The endothelial response to LPS is critical in the recruitment of leukocytes, thereby allowing the host to survive Gram-negative infection. Herein, we investigated the roles of soluble CD14 (sCD14) and membrane CD14 (mCD14) in the endothelial response to low level LPS (0.1 ng/ml), intermediate level LPS (10 ng/ml), and high level LPS (1000 ng/ml). Removal of sCD14 from serum and sCD14-negative serum prevented low level LPS detection and subsequent response. Addition of recombinant sCD14 back into the endothelial system rescued the endothelial response. GPI-linked mCD14 removal from endothelium or endothelial treatment with a CD14 mAb prevented responses to low-level LPS even in the presence of sCD14. This demonstrates essential nonoverlapping roles for both mCD14 and sCD14 in the detection of low-level LPS. At intermediate levels of LPS, sCD14 was not required, but blocking mCD14 still prevented endothelial LPS detection and E-selectin expression, even in the presence of sCD14, suggesting that sCD14 cannot substitute for mCD14. At very high levels of LPS, the absence of mCD14 and sCD14 did not abrogate TLR4-dependent, E-selectin synthesis in response to LPS. The MyD88 independent pathway was detected in endothelium (presence of TRIF-related adaptor molecule TRAM). The MyD88-independent response (IFN-β) in endothelium required mCD14 even at the highest LPS dose tested. Our results demonstrate an essential role for endothelial mCD14 that cannot be replaced by sCD14. Furthermore, we have provided evidence for a TRAM pathway in endothelium that is dependent on mCD14 even when other responses are no longer mCD14 dependent.

Lipopolysaccharide is a major constituent of the outer membrane of Gram-negative bacteria, shed during infection and harbors very proinflammatory properties. The LPS receptor is composed of at least three proteins: TLR4, MD-2, and CD14. The 55-kDa glycoprotein CD14 is present in the body in two main forms, a soluble form, (found for example in plasma) (1) and a membrane-bound form, demonstrated to be GPI-linked to the outer leaflet of the plasma membrane (2, 3). CD14 alone cannot cause cellular activation and hence the type 1 transmembrane receptor,TLR4 is required. The generally accepted scheme for LPS responses to monocytes (CD14- bearing cells) is that LPS-binding protein (LBP)3 in serum binds shed LPS and shuttles it to membrane CD14 (mCD14) which then associates with the TLR4-MD-2 complex to initiate a downstream signal, causing a proinflammatory response such as leukocyte recruitment. By contrast, cells thought to be devoid of mCD14 make use of soluble CD14 (sCD14) to respond to LPS.

Endothelium can detect LPS and the response is critical to the recruitment of leukocytes. Endothelium was generally accepted as being CD14 negative and sCD14 was thought to completely substitute for mCD14, thus allowing the endothelium to be activated (4, 5, 6, 7, 8). However, Jersmann et al. (9) suggested that CD14 was present on endothelium and that passaging of the endothelium reduced CD14 expression. This work was recently both confirmed and extended (10). Lloyd and Kubes (10) reported that endothelium can synthesize CD14 and that the CD14 was GPI-linked to the cell surface and by definition mCD14, not just sCD14 that was trapped to the endothelial surface as had been suggested by others. Furthermore, LPS detection by CD14 could be blocked by pretreatment of the endothelium with phosphatidylinositol-phospholipase C (PI-PLC) to cleave the GPI link and by pretreatment of the endothelium with a CD14-blocking mAb, MEM-18 (10). Blocking CD14 function on endothelium resulted in nonresponsiveness of the endothelium to LPS (as measured by E-selectin expression and leukocyte recruitment). In addition, a plasma-dependent response was noted at low LPS levels (<1 ng/ml) and a plasma-independent response at intermediate LPS levels (>1 ng/ml), suggesting potentially varying roles for plasma proteins including perhaps sCD14.

There is increasing complexity downstream of CD14 and TLR4. TLR4-MD-2 activation generates a bivalent signaling cascade consisting of a MyD88-dependent and MyD88-independent pathway (11, 12, 13, 14). MyD88, a Toll/IL-1 receptor (TIR) domain containing adaptor molecules, recruits IL-1R-associated kinase and TNFR-associated factor 6, which then amplify the signal. This pathway directly leads to activation of the nuclear factor NF-κB, which is essential to the production of inflammatory cytokines and adhesion molecules important in leukocyte recruitment (such as VCAM-1, ICAM-1, and E-selectin). The MyD88-independent pathway downstream of the TIR domains requires the adaptor molecules TIR domain-containing adaptor-inducing IFN-β (TRIF) and TRIF-related adaptor molecule (TRAM) (14, 15). TRAM is a bridging adaptor and recruits TRIF to TLR4 (16). This pathway is associated with maturation of dendritic cells, slow activation of NF-κB, and the induction of IFN-inducible genes (17). TRIF is also present in the TLR3 pathway but TRAM is unique to the TLR4 pathway (14). Saito et al. (18) reported a requirement for mCD14 for the induction of LPS-induced IFN-β and presumably the TRIF-TRAM pathway in mouse macrophages. Moreover, in macrophages, CD14 was essential for activation of the MyD88-independent pathway (19), although discrimination between mCD14 and sCD14 was not investigated. In endothelium, very little is known about the MyD88-independent pathway but Harari et al. (20) reported absence of TRAM in passaged endothelium. However, much like the loss of CD14 in passaged endothelium (9, 10), TRAM may also follow a similar fate.

Two objectives were examined in this study. We addressed the specific role for sCD14 and mCD14 in endothelial detection of LPS. Furthermore, we examined the importance of the TRIF-TRAM pathway downstream of CD14 in primary endothelium. Our results reveal an essential role for mCD14 that cannot be replaced by sCD14. In addition, we have identified a TRAM pathway in primary endothelium that was entirely dependent on the presence of mCD14.

CD14 mAb used for plasma depletion of sCD14 was clone UCHM1 supplied by Ancell. CD14 Ab used for ELISA and blocking studies was unconjugated clone MEM-18 mAb supplied by Hycult Biotechnology. TRAM polyclonal Ab used was ALX-210-914 (Alexis Biochemicals). The isotype control (IgG1) for E-selectin ELISA of MOPC-21 was obtained from Sigma-Aldrich. Protein G Plus-agarose for plasma depletion of sCD14 was supplied by Santa Cruz Biotechnology. Recombinant sCD14 was purchased from R&D Systems. Smooth Escherichia coli O111:B4 LPS was purchased from Calbiochem and had no detectable contamination and no detectable response in TLR4-deficient mice. The human monoblastic leukemic cell line U937 and human muscle cell line TE671, as positive and negative controls, were generously supplied by Dr. S. Robbins and Dr. D. Muruve (respectively; University of Calgary, Alberta, Calgary, Canada). Human umbilical cords were obtained from the Foothills Hospital and endothelium was isolated from human umbilical veins and used at first or third passage, as described previously by Lloyd and Kubes (10).

Blood was obtained from healthy human donors and plasma was isolated. CD14-specific mAb was added to the plasma for 1 h on a rotating platform at 4°C. Agarose beads were added and incubated overnight at 4°C. Plasma was removed after gentle centrifugation. The plasma underwent this depletion protocol five times and the final plasma was tested for sCD14 with a sCD14 sandwich ELISA obtained from BioSource International revealed no detectable sCD14. Plasma was frozen at −80°C until required.

Whole blood was collected by cardiac puncture from CD14-deficient mice and their wild-type counterparts using a 1-ml syringe precoated with heparin (21). The blood was centrifuged for 8 min at 268 × g, plasma was collected, and stored at −80°C until required.

Endothelium was isolated and grown in 6-well plates. Membrane CD14 was removed by two methods: first, to remove GPI-linked proteins from the cell surface, the endothelium was treated with PI-PLC (0.1 U/ml) as described in the study by Lloyd and Kubes (10). This concentration of PI-PLC was shown to remove all GPI-linked CD14 but did not affect transmembrane proteins. Non-LPS responses (e.g., TNF-α) remained intact in PI-PLC-treated endothelium. Second, mCD14 was inactivated with a concentration of CD14 Ab which effectively blocks CD14 without affecting TNF-α-induced E-selectin expression. The endothelium was washed to remove any unbound Ab. The E-selectin ELISA was conducted as described elsewhere (10). The laminar flow chamber assay to study LPS-induced leukocyte recruitment on HUVEC under shear conditions was used as previously described (10).

TE671 cells and matured U937 cells were washed with sterile PBS, total RNA was isolated (RNeasy kit; Qiagen), dissolved in diethyl pyrocarbonate water, and reverse transcribed (OmniScript; Qiagen). PCR analysis was performed with the following specific primers: TRAM as forward primer 5′-CAA GTC CAG GAT ATC AT-3′ and reverse primer 5′-CAA ACC ATC AAT GCC TT-3′; IFN-β as both forward and reverse primers, as described (22), and GAPDH primers, as described (23). The TRAM final product was 494 bp, IFN-β final product was 451 bp, and GAPDH final product was 306 bp. PCR conditions included a 94°C hot start for 4 min followed by 30 cycles of denaturing at 94°C for 1 min, annealing temperature 56°C for 1 min (TRAM), (62°C for IFN-β), and extension at 72°C for 2 min. Samples were run on a 1% agarose gel containing SYBR safe.

HUVEC monolayers were lysed in a buffer containing 1% Triton X-100. Soluble lysates were separated by SDS-PAGE and proteins were transferred to a polyvinylidene difluoride membrane. Membranes were probed with a TRAM-specific Ab and bands were visualized using a multi-imager (Fluor-S MAX; Bio-Rad).

Values are expressed as means ± SE. Where appropriate, results were analyzed for statistical significance by using Student’s t test. For multiple comparisons, ANOVA with Bonferroni’s multiple comparison test was conducted. A p < 0.05 was taken to be significant.

In a previous study, we demonstrated that low-level LPS (0.01–0.1 ng/ml) could stimulate endothelium in the presence but not the absence of plasma (10). To determine whether sCD14 was important for low levels of LPS response, sCD14 was depleted from human plasma. This sCD14-depleted plasma was then added to primary HUVEC in the presence or absence of 0.1 ng/ml LPS. The data demonstrate that in the presence of normal control plasma, 0.1 ng/ml LPS increased endothelial E-selectin levels 3-fold over basal levels (Fig. 1,A). In contrast, the sCD14-depleted plasma (as assessed by a sCD14 ELISA, data not shown) was not able to up-regulate E-selectin expression in the presence of 0.1 ng/ml LPS. For completeness, addition of LPS without any plasma elicited no activation of endothelium (data not shown). Although we used a CD14-specific Ab, it could be that the depletion process was removing other relevant plasma proteins. Because of the similarity of mouse and human CD14, we verified these results using mouse plasma from CD14 knockout mice and their wild-type counterparts. The mouse plasma was added to HUVEC in the presence or absence of 0.1 ng/ml LPS. Mouse wild-type plasma on HUVEC caused E-selectin expression to be significantly up-regulated >2-fold in the presence of 0.1 ng/ml LPS (Fig. 1 B). In contrast, CD14-negative plasma in the presence of 0.1 ng/ml LPS was unable to induce E-selectin expression compared with CD14-negative plasma alone. These data suggest that sCD14 is critical in the detection of low-level LPS even if all other plasma proteins are available.

FIGURE 1.

sCD14-depleted plasma cannot reconstitute an endothelial response to low-level LPS (0.1 ng/ml). Control human plasma and sCD14-depleted human plasma were added to human endothelium in the presence or absence of low-level LPS (0.1 ng/ml) and the amount of E-selectin expression was measured. This was conducted in duplicate and a representative experiment is shown. Data are expressed as arithmetic mean ± SD (A). Control mouse plasma and sCD14 knockout mouse plasma was added to human endothelium in the presence or absence of low-level LPS (0.1 ng/ml) (B). E-selectin was measured using cell surface ELISA. Data are expressed as arithmetic mean ± SEM from four separate experiments. *, p < 0.01.

FIGURE 1.

sCD14-depleted plasma cannot reconstitute an endothelial response to low-level LPS (0.1 ng/ml). Control human plasma and sCD14-depleted human plasma were added to human endothelium in the presence or absence of low-level LPS (0.1 ng/ml) and the amount of E-selectin expression was measured. This was conducted in duplicate and a representative experiment is shown. Data are expressed as arithmetic mean ± SD (A). Control mouse plasma and sCD14 knockout mouse plasma was added to human endothelium in the presence or absence of low-level LPS (0.1 ng/ml) (B). E-selectin was measured using cell surface ELISA. Data are expressed as arithmetic mean ± SEM from four separate experiments. *, p < 0.01.

Close modal

Next, we examined whether sCD14 in the absence of other plasma proteins could reconstitute the endothelial response to 0.1 ng/ml LPS. The data demonstrate that when 0.1 ng/ml LPS is added to HUVEC in the absence of plasma, there is no significant up-regulation of E-selectin. However, addition of 0.1 μg/ml sCD14 with LPS induced a 5-fold increase in E-selectin (Fig. 2). Interestingly, sCD14 alone (no LPS) was able to activate endothelium to a small degree (Fig. 2), which was greatly increased at higher sCD14 concentrations (data not shown). This occurred despite <1 ng/μg of contaminating LPS in the sCD14 preparation and may reflect direct effects of sCD14 as previously reported (24).

FIGURE 2.

The addition of sCD14 reconstitutes the endothelial response to low-level LPS (0.1 ng/ml). sCD14 was added to mCD14-positive endothelium in the presence of HBSS as the endothelial cell-activating medium and the endothelial proinflammatory response to low-level LPS was measured using an E-selectin cell surface ELISA. Data are expressed as arithmetic mean ± SEM from four separate experiments. *, p < 0.05.

FIGURE 2.

The addition of sCD14 reconstitutes the endothelial response to low-level LPS (0.1 ng/ml). sCD14 was added to mCD14-positive endothelium in the presence of HBSS as the endothelial cell-activating medium and the endothelial proinflammatory response to low-level LPS was measured using an E-selectin cell surface ELISA. Data are expressed as arithmetic mean ± SEM from four separate experiments. *, p < 0.05.

Close modal

We previously reported that endothelium expressed mCD14 and could be removed with PI-PLC without affecting transmembrane proteins (10). The increase in E-selectin following 0.1 ng/ml LPS in the presence of sCD14 was entirely abrogated following cleavage of mCD14 from the surface of endothelium with PI-PLC (Fig. 3,A). PI-PLC-treated endothelium responded normally to non-LPS stimuli, including TNF-α (data not shown). The importance of mCD14 was further examined when HUVEC was stimulated with 10 ng/ml LPS, a concentration that does not require plasma. In the presence of mCD14 and sCD14, there was an 8-fold increase in E-selectin expression compared with untreated HUVEC (Fig. 3,B). Following removal of mCD14 with PI-PLC, there was no significant increase in E-selectin in response to LPS (Fig. 3 B).

FIGURE 3.

mCD14-negative endothelium was unable to reconstitute an endothelial response to 0.1 and 10 ng/ml LPS even in the presence of sCD14. Human endothelium was either treated with a GPI anchor-removing enzyme (PI-PLC) to remove mCD14 or left untreated. Proinflammatory responses of the endothelium were then measured using an E-selectin cell surface ELISA in response to 0.1 ng/ml LPS (A), 10 ng/ml LPS (B), and 1000 ng/ml LPS (C) where the endothelial cell-activating medium was HBSS. Data are expressed as arithmetic mean ± SEM from three to four separate experiments. #, p < 0.01 and *, p < 0.05.

FIGURE 3.

mCD14-negative endothelium was unable to reconstitute an endothelial response to 0.1 and 10 ng/ml LPS even in the presence of sCD14. Human endothelium was either treated with a GPI anchor-removing enzyme (PI-PLC) to remove mCD14 or left untreated. Proinflammatory responses of the endothelium were then measured using an E-selectin cell surface ELISA in response to 0.1 ng/ml LPS (A), 10 ng/ml LPS (B), and 1000 ng/ml LPS (C) where the endothelial cell-activating medium was HBSS. Data are expressed as arithmetic mean ± SEM from three to four separate experiments. #, p < 0.01 and *, p < 0.05.

Close modal

For completeness, we show that sufficiently high LPS concentrations (1000 ng/ml) no longer required either soluble or membrane CD14 (Fig. 3,C). In the presence of mCD14 and sCD14, there was a 7-fold increase in E-selectin expression following 1000 ng/ml LPS. Following cleavage of mCD14, a 7-fold increase in E-selectin expression was still observed in the presence of sCD14. Furthermore, in the absence of both mCD14 and sCD14, a 7-fold increase in E-selectin expression was still observed (Fig. 3 C). These data suggest that 1000 ng/ml LPS can induce a proinflammatory endothelial response, independent of sCD14 or mCD14.

To ensure that 1000 ng/ml LPS was still functioning via the TLR4 pathway, the lipid A-derived TLR4 competitive inhibitor E5564 (also known as Eritoran) was added to HUVEC along with LPS. E5564 in a dose-dependent manner caused a significant reduction of E-selectin expression in the presence of 1000 ng/ml LPS (Fig. 4). At a concentration of 1000 nM E5564, the E-selectin expression was completely abrogated, although even at the lowest concentrations of E5564 significant inhibition was noted. These data suggest that 1000 ng/ml LPS still directly activates TLR4 to lead to an LPS-induced proinflammatory response.

FIGURE 4.

One thousand nanograms of LPS per milliliter can activate the TLR-4-MD-2 complex in the absence of sCD14 and mCD14. Endothelium was treated with E5564 followed by 1000 ng/ml LPS in the presence of HBSS and proinflammatory responses in endothelium were measured using an E-selectin cell surface ELISA. Data are expressed as arithmetic mean ± SEM from one to three experiments. *, p < 0.05.

FIGURE 4.

One thousand nanograms of LPS per milliliter can activate the TLR-4-MD-2 complex in the absence of sCD14 and mCD14. Endothelium was treated with E5564 followed by 1000 ng/ml LPS in the presence of HBSS and proinflammatory responses in endothelium were measured using an E-selectin cell surface ELISA. Data are expressed as arithmetic mean ± SEM from one to three experiments. *, p < 0.05.

Close modal

The data suggest that sCD14 is not sufficient to induce an endothelial response to LPS as measured by E-selectin expression. To determine whether E-selectin expression translates to leukocyte recruitment, studies were conducted using a laminar flow chamber. It also gave us the opportunity to validate the PI-PLC inhibitory data using a CD14-specific blocking mAb (MEM-18). The leukocyte recruitment parameters (leukocyte rolling and adhesion) were not increased when either LPS (0.1 ng/ml) or sCD14 was administered alone (Fig. 5, A and B). However, HUVEC in the presence of sCD14 and 0.1 ng/ml LPS caused a 6-fold increase in the number of rolling leukocytes in response to the low-dose LPS. Pretreating LPS-stimulated HUVEC with MEM-18 (blocks only mCD14 as sCD14 was added after HUVEC was washed of excess Ab) reduced the number of rolling leukocytes to levels seen in unstimulated HUVEC (Fig. 5,A). Fig. 5,B shows that HUVEC in the presence of sCD14 or 0.1 ng/ml LPS alone did not induce significant leukocyte adhesion. However, when added together, a 12-fold increase in the number of adhering leukocytes was observed compared with unstimulated HUVEC (Fig. 5,B). Inhibition with the CD14 Ab MEM-18 was sufficient to block all leukocyte adhesion in the presence of sCD14 (Fig. 5 B).

FIGURE 5.

Leukocyte rolling and adhesion responses to LPS. Endothelium was grown to confluence on glass coverslips, stimulated with LPS where the endothelial cell-activating medium was HBSS, and blood was perfused across the endothelium. Leukocyte recruitment was recorded. Data were analyzed for the mean number of rolling (A, C, and E) and adhering (B, D, and F) leukocytes. Data are expressed as arithmetic mean ± SEM from three to four experiments. *, p < 0.05.

FIGURE 5.

Leukocyte rolling and adhesion responses to LPS. Endothelium was grown to confluence on glass coverslips, stimulated with LPS where the endothelial cell-activating medium was HBSS, and blood was perfused across the endothelium. Leukocyte recruitment was recorded. Data were analyzed for the mean number of rolling (A, C, and E) and adhering (B, D, and F) leukocytes. Data are expressed as arithmetic mean ± SEM from three to four experiments. *, p < 0.05.

Close modal

Fig. 5,C demonstrates that HUVEC stimulated with 10 ng/ml LPS in the absence of sCD14 induced a 3-fold increase in the number of rolling leukocytes. Addition of sCD14 and 10 ng/ml LPS caused a 7-fold increase in rolling cells compared with unstimulated HUVEC. However, in the absence of mCD14, a complete loss of rolling leukocytes was observed, despite the presence of sCD14 (Fig. 5,C). A similar pattern was observed with the number of adhering leukocytes (Fig. 5 D). These leukocyte recruitment data support our E-selectin expression data and suggest that mCD14 is essential for HUVEC to respond to 0.1 and 10 ng/ml LPS stimulation, and furthermore sCD14 cannot substitute for mCD14. The existence and importance of mCD14 is underscored by our observation of mCD14 on human lung endothelium (data not shown) and a report of mCD14 on endothelium in unperturbed umbilical cords (9).

In the last series of in vitro experiments, a high-dose LPS (1000 ng/ml) was used to stimulate endothelium (Fig. 5, E and F). The data show that when mCD14 is blocked by pretreating endothelium with CD14-blocking Ab, rolling and adhesion still occur. This suggests a CD14-independent response to LPS and supports our CD14-independent E-selectin expression data when high doses of LPS are used to stimulate the endothelium. Although a further increase in rolling and adhesion was noted when sCD14 was added to LPS, it did not reach significance (p = 0.08 and p = 0.1, respectively).

TLR4 activates two signaling pathways, a MyD88-dependent pathway and a MyD88-independent pathway, but in the absence of CD14 only the former is activated in macrophages. The MyD88-independent pathway requires both TIR-containing adaptor molecules TRIF and TRAM. TRIF is common between TLR4 and TLR3 signaling pathways, whereas TRAM is unique to TLR4. Whether the MyD88-independent pathway is present in endothelium is controversial since one study describes endothelium as being TRAM negative and another TRAM positive (20, 25). We examined both primary and passaged endothelium for evidence of TRAM. A band corresponding to the expected 450-bp product was detected in both primary and tertiary endothelium (Fig. 6,A, lanes 4 and 5, respectively). For controls, TRAM-specific message was observed in monocytes, as expected, and the water control showed no contamination (Fig. 6,A, lanes 2 and 1, respectively). Furthermore, we examined both primary and tertiary endothelium for TRAM protein. A band corresponding to the expected product was detected in monocytes, primary endothelium, and tertiary endothelium (Fig. 6 B, lanes 1, 3, and 4, respectively).

FIGURE 6.

TRAM message and protein are present in primary and passaged endothelium. Endothelium was used at first passage or passaged three times and both a RT-PCR and a Western blot for TRAM were performed. A, Lane 1, water control; lane 2, monocytes; lane 3, M167 cells; lane 4, primary HUVEC untreated; and lane 5, tertiary passaged HUVEC. B, Lane 1, monocytes; lane 2, HEK293 cells; lane 3, primary HUVEC; and lane 4, tertiary passaged HUVEC. The RT-PCR was conducted three times and the Western blot was repeated twice.

FIGURE 6.

TRAM message and protein are present in primary and passaged endothelium. Endothelium was used at first passage or passaged three times and both a RT-PCR and a Western blot for TRAM were performed. A, Lane 1, water control; lane 2, monocytes; lane 3, M167 cells; lane 4, primary HUVEC untreated; and lane 5, tertiary passaged HUVEC. B, Lane 1, monocytes; lane 2, HEK293 cells; lane 3, primary HUVEC; and lane 4, tertiary passaged HUVEC. The RT-PCR was conducted three times and the Western blot was repeated twice.

Close modal

A major product of the MyD88-independent pathway (TRIF-TRAM pathway) is the induction of IFN-β. Primary endothelium was stimulated with intermediate levels of LPS and, after 2 h, the endothelium was examined for IFN-β-specific message. IFN-β-specific message was observed in primary endothelium after stimulation with LPS in the absence of sCD14 (Fig. 7). The water control was negative, indicating that no contaminants were present (Fig. 7, lane 1). Monocytes stimulated with LPS were positive for IFN-β as expected (Fig. 7, lane 2).

FIGURE 7.

The TRIF/TRAM pathway is functional in primary endothelium. Endothelium was used at first passage and left untreated or treated with 10 ng/ml LPS where the endothelial cell-activating medium was HBSS. After 2 h of stimulation, total RNA was isolated, cDNA was synthesized, and a RT-PCR for IFN-β was conducted. Lane 1, Water control; lane 2, monocytes treated with 10 ng/ml LPS; lane 3, monocytes untreated; lane 4, primary HUVEC untreated; and lane 5, primary HUVEC treated with 10 ng/ml LPS. This experiment was repeated twice.

FIGURE 7.

The TRIF/TRAM pathway is functional in primary endothelium. Endothelium was used at first passage and left untreated or treated with 10 ng/ml LPS where the endothelial cell-activating medium was HBSS. After 2 h of stimulation, total RNA was isolated, cDNA was synthesized, and a RT-PCR for IFN-β was conducted. Lane 1, Water control; lane 2, monocytes treated with 10 ng/ml LPS; lane 3, monocytes untreated; lane 4, primary HUVEC untreated; and lane 5, primary HUVEC treated with 10 ng/ml LPS. This experiment was repeated twice.

Close modal

To investigate the importance of mCD14 in the IFN-β response on endothelium, mCD14 was removed from primary endothelium. We tested the high dose of LPS, since it was a dose that did not require CD14 for the induction of E-selectin expression. Therefore, to test for an absolute requirement of mCD14 for the IFN-β response, this high dose was used (Fig. 8). Unstimulated CD14-bearing endothelium was negative for IFN-β, as expected. Primary endothelium stimulated with LPS produced strong IFN-β signal. Not surprisingly, this strong band was also present when sCD14 was added. We then removed mCD14 from the endothelium by three different methods: 1) PI-PLC to cleave the GPI-link, 2) blocking mCD14 function by treating only endothelium with a mAb-specific for CD14, and 3) passaging the endothelium which causes loss of mCD14 (Fig. 8). Unstimulated endothelium pretreated with PI-PLC produced a slight IFN-β-specific band. When this endothelium was stimulated with LPS, no LPS-induced IFN-β band was detectable. In addition, the mCD14-negative endothelium did not produce an IFN-β-specific band even in the presence of sCD14. Similarly, endothelium pretreated with blocking mAb specific for CD14 was also negative for LPS-induced IFN-β, even in the presence of sCD14. Lastly, an LPS-induced IFN-β band on passaged endothelium was not detectable above untreated, even in the presence of sCD14. The highest LPS dose was used to test for CD14 dependence of the TRIF-TRAM (MyD88-independent) pathway because this dose was able to bypass CD14 to activate the MyD88-dependent pathway, thus we knew that TLR4 was able to detect LPS. Despite using the highest dose of LPS which requires no CD14 to stimulate E-selectin and leukocyte recruitment (MyD88-dependent), IFN-β production (MyD88-independent) required mCD14.

FIGURE 8.

mCD14 is required for function of the TRIF/TRAM pathway. Endothelium was used either at first passage or passaged an additional three times. The primary endothelium was either pretreated with 0.1 μl of PI-PLC, 10 μg/ml CD14-blocking Ab MEM-18, or left untreated. Both treated and untreated endothelium and passaged endothelium were treated with 1000 ng/ml LPS alone or 1000 ng/ml LPS plus 0.1 μg/ml sCD14 or left untreated where the endothelial cell-activating medium was HBSS. A RT-PCR for IFN-β was conducted. Lane 1, Water control. This experiment was repeated twice.

FIGURE 8.

mCD14 is required for function of the TRIF/TRAM pathway. Endothelium was used either at first passage or passaged an additional three times. The primary endothelium was either pretreated with 0.1 μl of PI-PLC, 10 μg/ml CD14-blocking Ab MEM-18, or left untreated. Both treated and untreated endothelium and passaged endothelium were treated with 1000 ng/ml LPS alone or 1000 ng/ml LPS plus 0.1 μg/ml sCD14 or left untreated where the endothelial cell-activating medium was HBSS. A RT-PCR for IFN-β was conducted. Lane 1, Water control. This experiment was repeated twice.

Close modal

Soluble CD14 was demonstrated to be key for LPS detection by human endothelium (4, 5, 6, 7, 8). These studies used a mCD14-negative-endothelial model, passaged HUVEC. According to these previous studies sCD14 enables mCD14-negative endothelium to detect LPS concentrations as low as 1 ng/ml (7). Recently we demonstrated that nonpassaged human endothelium expressed GPI-linked mCD14 and could detect LPS concentrations equal to or greater than 1 ng/ml in the absence of serum, thus negating a need for sCD14 (10). However, we observed that endothelium could detect even lower LPS levels (<1 ng/ml) but only in the presence of plasma (10). In this study, we clearly demonstrate that sCD14 in the serum was key in the endothelial detection of low-level LPS. We have also demonstrated that mCD14 was essential for detection and could not be replaced by sCD14. Moreover, we have demonstrated in endothelium differential requirements for CD14 in its soluble and membrane form depending on the concentration of LPS present. In addition to low doses requiring both sCD14 and mCD14, intermediate levels of LPS required mCD14 only, whereas high levels of LPS could bypass a requirement for either membrane or soluble forms of CD14 to induce E-selectin but still required mCD14 for IFN-β production.

Our initial studies looked at the relative importance of sCD14 to the detection of low-level LPS (0.1 ng/ml) by mCD14-positive endothelium. Both depletion of sCD14 from human serum and using serum from CD14-deficient mice abrogated endothelial activation in response to low-level LPS. Furthermore, when we added back sCD14, the endothelial inflammatory response to low-level LPS (0.1 ng/ml) was regained in mCD14-positive endothelium but not in endothelium lacking mCD14. By contrast, numerous investigators have shown that sCD14 can be added to passaged endothelium and replace LPS detection (5, 7, 8). Whether the passaged endothelium still had sufficient mCD14 to detect LPS in the presence of sCD14 was not determined by these other groups. Taken together, these results demonstrate the essential role of sCD14 for endothelial response to low-level LPS but also suggest that mCD14 is required. We do not rule out that higher levels of sCD14 could substitute for mCD14 for endothelial activation. However, we were unable to show this because sCD14 alone directly activated the endothelium as has been observed by others (24). Although LPS contamination could not be detected in sCD14 samples, one cannot absolutely exclude this possibility. It also remains possible that some permutation of an untested concentration of LPS and sCD14 could replace mCD14 but that was not revealed herein.

Previously, it has been suggested that the plasma protein LBP may also be required for endothelium to respond to LPS. Indeed, lesser sensitivity to LPS in the LBP knockout mouse was observed; however, a potent response could still be elicited (26). We previously reported that plasma and by inference LBP was not required for LPS concentrations of 1 ng/ml and higher. In this study, a role for LBP in LPS responses as low as 0.1 ng/ml for CD14-positive endothelium was also not essential. Whether LBP is more important in mice than humans is not clear but our experiments also made use of CD14-deficient mouse plasma that contained LBP, yet no response was noted, and we do not discount the possibility that LBP, much like sCD14, could increase the sensitivity of endothelium to LPS in humans, but this requires further investigation.

High-level LPS added to nonpassaged mCD14-removed endothelium caused a response even in the absence of sCD14. This is the first time that CD14-independent activation has been demonstrated on primary endothelium. Previous studies have demonstrated LPS binding but not activation independent of CD14 on endothelium (27) and a CD14-independent LPS receptor cluster containing heat shock proteins, chemokine receptor 4, and growth differentiation factor 5 has been implicated in a HUVEC cell line (28). CD14-independent inflammatory responses were also observed at high LPS concentrations (>10 ng/ml) in monocytes (29, 30, 31). In addition, CD14-independent mechanisms have been observed in CD14-deficient mice in vivo (32, 33) and ex vivo in mouse macrophages (34). One could question what a reasonable pathologic dose of LPS might be, since 1000 ng/ml LPS is above previously reported plasma levels (35) but levels as high as 10 μg/ml LPS have been reported in abscesses (30). Andonegui et al. (34) showed that certain doses of LPS required CD14 on some vascular endothelium, but not others, for inflammatory responses, also supporting a CD14-independent LPS response in vivo. Moreover, we show that despite a lack of need for CD14, the pathway was still dependent upon TLR4. E5564 is a lipid A derivative that has been reported to bind MD-2 and block TLR4 responses to LPS (36, 37). Preincubation of endothelium with E5564 and addition of 1000 ng/ml LPS caused abrogation of leukocyte recruitment. This was not due to E5564 causing toxicity to the endothelium since TNF-α responses were normal in E5564-treated endothelium, (data not shown). We speculate that at high LPS concentrations, there is sufficient LPS present to allow activation of the TLR4-MD-2 complex and to enable inflammatory responses without the requirement for CD14. However, it is worth re-emphasizing that although high levels of LPS did not require mCD14 for MyD88-dependent endothelial activation, MyD88-independent (TRIF-TRAM) activation did require mCD14, suggesting some potential for differential activation of these downstream pathways.

It is noteworthy that a controversy exists regarding the presence of TRAM in endothelium that is reminiscent of the mCD14 issue. Indeed, we demonstrate that the TRAM pathway is functional in primary endothelium but not passaged endothelium. However, TRAM is present in both primary and passaged endothelium. In fact, the lack of function of TRAM in the passaged endothelium is likely due to a lack of mCD14 on the endothelium. Moreover, removal of mCD14 from the primary endothelium ablated the LPS-induced IFN-β response and this response could not be rescued with sCD14. These observations in endothelium are similar to findings in macrophage cell lines. Saito et al. (18) demonstrated that a mouse macrophage cell line lacking mCD14 expression but able to release sCD14 could not detect LPS and failed in IFN-β production. In addition, CD14-deficient mouse monocytes were able to induce the MyD88-dependent but not the MyD88-independent pathway as reported by Jiang et al. (19). We add further to this body of work by reporting that LPS can bypass the MyD88-dependent but not the MyD88-independent pathway in the absence of human endothelial mCD14.

Membrane CD14 has been demonstrated to be present in lipid rafts (38, 39). Furthermore, it is possible that sCD14, lacking GPI linkage, cannot incorporate into lipid rafts and therefore fails to replace mCD14. It is generally accepted that upon stimulation with LPS, there is a clustering of molecules, including TLR4 within the raft associated with GPI-linked CD14. Lipid rafts are associated with lipid-linked signal transduction molecules such as heterotrimeric G proteins and src family kinases. Recently, TRAM was described as being N-terminal myristoylated to the plasma membrane, a linkage similar to that of src family kinases (40). Perhaps TRAM and mCD14 are located in the same raft. This would suggest that LPS stimulation causes TLR4 to cluster to mCD14 and TRAM, enabling activation of the TRIF-TRAM pathway. However, evidence of microscopy overlays suggests that TRAM is constitutively linked to TLR4 (40). TLR4 is known to be closely associated with mCD14 and if an overlay of mCD14 and TRAM expression was conducted this could also suggest linkage. In addition, Oshiumi et al. (16) suggest TRAM is constitutively bound to TLR4 from immunoprecipitation studies in HEK293 cells but LPS contamination could also explain the constitutive association.

Pugin et al. (39) have studied the importance of the CD14 GPI link. A transmembrane chimera of CD14 (using the transmembrane domain of tissue factor) and GPI-linked CD14 were both activated with LPS and responses were examined. They found that there was no difference in the NF-κB response to LPS (as measured by EMSA) if CD14 was linked to the membrane by a GPI link or by a transmembrane segment, concluding that the GPI link is not a requirement for CD14 function to respond to LPS (39). However, responses were focused on the MyD88-dependent pathway and perhaps a study of TRIF-TRAM pathway products would suggest a critical role for GPI-linked CD14. Alternatively, it could be that mCD14 can form larger clusters than the reported dimer formation of sCD14 and hence have a higher affinity to LPS (41).

This is the first study of differential roles of mCD14 and sCD14 in the CD14-bearing endothelial response to LPS looking at products from both signaling pathways of TLR4. Our findings demonstrate five important observations: 1) mCD14 and sCD14 are both important in the detection of low-level LPS (0.1 ng/ml), a level of detection previously not appreciated in endothelium. 2) Membrane CD14 is important in the detection of intermediate levels of LPS as well, and sCD14 cannot substitute for mCD14 in this regard. 3) The requirement of mCD14 can be overcome by a high level of LPS still signaling through TLR4, thus suggesting the role of mCD14 and sCD14 are to intensify the LPS signal with respect to E-selectin expression and leukocyte recruitment. 4) We demonstrate that the TRIF-TRAM pathway is present and functional in endothelium. 5) Moreover, mCD14 and not sCD14 is critical in the activation of the endothelial TRIF-TRAM pathway and even a high level of LPS cannot overcome this requirement. This suggests that mCD14 is not present to just intensify the signal for the TRIF-TRAM pathway, but perhaps has an essential role in bringing in other signaling molecules required for this pathway to be activated.

We thank Krista McRae and Erin McAvoy for technical assistance with the mouse cardiac punctures. We thank the voluntary healthy blood donors and Foothills Hospital Ward 52 for umbilical cord donations.

The authors have no financial conflict of interest.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1

This work was supported by a Canadian Institutes for Health Research operating grant and group grant. P.K. is a Canadian Research Chair and an Alberta Heritage Foundation for Medical Research Scientist and the Snyder Chair in Critical Care Medicine.

3

Abbreviations used in this paper: LBP, LPS-binding protein; mCD14, membrane CD14; sCD14, soluble CD14; PI-PLC, phosphatidylinositol-phospholipase C; TIR, Toll/IL-1 receptor; TRIF, TIR domain-containing adaptor-inducing IFN-β; TRAM, TRIF-related adaptor molecule.

1
Bazil, V., V. Horejsi, M. Baudys, H. Kristofova, J. L. Strominger, W. Kostka, I. Hilgert.
1986
. Biochemical characterization of a soluble form of the 53-kDa monocyte surface antigen.
Eur. J. Immunol.
16
:
1583
-1589.
2
Haziot, A., S. Chen, E. Ferrero, M. G. Low, R. Silber, S. M. Goyert.
1988
. The monocyte differentiation antigen, CD14, is anchored to the cell membrane by a phosphatidylinositol linkage.
J. Immunol.
141
:
547
-552.
3
Simmons, D. L., S. Tan, D. G. Tenen, A. Nicholson-Weller, B. Seed.
1989
. Monocyte antigen CD14 is a phospholipid anchored membrane protein.
Blood
73
:
284
-289.
4
Frey, E. A., D. S. Miller, T. G. Jahr, A. Sundan, V. Bazil, T. Espevik, B. B. Finlay, S. D. Wright.
1992
. Soluble CD14 participates in the response of cells to lipopolysaccharide.
J. Exp. Med.
176
:
1665
-1671.
5
Golenbock, D. T., R. R. Bach, H. Lichenstein, T. S. Juan, A. Tadavarthy, C. F. Moldow.
1995
. Soluble CD14 promotes LPS activation of CD14-deficient PNH monocytes and endothelial cells.
J. Lab. Clin. Med.
125
:
662
-671.
6
Haziot, A., G. W. Rong, J. Silver, S. M. Goyert.
1993
. Recombinant soluble CD14 mediates the activation of endothelial cells by lipopolysaccharide.
J. Immunol.
151
:
1500
-1507.
7
Pugin, J., C. C. Schurer-Maly, D. Leturcq, A. Moriarty, R. J. Ulevitch, P. S. Tobias.
1993
. Lipopolysaccharide activation of human endothelial and epithelial cells is mediated by lipopolysaccharide-binding protein and soluble CD14.
Proc. Natl. Acad. Sci. USA
90
:
2744
-2748.
8
Read, M. A., S. R. Cordle, R. A. Veach, C. D. Carlisle, J. Hawiger.
1993
. Cell-free pool of CD14 mediates activation of transcription factor NF-κB by lipopolysaccharide in human endothelial cells.
Proc. Natl. Acad. Sci. USA
90
:
9887
-9891.
9
Jersmann, H. P., C. S. Hii, G. L. Hodge, A. Ferrante.
2001
. Synthesis and surface expression of CD14 by human endothelial cells.
Infect. Immun.
69
:
479
-485.
10
Lloyd, K. L., P. Kubes.
2006
. GPI-linked endothelial CD14 contributes to the detection of LPS.
Am. J. Physiol.
291
:
H473
-H481.
11
Kawai, T., O. Adachi, T. Ogawa, K. Takeda, S. Akira.
1999
. Unresponsiveness of MyD88-deficient mice to endotoxin.
Immunity
11
:
115
-122.
12
Kawai, T., O. Takeuchi, T. Fujita, J. Inoue, P. F. Muhlradt, S. Sato, K. Hoshino, S. Akira.
2001
. Lipopolysaccharide stimulates the MyD88-independent pathway and results in activation of IFN-regulatory factor 3 and the expression of a subset of lipopolysaccharide-inducible genes.
J. Immunol.
167
:
5887
-5894.
13
Takeuchi, O., K. Takeda, K. Hoshino, O. Adachi, T. Ogawa, S. Akira.
2000
. Cellular responses to bacterial cell wall components are mediated through MyD88-dependent signaling cascades.
Int. Immunol.
12
:
113
-117.
14
Yamamoto, M., S. Sato, H. Hemmi, K. Hoshino, T. Kaisho, H. Sanjo, O. Takeuchi, M. Sugiyama, M. Okabe, K. Takeda, S. Akira.
2003
. Role of adaptor TRIF in the MyD88-independent toll-like receptor signaling pathway.
Science
301
:
640
-643.
15
Fitzgerald, K. A., D. C. Rowe, B. J. Barnes, D. R. Caffrey, A. Visintin, E. Latz, B. Monks, P. M. Pitha, D. T. Golenbock.
2003
. LPS-TLR4 signaling to IRF-3/7 and NF-κB involves the Toll adapters TRAM and TRIF.
J. Exp. Med.
198
:
1043
-1055.
16
Oshiumi, H., M. Sasai, K. Shida, T. Fujita, M. Matsumoto, T. Seya.
2003
. TIR-containing adapter molecule (TICAM)-2, a bridging adapter recruiting to Toll-like receptor 4 TICAM-1 that induces interferon-β.
J. Biol. Chem.
278
:
49751
-49762.
17
Hoebe, K., E. M. Janssen, S. O. Kim, L. Alexopoulou, R. A. Flavell, J. Han, B. Beutler.
2003
. Upregulation of costimulatory molecules induced by lipopolysaccharide and double-stranded RNA occurs by Trif-dependent and Trif-independent pathways.
Nat. Immunol.
4
:
1223
-1229.
18
Saito, S., M. Matsuura, K. Tominaga, T. Kirikae, M. Nakano.
2000
. Important role of membrane-associated CD14 in the induction of IFN-β and subsequent nitric oxide production by murine macrophages in response to bacterial lipopolysaccharide.
Eur. J. Biochem.
267
:
37
-45.
19
Jiang, Z., P. Georgel, X. Du, L. Shamel, S. Sovath, S. Mudd, M. Huber, C. Kalis, S. Keck, C. Galanos, et al
2005
. CD14 is required for MyD88-independent LPS signaling.
Nat. Immunol.
6
:
565
-570.
20
Harari, O. A., P. Alcaide, D. Ahl, F. W. Luscinskas, J. K. Liao.
2006
. Absence of TRAM restricts Toll-like receptor 4 signaling in vascular endothelial cells to the MyD88 pathway.
Circ. Res.
98
:
1134
-1140.
21
Andonegui, G., S. M. Kerfoot, K. McNagny, K. V. Ebbert, K. D. Patel, P. Kubes.
2005
. Platelets express functional Toll-like receptor-4.
Blood
106
:
2417
-2423.
22
Andreakos, E., S. M. Sacre, C. Smith, A. Lundberg, S. Kiriakidis, T. Stonehouse, C. Monaco, M. Feldmann, B. M. Foxwell.
2004
. Distinct pathways of LPS-induced NF-κB activation and cytokine production in human myeloid and nonmyeloid cells defined by selective utilization of MyD88 and Mal/TIRAP.
Blood
103
:
2229
-2237.
23
Silva, C., K. Zhang, S. Tsutsui, J. K. Holden, M. J. Gill, C. Power.
2003
. Growth hormone prevents human immunodeficiency virus-induced neuronal p53 expression.
Ann. Neurol.
54
:
605
-614.
24
Landmann, R., S. Link, S. Sansano, Z. Rajacic, W. Zimmerli.
1998
. Soluble CD14 activates monocytic cells independently of lipopolysaccharide.
Infect. Immun.
66
:
2264
-2271.
25
Sacre, S. M., A. M. Lundberg, E. Andreakos, C. Taylor, M. Feldmann, B. M. Foxwell.
2007
. Selective use of TRAM in lipopolysaccharide (LPS) and lipoteichoic acid (LTA) induced NF-κB activation and cytokine production in primary human cells: TRAM is an adaptor for LPS and LTA signaling.
J. Immunol.
178
:
2148
-2154.
26
Wurfel, M. M., B. G. Monks, R. R. Ingalls, R. L. Dedrick, R. Delude, D. Zhou, N. Lamping, R. R. Schumann, R. Thieringer, M. J. Fenton, et al
1997
. Targeted deletion of the lipopolysaccharide (LPS)-binding protein gene leads to profound suppression of LPS responses ex vivo, whereas in vivo responses remain intact.
J. Exp. Med.
186
:
2051
-2056.
27
Triantafilou, M., K. Triantafilou, N. Fernandez.
2000
. Rough and smooth forms of fluorescein-labelled bacterial endotoxin exhibit CD14/LBP dependent and independent binding that is influenced by endotoxin concentration.
Eur. J. Biochem.
267
:
2218
-2226.
28
Triantafilou, K., M. Triantafilou, R. L. Dedrick.
2001
. A CD14-independent LPS receptor cluster.
Nat. Immunol.
2
:
338
-345.
29
Cohen, L., A. Haziot, D. R. Shen, X. Y. Lin, C. Sia, R. Harper, J. Silver, S. M. Goyert.
1995
. CD14-independent responses to LPS require a serum factor that is absent from neonates.
J. Immunol.
155
:
5337
-5342.
30
Lynn, W. A., Y. Liu, D. T. Golenbock.
1993
. Neither CD14 nor serum is absolutely necessary for activation of mononuclear phagocytes by bacterial lipopolysaccharide.
Infect. Immun.
61
:
4452
-4461.
31
Wright, S. D., R. A. Ramos, P. S. Tobias, R. J. Ulevitch, J. C. Mathison.
1990
. CD14, a receptor for complexes of lipopolysaccharide (LPS) and LPS binding protein.
Science
249
:
1431
-1433.
32
Haziot, A., E. Ferrero, F. Kontgen, N. Hijiya, S. Yamamoto, J. Silver, C. L. Stewart, S. M. Goyert.
1996
. Resistance to endotoxin shock and reduced dissemination of Gram-negative bacteria in CD14-deficient mice.
Immunity
4
:
407
-414.
33
Haziot, A., X. Y. Lin, F. Zhang, S. M. Goyert.
1998
. The induction of acute phase proteins by lipopolysaccharide uses a novel pathway that is CD14-independent.
J. Immunol.
160
:
2570
-2572.
34
Perera, P. Y., S. N. Vogel, G. R. Detore, A. Haziot, S. M. Goyert.
1997
. CD14-dependent and CD14-independent signaling pathways in murine macrophages from normal and CD14 knockout mice stimulated with lipopolysaccharide or Taxol.
J. Immunol.
158
:
4422
-4429.
35
van Deventer, S. J., H. R. Buller, J. W. ten Cate, A. Sturk, W. Pauw.
1988
. Endotoxaemia: an early predictor of septicaemia in febrile patients.
Lancet
1
:
605
-609.
36
Kim, H. M., B. S. Park, J. I. Kim, S. E. Kim, J. Lee, S. C. Oh, P. Enkhbayar, N. Matsushima, H. Lee, O. J. Yoo, J. O. Lee.
2007
. Crystal structure of the TLR4-MD-2 complex with bound endotoxin antagonist Eritoran.
Cell
130
:
906
-917.
37
Visintin, A., K. A. Halmen, E. Latz, B. G. Monks, D. T. Golenbock.
2005
. Pharmacological inhibition of endotoxin responses is achieved by targeting the TLR4 coreceptor, MD-2.
J. Immunol.
175
:
6465
-6472.
38
Wang, P. Y., R. L. Kitchens, R. S. Munford.
1995
. Bacterial lipopolysaccharide binds to CD14 in low-density domains of the monocyte-macrophage plasma membrane.
J. Inflamm.
47
:
126
-137.
39
Pugin, J., V. V. Kravchenko, J. D. Lee, L. Kline, R. J. Ulevitch, P. S. Tobias.
1998
. Cell activation mediated by glycosylphosphatidylinositol-anchored or transmembrane forms of CD14.
Infect. Immun.
66
:
1174
-1180.
40
Rowe, D. C., A. F. McGettrick, E. Latz, B. G. Monks, N. J. Gay, M. Yamamoto, S. Akira, L. A. O'Neill, K. A. Fitzgerald, D. T. Golenbock.
2006
. The myristoylation of TRIF-related adaptor molecule is essential for Toll-like receptor 4 signal transduction.
Proc. Natl. Acad. Sci. USA
103
:
6299
-6304.
41
Kim, J. I., C. J. Lee, M. S. Jin, C. H. Lee, S. G. Paik, H. Lee, J. O. Lee.
2005
. Crystal structure of CD14 and its implications for lipopolysaccharide signaling.
J. Biol. Chem.
280
:
11347
-11351.