Bacterial Lipopolysaccharide Can Enter Monocytes Via Two CD14-Dependent Pathways¹

Sensitive immune recognition of Gram-negative bacterial LPS (LPS or endotoxin) requires CD14, a GPI³-anchored protein expressed by monocytes, macrophages, and neutrophils, and is enhanced by LPS-binding protein (LBP), a soluble serum protein (1, 2, 3, 4, 5). LBP binds LPS aggregates and rapidly transfers them to GPI-anchored CD14 (mCD14) or to soluble CD14 (sCD14), which are thought to promote signal responses by facilitating the interaction of LPS with as yet unidentified signaling molecules (6, 7, 8, 9). mCD14 may also internalize LPS monomers or aggregated LPS-LBP complexes (10, 11, 12, 13, 14). Although LPS internalization by phagocytes has generally been regarded as a disposal function (12), several recent studies have suggested that internalization may be required for LPS signaling (13, 15, 16, 17). Clarification of the signaling versus disposal functions of LPS internalization should be fostered by a more complete understanding of the membrane structures and biochemical mechanisms that mediate the internalization process.

The most thoroughly characterized pathway for receptor-mediated endocytosis involves the entry of ligand-receptor complexes into clathrin-coated pits, a process that largely involves transmembrane proteins that have a specific cytoplasmic domain-targeting signal (18, 19). Although GPI-anchored proteins lack coated pit-targeting signals, certain of these proteins have been shown to enter coated pits (20, 21) presumably by associating with another protein. Some GPI-anchored proteins, when bound by anti-receptor Abs (22, 23, 24) or extracellular ligands (25), may also enter caveolae or similar noncoated invaginations. Caveolae, the best-characterized noncoated membrane invaginations, are thought to be involved in potocytosis, transcytosis, and signal transduction (20, 26, 27, 28). Nonclathrin-mediated internalization pathways may be taken by protein toxins, such as ricin (29, 30) or cholera toxin (31), that bind to membrane glycolipids, and by other proteins, such as IL-2 (32), that bind to transmembrane receptors lacking coated pit-targeting signals. The biochemical and functional properties of noncoated invaginations found in cells that do not express caveolin (e.g., resting monocyte/macrophages and lymphocytes) are not well understood, however, and it is unclear whether vesicles derived from these invaginations recycle or move to late endosomes and lysosomes.

Other noncoated structures may also mediate endocytosis in macrophages. Nichols (33) noted that alveolar macrophages internalize horseradish peroxidase into tubular invaginations of the plasma membrane (tubular pinosomes). Myers et al. (34) subsequently showed that multivalent β-very low-density lipoprotein (β-VLDL) particles enter larger surface-connected tubules (surface tubules for entry into macrophages, or STEMs (35)) in murine macrophages. Whereas tubular pinosomes could acquire acid phosphatase, presumably from fusion with lysosomes (33), the surface-connected tubules noted by Myers et al. were thought to detach from the surface and transport VLDL to perinuclear lysosomes. More recently, Zhang et al. (36) have described even larger surface-connected compartments induced by aggregated LDL in human monocyte/macrophages. Although the three reported surface-connected tubular structures differ in important ways, both STEMs and surface-connected compartments appear to take up multivalent or highly aggregated ligands. We show in this study that LPS molecules that bind GPI-anchored CD14 on monocytic THP-1 cells are internalized predominantly by a nonclathrin-mediated pathway that involves noncoated tubular membrane invaginations and intracellular tubular and vacuolar structures, while a minority of LPS molecules enter the cells via coated pits. Aggregation of LPS, which enhances both the rate and extent of LPS internalization (14), accelerates its entry into the nonclathrin-mediated pathway.

cDNA-encoding wild-type human CD14 (CD14-GPI) was a gift from Douglas T. Golenbock (Boston University, Boston, MA). cDNA encoding the transmembrane and cytoplasmic domains of the human LDLR (37) was provided by Steve Lacey (University of Texas Southwestern Medical School, Dallas, TX). We generated a CD14-LDLR chimeric receptor by replacing the C-terminal 21 amino acids of CD14 with the transmembrane and cytoplasmic domains of the LDLR (Fig. 1) by patch PCR. The CD14 cDNA template was amplified using primer A (5′-GGA ATT CAA GCT TAT GGA GCG CGC GTC CTG-3′) and primer B (5′-ACG CTA CTG GGC TTC TTC TCA CGT GCA CAG GCT GGG AC-3′) to yield a product that contained a 5′ HindIII restriction site. In a separate reaction, LDLR cDNA was amplified using primer C (5′-GTC CCA GCC TGT GCA CGT GAG AAG AAG CCC AGT AGC GT-3′) and primer D (5′-GCT CTA GAT CAC GCC ACG TCA TCC TCC-3′) to yield a product that contained a 3′ XbaI site. The two isolated PCR products were then mixed and amplified for five cycles without primers to generate a full CD14-LDLR template. The chimeric construct was then amplified with primers A and D and isolated on an agarose gel. The CD14-GPI and CD14-LDLR cDNAs were cloned into HindIII and XbaI restriction sites in the pRc/RSV expression vector (Invitrogen, San Diego, CA), and their structures were confirmed by automated DNA sequencing.

THP-1 cells were obtained from D. Altieri (Scripps Research Institute, La Jolla, CA) and cultured, as previously described (6). For transfections, we used either bulk populations of the parental cells, or we obtained a single clone by limiting dilution to minimize variation among the transfectants. To increase mCD14 expression above the virtually undetectable level seen in undifferentiated THP-1 cells (7), the cells were transfected with pRC/RSV containing either wild-type human CD14 (CD14-GPI) or the CD14-LDLR chimera (CD14-LDLR). Bulk populations of stably transformed cells were selected in 0.5 mg/ml G418. The cells expressing CD14 were isolated using FACStar^PLUS (Becton Dickinson Immunocytometry, San Jose, CA) and expanded in culture. Chinese hamster ovary (CHO) cells that were stably transfected with recombinant human LPS-binding protein (rLBP) or empty vector control (pRc/RSV) were kindly provided by Peter 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), and the media were tested and used as previously described (14). Human peripheral blood monocytes were isolated from heparinized blood by centrifugation over Histopaque 1077 (Sigma, St. Louis, MO). The mononuclear cell fraction was harvested from the interface, washed in RPMI 1640 medium, and allowed to adhere to 22 × 22-mm glass coverslips in six-well plates containing RPMI 1640 with 8.5% FCS and 10% autologous serum. After a 1–3-h incubation, the nonadherent cells were removed by aspiration, and the adherent cells were used in the experiments.

Purified recombinant human soluble CD14_1–356 (sCD14) was a generous gift of R. Thieringer (Merck, Rahway, NJ). Anti-DNP mAb HDP-1 was generously provided by Drs. J. Goldstein and M. Brown (University of Texas Southwestern Medical Center). Goat anti-mouse IgG (H+L) 10-nm gold conjugate and normal heat-inactivated goat serum were obtained from BB International (Cardiff, U.K.) through Goldmark Biologics (Phillipsburg, 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 from Zymed Laboratories (South San Francisco, CA). Rabbit antifluorescein (Texas Red conjugate) and BODIPY FL-Escherichia coli were from Molecular Probes (Eugene, OR). RPMI 1640, Cellgro Complete serum-free medium, and G418 were from Mediatech (Herndon, VA). Cytochalasins H and D were from Aldrich Chemical (Milwaukee, WI). PMA, dimethylamiloride, lucifer yellow (LY), proteinase K (from Tritirachium album), cell culture-tested BSA, sucrose, PMSF, and 1,4-diazabicyclo (2, 2, 2) octane (DABCO) were from Sigma. Phosphatidylinositol-specific phospholipase C from Bacillus cereus was from Boehringer Mannheim (Indianapolis, IN). Glutaraldehyde (25%), picric acid (2, 4, 6 trinitrophenol), osmium tetroxide (4%), tannic acid, uranyl acetate, propylene oxide, and EMbed-812 plastic-embedding medium were obtained from Electron Microscopy Sciences (Ft. Washington, PA).

Human holo-transferrin (Tf) was obtained from Sigma and radioiodinated by incubating 100 μg in 0.1 ml of 50 mM sodium phosphate, pH 7.4, with 1 mCi Na¹²⁵I over Iodogen (Pierce Chemical, Rockford, IL). The sp. act. was 3.4 × 10⁶ cpm/μg Tf. BODIPY FL-Tf and Texas Red-X-Tf conjugates were made using FluoReporter Protein Labeling Kits (F-10232 and F-6162, respectively; Molecular Probes), according to the manufacturer’s instructions. The resulting molar ratios of dye to Tf were 1.8 for BODIPY and 7.2 for Texas Red.

E. coli LCD25 [³H]LPS (1.5 × 10⁶ dpm/μg) was biosynthetically labeled and isolated as previously described (38). For derivatization, unlabeled LCD25 LPS was obtained from List Biologic Laboratories (Campbell, CA) and repurified (39) 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 per lane (not shown). Tracer quantities of [³H]LPS were added to aliquots of unlabeled LPS to determine recovery after derivatization reactions. DNP-LPS was made by the method of Rietschel et al. (40), except that 50 μg of repurified LPS in 100 μl of 1% triethylamine adjusted to pH 10.5 with boric acid was mixed with 200 μl of 5% 1-fluoro-2,4-dinitrobenzene in ethanol. The DNP/LPS molar ratio was 1.5, and the bioactivity of the DNP-LPS was equivalent to that of underivatized LPS, as determined by the threshold concentration of LPS required to stimulate IL-8 in THP-1 cells (7) (not shown). FITC-LPS was prepared from repurified LCD25 LPS, as previously described (14). The molar ratio of FITC/LPS was 0.36. BODIPY FL-LPS or Texas Red-X-LPS conjugates were prepared from the same repurified LPS using reagents from the FluoReporter Labeling Kits (above). We mixed 100 μg of LPS with 500 μg BODIPY FL or 80 μg Texas Red-X succinimidyl esters in 200 μl of a buffer containing 0.1 M sodium bicarbonate, 0.3% sodium deoxycholate (Sigma; Ultra pure), and 1 mM EDTA. The derivatized LPS was dialyzed against 0.9% NaCl containing 10 mM Tris (Cl), pH 7.5, at 0–4°C. The BODIPY/LPS molar ratio was 0.14, and the Texas Red/LPS molar ratio was 0.23.

Partially disaggregated LPS (DAg-LPS) and monomeric LPS-sCD14 complexes were made as previously described (14).

Internalization of [³H]LPS was measured by protease protection, as previously described (14). To test the effect of hypertonic medium on LPS internalization, the cells were preincubated in SFM or SFM containing 0.45 M sucrose for 15 min at 37°C before adding LPS.

Internalization of FITC-LPS by nonadherent monocytes was measured by flow cytometry by quenching surface-exposed FITC-LPS with rabbit antifluorescein IgG (Texas Red conjugate), as previously described (14). Briefly, serum-free PBMC (23 × 10⁶ cells in 90 μl SFM) were warmed to 37°C for 15 min in the presence or absence of 0.45 M sucrose, 10 μl FITC-LPS in CHO-rLBP, or CHO-RSV supernatant were then added to make a final LPS concentration of 100–200 ng/ml, and the incubation was continued for an additional 10 min. The cells were washed with ice-cold PBS and analyzed by flow cytometry, as described (14). The mean fluorescence intensities (MFI) of the monocyte populations (gated by forward and side angle light scatter) were analyzed.

Internalization of BODIPY FL-LPS by THP-1 cells was measured by flow cytometry either by removing surface-exposed BODIPY-LPS with proteinase K or by quenching its fluorescence with trypan blue. The cells (6.3 × 10⁵ cells in 90 μl SFM) were incubated with 100–200 ng/ml BODIPY-LPS in the presence of rLBP and washed with cold PBS, as described above. Some aliquots of cells were stripped with ice-cold proteinase K, and other aliquots were resuspended in cold PBS. The unfixed cells were then analyzed by flow cytometry; then the cells were mixed with an equal volume of ice-cold 0.2% trypan blue in PBS for 1–2 min and reanalyzed. Data analysis was performed as described for FITC-LPS (14).

In all assays, the nonspecific (or non-CD14) binding of LPS to the cells was determined by incubating the cells with labeled LPS in the absence of rLBP or sCD14. Internalization of ¹²⁵I-Tf was measured by removing surface-exposed Tf by exposure to a low pH buffer, as previously described (41). Briefly, the cells were prepared in SFM as described above and incubated with 0.33 μg/ml ¹²⁵I-Tf for 1 h on ice. The cells were then warmed to 37°C for the indicated times and washed in ice-cold PBS, and surface-exposed ¹²⁵I-Tf was removed by washing the cells in ice-cold 0.2 M acetic acid in 0.5 M NaCl, pH 2.7, for 6 min. Radioactivity in acid supernatants and cell lysates was measured by liquid scintillation, counted as described (14). Nonspecific binding was determined by adding 100 μg/ml of unlabeled Tf. (SFM did not contain unlabeled Tf.)

DNP-LPS (500 ng LPS/ml SFM) was allowed to bind to nonadherent THP-1 cells or adherent monocytes on glass coverslips for 1.5 min at 37°C in the presence of CHO-rLBP or CHO-RSV supernatant. The cells were then washed with ice-cold SFM and kept on ice for the following incubations: 100 μl of blocking medium (SFM containing 1% BSA, 5% normal heat-inactivated goat serum, and 5% human serum (heat inactivated) was added to the cells to block nonspecific and FcR-mediated binding of subsequently added Abs. The cells were then incubated for 30 min with 10 μg/ml anti-DNP mAb HDP-1 in blocking medium, washed with SFM, and incubated for 30 min with goat anti-mouse IgG 10-nm gold conjugate diluted 1/10 in blocking medium. The THP-1 cells were gently washed by adding 10 ml cold SFM and centrifuged for 10 min at 52 × g (adherent monocytes were washed with 2 ml of cold SFM). The cells were mixed with 100 μl of SFM and warmed for 0–2.5 min in a 37°C water bath to allow LPS internalization. The cells were placed on ice, washed twice with PBS, and fixed for 1 h in a solution containing 2% glutaraldehyde and 3 mM picric acid in NaPi buffer (100 mM sodium phosphate, pH 7.4, containing 3 mM KCl and 3 mM MgCl₂). Approximately 20 min after adding the fixative, the adherent monocytes were scraped off the coverslips with a rubber policeman, and the monocytes or THP-1 cells were centrifuged for 1 min at 12,000 × g in a 1.5-ml microfuge tube to fix the cells together in a small pellet. The cells were postfixed for 1 h at room temperature in 2% osmium tetroxide and 1.5% potassium ferrocyanide in NaPi buffer, followed by 0.05% tannic acid in NaPi buffer for 30 min. They were then washed with distilled water and dehydrated by exposure to increasing concentrations of ethanol (30, 50, 70, 90, and 100%). The cells were stained for 1 h with 0.25% uranyl acetate in 70% ethanol during the dehydration procedure. The dehydrated cells were washed with propylene oxide and embedded in plastic (EMbed 812), according to the manufacturer’s protocol. Thin (90 nm) and semithick (300 nm) sections were cut with a diamond knife and mounted on uncoated nickel or copper grids (200 mesh). The 90-nm thin sections were stained with uranyl acetate and lead citrate. The sections were viewed with a Jeol JEM-100SX electron microscope.

THP-1 cells (3 × 10⁵ cells in 40 μl SFM) were placed on ice and mixed with 5 μl BODIPY-Tf (30 μg/ml final concentration) for 1 h. A total of 5 μl of Texas Red-LPS (complexed with rLBP or sCD14, as described above; 30–100 or 200 ng LPS/ml final concentration, respectively) was added, and the incubation was continued for an additional 15 min. The cells were then warmed in a 37°C water bath for 0–5 min to allow internalization of the bound ligands. In some experiments, the cells were washed with cold SFM and reincubated at 37°C for 5–15 min. The cells were washed with cold PBS, and surface-exposed LPS and Tf were removed by exposing the cells to 1 ml of ice-cold 0.02% proteinase K for 30 min (to remove LPS) (14), followed by 0.2 M acetic acid with 0.5 M NaCl, pH 2.7, for 6 min (to remove Tf). The cells were fixed for 30 min in cold 4% paraformaldehyde in NaPi buffer containing 0.5 mM PMSF, centrifuged onto poly(l-lysine)-coated slides, and mounted in under No. 1 glass coverslips, as previously described (14). The cells were viewed with an MRC-1024 laser confocal imaging system (Bio-Rad, Hercules, CA) using a ×63 objective lens. Sequential optical sections (0.8 μm) were collected digitally with a resolution of 0.155 μm per pixel. When only one fluorescent ligand was applied to the cells, its fluorescence did not overlap detectably into the fluorescence channel used for the other fluorophore. Nonspecific binding was virtually undetectable when the cells were incubated with labeled LPS in the absence of rLBP or sCD14 or with labeled Tf in the presence of 3 mg/ml unlabeled Tf.

PMA-stimulated pinocytosis (fluid-phase uptake) of LY (42) was measured in THP-1 cells (6.3 × 10⁵ cells/90 μl SFM) after incubating the cells for 10 min at 37°C in the presence or absence of the inhibitor, 300 μM dimethylamiloride, or control medium containing an equivalent amount of dimethlysulfoxide carrier. LY (0.5 mg/ml) was then added with or without PMA (100 nM), and the incubations were continued for an additional 30 min. The cells were washed thoroughly with PBS, and cell-associated LY was measured by flow cytometry (see above) using excitation and emission wavelengths of 457 and 530 nm, respectively. The MFI of the THP-1 cell populations (gated by forward and side angle light scatter) were determined, and the MFI of cells that had been exposed to LY without warming to 37°C was subtracted. Phagocytosis of BODIPY FL-labeled E. coli by THP-1 cells was measured by flow cytometry in the presence of trypan blue, essentially as described by Schiff et al. (43). Briefly, THP-1 cells in SFM were incubated in the presence or absence of 10 μM cytochalasin H or D for 30 min at 37°C, mixed with LBP-opsonized BODIPY-E. coli (BODIPY-E. coli was preincubated with CHO-rLBP supernatant for 30 min at 37°C), and the incubation was continued for an additional 60 min. The MFI of gated THP-1 cell populations were measured before and after mixing the cells with trypan blue.

We first tested whether agents that disrupt coated pits could block CD14-mediated endocytosis of LPS. For these experiments, we used CD14-transfected THP-1 cells or normal human monocytes. Surface-exposed [³H]LPS was removed by incubating the cells with ice-cold proteinase K (14), and the remaining protease-resistant, cell-associated LPS was considered to be internal. Sequestration of LPS from proteinase K was strongly inhibited by agents that deplete intracellular ATP (14) or by keeping the cells on ice (see below).

Brief incubation of cells in hypertonic medium disrupts clathrin-coated pits in diverse cell types (44, 45). We found that exposure to hypertonic medium (0.45 M sucrose) eliminated coated pits on THP-1 cells. Using thin-section (90-nm) electron microscopy, we found 51 coated pits per millimeter (0.901 mm analyzed) of cell surface on untreated cells, whereas no coated pits were found on cells that had been treated with hypertonic sucrose (0.482 mm analyzed). In addition, we measured the receptor-mediated endocytosis of Tf, a protein that enters cells via coated pits (41), and found that hypertonic sucrose treatment reduced the internalization of cell-associated¹²⁵I-Tf from 74 ± 2% (SD, control cells, n = 4) to 12 ± 3% (treated cells, n = 4) (Fig. 2). In contrast, CD14-dependent internalization of [³H]LPS aggregates and monomers was only slightly inhibited under the same conditions (Fig. 3). As we have previously shown (14), monomeric [³H]LPS that binds to mCD14 was internalized more slowly than aggregated LPS, and a considerably smaller percentage of the total cell-associated LPS was internalized. This accelerating effect of LPS aggregation also occurred in the presence of hypertonic medium (Fig. 3, B and D).

Exposure of the cells to chlorpromazine, a partial inhibitor of coated pit function (46), decreased ¹²⁵I-Tf internalization by 30–50%, but had no effect on LPS internalization (not shown).

The kinetics of LPS internalization by human monocytes in suspension were similar to those found for THP-1 cells (not shown). The data in Table I show that in monocytes, as in THP-1 cells, hypertonic medium did not inhibit the internalization of [³H]LPS by adherent cells or of FITC-LPS by cells in suspension (measured by Ab quenching of surface-exposed FITC-LPS (14)). Electron microscope (EM) thin-section analysis of monocytes showed that exposure to hypertonic medium reduced coated pits from 71 per mm of cell surface (0.69 mm analyzed) to none (0.65 mm analyzed).

To divert LPS into coated pits, we transfected THP-1 cells with a chimeric CD14 receptor (CD14-LDLR) that contained the transmembrane and cytoplasmic domains of the LDLR (Fig. 1). Unlike CD14-GPI, CD14-LDLR was not released from the cell surface by treatment with phosphatidylinositol-specific phospholipase C (not shown). CD14-GPI was also largely insoluble in Triton X-100, whereas CD14-LDLR was completely solubilized by this treatment (47) (not shown).

The rate and extent of LPS internalization were increased significantly in cells expressing CD14-LDLR (Fig. 4) compared with those expressing CD14-GPI (Fig. 3), whether the LPS was in aggregated (Fig. 4,B vs Fig. 3,B) or monomeric form (Fig. 4,D vs Fig. 3,D). Hypertonic treatment of cells expressing CD14-LDLR resulted in a sharp (50–75%) decrease in LPS internalization (Fig. 4, A and C). This decrease could be accounted for by a loss of LPS binding due to the loss or sequestration of CD14 receptors, as determined by FACS analysis of cell surface CD14 (not shown). Unexpectedly, the rate of internalization of both aggregated and monomeric [³H]LPS bound by the remaining CD14-LDLRs (Fig. 4, B and D) was similar to that in the untreated cells, indicating that CD14-LDLR can internalize LPS in the absence of coated pits. To confirm that coated pits were disrupted by hypertonic treatment of these cells, we measured the internalization of ¹²⁵I-Tf and [³H]LPS in aliquots of cells from the same experiment and found that Tf internalization was strongly inhibited (not shown).

To observe the morphology of endocytic structures that internalized LPS aggregates, we bound DNP-LPS to cell surface CD14 in the presence of LBP and observed the internalization of gold particles that were attached to the DNP-LPS by an anti-DNP mAb.

No gold particles were found on sections of CD14-transfected THP-1 cells or monocytes that had been incubated with DNP-LPS in the absence of LBP or on THP-1 cells that were not transfected with CD14 (not shown), indicating that neither the DNP-LPS nor the gold conjugates bound nonspecifically to the cells. In the presence of LBP, cells that expressed GPI-anchored CD14 accumulated DNP-LPS in noncoated invaginations, tubules, and vacuolar structures of various shapes and sizes (Fig. 5,, A–G). The noncoated invaginations were usually tubular in structure (Fig. 5, A, B, C, and F), and the omega or flask-shape morphology that is characteristic of caveolae (20, 22, 48) was rarely seen. The tubular invaginations that contained LPS had an average diameter of 57 nm ± 28 SD (n = 22), and their lengths varied in the planes of the sections from 74 to 850 nm. The diameters of these invaginations were much more variable that those of coated pits, which were 66 nm ± 9 SD (n = 10). LPS was also found in tubules (27–133 nm diameter) and electron-lucent vacuoles (100–500 nm) that appeared to be intracellular (not connected to the plasma membrane in the plane of section). The fact that very few of these structures were labeled when the cells were not warmed to 37°C (Table II) indicates that they were either intracellular or connected to the surface by narrow openings that restricted the movement of gold particles. The vacuoles frequently had an irregular or convoluted structure, suggesting that they were probably cross-sections of tubular inclusions, and some of the vacuoles were found to be connected to the surface by tubules or narrow invaginations (Fig. 5, C and D). LPS was also found in coated pits and vesicles, but much less frequently than in noncoated structures.

In cells that expressed CD14-LDLR, relatively little LPS was found in coated pits before warming. After warming, most of the coated pits that contained gold particles were found at the cell surface (Fig. 5,I), but coated pits were frequently found as extensions of tubular invaginations (Fig. 5,H), internal tubules, and vacuoles (Fig. 5 K) in these cells.

The data in Table II show a quantitation of the EM data. We counted the gold particles associated with DNP-LPS and determined the percentages of the total cell-associated gold particles (total LPS) contained in the various structures in cells expressing either CD14-GPI or CD14-LDLR. The striking differences in the locations of the LPS-Ab-gold complexes in coated versus noncoated structures in the two transfected cell lines suggest strongly that the dominant localizing factor was the CD14 anchor, and that the presence of Abs or gold in the complexes was not determinative. In cells expressing CD14-GPI, only 0.6% (1 min) to 2.7% (2.5 min) of the gold was found in coated pits and vesicles, and 10-fold more LPS-associated gold particles entered noncoated than coated structures. In cells expressing CD14-LDLR, in contrast, we found approximately 10% of the gold particles in coated pits or vesicles. Although in these cells a large fraction of the internalized gold was also found in tubular invaginations, tubules, or vacuoles, at least 12% of the gold in these noncoated structures was in coated pits that appeared to bud from them (Fig. 5 K). In these cells, LPS may thus enter coated pits either at the cell surface or after it is internalized into tubular structures.

Peripheral blood monocytes internalized LPS into similar structures (Fig. 6). In monocytes, we found LPS more frequently in tubular invaginations and in apparently intracellular tubular and vacuolar structures, whereas relatively little LPS was found in coated pits and vesicles (Table II). In these cells, the coated pits that contained LPS were found in tubular invaginations and other tubular or vacuolar structures, and 3.5% of the gold particles in these noncoated structures were in coated pits.

Surface-connected tubules have been shown to sequester β-VLDL in murine macrophages (34, 35). Before complete internalization, the β-VLDL is poorly accessible to Abs, whereas it is readily accessible to trypan blue, a low m.w. fluorescence quencher. As shown in Figs. 5 and 6, we found internalized LPS in surface-connected tubular structures. Accordingly, we tested whether cell-associated BODIPY FL-LPS was more accessible to trypan blue (M_r = 961) than to proteinase K (M_r = 28,900). As shown in Table III, the ability of proteinase K or trypan blue to remove or quench BODIPY-LPS that was bound to mCD14 on the cell surface at 0–4°C was equivalent. However, after warming the cells to 37°C, 22–28% of the BODIPY FL-LPS that was not removed by proteinase K was accessible to trypan blue (Table III). These findings are consistent with the conclusion that, like β-VLDL, LPS passes through intracellular structures that, while open to the surface, are relatively inaccessible to large extracellular probes; these structures are most likely the tubular invaginations visualized using electron microscopy. In each cell line, the effect of hypertonic medium on LPS internalization was similar whether surface-exposed LPS was detected using proteinase K or trypan blue (data not shown).

We next asked whether CD14-GPI cells internalize LPS into endosomes that also contain Tf, a well-established marker for early sorting endosomes that accumulate the contents of coated pits and coated vesicles (35). We bound Texas Red-DAg-LPS and BODIPY-Tf to THP-1 cells that expressed either CD14-GPI or CD14-LDLR, warmed the cells to 37°C for 3 or 5 min, and viewed the cells with a laser confocal microscope. In cells that were not rewarmed, both ligands uniformly stained the surfaces of receptor-positive cells without punctate focal accumulations (not shown). Focal accumulations of both LPS and Tf appeared after warming the cells to 37°C, and these accumulations increased in intensity and apparent size over time. We removed the surface-exposed ligands so that the locations of the internalized ligands could be clearly evaluated. In cells that expressed CD14-GPI, Texas Red-LPS and BODIPY-Tf (Fig. 7) (or BODIPY-LPS and Texas Red-Tf; not shown) accumulated predominantly in different locations after 3 min of internalization (Fig. 7,, A–C). After 5 min, LPS and Tf partially colocalized (Fig. 7,, G–I), but extended incubation (up to 20 min) did not increase the coincidence of the two ligands (not shown). In contrast, in cells expressing CD14-LDLR, Texas Red-LPS and BODIPY-Tf nearly always accumulated in the same locations after 5 min at 37°C (Fig. 7,, J–L). At earlier (3 min) time points (Fig. 7,, D–F), however, there were differences in the locations of LPS and Tf; this finding is consistent with the conclusion that some of the LPS internalized by CD14-LDLR cells can be internalized rapidly into noncoated structures before it enters coated pits (see Table II).

These observations were made using LPS aggregates (DAg-LPS). When we incubated cells expressing CD14-GPI with monomeric Texas Red-LPS for 5 min at 37°C, the intracellular LPS was diffusely localized along the cell periphery (Fig. 8,A). The location of internalized LPS monomers did not correspond to the location of Tf-containing vesicles (Fig. 8,B). In contrast, in cells that expressed CD14-LDLR, monomeric LPS accumulated in discrete foci that usually overlapped with accumulations of Tf (Fig. 8, C and D).

We next asked whether nonclathrin-mediated LPS internalization involves macropinocytosis or phagocytosis. The data presented above make it unlikely that LPS internalization requires macropinocytosis, since fluid-phase uptake of LPS is quantitatively trivial compared with CD14-mediated uptake (Fig. 3, compare uptake with or without LBP), and LPS stimulation of THP-1 cells in the presence of LBP does not greatly enhance fluid-phase uptake (not shown). Moreover, in contrast to previous studies that used higher concentrations of LPS (49, 50), our partially disaggregated LPS was not visible as LPS bilayers by thin-section EM analysis (Figs. 5 and 6), suggesting that these LPS aggregates are probably too small to require engulfment by the membrane “zipper” mechanism of phagocytosis. Nevertheless, the cellular mechanisms that are essential for these processes (e.g., membrane ruffling and actin polymerization) may be important for nonclathrin-dependent LPS internalization. To test this hypothesis, we first treated THP-1 cells with dimethylamiloride, a potent inhibitor of membrane ruffling and macropinocytosis (42, 51), and found that the drug had virtually no inhibitory effect on [³H]LPS internalization, whereas it strongly inhibited PMA-stimulated uptake of LY (Fig. 9). These data suggest that membrane ruffling is not responsible for the formation of noncoated invaginations that internalize LPS. We also treated the cells with inhibitors of actin polymerization and phagocytosis, cytochalasins H and D, and found that whereas CD14-dependent phagocytosis of BODIPY-E.coli (43) was almost completely inhibited by these drugs, CD14-dependent internalization of [³H]LPS aggregates was only partially inhibited (Fig. 9). We also found that cytochalasin D partially inhibited BODIPY-LPS internalization, as measured by quenching surface LPS with trypan blue (see Materials and Methods). In three experiments, the internalized fraction of BODIPY-LPS in cytochalasin-treated cells was 68 ± 13% SD (n = 6) of that of untreated control cells. (Control cells internalized 22 ± 3% of the total cell-associated LPS, MFI = 42 ± 8, in 10 min). These data suggest that while actin polymerization may have a role in LPS internalization, the endocytic mechanism appears to be distinct from that of phagocytosis.

Although previous studies showed that LPS signaling occurs normally in cells that express CD14, regardless of the structure of its membrane anchor (52, 53), none of the anchors used for those studies contained specific targeting signals. In this study, we tested whether an anchor that has a coated pit-targeting signal (i.e., CD14-LDLR) would alter the ability of the cells to respond to LPS. As noted above, however, CD14-LDLR does not target LPS exclusively to coated pits: it often directs LPS to coated pits that are attached to noncoated invaginations. Moreover, CD14-GPI could also direct some LPS to coated pits. Although similar LPS dose responses for nuclear factor-κB translocation and IL-8 production were found in cells that expressed equivalent amounts of either CD14-GPI or CD14-LDLR (not shown), the lack of localization specificity frustrates interpretation of these results. On the other hand, we found that cells that internalize LPS via CD14-GPI or CD14-LDLR perform LPS deacylation at closely similar rates, suggesting that both clathrin-coated and noncoated structures can target LPS to the endosomes, where deacylation is presumed to occur (54) (not shown).

We used three complementary strategies to study the initial steps in LPS internalization by monocytic cells. The kinetics of LPS internalization were quantitated by measuring the disappearance of the ligand from the cell surface, which we defined operationally as the loss of accessibility to extracellular proteinase K or trypan blue. Disruption of coated pits with hypertonic media was used to determine whether LPS endocytosis was predominantly clathrin mediated. The movement of LPS-Ab-gold complexes into the cell was then tracked using thin- and thick-section transmission electron microscopy. Finally, laser confocal microscopy was used to study the uptake of fluorescent Tf and LPS into intracellular vesicles. In each of these approaches, we compared LPS internalization in cells that express GPI-anchored CD14 with that in cells expressing CD14 modified so as to favor cell entry via clathrin-coated pits (CD14-LDLR).

We also limited our analysis to LPS that bound CD14, the major receptor for LPS on phagocytes, by using relatively low concentrations of LPS and by binding LPS to cells using LBP or sCD14. Our approach therefore differed substantially from previous ultrastructural analyses of LPS internalization, which used much greater concentrations of LPS and/or LBP- and sCD14-free conditions (49, 50, 55, 56, 57).

Three lines of evidence support the conclusion that the predominant pathway of LPS internalization mediated by GPI-anchored CD14 is nonclathrin mediated: 1) LPS internalization was relatively insensitive to the effects of hypertonic medium (Fig. 3), which destroyed coated pits and strongly inhibited Tf internalization (Fig. 2). 2) EM analysis showed that gold particles attached to DNP-LPS accumulated rapidly (1 min) in noncoated invaginations and intracellular vesicles in THP-1 cells (Table II and Fig. 5) and monocytes (Fig. 6), whereas at least 10-fold less DNP-LPS was found in clathrin-coated pits and vesicles. 3) Confocal microscope analysis of internalized Texas Red-LPS and BODIPY-Tf showed that these ligands accumulated in different intracellular locations immediately (3 min) after internalization by CD14-GPI cells (Fig. 7,, A–C), and that lack of colocalization persisted to a significant extent for many minutes thereafter (Fig. 7 , G–I).

EM analysis of cells expressing CD14-GPI showed that a small percentage of cell-associated DNP-LPS entered coated pits (Fig. 5,G and Table II). This observation is consistent with the apparent colocalization of some intracellular Texas Red-LPS with BODIPY-Tf (Fig. 7 , G–I). Intracellular accumulations of LPS and Tf occurred mostly in different locations after 3 min of internalization, but appeared to converge partially by 5 min. Colocalization of the LPS- and Tf-containing compartments was never so complete in these cells as in cells in which LPS was bound to CD14-LDLR.

Although the bulk of the LPS entered cells expressing CD14-GPI by a nonclathrin pathway, the failure of fluorescent LPS and Tf to colocalize completely in intracellular foci was somewhat surprising in view of previous studies of nonclathrin-mediated endocytosis (31, 58, 59). For example, Hansen et al. (58) showed that in potassium-depleted HEp-2 cells, Con A-gold is taken up by nonclathrin-mediated endocytosis and moves to early endosomes that contain Tf receptors, although the internalized Con A-gold is ultimately excluded from late endosomes and lysosomes. Perhaps the noncoated invaginations that internalize LPS turn over more slowly than coated pits and coated vesicles (34), or they may fail to fuse efficiently with early endosomes. Whether endosomes from the clathrin-mediated pathway fuse with those from nonclathrin-mediated pathways may differ in various cell types (35).

We produced the CD14-LDLR chimeric receptor to direct CD14-bound LPS into coated pits. As shown in the electron-microscope images (Fig. 5, H and J, and Table II), however, LDLR-CD14 cells also internalize LPS via nonclathrin structures. Monocyte/macrophages, like hepatocytes and many epithelial cells (60), evidently do not restrict the membrane location of LDLR to coated pits. Indeed, β-VLDL, which binds to LDL (apoE/B) receptors (61), also moves into noncoated tubular invaginations in macrophages (34). Treatment of CD14-LDLR cells with hypertonic sucrose blocked Tf internalization and diminished the number of cell surface CD14 receptors without significantly diminishing the rate of LPS internalization (Fig. 4). This observation suggests that these cells can divert LPS almost entirely into noncoated structures, in keeping with previous observations that inhibition of coated pit function may up-regulate nonclathrin-mediated endocytosis in other cell types (62).

We also found that noncoated tubular invaginations, tubules, and vacuoles can contain coated pits (Fig. 5, H and K). In THP-1 cells expressing CD14-LDLR (and to a lesser extent in human monocytes), the immunogold-LPS found in tubular invaginations was often in, or near, these coated structures. In these cells, therefore, the LDLR anchor may target CD14 to coated pits that exist either on the cell surface or within surface-connected membrane invaginations. This phenomenon may account for the observations that, in cells expressing CD14-LDLR, 1) similar amounts of immunogold-LPS were found in coated and noncoated structures (Table II), yet 2) after internalization, Texas Red-LPS overlapped substantially with BODIPY-Tf, even at early time points (Fig. 7). Presumably, coated vesicles derived from different membranes (plasma membrane and noncoated tubular invaginations or vesicles) fuse with early endosomes.

The results of the EM analysis of LPS internalization (Fig. 5, Table II) suggest that nonclathrin-mediated endocytosis of LPS aggregates occurs via tubular invaginations of the plasma membrane. The diameters of the tubular invaginations (57 ± 28 nm) were similar to those of coated pits (66 ± 9 nm) and those reported for caveolae (50–80 nm (27)). Morphologically, the tubular invaginations resemble the tubular pinosomes, noted in alveolar macrophages, that take up horseradish peroxidase and contain acid phosphatase (33). They also resemble the surface-connected tubules (STEMs), described in murine macrophages, that internalize and partially process large β-VLDL particles (34, 35, 63), although STEMs are significantly larger in diameter (∼250 nm). The roles of noncoated vesicles in LPS signaling and intracellular processing and their relationship to low-density, lipid-enriched membrane microdomains that bind CD14-bound LPS (64) are under investigation.

The kinetics of LPS internalization by cells expressing either CD14-GPI or CD14-LDLR is strongly influenced by the LPS aggregation state. Regardless of the CD14 anchor, aggregation greatly accelerates LPS movement into noncoated structures and/or the endosomes derived from them. Monomeric LPS enters these structures much more slowly in cells that express CD14-GPI, so that much of the cell-associated LPS remains on the surface over time. In cells that express CD14-LDLR, however, monomeric LPS that binds CD14 is targeted to coated pits, so that it disappears more rapidly from the cell surface into Tf-containing endosomes. This formulation is consistent with the reported pattern of β-VLDL internalization by macrophages, since the presence of multiple apolipoprotein E molecules in large β-VLDL particles also promotes movement into surface-connected tubules rather than coated pits (34, 63, 65). In keeping with these results, the internalization-accelerating effect of LPS aggregation occurred even in cells that had no functional coated pits (Figs. 3 and 4).

While some LPS internalization is mediated by clathrin-coated pits, most occurs via nonclathrin-coated membrane invaginations and tubules. Aggregation promotes entry by accelerating uptake via the noncoated pathway. Like the plant-derived protein toxin, ricin (29), LPS thus has a complex pattern of cell entry, and it seems reasonable to expect similarly complex pathways of intracellular movement. Moreover, the fate of the LPS studied by various techniques may be different from that of a much smaller population of LPS molecules that has important or different biologic consequences. Understanding the role of internalization in LPS signaling centers on this issue; it is entirely possible that the “bulk” LPS, followed by virtue of its radioactivity or a visual tag, does not include a small population of molecules that triggers cellular responses. Another possibility, given the pleiotropic nature of responses to LPS, is that different cellular reactions are initiated by the LPS that finds its way into different surface domains or intracellular compartments. Sorting out these possibilities will be a major challenge for future research.

We thank Drs. Richard Anderson, Stephen Lacey, Michael Roth, and Leon Eidels for helpful advice; Stephen Lacey for providing the low-density lipoprotein receptor cDNA; and Rolf Thieringer for providing soluble CD14.