE-Fab probe, in which three major peaks are shown corresponding to PE-Fab (1), free PE (2) and unlabelled Fab (3). The arrow indicates the 1:1 PE-Fab fractions collected for flow cytometric and FRAP experiments.

The 1:1 PE-Fab was then tested for specificity and binding to the CD14 receptor on CHO-CD14 cells (Fig. 1B-E). Flow cytometric analysis revealed a high peak fluorescence intensity when 4 pmol of the 1:1 PE-Fab was used (Fig. 1C), when compared with the fluorescence intensity of unlabelled cells (Fig. 1B). Nonspecific binding of the PE was checked by incubating cells with 4 pmol of unconjugated PE, and peak fluorescence intensity was reduced by 94% (Fig. 1D). Using flow cytometry, the binding of the PE-Fab conjugate was also compared with that of the PE-Fab conjugate in the presence of unlabelled LPS. CHO-CD14 cells were incubated with 10 ng ml⁻¹ unlabelled LPS prior to labelling with the 26ic PE-Fab probe. It was shown that the binding of the PE-Fab probe in the presence of unlabelled LPS was the same as that of the PE-Fab conjugate (Fig. 1E), demonstrating that the binding of LPS to the CD14 molecule does not interfere with the binding of the PE-Fab probe.

In order to investigate the mobility of CD14 before and after LPS binding, FRAP experiments were performed using the 26ic PE-Fab probe that we had constructed. Initially, we investigated the mobility of CD14 before LPS stimulation on CHO-CD14 cells. It was found that, at 22°C, CD14 had a diffusion coefficient of (1.6±0.9)×10⁻⁹ cm² s⁻¹ with 56±15% of the fluorescence recovered after photobleaching (Fig. 2A). At physiological temperature (37°C) the diffusion coefficient of CD14 was (4.5±3.0)×10⁻⁹ cm² s⁻¹ with 75±10% of the fluorescence recovered after photobleaching.

Because monocytes are of particular importance in the host’s response to LPS during endotoxaemia, MonoMac 6, a human monocytic cell line that expresses mCD14, was also used to study the mobility of CD14 before and after LPS stimulation. It was found that, at 22°C, CD14 had a diffusion coefficient of (1.35±0.6)×10⁻⁹ cm² s⁻¹ (Fig. 2B), almost identical to that obtained with the CHO transfectants, although the fluorescence recovery was somewhat higher.

After CHO-CD14 cells were stimulated with 5 ng ml⁻¹ of unlabelled LPS in 5% HPS, the diffusion coefficient of CD14 remained virtually unchanged, (1.2±0.7)×10⁻⁹ cm² s⁻¹ with 51±13% of the fluorescence recovered after photobleaching (Table 1). Even when higher concentrations of LPS were used to stimulate the cells (250 ng ml⁻¹), the diffusion coefficient of CD14 remained essentially the same, (1.5±0.7)×10⁻⁹ cm² s⁻¹ with 50±8% fluorescence recovery (Fig. 2C). Similar results were obtained when MonoMac 6 cells were stimulated with 250 ng ml⁻¹ of LPS. The diffusion coefficient of CD14 on MonoMac 6 cells after stimulation with LPS was (1.5±0.6)×10⁻⁹ cm² s⁻¹ with 70±20% of the fluorescence recovered after photobleaching (Fig. 2D), very similar to the diffusion coefficient obtained prior to LPS stimulation. Furthermore, FRAP experiments performed on CHO-CD14 and MonoMac 6 cells with the 26ic Fab-PE probe in the presence and absence of unlabelled LPS revealed almost identical results (data not shown), suggesting that the binding of LPS did not affect the lateral diffusion of CD14 in the lipid bilayer.

It has been suggested that CD14 might also function as a cellular receptor for peptidoglycan or LTA from Gram-positive bacteria, and so we investigated whether stimulation with LTA would affect the mobility of CD14. Before and after LTA stimulation (250 ng ml⁻¹), the CD14 lateral diffusion coefficient remained essentially the same. On CHO-CD14 cells, CD14 exhibited diffusion coefficients of (1.9±0.8)×10⁻⁹, with 80±16% of the fluorescence recovered after photobleaching (Fig. 2E). After LTA stimulation on MonoMac 6 cells, CD14 exhibited diffusion coefficients of (1.6±0.5)×10⁻⁹, with 75±12% of the fluorescence recovered after photobleaching (Fig. 2F). Table 1 summarizes the results obtained from FRAP experiments using the 26ic Fab-PE probe.

In order to test whether LPS remains bound to CD14, we decided to label the ligand itself and to use it as a probe for FRAP measurements. Using a protocol described previously (Troelstra et al., 1997), we prepared an FITC-conjugated LPS of high fluorescent yield that enabled us to detect low concentrations of LPS. This conjugate had previously been shown to be biological active, by testing its ability to induce TNF-α production (Triantafilou et al., 2000a).

The FITC-LPS probe was then tested for specificity and binding to CHO-CD14 cells (Fig. 3). Flow cytometric analysis revealed a high peak fluorescence intensity when 10 ng ml⁻¹ of FITC-LPS were used (Fig. 3B) compared with the fluorescence intensity of unlabelled cells (Fig. 3A). Nonspecific binding of the FITC-LPS probe was checked by incubating cells with a 100-fold excess of unlabelled LPS, in which peak fluorescence intensity was almost identical to that of the unlabelled cells (Fig. 3C).

In order to examine the mobility of LPS by FRAP, CHO-CD14 cells were labelled with 10 ng ml⁻¹ of FITC-LPS in serum-free medium supplemented with either 5% HPS or 1 μg ml⁻¹ of rLBP and 1 μg ml⁻¹ rsCD14. Surprisingly, it was found that, in 60% of the cells examined, LPS was completely immobile and, in 40% of the cells examined, LPS had a diffusion coefficient of (1.1±0.3)×10⁻⁸ cm² s⁻¹ with 53±10% recovery, much different to that observed for CD14 (Fig. 4A). In order to verify that the fast diffusion coefficient observed in 40% of the cells was not free FITC-LPS in solution, FRAP measurements of the background were performed. It was shown that there was almost no free FITC-LPS in solution and there was no photobleaching, suggesting that the background did not contribute to the measurement (data not shown). Furthermore, when the lateral diffusion of FITC-LPS bound to CHO-CD14 was measured at physiological temperature, it was found that the FITC-LPS was completely immobile (Fig. 4B). Overall, these results suggested that the FITC-LPS was not bound to CD14 but to some other receptor or complex of receptors with completely different diffusion characteristics.

Similar results were also obtained with MonoMac 6 cells. At 22°C, in 10% of the cells examined, LPS was completely immobile, whereas the remaining 90% of the cells examined exhibited a diffusion coefficient of (3.5±2.0)×10⁻⁹ cm² s⁻¹, with 52±18% of the fluorescence recovered after photobleaching (Fig. 4C), a diffusion coefficient similar to that of CD14. This suggests that, in 90% of the cells, FITC-LPS might still be associated with CD14, whereas FITC-LPS had passed from CD14 to another receptor or a complex of receptors in 10% of the cells examined. At physiological temperature, in 50% of the cells examined, LPS was completely immobile, whereas the remaining 50% of the cells examined exhibited a diffusion coefficient of (4.7±2.0)×10⁻⁹ cm² s⁻¹, with 39±8% of the fluorescence recovered after photobleaching (Fig. 4D).

In order to confirm that in the population of cells that LPS was mobile, the diffusion coefficient obtained was due to the binding of FITC-LPS to a receptor other than CD14, we set out to investigate the lateral mobility of LPS once bound to cells which lacked mCD14. Human umbilical vein endothelial cells (ECV-304), which lack mCD14, were used for these studies. ECV-304 cells were labelled with 10 ng ml⁻¹ of FITC-LPS probe in serum free medium supplemented with either 5% HPS or 1 μg ml⁻¹ LBP and 1 μg ml⁻¹ rsCD14 prior to FRAP measurements. We found that at 22°C the diffusion coefficient of FITC-LPS bound to ECV-304 cells was (4.4±2.0)×10⁻⁹ cm² s⁻¹ with a 48±7% recovery of fluorescence after photobleaching (Fig. 4E). By contrast, at physiological temperature, in 40% of the cells, LPS was completely immobile, whereas in the remaining 60%, LPS had a diffusion coefficient of (6.1±2.6)×10⁻⁹ cm² s⁻¹ with a recovery of 62±3% (Fig. 4F). These results indicate that the immobilization of LPS observed earlier on the CHO-CD14 and MonoMac 6 cells at 37°C, was a common event that occurs after LPS ligation whether or not mCD14 is present.

Even though it has been previously demonstrated by a variety of experimental tests that the FRAP technique does not generate photobleaching artefacts (Wolf et al., 1980), we decided to perform experiments in order to rule out the possibility that the immobility observed was due to potential FRAP artefacts, such as damage to the specimen by the photobleaching. To this end, we carried out experiments in which a second FRAP experiment was performed on the same spot. We found that the second time resulted in a recovery curve that had almost no immobile fraction, suggesting that the immobilization previously observed was not due to bleaching damage.

The FRAP measurements obtained with the 26ic PE-Fab probe and the FITC-LPS suggested that LPS was probably binding to an immobile receptor or to a complex of receptors. In order to find out whether LTA from Gram-positive cells also binds to the same receptor, we constructed a FITC-LTA probe. Prior to using this probe for FRAP measurements, we tested its binding and specificity using flow cytometry. We found that, when CHO-CD14 cells were incubated with 10 ng ml⁻¹ of FITC-LTA, a high fluorescence intensity was observed (Fig. 5B) compared with the unlabelled cells (Fig. 5A). The specificity of the probe was further tested by incubating the cells with 100× excess of unlabelled LTA, when it was observed that the binding was abolished (Fig. 5C).

Once the specificity of the probe was established, CHO-CD14 cells were labelled with 10 ng ml⁻¹ of FITC-LTA prior to FRAP experiments. It was shown that, when FITC-LTA was bound on CHO-CD14 cells, its diffusion coefficient at 22°C was (1.6±0.2)×10⁻⁸ cm² s⁻¹ with a 47±7% fluorescence recovery after photobleaching (Fig. 5D).

In 10% of the MonoMac 6 cells examined at 22°C, LTA was completely immobile, whereas in 90% of the cells examined, LTA exhibited a diffusion coefficient of (9.4±4.0)×10⁻⁹ cm² s⁻¹ with 65±8% of the fluorescence recovered after photobleaching (Fig. 5E). The diffusion coefficient of FITC-LTA when bound to ECV-304 cells at 22°C was (7.1±3.6)×10⁻⁹ cm² s⁻¹ with 73±5% of the fluorescence recovered after photobleaching (Fig. 5F).

At physiological temperature, FITC-LTA was completely immobile in 70% of the CHO-CD14 cells examined. In the remaining 30% of the cells examined LTA exhibited a diffusion coefficient of (1.2±0.2)×10⁻⁹ cm² s⁻¹ with 59±2% of the fluorescence recovered after photobleaching. When MonoMac 6 cells were used, LTA was completely immobile in 60% of the cells examined, whereas LTA displayed a diffusion coefficient of (1.4±0.2)×10⁻⁹ cm² s⁻¹ in 40%, with 62±3% of the fluorescence recovered after photobleaching. These results are similar to those obtained with FITC-LPS on CHO-CD14 and MonoMac 6 cells, suggesting that both LPS and LTA transfer to the same receptor.

FRAP measurements of the background in both cell types (i.e. away from the cells) showed much weaker fluorescence intensities and no photobleaching, verifying that the diffusion coefficients obtained were not free FITC-LTA probe (data not shown). FRAP data obtained using FITC-LPS/LTA probes are summarized in Table 2.

Our data indicated that FITC-LPS/LTA were probably binding to either an immobile receptor or a complex of receptors, and so we set out to reveal the identity of this receptor. Probes against integrin CD11c/CD18, hsp70 and hsp90 were generated by conjugating antibodies against these molecules to OG 488. All probes were characterized and tested for specificity by flow cytometry prior to using them for FRAP experiments (data not shown). Integrins such as CD11c/CD18, which had previously been suggested as LPS ‘signal transducers’, were found to be diffusing on the plasma membrane before and after LPS stimulation (data not shown). By contrast, hsp70 (Fig. 6A) and hsp90 (Fig. 6B), which had recently been implicated in mediating macrophage activation by taxol (Byrd et al., 1999), a plant-derived antitumour agent that elicits cell signals indistinguishable to LPS, were found to have either very low mobility of were found to be immobile on MonoMac 6 cells, thus suggesting that these proteins were probably the immobile receptor(s) that we had observed. The lateral diffusions of hsp70 and hsp90 were also measured on ECV-304 cells, which lack mCD14, yielding similar results (Fig. 6C,D). In order to determine whether hsp70 and hsp90 were the immobile receptors that LPS was transferred to, we investigated the effect of anti-hsp antibodies on the immobilization of LPS using FRAP. It was revealed that in the presence of anti-hsp70 and anti-hsp90 antibodies, LPS remained mobile at 37°C, with a diffusion coefficient of (3.3±0.1)×10⁻⁹ cm² s⁻¹ and recovery of 78±4%, suggesting that hsp70 and hsp90 were the immobile receptors that LPS was transferring to.

Our FRAP data suggested that the heat shock proteins were the immobile receptor that we were observing, and so antibody inhibition experiments of IL-6 induction were performed in order to determine the roles of hsp70 and hsp90 in the biological response to LPS. IL-6 secretion was measured after LPS stimulation. MonoMac 6 cells were pre-incubated with MY4 (CD14 specific blocking mAb), anti-hsp70, anti-hsp90 or an irrelevant polyclonal serum and an isotype control. It was found that antibodies against hsp70 and hsp90 inhibited the response by 85%, whereas a combination of hsp70 and hsp90 antibodies completely inhibited IL-6 secretion (Fig. 7A), strongly suggesting that hsp70 and hsp90 are involved in LPS signal transduction. Similar results were obtained with ECV-304 cells, which lack mCD14 (Fig. 7B).

The recognition of foreign antigens is the first and most crucial step in both innate and adaptive immunity. Even though the innate immune recognition is not well understood, the molecular interactions involved in the acquired antigen recognition have been revealed. It is generally accepted that the acquired immune response involves the engagement of two types of cells, APCs and T cells. Recently it was revealed that the interaction between the two is more complex than previously imagined. The APC-T cell interaction was shown to be an elaborate choreography involving several receptors that rapidly move to different microdomains on the cell surface in order to initiate signal transduction (Monks et al., 1998; Grakoui et al., 1999).

Such an elaborate choreography must have taken many years of evolution in order to achieve perfect synchronization and precision. Somewhere along the evolutionary trail, the ‘building blocks’ of this interaction must still be conserved as components of the most ancient system of host defence - the innate immune recognition. If the ‘building blocks’ of the acquired immune recognition lie in the innate, then this must involve choreography of several receptor molecules and microdomains, possibly in a more simplified form.

This hypothesis becomes more probable because CD14, a key molecule in the innate bacterial recognition, is a GPI-linked protein. A characteristic feature of the GPI-anchored proteins is that they are found in microdomains (or lipid rafts). These microdomains can be viewed, with some simplification, as small semiliquid islands floating in the more liquid, phospholipid-rich cell membrane (Horejsi et al., 1999). It has been shown that these GPI domains also contain prominent signalling molecules. Because CD14 is found in such microdomains on the cell surface, it is possible that the entire bacterial recognition is based around the ligation of CD14 by bacterial components and the recruitment of multiple signalling molecules at the site of CD14-LPS ligation.

In this study, we set out to reveal the first dynamic picture of bacterial recognition, by investigating the lateral diffusion of immune receptors involved in the innate recognition of bacteria. FRAP was used to observe the lateral diffusion of CD14 before and after LPS stimulation, as well as the lateral diffusion of receptor molecules to which FITC-LPS and FITC-LTA are bound. To achieve this, we constructed two types of probes. One used an antibody against CD14 conjugated to PE, the other used the ligand itself (either LPS or LTA) labelled with fluorescein.

We found that the lateral diffusion of CD14 on the lipid bilayer of the cell membrane of CHO-CD14 and MonoMac 6 cells remained virtually unchanged after LPS stimulation, suggesting that the binding of its ligand does not affect its mobility. However, when the labelled ligand (FITC-LPS) was used as a probe, its lateral diffusion was distinct from that of CD14. At 22°C, FITC-LPS was immobile on 60% of CHO-CD14 cells but, on the remaining 40% of cells, the diffusion coefficient was much higher than for CD14. At 37°C, FITC-LPS was immobile on all cells. This result shows that at the time of the measurement, typically 30 minutes after the incubation of cells with the probe, FITC-LPS was no longer associated with CD14. This explains why a change in CD14 mobility was not observed upon incubation with LPS. Similar results were obtained with MonoMac 6 cells but the fraction of cells on which FITC-LPS was completely immobile was less than for CHO-CD14 cells, and the change in diffusion coefficient in the remaining cells was less marked. The high diffusion coefficient of 1.1×10⁻⁸ cm² s⁻¹ observed at 22°C in a fraction of CHO-CD14 cells is particularly interesting because it is a typical value for lipid diffusion. The results suggest that LPS detaches from CD14 and freely diffuses through the lipid bilayer before binding to a different immobile receptor or receptor complex. At 22°C, this process is only partially completed during the time taken to perform the FRAP measurements, but is fully completed at 37°C. In MonoMac 6 cells, the process may be slower so that complete immobilization is not observed at 37°C.

The experiments with FITC-LTA support the above conclusions. The transfer of LTA to another receptor appears to be slower on CHO-CD14 cells than for FITC-LPS but the results are qualitatively similar. The mobile fraction of FITC-LTA again exhibits high mobility typical of lipid diffusion.

FRAP analysis of FITC-LPS probe bound to ECV-304 cells revealed that the diffusion coefficient of the mobile component was similar to that observed in MonoMac 6 cells but slower than that observed on CHO-CD14 cells at 22°C. At 37°C, an immobile fraction had begun to be formed (40% of the cells). This suggested that we were observing the same events as the ones on CHO-CD14 and MonoMac 6 cells, but at a slower rate.

It is well established that CD14 is the initial receptor (Wright et al., 1990). Our observations however, suggest that LPS only briefly associates with CD14 before transferring to another receptor. Because integrins, such as CD11c/CD18 and hsps had previously been suggested as LPS signal transduction molecules, we decided to directly test the hypothesis that one of them could be the immobile receptor to which LPS and LTA are transferred. FRAP measurements of CD11c/CD18 revealed that it was diffusing on the plasma membrane (data not shown), whereas hsp70 and hsp90 were found to be completely immobile on the cell surface. This suggests that the latter could well be the immobile receptors to which LPS/LTA is transferred. Therefore, we decided to investigate the effect of anti-hsp70 and anti-hsp90 antibodies on the immobilization of LPS. We found that, once hsp70 and hsp90 were blocked, FITC-LPS remained mobile, suggesting that these were the receptors that LPS was transferring to. Furthermore, we decided to test whether hsp70 and hsp90 were involved in signal transduction by performing antibody inhibition experiments of IL-6 induction. It was found that incubation with polyclonal sera against these hsps, prior to LPS stimulation, completely inhibited IL-6 secretion, suggesting that hsp70 and hsp90 are part of the LPS multimeric receptor and are involved in LPS signal transduction.

Previous investigations have shown that a variety of molecules including CD11c/CD18 (Ingalls and Golenbock, 1995), the Toll-like family of receptors (Kirschning et al., 1998), moesin (Tohme et al., 1999) and heat shock proteins (Byrd et al., 1999) are involved in LPS signal transduction. The events that occur between binding of LPS to CD14 and signal transduction have, however, not been established. Our experiments show that LPS transfers from CD14 to immobile receptors that are very probably hsp70 and hsp90. The lack of mobility suggest that these heat shock proteins could be part of a large multimeric complex that includes other molecules that have previously been implicated in signal transduction but does not include CD14.

In conclusion our data suggest that the heat shock family of proteins might be part of an LPS multimeric receptor but that they are not physically associated with the CD14 receptor. Therefore, we propose a model based on our findings that, following CD14 ligation by bacterial components, a complex of receptors is probably formed within the lipid rafts involving several signalling molecules, including hsp70 and hsp90 (Fig. 8). CD14 releases LPS, catalysing the transfer of LPS to a signal transducing complex that involves hsp70 and 90, leading to cell activation. In the absence of CD14, this receptor complex is still formed.

This work was supported by the BBSRC. We would like to thank Keith Wilson for helpful discussions. We would also like to thank Theodore Kirkland and Suganya Viriyakosol for kindly providing their CHO-CD14 cells.