Endocytic pathways regulate Toll‐like receptor 4 signaling and link innate and adaptive immunity

During the early hours of incubation with LPS, a gradual decrease in plasma membrane TLR4 was observed in monocytes (Figure 1A). This occurred concomitantly with formation of endosome‐like structures containing LPS^Cy5 (Figure 1B, upper panel). At later time points, LPS accumulated in the perinuclear area (Figure 1B, lower panels). For the detection of TLR4, a TLR4 antibody (anti‐TLR4^A546) was added simultaneously with fluorescent LPS (LPS^OG) to monocytes. Colocalization of TLR4 and LPS was observed in microdomains on the endosomal limiting membrane 1 h after stimulation (Figure 1C).

We went on to study the localization and transport of TLR4 fused to YFP (TLR4^YFP) in live HEK293 cells expressing the co‐receptors MD‐2 and CD14 using confocal microscopy and LPS^CY5 for stimulation. In unstimulated cells, TLR4^YFP was found in the endoplasmic reticulum, at the plasma membrane, as a large pool in the Golgi area and sporadically on endosomal structures (Figure 2A). Following LPS stimulation, endosomes showing LPS and TLR4 colocalization appeared within 15 min, suggesting active endocytosis (Figure 2A and B). Addition of LPS increased the number of TLR4‐containing endosomes seven‐fold within the first 40 min of stimulation and this increase was dependent on the co‐receptor MD‐2 (Figure 2B and C). Although we found endosomes containing TLR4 and LPS in the absence of MD‐2, the presence of MD‐2 enhanced the process, and also promoted aggregation of TLR4 in patches on the endosomes, as seen in Figure 1C and upper right panel of Figure 2A. CD14 greatly increased both binding to the surface and uptake of LPS, as has previously been reported (Kitchens et al, 1998; Latz et al, 2002).

We further characterized the compartments into which LPS and TLR4 were sorted. Since LPS binding to the surface was only detected in cells expressing CD14, we subsequently used LPS binding as a marker for cells that had been successfully transfected with the CD14 plasmid. To characterize the TLR4‐positive endosomes, cells were transfected with plasmids encoding CD14 and a CFP chimera containing tandem FYVE domains from the mouse Hrs (2 × FYVE^CFP). This construct binds phosphatidylinositol‐3‐phosphate (PI3P) that is highly enriched on early/sorting endosomes (Gaullier et al, 1998; Simonsen et al, 1998; Gillooly et al, 2000). Within the first hour of LPS stimulation, most endosomes containing TLR4 and LPS were FYVE positive (Figure 2D). Endosomes containing TLR4 and LPS, and identified by 2 × FYVE^CFP, were also positive for transferrin (Figure 2E, Supplementary Figure 2). Transferrin is endocytosed in a clathrin‐dependent manner and sorted on early/sorting endosomes. In cells stimulated for longer periods, TLR4 and LPS appeared on endosomes being FYVE negative (data not shown). TLR4 showed extensive colocalization with LPS in all compartments. The localization of TLR4 to early endosomes was also confirmed by staining fixed cells with an antibody to the early endosomal marker EEA1 (data not shown) or by cotransfecting the cells with a truncated but functional EEA1 (Stenmark et al, 1996) fused to CFP (EEA1^CFP). In unstimulated cells, TLR4 was present on small EEA1^CFP‐positive endosomes that grew significantly larger in size after LPS stimulation (Supplementary Figure 1A). These EEA1‐positive endosomes, like the FYVE‐positive endosomes, contained a substantial amount of LPS and TLR4 on the limiting membrane. In addition, images were obtained demonstrating apparent fusion events between LPS‐positive endosomes (Supplementary Figure 1B and Supplementary Movie). Taken together, our data show that LPS and TLR4 are trafficked to early/sorting endosomes and that LPS‐induced signaling increases the size of these endosomes probably by endosome–endosome fusions.

Receptor‐mediated endocytosis is often classified by being clathrin‐dependent (through coated pits as seen with transferrin) or clathrin‐independent (through caveoli or lipid rafts as seen for many glycosylphosphatidylinositol (GPI)‐anchored proteins). Both mechanisms are dependent on dynamin, whereas (macro)pinocytosis is dependent on actin (Nichols, 2003). Dynamin proteins are GTPases that are essential for budding of vesicles from the plasma membrane and from recycling endosomes. Whereas transmembrane proteins contain sorting information in the cytosolic domain, GPI‐anchored proteins (like CD14) have, in general, been shown to either be coendocytosed with the transmembrane proteins or pinocytosed together with fluid phase contents (Mayor and Riezman, 2004). We next investigated whether LPS endocytosis was dependent on dynamin using constructs of dynamin II wild type (WT) and dominant negative dynamin II (Dyn K44A) (Damke et al, 1994). Cells expressing TLR4^YFP/MD‐2 were cotransfected with CD14 and dynamin WT or Dyn K44A, and investigated by confocal microscopy (Figure 3A). Fluorescent transferrin and LPS were added together for selection of cells with a dynamin knock down phenotype. LPS and transferrin uptake were not affected in the dynamin WT transfectants (Figure 3A, left panel), whereas in the Dyn K44A transfectants (Figure 3A, right panel) a clear accumulation of transferrin on the plasma membrane was observed and, as a consequence, reduced transferrin uptake (Damke et al, 1994). In DynK44A transfectants, total LPS uptake was reduced to about 18% of the uptake in dynamin WT transfectants (Figure 3B), supporting that LPS is endocytosed by a receptor‐mediated vesicular mechanism.

Clathrin is a scaffold protein involved in assembly of selected transmembrane proteins in invaginations called clathrin‐coated pits. We knocked down clathrin heavy chain by short interfering (si) RNA technology (Motley et al, 2003) in HEK293 cells expressing TLR4^YFP/MD‐2 including a noninterfering scrambled RNA duplex (Bache et al, 2003) as a control. Following siRNA treatment, the cells were transfected with CD14 and the uptake and distribution of LPS and transferrin were investigated (Figure 3C). The introduction of control siRNA did not affect LPS and transferrin uptake (Figure 3C, left panel), whereas clathrin siRNA resulted in significant accumulation of both transferrin and LPS at the plasma membrane in these cells (Figure 3C, right panel). Cells showing accumulation of transferrin at the plasma membrane were scored as clathrin knock down transfectants. In these cells, LPS did not appear on endosomes during the first 40 min of stimulation (Figure 3C and D). At later time points, LPS appeared on large endosomes also positive for transferrin (data not shown), suggesting an additional but slower clathrin‐independent uptake mechanism. As we found siRNA against clathrin to be very effective in inhibiting both transferrin and LPS uptake, our results show clathrin to be an essential component for the early endocytosis of the LPS receptor complex.

The targeting of transmembrane proteins to lysosomes is determined by tyrosine phosphorylation and di‐leucin motifs of which there are candidates in TLR4. Tyrosine‐based motifs (YXXΦ) are found in positions 587 (YDAF), 622 (YRDF) and 707 (YLEW), whereas a di‐leucine motif ([DE]XXXL[LI]) is found in position 697 (ELYRLL), both recognized by adaptor protein (AP) complexes involved in the initial phase of endocytosis (Bonifacino and Traub, 2003). A major additional mechanism for endocytosis and lysosomal targeting of transmembrane proteins is by the covalent attachment of one or more ubiquitins to lysines to the cytosolic domain (Hicke, 2001; Raiborg et al, 2003). A recent report showed that the E3 ubiquitin ligase Triad3A was involved in degradation of TLRs (except TLR2) (Chuang and Ulevitch, 2004). Thus, we wanted to address if LPS induced ubiquitination of TLR4 and if this had any effect on endocytosis and degradation of TLR4. We found that TLR4 ubiquitination was both constitutive and enhanced by LPS in HEK293 cells expressing TLR4^YFP/MD‐2 (Figure 4A). In monocytes, a weak increase in ubiquitination of TLR4 was observed at 1 h of LPS stimulation, followed by a reduction after 3 h (Figure 4B). TLR4 was ubiquitinated in cells transfected with c‐myc Ub wild type (UbRGG) but not in cells transfected with a c‐myc nonfunctional Ub (UbR) lacking the two C‐terminal glycines required for covalent conjugation to proteins (data not shown) (Stang et al, 2004). There was, however, no effect of UbR on uptake of LPS within the first hour of stimulation (data not shown), suggesting that LPS uptake is Ub‐independent. Ubiquitin ligases are recruited by their SH2 domains to tyrosine‐phosphorylated targets, and we found that TLR4 was indeed tyrosine phosphorylated in response to LPS (data not shown). These data suggest that although TLR4 is ubiquitinated as a response of LPS stimulation, ubiquitination is not required for the initial endocytosis of the LPS receptor complex.

Hrs is located to clathrin‐coated microdomains of the early/sorting endosomal limiting membrane where it is involved in the recognition and targeting of ubiquitinated protein cargo to the lysosomal degradation pathway by promoting translocation of the target proteins to the lumen of endosomes forming multivesicular bodies (MVB) (Gruenberg and Stenmark, 2004). Hrs has recently also been shown to associate with phagosomes and aid in fusion with lysosomes (Vieira et al, 2004). We investigated the involvement of Hrs in TLR4 trafficking in monocytes and in HEK293‐TLR4^YFP/MD‐2 cells transiently transfected with c‐myc‐tagged Hrs. In monocytes Hrs constitutively associated with TLR4; however, the association increased transiently after 1 h of LPS stimulation (Figure 4B). Control IgG did not co‐precipitate TLR4 (Figure 4B) or Hrs (data nor shown). The TLR4 immunoprecipitated with Hrs was slightly larger and appeared more smeared than the mature form of TLR4 (Figure 4B, middle), possibly because Hrs interacts mainly with multiple ubiquitins on the intracellular domain of TLR4. Also in HEK293 cells overexpressing c‐myc‐Hrs, we found that TLR4 co‐precipitated with Hrs (Figure 4C) in an LPS‐inducible manner, whereas control IgG did not (control C1, Figure 4C). The upper band in Figure 4C represents the c‐myc‐tagged Hrs and the lower band is the endogenous Hrs as shown by comigration with the endogenous Hrs from nontransfected HEK293 cells (control C2, Figure 4C). Importantly, a moderate overexpression of c‐myc Hrs (here estimated to be less than three times the endogenous level) lead to a rapid decrease in the amount of total cellular TLR4 after addition of LPS, whereas the amounts of Hrs and tubulin were unaffected (Figure 4C). This result is in accordance with a recent paper (Scoles et al, 2005), which demonstrated that overexpression of Hrs (e.g. less than 10 times the endogenous level) results in a rapid decrease in the EGFR levels upon stimulation with EGF. Furthermore, TLR4^CFP and Hrs^YFP were colocalized on discrete microdomains on the endosomal limiting membrane following LPS stimulation (Figure 4D).

Next, we investigated the effect of chloroquine and lactacystin on TLR4 levels following LPS stimulation in HEK293 cells expressing TLR4^YFP/MD‐2 and transfected with CD14. Chloroquine inhibits maturation of early endosomes to lysosomes by inhibiting acidification and thus preventing lysosomal degradation (Mellman et al, 1986). Lactacystin is an irreversible proteasome inhibitor. Previous studies have shown that cellular TLR4 is present in two different sizes, the larger is a 130 kDa heavily glycosylated, MD‐2‐dependent, surface‐translocated form and the smaller is a 110 kDa partially glycosylated form found predominantly in Golgi (Nagai et al, 2002; Ohnishi et al, 2003). LPS seemed to increase the total cellular TLR4 after 10 h which may be due to increased transcription of the CMV‐promoted TLR4‐YFP constructs by LPS (Lee et al, 2004). However, 1 h of LPS stimulation did not increase the TLR4 level. Treatment with chloroquine resulted in a significant cellular accumulation of both forms of TLR4 as early as 1 h after LPS (Figure 5A). Interestingly, lactacystin reported to block Triad3A‐mediated degradation of TLR9 (Chuang and Ulevitch, 2004) did not cause the same accumulation of TLR4^YFP in LPS‐stimulated cells (Figure 5A). LY294002 (LY) inhibits the PI3‐kinase enzymes responsible for the formation of the cellular PI3P controlling endosome fusions (Simonsen et al, 1998) and the formation of intraluminal vesicles in late endosomes/MVBs (Futter et al, 2001). PI3‐kinase activity is also required for the recruitment of Hrs to endosomal membranes (Komada and Soriano, 1999; Urbe et al, 2000). Cells treated with LY and stimulated with LPS showed a dramatic increase in the level of TLR4^YFP after 10 h of LPS stimulation compared to LPS‐stimulated cells treated with control vehicle (Figure 5B). Altogether, these data suggest that endosomal sorting and acidification are required for degradation of TLR4.

Furthermore, we investigated the involvement of Hrs in TLR4 degradation by knocking down the endogenous level of Hrs by siRNA technology (Bache et al, 2003). Hrs siRNA‐treated HEK293 cells expressing TLR4^YFP/MD‐2 and transfected with CD14 showed an increased levels of both the large fully glycosylated and the partially glycosylated TLR4^YFP form compared to the cells treated with control siRNA (Figure 5C, upper panel). Our results suggest that ubiquitinated TLR4 associates with Hrs directing the activated TLR4 to lysososomes for degradation. This result is in contrast to TLR9 that has been reported to be degraded by proteasomes after ubiquitination by the Triad3A ubiquitin ligase (Chuang and Ulevitch, 2004).

After describing mechanisms of the TLR4 sorting through the endosomal pathway, we investigated the role of this pathway on LPS‐induced NFκB activation. By comparing the effect of wild‐type versus dominant negative constructs for the trafficking components dynamin II (dynamin WT versus Dyn K44A) and ubiquitin (UbRGG versus UbR), we found that the reduced endocytic activity caused by Dyn K44A, or saturation of Ub‐interacting proteins with the nonfunctional UbR, caused increased NFκB activation compared to overexpression of the WT constructs (Figure 5D and Supplementary Figure 3). This result strongly supports that signaling is initiated at the plasma membrane (Ahmad‐Nejad et al, 2002; Latz et al, 2002), and that endocytosis and lysosomal targeting of TLR4 limit NFκB activation. Additional arguments for this conclusion was also obtained in knock down experiments of Hrs with siRNA showing a 70% increase in LPS‐induced NFκB activation compared to control (Figure 5E and Supplementary Figure 3). Furthermore, by knocking down an essential component (HCRP1) of the endosomal sorting complexes required for transport‐I (ESCRT‐I) with siRNA (Bache et al, 2004), the LPS (10 ng/ml)‐induced NFκB activation increased by 245% compared to the siRNA control (Figure 5E). These data suggest that dynamin, ubiquitin, Hrs and HCRP1 (i.e. ESCRT‐I) are involved in desensitizing the cell after LPS activation, most likely through endocytosis and lysosomal degradation of the activated LPS receptor complex.

Time‐lapse series of LPS‐stimulated monocytes showed that LPS overlapped with dextran in dynamic tubular structures protruding from the perinuclear area after 3–5 h of incubation (data not shown). These protrusions were directed towards the plasma membrane as has been described for MHC class II containing lysosome‐like tubulae after LPS stimulation in dendritic cells (Boes et al, 2002), a phenotype not observed in HEK293 cells. By counterstaining HEK293 cells expressing TLR4^YFP/MD‐2 transfected with CD14 with an antibody against the late endosomal/lysosomal marker protein LAMP‐1, we observed that TLR4^YFP was present in the lumen of late endosomes/MVBs following LPS stimulation (Figure 6A). Treatment of stimulated live cells with LysoTracker Red showed that TLR4^YFP and LPS^CY5 colocalized at the limiting membrane and in the lumen of nonacidic endosomes (Figure 6B). Intraluminal TLR4^YFP was lost following endosomal acidification, whereas LPS^CY5 was still detectable. Since TLR4 and LPS/CD14 are directed through the endocytic compartments to lysosomes for degradation, associated antigens may be presented on HLA class II to CD4⁺ T cells. We performed antigen presentation assays using mouse antibodies to TLR4 (HTA125), CD14 (5C5) and MD‐2 (IIC2) and proliferation of a CD4⁺ T cell specific for the mouse κ‐light chain‐derived peptide Cκ^40–48 presented on HLA‐DR4 (Schjetne et al, 2002) in an LPS‐free system. This system is designed for functional testing of the class II presentation pathway by providing a read‐out for lysosomal antigen processing, loading and presentation in primary human cells. We found efficient presentation of the antibody‐derived peptide (Figure 6C) suggesting that TLR4, MD‐2 and CD14 were directed to a functional class II loading compartment, most commonly ascribed to MVB (Peters et al, 1991). This result implies that the LPS receptor complex, TLR4/MD‐2/CD14, not only signals the presence of microbes but also assists in directing the endocytosis of antigens on class II molecules for presentation to T helper cells.

The TLR4 agonist, LPS, is a potent immune stimulator which can cause sepsis. TLR4 signaling is under tight regulatory control at multiple levels and several inhibitors of TLR4 signaling have now been described. Internalization of receptors regulates signal transduction (Di Guglielmo et al, 2003); however, little attention has been devoted to mechanisms and functional consequences of the endocytic uptake of the LPS receptor complex. In this study, we have characterized the uptake of LPS and the sorting of the LPS receptor complex from the plasma membrane to the late endosomes/lysosomes and explored functional consequences of this endosomal pathway.

Internalization of membrane receptors occurs both through clathrin‐ and caveolae/raft‐mediated pathways. The EGF receptor is internalized within minutes after ligand addition by a clathrin‐mediated pathway (Le and Wrana, 2005), whereas the IL‐2 receptor appears to be taken up by a lipid raft‐dependent manner (Lamaze et al, 2001). Previous reports have shown that CD14 is located in lipid rafts and that a fraction of TLR4 moves into the lipid raft domain after LPS stimulation, a result suggesting a clathrin‐independent uptake of the LPS receptor complex (Triantafilou et al, 2004). Furthermore, Kitchens et al (1998) reported that the CD14‐mediated uptake of LPS occurs by both clathrin‐dependent and ‐independent pathways. Knocking down clathrin by siRNA results in a significant inhibition of EGF uptake at low EGF doses and less effective at higher EGF doses (Sigismund et al, 2005). Our results using siRNA for clathrin clearly demonstrate that the early phase (up to 40 min) of LPS internalization is predominantly clathrin‐dependent, whereas clathrin‐independent pathways operate at later time points.

Monoubiquitination serves an important role in sorting membrane proteins through the endosomal pathway (Raiborg et al, 2003), whereas polyubiquitination is the recognition signal for proteasomal degradation (Marmor and Yarden, 2004). As mentioned above, Triad3A has been reported to be an E3 ubiquitin ligase that promotes proteasomal degradation of TLR9 (Chuang and Ulevitch, 2004). Triad3A‐mediated ubiquitination does not seem to play a role in the lysosomal degradation pathway as the effect of this ligase on TLR9 degradation is dependent on proteasomal, but not lysosomal activity (Chuang and Ulevitch, 2004). We observed increased ubiquitination of TLR4 upon LPS activation both in monocytes and in genetically engineered cells. Hrs has an essential role in sorting ubiquitinated proteins into clathrin‐coated microdomains of early endosomes (Raiborg et al, 2002) and Hrs has interacting domains for PI3P (the FYVE domain), clathrin and ubiquitin (Bache et al, 2003). We provide evidence for Hrs interaction with TLR4 on endosomes and that this interaction was increased upon LPS activation. Thus, Hrs seems to play an important role in the initial steps of TLR4 sorting to the endosomal pathway. A part of the LPS receptor complex also traffics from the plasma membrane to the Golgi (Latz et al, 2002), and this pathway may either go directly from the plasma membrane or through the early/sorting endosomes. Whether Hrs also plays a role in TLR4 trafficking to the Golgi apparatus is not known. Another protein, Tollip, has been shown to inhibit LPS‐induced NFκB activation (Zhang and Ghosh, 2002) by a mechanism requiring localization to endosomes or activated plasma membranes and through binding of ubiquitinated cargo and recruitment to clathrin‐coated microdomains (Katoh et al, 2004), not unlike Hrs. It is thus likely that redundancy exists in the machinery that ensures that inflammatory signals are shut off to avoid hyperinflammation and the accompanied adverse effects.

Endosomal trafficking of the LPS receptor complex leading to lysosomal degradation and signal termination was consistently demonstrated in the HEK293‐TLR4 cells. Our results suggest that a similar pathway of LPS receptor endocytosis and degradation also operates in monocytes. Firstly, LPS induced a clear downregulation of TLR4 on the plasma membrane with subsequent localization of LPS and TLR4 to endosomes. Secondly, ubiquitination of TLR4 was constitutive, weakly increased after 1 h of LPS stimulation and markedly decreased after 3 h. Finally, TLR4 associated with Hrs and this association was transiently increased at 1 h of LPS stimulation. Direct evidence for LPS‐induced lysosomal degradation of TLR4 in monocytes was difficult to obtain due to low TLR4 signals in immunoblots and probably a high turnover rate for TLR4. However, antibodies towards TLR4, MD‐2 and CD14 were degraded in MVBs/lysosomes since an antibody‐derived peptide was loaded onto HLA class II (Figure 6C). Thus, it is likely that the pathway of endosomal degradation of TLR4 that we demonstrated in the HEK293‐TLR4 cells also, at least to some extent, takes place in monocytes.

Our studies provide evidence that sorting the LPS receptor complex through the endosomal pathway has important functional consequences. We observed a striking increase in LPS signaling when proteins involved in endosomal trafficking and maturation were inhibited. By reducing the amount of functional ESCRT‐I, which associates with Hrs and is involved in the sorting of ubiquitinated proteins into multivesicular bodies (Bache et al, 2002), a more than two‐fold increase in the LPS signaling was observed. Inhibition of ubiquitination and depletion of Hrs or ESCRT‐I resulted in increased LPS signaling. Furthermore, overexpression of Hrs lead to a significant increase in LPS‐induced TLR4 degradation. From these results it is apparent that LPS signaling starts at the plasma membrane; however, signaling is likely to continue along the endosomal pathway until an Hrs‐ and ESCRT‐I‐mediated translocation of TLR4 to the luminal side of endosomes has taken place. This view is also supported by confocal imaging showing the presence of the signaling adapters MyD88 and TRAM at the plasma membrane and on endosomes in LPS‐stimulated cells (H Husebye and T Espevik, unpublished results). Signaling on endosomes is also reported for TGF‐β which is taken up by a clathrin‐dependent mechanism and trafficked into early endosomes where the signaling component SARA is enriched (Di Guglielmo et al, 2003).

A very important consequence of the trafficking of the LPS receptor complex described in this paper is that antigens associated with the LPS receptor complex are directly transported to the antigen presentation machinery for presentation to CD4⁺ T helper cells. The loading of antigens on class II molecules occurs by fusion of limiting and intraluminal membranes in MVBs (Peters et al, 1991), resulting in exchange of the class II inhibitory peptide (CLIP) with lysosomal peptides derived from processed exogenous proteins. This is also highly relevant for other suggested ligands of TLR4, like heat shock proteins, viral proteins (Akira and Takeda, 2004) and cotrafficked antigens, by providing a direct link between activation of innate and acquired immune cells. This may be exploited in design of synthetic vaccines aimed at the CD4⁺ T‐helper cell population, as has been shown for conjugated complexes of lipoproteins and immunogenic peptides acting through TLR2 (Jackson et al, 2004).

Altogether, our results strongly support that distinct components of the endocytic pathway inhibit inflammatory signals induced by TLR4 and that long‐lived information is simultaneously passed on to the adaptive branch of the immune system.

LPS (0111:B4) from Escherichia coli was purchased from Invivogen. Cy5‐ and OregonGreen (OG) labeling of LPS was performed as previously described (Latz et al, 2002). Alexa labeling of antibodies was performed according to the manufacturer's protocol (Invitrogen). Alexa⁵⁴⁶‐labeled transferrin and LysoTracker Red were purchased from Invitrogen. LY294002 and lactacystin were purchased from Calbiochem. Chloroquine was purchased from Sigma. Mouse antibodies used: anti‐GFP (JL‐8) (Clontech), anti‐TLR4 (HTA125) was kindly provided by Dr Kensuke Miyake (Saga Medical School, Japan) (Yang et al, 2000), anti‐c‐myc (9E10) (BD Bioscience), anti‐ubiquitin‐protein conjugates (FK2) (Affiniti Research), anti‐LAMP1 (R&D Systems), anti‐CD14 (5C5) (Lien et al, 1998) and anti‐MD‐2 (IIC1). The mouse monoclonal antibody IIC1 (IgM) detects MD‐2 in CHO and HEK293 cells expressing TLR4/MD‐2 as well as on the surface of human monocytes (T Espevik and L Ryan, unpublished results). Rabbit polyclonal antibodies used: anti‐TLR4 (eBioscience), anti‐GFP (Clontech) and anti‐Hrs (Raiborg et al, 2001). The following mammalian expression vectors were used: pcDNA3 (Invitrogen), CD14 (Latz et al, 2002), c‐myc‐Hrs, c‐myc‐ubiquitin wild‐type UbRGG (Stang et al, 2004) and c‐myc‐ubiquitin DN UbR (Stang et al, 2004) were in pcDNA3. Dynamin‐II wild type and Dynamin‐II K44A in pcDNA3 were kindly provided by Dr Sandy Schmid (Scripps, USA). MD‐2 in pEF‐BOS was kindly provided by Dr Miyake (Yang et al, 2000). Hrs^YFP was in pEYFP‐C1 (Clontech). The ECFP‐2 × FYVE construct was made by substituting GFP from EGFP‐2 × FYVE (Gillooly et al, 2000) with CFP using pECFP (Clontech) as a template and PCR primers 5′‐CCA AGT ACG CCC CCT ATT GA‐3′ and 5′‐ATT TAA GCT TGT ACA GCT CGT CCA TGC‐3′. The CFP encoding fragment was inserted as an NheI/XhoI fragment.

HEK293 cell lines that stably expressed human TLR4 (Yang et al, 2000), TLR4^YFP, TLR4^YFP/MD‐2 (Latz et al, 2002) or TLR4^CFP/MD‐2 were used and cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS) and 0.5 mg/ml G418 and transfected using GeneJuice™ transfection reagent (Novagene) according to the manufacturer's protocol. Human monocytes were isolated by plastic adherence from Lymphoprep (Axis‐Shield) separated buffycoats (The Blood Bank, St Olavs Hospital, Trondheim, Norway) and grown in RPMI/10% pooled A+ serum (The Blood Bank). LPS was sonicated for 5 min and preincubated in serum‐containing medium at 37°C for 5 min before being added to cells. For confocal imaging, the cells were seeded on 35 mm glass bottom γ‐irradiated tissue cell dishes (MatTek Corporation) and buffered with 25 mM HEPES buffer just before stimulation. LPS^CY5 (250 ng/ml) or LPS^OG (600 ng/ml) was used in LPS uptake studies.

Adherent monocytes were incubated with or without LPS (100 ng/ml) in medium/10% A+ serum for 2 h, detached with EDTA and stained with 10 μg/ml anti‐TLR4 (HTA125) or mouse IgG2a for 30 min on ice, washed and incubated with 10 μg/ml secondary anti‐mouse Ig‐FITC (Becton Dickinson) for 30 min on ice and washed and analyzed by flow cytometry (Coulter).

SiRNA duplexes targeting the coding region of clathrin heavy chain (Motley et al, 2003), the coding region of Hrs (Bache et al, 2003) and the coding region of the ESCRT‐I component HCRP1 (Bache et al, 2004) were used. A scrambled RNA duplex (Bache et al, 2003) was used as a noninterfering control. All siRNA duplexes used were synthesized by Dharmacon. HEK293 cells and cells expressing TLR4^YFP/MD‐2 were transfected with 20 nM siRNA for 48 h using Oligofectamine (Invitrogen) according to the manufacturer's protocol. For the silencing of Hrs and HCRP1, the cells were replated and subjected to a second round of siRNA treatment. For the introduction of plasmid DNA into the siRNA‐transfected cells, the cells were replated and transfected for 24 h using the GeneJuice transfection reagent.

HEK293 cells expressing TLR4^YFP were plated on 35 mm dishes and transfected 24–48 h before two washes in PBS and lysed in 100–300 μl of lysis buffer (20 mM Tris–HCl, 1 mM EDTA, 1 mM EGTA, 137 mM NaCl, 1% Triton X‐100, 1 mM sodium deoxycholate, 10% glycerol, 1 mM Na₃VO₄, 50 mM NaF and Complete protease inhibitor (Roche)). For Western blotting, the samples were denatured in 1 × NuPAGE LDS sample buffer supplemented with 25 mM DTT for 10 min at 70°C. Electrophoresis was performed using NuPAGE novex 7% Tris‐Acetate polyacrylamide or 10% Bis‐Tris polyacrylamide gels (Invitrogen) and the samples blotted onto 0.45 μm nitrocellulose filters (Biorad laboratories) and developed using ECL (Amersham Biosciences). For the analysis of immune complexes, lysates were precleared for 1 h with protein G Sepharose and subsequently incubated for 2 h with the appropriate antibody (2–3 μg/ml). Immune complexes were harvested with protein G Sepharose for 2 h before four washes in lysis buffer. Elution of the immune complexes was performed at 95°C for 10 min in 2 × LDS sample buffer supplemented with 50 mM DTT.

Images of live cells were captured at 37°C using an Axiovert 100‐M microscope with a heated stage equipped with a Zeiss LSM 510 META scanning unit and a 1.4 NA × 63 plan apochromat objective. For intracellular staining, the cells were permeabilized in 0.05% saponin, fixed in 3% paraformalaldehyde and incubated with the appropriate antibody as described previously (Simonsen et al, 1998). For LPS uptake studies in monocytes, fluorescently labeled LPS was added to the cells alone or simultaneously with 2 μg/ml Alexa⁵⁴⁶‐labeled anti‐TLR4 (HTA125) or 2 μg/ml Alexa546‐labeled isotype control. Only ring formed endosomes positive for TLR4 and/or LPS on the limiting membrane (minimal cutoff size 0.5 μm) were counted. The total intracellular LPS fluorescence in Figure 3B was quantified by manually drawing a region of interest that covered the cytoplasm and using the LSM510 software to calculate total fluorescence.

The human CD4⁺ Th1 cell clone (T18) specific for mouse antibody constant κ‐chain‐derived peptide Cκ^40–48 and restricted by HLA‐DR4 (DRA1, B1^*0401) was used to measure T helper‐cell responses (Schjetne et al, 2003). Antigen‐presenting cells used were HLA‐matched adherent peripheral blood monocytes (CD14⁺). Tissue culture medium was RPMI 1640 with glucose and l‐glutamine (Gibco, Paisley, Scotland) supplemented with pooled 10% human serum (HS) from blood donors.

NFκB activation was determined by an NFκB luciferase reporter assay as described (Latz et al, 2002).

Supplementary data are available at The EMBO Journal Online.