CD14-Dependent Internalization of Bacterial Lipopolysaccharide (LPS) Is Strongly Influenced by LPS Aggregation But Not by Cellular Responses to LPS¹ Free

Animal cells recognize Gram-negative bacterial LPS (LPS or endotoxin) by innate immune mechanisms, the most sensitive of which involves binding of LPS to CD14 (1, 2). CD14, expressed as a cell surface receptor on phagocytes (mCD14)³ and as a soluble protein in serum (sCD14), rapidly binds LPS in the presence of LPS binding protein (LBP), another serum component. mCD14 or sCD14 can support cellular responses to very low concentrations of LPS, and mCD14 can also bind and facilitate the internalization of large amounts of LPS from the cellular milieu (3, 4, 5, 6). Despite these prominent roles in LPS uptake by cells, however, mCD14 is not thought to transduce LPS signals across the plasma membrane (7, 8, 9, 10).

The mechanism of internalization of mCD14-bound LPS has not been described. It may enter cells via clathrin-coated pits (11, 12) or possibly via the noncoated invaginations reported to internalize other glycosylphosphatidylinositol-anchored proteins (13, 14, 15, 16). LPS then moves into acidic intracellular compartments where the lipid A moiety may be partially degraded in ways that modify its biologic activity (17). LPS has been shown to move from the plasma membrane to intracellular vesicles (4) that, in macrophages, may reach a perinuclear location (18). There is also evidence that LPS may remain associated with the plasma membrane for extended periods of time (19) and that cells may release bioactive LPS (20).

The biologic importance of LPS internalization and its relationship to signaling are disputed. Much experimental evidence suggests that internalized LPS undergoes enzymatic degradation or physical sequestration (17), processes that should reduce its signal potency (5). LPS-induced signals may regulate these processes. For example, LPS can down-regulate its own dephosphorylation by macrophages (21), and according to one recent report (18), C3H/HeJ macrophages, which fail to respond to LPS, show alterations in the movement of intracellular LPS-containing vesicles. While some evidence suggests that LPS must be present in intracellular vesicles before certain signals are transduced (6, 18, 22), other data argue that the LPS that undergoes internalization by monocytes is not involved in signaling (5). Clarification of these issues will require knowledge of the mechanisms by which LPS is internalized and transported within the cell.

In this report we show that the kinetics with which mCD14-bound LPS is internalized are influenced prominently by the initial LPS aggregation state, but not by cellular responses to the LPS.

THP-1 cells (23) were obtained from D. Altieri (Scripps Research Institute, La Jolla, CA) and cultured as previously described (8). To induce CD14 expression, the cells were either differentiated by culture in 0.05 μM 1,25-dihydroxyvitamin D₃ for 48 to 96 h (7, 8) or stably transfected by electroporation with human CD14 cDNA (a gift from D. Golenbock, Boston University, Boston, MA) cloned into pRc/RSV (Invitrogen, San Diego, CA). Bulk populations of stably transformed cells were selected in 0.5 mg/ml G418, and cells expressing CD14 were isolated using a fluorescence-activated cell sorter (FACStar Plus, Becton Dickinson Immunocytometry, San Jose, CA) and expanded in culture. Thioglycolate-elicited peritoneal macrophages were isolated as previously described (24) from C3H/HeJ and C3H/HeN mice (The Jackson Laboratory, Bar Harbor, ME). CHO cells stably transfected with recombinant human LPS binding protein (rLBP) or empty vector (pRc/RSV) were provided by P. Tobias (Scripps Research Institute). The cells (CHO-rLBP or CHO-RSV) were cultured in serum-free medium (CHO-S-SFM II, Life Technologies, Grand Island, NY).

CHO-rLBP and CHO-RSV culture supernatants were tested for their ability to promote [³H]LPS (100 ng/ml) binding to CD14 on THP-1 cells, and a dilution of the CHO-rLBP supernatant that rapidly enhanced LPS binding was chosen for subsequent experiments. CHO-RSV supernatant had no LPS transfer activity. Purified recombinant human soluble CD14_1–356 (sCD14) was a gift from R. Thieringer (Merck, Rahway, NJ). Anti-CD14 mAb 26ic (IgG2b) was provided by D. Golenbock (Boston University). FITC-conjugated goat anti-mouse IgG (H+L) F(ab′)₂ was obtained from Tago (Burlingame, CA). RPMI 1640, Cellgro Complete serum-free medium, and G418 were purchased from Mediatech (Herndon, VA). Proteinase K (from Tritirachium album), cell culture-tested BSA, nonimmune IgG2b control Ab MOPC-141, PMSF, and 1,4-diazabicyclo-(2, 2, 2)octane were obtained from Sigma Chemical Co. (St. Louis, MO). Phosphatidylinositol-specific phospholipase C (PI-PLC) from Bacillus cereus was purchased from Boehringer Mannheim (Indianapolis, IN).

Escherichia coli LCD25 [³H]LPS (1.5 × 10⁶ dpm/μg) was biosynthetically labeled and isolated as previously described (25). For derivatization, unlabeled LCD25 LPS was obtained from List Biological Laboratories (Campbell, CA) and repurified (26) to remove trace protein contamination. After repurification, contaminating protein could not be detected on silver-stained SDS-PAGE gels after loading 10 μg of LPS/lane. The LPS was fluoresceinated with FITC (Molecular Probes, Eugene, OR) (27) as follows. One hundred micrograms of LPS containing tracer amounts of [³H]LPS were mixed with 0.1 M borate buffer, pH 10.5, containing 0.3% sodium deoxycholate (Sigma Ultrapure), 1 mM EDTA, and 10 mg/ml FITC, and the mixture was incubated in the dark for 3 h at 37°C. Unbound FITC was removed by extensive dialysis against 0.9% NaCl with 10 mM Tris (Cl), pH 7.5, at 4°C. The molar ratio of FITC to LPS was 0.36.

Aggregated LPS (Ag-LPS) was prepared by diluting [³H]LPS stock suspensions to a concentration of 10 to 20 μg/ml in RPMI-HB and dispersing with a probe sonicator for 1 to 2 s (model 450, Branson Ultrasonics, Danbury, CT). Partially disaggregated LPS (DAg-LPS) was made in the same way, except that the RPMI-HB was replaced by HNEB. Monomeric LPS-sCD14 complexes (28) were prepared by mixing DAg-LPS with a 25-fold excess (by weight) of sCD14 (usually 300–500 μg sCD14/ml, final concentration) and incubating overnight at 37°C. For some experiments, LPS-sCD14 complexes were made in the presence of LBP by mixing LPS and sCD14 as described above with an equal volume of CHO-rLBP supernatant for 10 min at 37°C. LPS-sCD14 complexes generated by either method were filtered through a 100-kDa cut-off membrane (Microcon-100, Amicon, Beverly, MA) to minimize the presence of aggregates.

Sucrose gradients were made as previously described (29). Samples (0.2 ml) containing [³H]LPS were loaded on top of 4.7-ml gradients and centrifuged at 150,000 × g for 4.5 h at room temperature. Fractions (0.35 ml) were collected from the bottom with an 18-gauge needle, and ³H disintegrations per minute were measured in the presence of SDS and EDTA as previously described (7).

THP-1 cells were washed three times with cold RPMI 1640 and resuspended in SFM (Cellgro Complete serum-free medium containing 20 mM HEPES buffer, pH 7.4, and 0.3 mg/ml BSA) at 5 to 7 × 10⁶ cells/ml. LPS was diluted into CHO-rLBP or CHO-RSV supernatant and incubated for 5 min at 37°C before addition to the cells. The cells (90 μl) were warmed to 37°C for 5 min, 10 μl of LPS were added, and the incubation was continued at 37°C for 1 or 2 h. The cells were chilled on ice, the suspensions were centrifuged, and the culture supernatants were assayed for IL-8 (DuoSet ELISA Development System, Genzyme, Cambridge, MA). NF-κB was measured by gel-shift assay in nuclear extracts as previously described (8). Relative amounts of radioactivity in the gel-shifted bands were measured using a PhosphorImager SF (Molecular Dynamics, Sunnyvale, CA) and expressed in arbitrary units.

DAg-LPS ([³H]LPS or FITC-LPS) was used unless otherwise stated. The LPS in HNEB was diluted 1/10 with CHO-RSV or CHO-rLBP supernatant and incubated for 5 min at 37°C immediately before addition to the cells. LPS-sCD14 complexes were diluted in PBS with 0.3 mg/ml BSA. The cells were harvested in suspension, washed three times with cold RPMI 1640, and resuspended in SFM at 5 to 7 × 10⁶ cells/ml. Cells (90 μl) were warmed to 37°C, 10 μl of LPS were added, and the mixtures were incubated at 37°C for 1 to 60 min. For binding LPS to cells on ice, the DAg-LPS-CHO LBP supernatant mixtures were incubated for 5 min at 37°C, chilled on ice, mixed with the cells, and allowed to bind for 15 min on ice. The cells were then washed, resuspended in 100 μl of SFM, and warmed to 37°C for various times. Internalization was stopped by adding 1 ml of ice-cold PBS. The cells were pelleted by centrifugation (750 × g for 2 min at 0–4°C) and washed with 1 ml of cold PBS. Adherent macrophages were isolated in 24-well culture plates (Costar, Cambridge, MA), washed three times with RPMI 1640, and warmed to 37°C for 5 min in 270 μl of SFM. [³H]LPS in CHO-LBP or CHO-RSV supernatant (30 μl) was added to the cells, and the incubation was continued at 37°C for various times. The cells were then placed on ice, aspirated, and washed twice with 1 ml of ice-cold PBS. Surface-exposed and internal LPS were measured as described below.

The washed cells were resuspended in 1 ml of ice-cold 0.02% proteinase K and kept on ice for 30 min. The cells were pelleted by centrifugation, and 300 μl of supernatant were removed for ³H counting. The cells were resuspended in 200 μl of PBS containing 0.5 mM PMSF, and 100 μl of cells were counted. The protein content of each cell suspension was measured using Bio-Rad protein assay reagent (Bio-Rad, Hercules, CA). Proteinase K treatment did not increase cell permeability to trypan blue or decrease cell number (not shown), although some of the cells that were already permeable to trypan blue were digested by the protease. For removal of surface-exposed LPS with PI-PLC instead of proteinase K, the washed cells were resuspended in 100 μl of HNE buffer (20 mM HEPES (pH 7.4), 150 mM NaCl, and 1 mM EDTA) containing 0.5 U of PI-PLC. The mixture was incubated for 1 h on ice, 900 μl of cold PBS were added, and the cells were pelleted by centrifugation. Cell-associated and released [³H]LPS were measured as described above. For adherent murine macrophages, the washed cells were incubated on ice with 500 μl of 0.02% proteinase K in PBS. The protease solution was saved, and the cells were resuspended in 200 μl of PBS using a rubber policeman. [³H]LPS and cell protein content were measured as described above.

After incubating cells with FITC-LPS in suspension, the cells were washed as described above, resuspended in 100 μl of SFM, and split into 50-μl aliquots. Rabbit anti-fluorescein (Texas Red conjugate; Molecular Probes, Eugene, OR) was added to one aliquot (50 μg/ml, final concentration), and the mixtures were incubated on ice for 30 min. The cells were then washed with cold PBS. To prevent quenching of fluorescein, the intracellular pH was raised by fixing the cells in 1 ml of 3% paraformaldehyde in 100 mM sodium phosphate, pH 7.4, for 30 min on ice, centrifuging, and resuspending the cells in cold PBS containing 20 mM Tris, pH 8.0. The mean fluorescence intensity (MFI) of FITC-LPS in each cell population was measured by a flow cytometer (FACScan, Becton Dickinson Immunocytometry, San Jose, CA). We found the same relationship between green (530 nm) and orange (585 nm) fluorescence emissions in cell populations that had internalized FITC-LPS (not shown) as in those that had not internalized FITC-LPS, providing evidence that intracellular fluorescein was not quenched. In some experiments, anti-fluorescein was replaced by 1 ml of 0.02% proteinase K in PBS to remove surface-exposed FITC-LPS.

THP-1 cells were washed twice with cold PBS containing 0.5 mM PMSF and resuspended in 90 μl of blocking buffer (PBS with 0.3 mg/ml BSA, 10% heat-inactivated normal goat serum, and 0.5 mM PMSF) for 10 min on ice. The mCD14 was stained with anti-CD14 mAb 26ic or control IgG2b (10 μg/ml) for 45 min on ice followed by FITC-conjugated goat anti-mouse IgG F(ab′)₂ (1/200) for 45 min. The MFI in each cell population was measured by FACS analysis.

Cells were incubated with 200 ng/ml FITC-LPS prepared in CHO-LBP or CHO-RSV supernatant for various times at 37°C in SFM as described above. The cells were washed twice with cold PBS and resuspended in 200 μl of 4% paraformaldehyde in 100 mM sodium phosphate, pH 7.4, for 30 min on ice. The cells were then centrifuged onto poly-l-lysine-coated slides for 5 min at room temperature and washed in cold PBS. The slides were rinsed in distilled water, aspirated dry, and mounted with one drop of mounting solution (9 vol of glycerol, 1 vol of 1 M Tris (Cl), pH 8.6, and 2.5% 1,4-diazabicyclo-(2, 2, 2)octane) with a 22- × 22-mm glass coverslip (no. 1, Fisher Scientific, Pittsburgh, PA). The cells were viewed with an MRC-1024 laser confocal imaging system (Bio-Rad). Sequential optical sections (1 μm) were observed, and the resolution of the digital images was 512 × 512 pixels (1 pixel = 0.155 μm).

Binding and internalization measurements in this study were restricted to LPS that binds initially to mCD14 in the presence of either LBP or sCD14. We (7) and others (30) have previously shown that LBP promotes LPS binding to mCD14 in THP-1 cells. Hailman et al. (28) showed that [³H]LPS-sCD14 complexes can rapidly transfer LPS to unoccupied sCD14 molecules in the absence of LBP and further showed indirect evidence that sCD14 transfers LPS to mCD14. Using CD14-transfected THP-1 cells, we found that [³H]LPS from [³H]LPS-sCD14 complexes binds to mCD14 in the absence of LBP (Fig. 1). In contrast, mCD14-negative cells that were transfected with empty vector did not measurably bind LPS from LPS-sCD14 complexes. The LPS bound to mCD14 saturably and with high apparent affinity. The apparent equilibrium dissociation constant (K_d) of 31 ± 3 ng/ml (n = 2) or 7.8 nM was very similar to that obtained for [³H]LPS-LBP complexes (K_d = 33 ng/ml or 8.2 nM) (Ref. 7 and data not shown). The actual K_d (not determined) may differ from the apparent K_d reported here, since the affinity of the LPS-sCD14 interaction and the equilibrium between LPS-mCD14 and LPS-sCD14 were not taken into account. The binding capacity (6.2 × 10⁵ molecules of LPS/cell) is similar to the number of mCD14 molecules expressed per cell (Ref. 7 and data not shown).

The protease protection assay is based on the ability of proteinase K to remove surface-exposed LPS at the temperature of ice while leaving the plasma membrane intact. As shown in Table I, proteinase K removed nearly all the [³H]LPS that bound to the cells whether they had been incubated with [³H]LPS-LBP or with [³H]LPS-sCD14 complexes. Likewise, proteinase K removed virtually all the mCD14 from the cells. Treatment of the cells with PI-PLC, which cleaves CD14 from its glycosylphosphatidylinositol anchor, removed less [³H]LPS than treatment with proteinase K. However, PI-PLC also removed significantly less mCD14 from the cells; this may account for its lower efficiency at removing [³H]LPS. As shown in Table I, the percentages of the total mCD14 (73%) and [³H]LPS (derived from sCD14 complexes; 69%) that were removed by PI-PLC were virtually the same. PI-PLC removed significantly less [³H]LPS (from LBP complexes) than mCD14, however, suggesting that some of the LBP-delivered LPS may interact with molecule(s) other than mCD14 on the cell surface. Although the [³H]LPS was labeled entirely in its fatty acyl chains (25), deacylation of the cell-associated LPS occurred very slowly in these cells (<1.5% in 1 h; data not shown). This minor loss of ³H from the [³H]LPS should not significantly affect the measurements described here.

The fluorescence quenching assay is based on the ability of rabbit anti-fluorescein (Texas Red conjugate) to bind and efficiently quench the fluorescence of surface-exposed FITC-LPS. Cell-associated FITC-LPS was quantitated by FACS analysis to determine the MFI of the cell population. Figure 2 shows representative fluorescence histograms from an experiment in which THP-1 cells were exposed to FITC-LPS for 10 min at 37°C. MFI data were corrected (ΔMFI) for the autofluorescence of the cells (a; MFI = 4.6). LPS binding in the absence of LBP (b; ΔMFI = 1.2) was only slightly higher than the autofluorescence of the cells. After incubation with FITC-LPS in the presence of LBP, internalized LPS was measured by exposing the cells to the anti-fluorescein Ab (d; ΔMFI = 13.8). Surface-exposed LPS was determined by measuring the total cell-associated LPS (c; ΔMFI = 29.3) and subtracting internal LPS (c − d; 29.3 − 13.8 = 15.5). At least 90% of the FITC-LPS that bound to metabolically inhibited cells was quenched by the Ab (not shown).

Figure 3 shows LPS binding and internalization measured by protease protection (A) and by fluorescence quenching (B) in cells that were exposed to LPS for the indicated times at 37°C. MFI data for FITC-LPS are shown on a linear scale. LPS bound rapidly to mCD14 (maximal in 2 min) under these conditions. The kinetics of internalization were biphasic. Most of the LPS internalization occurred rapidly (within 5 min), followed by markedly slower internalization with very little additional accumulation of LPS inside the cells. Similar results were obtained for FITC-LPS internalization by freshly isolated human monocytes (not shown).

The protease protection and fluorescence quenching assays should give identical results only if cell surface LPS remains bound to mCD14 or another protease-sensitive molecule. We anticipated that if the FITC-LPS bound to a protease-resistant protein or inserted into the lipid bilayer, it would be quenched by the anti-fluorescein Ab yet be resistant to proteinase K treatment. Table II shows that after exposing THP-1 cells to FITC-LPS for 20 min at 37°C, protease treatment and fluorescence quenching gave virtually identical estimates of surface-exposed FITC-LPS. The results indicate that the LPS was not measurably transferred from mCD14 to a protease-resistant site. They also confirm that proteinase K treatment does not remove internal or sequestered LPS that is inaccessible to the gentler Ab quenching method.

During a long incubation period, intracellular CD14 may emerge on the surface and bind LPS. To exclude this phenomenon, we bound LPS to the cells on ice, washed away the unbound LPS, warmed the cells to 37°C for various times, and measured LPS internalization. Figure 4 shows the proportions of the total cell-associated [³H]LPS that are internal (protease resistant) or on the cell surface (protease sensitive) at each time point. Internalization was rapid during the first 5 min of rewarming and slowed thereafter. The biphasic internalization kinetics were similar to those observed in Figure 3. Rewarming the cells for 20 min resulted in <10% release of surface-bound LPS into the medium (not shown). Similar results were obtained whether CD14-transfected cells (Fig. 4) or differentiated THP-1 cells (Fig. 8) were used. We found that efficient binding of LPS to cells on ice required binding the LPS to LBP at 37°C before chilling on ice (5). Binding under these conditions was nearly maximal in 15 min (not shown).

We used laser confocal microscopy to visualize surface and internal FITC-LPS after exposure to THP-1 cells for various times at 37°C (Fig. 5). After 1 min, virtually all the LPS was diffusely distributed on the plasma membrane (Fig. 5,A). Treatment of the cells with proteinase K almost completely removed the FITC-LPS, although a few focal accumulations could be found (Fig. 5 E). These foci became more numerous and larger with increasing incubation times and were observed in proteinase K-treated cells, suggesting that they represent internalized LPS. We analyzed a series of 10 optical sections of each sample at 1-μm intervals and confirmed the intracellular location of many of the fluorescent foci (not shown). The resolution of this method, however, is not sufficient to distinguish whether the foci at the cell periphery are on the outer or the inner surface of the plasma membrane, and we cannot exclude the possibility that some of these foci represent LPS that accumulates in membrane invaginations. Most of the internalized (protease-resistant) LPS was found near the cell surface at early time points (1–5 min; A and B,and E and F), but at later time points (20–45 min; C and D, and G and H), much of the LPS was clearly separated from the plasma membrane. Although we cannot quantitate LPS internalization kinetics by this method, the confocal images confirm that FITC-LPS is internalized by the cells and that much of the LPS becomes well separated from the plasma membrane.

[³H]LPS was prepared in various aggregation states and subjected to sucrose density gradient analysis (Fig. 6). LPS stock suspensions that were diluted directly into Ca²⁺- and Mg²⁺-containing medium (RPMI 1640) remained highly aggregated even after probe sonication and exposure to LBP, and this LPS (Ag-LPS) migrated through the gradient to the bottom of the centrifuge tube (fraction 0). Monomeric LPS, prepared by incubating [³H]LPS overnight with an excess of purified soluble CD14 (28) in the presence or the absence of LBP, remained at the top of the sucrose gradient. DAg-LPS (used in the experiments above), prepared by diluting and sonicating stock suspensions in the presence of EDTA, moved slightly farther down the gradients than monomeric LPS, suggesting that the molecules were somewhat aggregated. When DAg-LPS was diluted in Ca²⁺- and Mg²⁺-containing medium, the aggregation state increased to a variable extent but did not revert to the highly aggregated state of Ag-LPS. Exposure to LBP under these conditions did not disaggregate the LPS to the extent observed with LPS-sCD14 complexes (not shown).

Monomeric LPS bound to mCD14 in the presence (not shown) or the absence (Fig. 7,A) of LBP at a rate similar to that at which DAg-LPS bound in the presence of LBP (Fig. 7,B; maximal in 1–2 min). While nearly half the maximal amount of Ag-LPS bound to the cells in 1 to 2 min, its binding reached a maximum more slowly (10–30 min). In contrast, the cells internalized aggregated LPS extremely rapidly (70–80% of the total LPS that bound to the cells was internalized in 1 min; Fig. 7,C), whereas LPS monomers were internalized very slowly (Fig. 7,A). Partially disaggregated LPS was internalized at an intermediate rate (Fig. 7,B). Although the initial rate of monomeric LPS internalization was markedly slower than that of partially disaggregated LPS, the second phase kinetics were similar (Fig. 7,D). Compared with the LPS monomers, approximately sixfold more LPS from the aggregated preparation was internalized during the first 20 min of incubation (Fig. 7, A and C).

The rate and extent of aggregated FITC-LPS internalization (not shown) were higher than those of the partially disaggregated FITC-LPS (Fig. 3,B), but they did not exhibit the dramatic differences seen with aggregated and disaggregated [³H]LPS preparations. This may be explained by the fact that the derivatization procedure for FITC-LPS involves disaggregating the LPS with deoxycholate. LPS treated in this way may not completely reaggregate to its original state. THP-1 cells internalized much less FITC-LPS from monomeric FITC-LPS-sCD14 complexes (not shown), in keeping with the results of the protease protection assay for [³H]LPS (Fig. 7 A).

The stimulatory potencies of the LPS preparations were assessed by their abilities to induce IL-8 production and NF-κB translocation in differentiated THP-1 cells. When incubated with cells at the submaximal stimulatory concentration of 1 ng/ml, the three preparations exhibited similar responses (IL-8 production and NF-κB translocation; Table III). The experiments were repeated with similar results.

Having previously shown that a 3-h pre-exposure to LPS desensitized THP-1 cells to restimulation of p42 protein tyrosine phosphorylation by LPS (31), we tested whether LPS internalization kinetics could be altered by any of the stimulatory or desensitizing effects of LPS pre-exposure. THP-1 cells were incubated with unlabeled LPS in the presence of LBP at 37°C before adding [³H]LPS plus LBP or monomeric [³H]LPS-sCD14 complexes. The rate and the extent of internalization of LPS aggregates (Fig. 8,A) or LPS monomers (not shown) were not significantly altered by pre-exposure to 2 or 100 ng/ml LPS for 3 h, although LPS binding was slightly decreased in some experiments (not shown). Similar results were obtained after short (5-min) pre-exposure to 50 ng/ml LPS (not shown). We next asked whether an LPS-specific inhibitor would alter LPS internalization. We previously showed that the tetra-acyl lipid A analogue, LA-14-PP, can inhibit LPS responses without inhibiting LPS uptake by mCD14 (8). As shown in Figure 8 B, a concentration of LA-14-PP that strongly inhibited LPS-induced IL-8 production did not alter [³H]LPS internalization. IL-8 production measured at 2 h was 1.55 ± 0.12 ng/ml (no LPS), 6.49 ± 0.19 (LPS), 2.78 ± 0.17 (LA-14-PP plus LPS), and 1.78 ± 0.09 (LA-14-PP alone). The experiment was repeated with similar results. LA-14-PP also did not inhibit the internalization of monomeric LPS from [³H]LPS-sCD14 complexes (not shown).

Finally, we measured internalization by macrophages from LPS-responsive (C3H/HeN) and LPS-hyporesponsive (C3H/HeJ) mice. Figure 9 shows that the macrophages internalized LPS with identical kinetics whether they were derived from HeN or HeJ mice, suggesting that LPS-induced responses do not alter the ability of cells to internalize LPS. The LPS hyporesponsiveness of the HeJ macrophages was confirmed by stimulating the HeN and HeJ macrophages for 4 h with 10 ng/ml LPS and measuring IL-6 production by ELISA. Using cells from the same experiment (Fig. 9), the HeN macrophages released 67 ± 3 ng/ml IL-6 into the culture supernatant, whereas the HeJ macrophages released <7 ng/ml. Taken together, the data in Figures 8 and 9 suggest that mCD14-bound LPS is internalized by a constitutive cellular process that is not affected by LPS-induced responses or desensitization.

Numerous studies have used fluorescence (3, 6) or electron microscopy (11, 32, 33, 34, 35, 36) to document the movement of LPS from the plasma membrane to various sites within animal cells. Others used resistance to extrinsic proteolytic stripping (5) or subcellular fractionation (4) to follow the intracellular migration of radiolabeled LPS from the plasma membrane. Relatively few studies of CD14-mediated LPS internalization have been reported (3, 4, 5, 6), however, and previous measurements of internalization kinetics have given variable results. For example, laser confocal microscopy was used to show that monocytes rapidly internalize most of the FITC-LPS that binds to them in the presence of LBP (3), whereas neutrophils demonstrated considerably slower internalization of BODIPY-labeled LPS complexed with sCD14 (6). Reliable quantitative methods for measuring LPS internalization are needed to resolve these issues and to study the mechanism(s) by which LPS enters cells.

Our initial goal was therefore to develop and test two independent, quantitative methods for measuring LPS internalization. We found that the protease protection and fluorescence quenching assays gave remarkably similar results, and laser confocal microscope images confirmed that the cells had internalized the LPS. These methods should be useful for quantitating LPS internalization in a variety of cell types.

It is important to note that the term internalization, as used here, refers to the movement of surface-exposed LPS to sites that are inaccessible to proteinase K or anti-fluorescein. The kinetics of this process may be quite different from the kinetics of subsequent processes that move LPS to various intracellular compartments. For example, Ward et al. (37, 38) showed that different ligands were internalized by their respective receptors on macrophages at vastly different rates, yet the kinetics of intracellular traffic were the same for each ligand. In neutrophils, LPS was shown to move slowly to an endocytic compartment that comigrated on Percoll gradients with vesicles that contain acyloxyacyl hydrolase, an enzyme that partially degrades LPS (4). The process by which internalized LPS moves into this compartment, however, might have very different kinetics from the process by which LPS enters the cell. Likewise, the recently reported (18) defect in intracellular LPS traffic in C3H/HeJ macrophages appears to be distal to LPS internalization. As we show in Figure 9, LPS internalization occurs with identical kinetics in C3H/HeN and C3H/HeJ macrophages.

The aggregation state had a dramatic influence on the initial rate and extent of CD14-dependent internalization of this amphipathic ligand. Highly aggregated LPS was internalized extremely rapidly, whereas monomeric LPS was internalized slowly. DAg-LPS was internalized at an intermediate rate. The initial phase of internalization of large LPS aggregates (70% of the cell-associated LPS within 1 min) occurred at a rate similar to that reported for the internalization of polyvalent mannosylated ligands by the mannose receptor on sinusoidal endothelial cells (t_1/2 = 10 s) (39) and on rabbit alveolar macrophages (t_1/2 = 1 min) (37). Such rapid internalization could be promoted by the ability of each ligand aggregate to cluster many receptor molecules, which may result in binding to endocytic structures with greater avidity. We cannot rule out the possibility, however, that the rapid sequestration of LPS aggregates from the actions of proteinase K or anti-fluorescein may be due to the inability of these proteins to access some ligand-receptor aggregates that are still surface exposed. Evidence that the assays measure LPS internalization includes the observations that 1) rapid internalization of LPS aggregates is energy dependent (Table I and data not shown), whereas LPS aggregation and binding to CD14 are not; 2) proteinase K and anti-fluorescein act by very different mechanisms (removal of LPS by degradation of mCD14 and possibly of LBP in contrast to quenching of exposed fluorescein groups on the LPS), but yield identical measurements of surface-exposed LPS; and 3) as mentioned above, the rate of internalization of aggregated LPS is similar to that of polyvalent mannosylated ligands, which were released from cell surfaces by the actions of very small molecules (EGTA or EDTA) (37, 39).

The LBP that is bound to LPS aggregates may also contribute to their internalization by interacting with other cell surface molecules. Gegner et al. (5) showed that LBP is internalized with LPS as LPS-LBP-CD14 ternary complexes and proposed that weak interactions between LBP and other cell surface molecules stabilize these complexes. We recently reported that in the presence of LBP, LPS preferentially binds to mCD14 in low density, lipid-enriched plasma membrane domains (40). These domains may harbor the molecules that stabilize LPS-LBP-CD14 complexes and contribute to their rapid internalization.

Since native LPS is found in bacterial cell walls or membrane fragments, where its physical state should be most similar to that of highly aggregated LPS, the accelerating effect of aggregation on internalization may be seen as a mechanism for cells to take up (clear) (5) native LPS from their environment. Indeed, Grunwald and others (41) have reported recently that CD14 can promote phagocytosis of Gram-negative bacteria. The relationship between LPS aggregation state and its ability to induce cellular responses is less clear. Others have reported that the aggregation state of LPS can have a significant influence on its stimulatory potency, with less aggregated forms (or monomers) usually having greater activity (42, 43, 44, 45, 46). In these studies, however, the cell stimulation assays were performed in the absence of serum, so that the incubation mixtures lacked the transfer proteins that promote interactions with mCD14. We found that aggregated, partially disaggregated, and monomeric forms of LPS had approximately equal stimulatory potencies under conditions that promote their binding to mCD14 on THP-1 cells and monocytes (Table III). In keeping with these observations, Hailman et al. (28) found that LPS-LBP aggregates and monomeric LPS-sCD14 complexes made from R-form LPS (similar to the LPS used in the present study) had similar potencies for stimulating mCD14-dependent cytokine responses in cultured human macrophages and neutrophils. Aggregated S-form LPS, however, was less stimulatory than monomeric S-form LPS for reasons that are unclear. Experiments using LBP mutants (47) or Abs (5) have dissociated signaling from bulk binding of LPS aggregates, suggesting that the binding of aggregates to mCD14 is not associated with cell stimulation. To reconcile these data, one might conclude that mCD14 must monomerize LPS aggregates to some extent to support signaling. LBP and sCD14 catalyze very rapid monomerization of LPS, resulting in the formation of monomeric LPS-sCD14 (28), and by inference, this process should also occur by the actions of LBP and mCD14. This mechanism could explain why LBP is required for sensitive responses to LPS aggregates (48, 49) but not for LPS monomers (28).

It has been proposed that mCD14 may transfer LPS to the lipid bilayer of the plasma membrane (50). We addressed this issue first by comparing the time course of LPS internalization using the fluorescence quenching and proteinase protection assays. As shown in Figure 3, these assays gave virtually identical estimates of the relative proportions of internal and surface LPS over time. We then measured surface-exposed FITC-LPS before and after treating the cells with proteinase K or an anti-fluorescein Ab. We found that protease treatment and Ab quenching gave virtually identical estimates of surface FITC-LPS (Table II). Assuming that all surface-exposed FITC-LPS can be quenched by the anti-fluorescein Ab, including FITC-LPS that is not susceptible to removal by proteinase treatment (i.e., inserted into the lipid bilayer), this result suggests that the LPS that remains on the cell surface after incubation at 37°C is almost entirely protein bound. If a large fraction of bound LPS is transferred to the lipid bilayer of the cellular membrane(s) (50), this transfer probably occurs in an internal compartment.

Although certain agonists can regulate their own internalization by eukaryotic cells (51, 52, 53), stimulus-induced modulation of LPS internalization has not been reported. LPS can down-regulate its own catabolism by macrophages (21), however, and a recent report suggests that LPS-induced signals may direct the intracellular traffic of LPS-containing endocytic vesicles (18). Moreover, in keeping with previous reports using other cell types (3, 4, 5), we found that the rate at which LPS was internalized by THP-1 cells slowed significantly after the first few minutes of exposure to LPS (Figs. 3 and 7). Does the cellular response to LPS decrease its internalization rate? If so, one would expect prior exposure to LPS to alter the rate at which newly presented LPS is internalized. We found that pre-exposure to LPS for 5 min or 3 h or inhibition of LPS signals using a lipid A analogue did not alter either the initial or the secondary rate of LPS internalization (Fig. 8, A and B). In addition, LPS-hyporesponsive and LPS-responsive murine macrophages internalized LPS with virtually identical kinetics (Fig. 9). We conclude, therefore, that LPS enters these cells by a constitutive mechanism that is insensitive to its own stimulatory effects.

We thank Peter Tobias, Rolf Thieringer, and Samuel Wright for providing important reagents; Leon Eidels and Samuel Wright for critical comments on the manuscript; Ping-yuan Wang for providing transfected THP-1 cells; and Steven Geiszler for performing IL-6 assays.