The crystal structure of human soluble CD14 reveals a bent solenoid with a hydrophobic amino-terminal pocket

Abstract

Human monocyte differentiation antigen CD14 is a pattern recognition receptor that enhances innate immune responses to infection by sensitizing host cells to bacterial lipopolysaccharide (LPS; endotoxin), lipoproteins, lipoteichoic acid and other acylated microbial products. CD14 physically delivers these lipidated microbial products to various Toll-like receptor signaling complexes that subsequently induce intracellular proinflammatory signaling cascades upon ligand binding. The ensuing cellular responses are usually protective to the host, but can also result in host fatality through sepsis. In this work, we have determined the X-ray crystal structure of human CD14. The structure reveals a bent solenoid typical of leucine rich repeat proteins with an amino terminal pocket that presumably binds acylated ligands including LPS. Comparison of human and mouse CD14 structures show great similarity in overall protein fold. However, compared to mouse CD14, human CD14 contains an expanded pocket and alternative rim residues that are likely to be important for LPS binding and cell activation. The X-ray crystal structure of human CD14 presented herein may foster additional ligand bound structural studies, virtual docking studies, and drug design efforts to mitigate LPS induced sepsis and other inflammatory diseases.

1. Introduction

CD14 was first characterized as a membrane cell surface differentiation marker of myeloid lineage cells (1). Membrane CD14 is a GPI anchored 55 kDa glycoprotein (2, 3) that is highly expressed on monocytes, macrophages and neutrophils with lower surface expression observed on a variety of other hematopoietic and stromal cells. Soluble forms of CD14 exist in serum, cerebrospinal, and other body fluids. Soluble CD14 is generated by at least three different mechanisms which include bypassing of GPI addition, cleavage of the GPI anchor by phospholipase D, or direct proteolytic cleavage from the cell surface (reviewed in (4)).

CD14 is best known as a pattern recognition receptor of the innate immune system that plays a prominent role in sensitizing cells to the presence of Gram-negative bacterial lipopolysaccharide (LPS; endotoxin) (4). CD14 is thought to sensitize cells to LPS by delivering this agonist to the Toll-like receptor 4 (TLR4) receptor signaling complex. Membrane CD14 shuttles LPS to TLR4 complexes in a two dimensional search (4); whereas, soluble CD14 enables cells that lack endogenous CD14, including most epithelial and endothelial cells, to respond to LPS (reviewed in (5)). LPS-induced TLR4 activation initiates host inflammation by driving cellular production of pro-inflammatory cytokines, chemokines and cell adhesion molecules. At the local site of infection, pro-inflammatory signaling resulting from CD14 shuttling is protective, leading to local clearance of invading bacteria. However, widespread infection and activation of this pro-inflammatory signaling system can cause fatality through sepsis (reviewed in (6, 7)). In response to LPS, CD14 deficient mice have blunted pro-inflammatory cytokine production and lack the lethargy, respiratory, and ruffled fur symptoms associated with murine septic shock. CD14 deficient mice also survive challenge with otherwise lethal doses of LPS (8).

Lipopolysaccharide is the major lipid present in the outer membrane of Gram-negative bacteria (9). LPS is an amphipathic molecule and in enteric bacteria is comprised of a hexaacylated lipid A region with a di-glucosamine backbone flanked by 1 and 4′ phosphates. The length, number, and saturation of the fatty acid tails in LPS can vary between different species of bacteria (10) as can the length and sugar content of the core and O-antigen regions (11). The binding of LPS by CD14 is slow unless this binding reaction is catalyzed by LPS binding protein (LBP) (12, 13). LBP is a 60 kDa human serum glycoprotein produced by liver hepatocytes (14, 15) that disaggregates and catalytically transfers LPS to CD14 (16). In concert with LBP, CD14 binds monomeric LPS (21) and shuttles it to the TLR4 complex. This effectively concentrates low levels of LPS and increases the sensitivity of the system (4, 12, 13, 17, 18). The major LPS acceptor of the TLR4 receptor complex is a 25kDa co-receptor called MD-2, which is physically associated with TLR4. The binding of LPS to the TLR4-MD2 complex facilitates dimerization resulting in a homodimeric receptor signaling complex composed of two TLR4 monomers, two MD-2 proteins and two LPS molecules (19). Cell signaling is facilitated by the intracellular signaling domain of TLR4 which upon dimerization creates a platform for intracellular signaling adaptors (20). Small gel filtration mixing studies have demonstrated the catalytic role of CD14 in delivering LPS to the TLR4-MD2 complex (21-24).

As a pattern recognition receptor, CD14 is capable of binding a wide variety of natural and synthetic acylated ligands in addition to LPS. For example, CD14 and LBP can work together to opsonize whole bacteria and apoptotic cells, clearing infection and reducing inflammation (25-29). CD14 is also able to bind and shuttle certain host phospholipids (30-32). Further, CD14 has been shown to enhance cellular inflammatory responses to a variety of acylated bacterial agonists of TLR2 including lipoteichoic acid, peptidoglycan, mycobacterial lipoarabinomannan, atypical lipopolysaccharides, and lipoproteins (33-40). Given this central role in TLR agonist delivery, CD14 is an obvious drug target for the treatment of sepsis (reviewed in (6, 7)).

The molecular interactions involved in the binding and delivery of structurally diverse ligands presumably requires multiple protein-ligand and protein-protein interaction sites on CD14 that are currently undefined. Additionally, the mechanisms involved in delivery of these ligands to multiple Toll-like receptor complexes are also unresolved. To address this gap in knowledge, many groups have purified various recombinant forms of soluble CD14 using bacteria, yeast, insect, and human cellular expression systems (12, 41-51), often with the goal of structure determination (49, 50). Currently, only the unliganded crystal structure of mouse CD14, purified from SF9 insect cells, is known (52). Mouse CD14 possesses an N-terminal hydrophobic cavity that provides a putative binding site for LPS and other acylated ligands. Despite previous efforts, no structure of human CD14 has been determined and the structural similarity between mouse and human CD14 is unresolved. As a first step towards this goal, we report the crystal structure of human CD14 (4GLP.pdb). Since septic shock is the most common cause of death in intensive care units (53), information garnered from the structure of human CD14 may foster ligand binding structural studies or drug development efforts.

2. Materials and methods

2.1. Expression

Human CD14 (aa 1-336) was amplified from genomic DNA by PCR using a sense primer, 5′-TTGGAATTCGCCGCCACCATGGAGCGCGCGTCCTGCTTGTTGC- 3′containing an EcoRI site, Kozak sequence, and a portion of the N-terminal secretion signal of human CD14, and an antisense primer, 5′- TTGTCTAGAACTACCGCGGGGGACGAGGGCAGTTCCAG GGACCAGGAAGG- 3′ containing a XbaI cleavage site followed by a thrombin digestion site. The PCR products were digested and cloned into a modified pDisplay vector via EcoRI and NheI sites (a kind gift from Dr. David Kranz, University of Illinois, Urbana-Champaign). This vector contains the coding sequence of the Fc domain of human immunoglobulin G1 downstream of a NheI restriction enabling generation of an Fc fusion protein. The antisense PCR primer was designed to include a thrombin cleavage site (LVPRGS) allowing for removal of the Fc fusion during purification. Site directed mutagenesis was completed through primer extension to yield the C306S mutation. The final construct sequence was confirmed by automated sequencing (UIUC Sequencing Center). After DNA amplification in E. coli DH5α cells, HEK 293F cells (Invitrogen) were transfected, cultured, and stably selected as previously described (54).

2.2. Purification

Human soluble CD14 (aa 1-336, C306S) was purified in four chromatographic steps. First, protein G affinity purification was used to harvest the human CD14 Fc fusion (CD14-Fc) from HEK 293F culture supernatant as previously described with the following exceptions (54). Two liters of HEK 293F supernatant from stably transfected CD14-Fc expressing cells was harvested seven days following seeding to 0.3 × 10⁶ cells mL⁻¹ in serum-free Freestyle 293F media (Invitrogen Life Technologies) under G418 (0.25 mg mL⁻¹) selection. Recombinant protein G sepharose beads (GE Healthcare; 2 mL 50% slurry) were added to the filtered supernatant with stirring overnight at 4°C. Protein G beads bound to CD14-Fc were harvested by centrifugation at 2,500 × g, 15 min, 4°C and packed into a disposable PD-10 column (GE Healthcare). The column was washed with 0.02 M sodium phosphate, pH 7.0 and eluted in ten column volumes of 0.1 M glycine-Cl, pH 2.3 with 1 mL neutralizing 1 M Tris-HCl, pH 9 buffer. Thrombin (Novagen) was used to remove the Fc tag by overnight incubation at 22°C.

Next, the products of the thrombin cleavage reaction were separated by passing the sample through an ÄKTAprime™ plus FPLC fitted with two tandem 1mL Hi-Trap Protein A high performance columns (GE Healthcare). The columns were run at a 1ml min⁻¹ flow rate in 0.02 M Tris HCl, pH 8.5. The flow through fractions containing free human CD14 were collected and injected on the two tandem Protein A columns three consecutive times to remove Fc. Flow through fractions containing soluble CD14 were pooled and concentrated to 1mL using an Amicon Ultra-15 unit (Millipore).

Since glycosylation contributes to protein heterogeneity, we removed N-linked glycans from CD14 by a one hour incubation at 37°C in 1xG7 buffer (0.05 M sodium phosphate, pH 7.5) containing 1,000 U peptide N-glycosidase F (PNGaseF) (New England Biolabs). This deglycosylated CD14 was further purified by anion exchange chromatography using two tandem 5 mL Hi-Trap Q anion exchange columns (GE Healthcare) and a linear NaCl gradient (0.02 M - 1 M) in 0.02 M Tris-HCl, pH 8.5 at 4 °C using a 1 mL min⁻¹ flow rate. CD14 containing fractions were concentrated using an Amicon Ultra-15 unit (Millipore) to 0.5 mL. Finally, size-exclusion chromatography was performed using a Superdex 200 column (GE Healthcare) equilibrated with 0.02 M Tris-HCl, pH 8.5 and 0.1 M NaCl at a flow rate of 0.4 mL min⁻¹. The fractions containing CD14 were pooled and concentrated to 10 mg ml⁻¹ using an Amicon Ultra-4 unit (Millipore) as measured by Pierce BCA assay (Rockford, IL). Buffer conditions were determined using a solubility screen (55). This four column purification process yielded ~2 mg L⁻¹ soluble, human CD14 with >90% purity by SDS-PAGE gel (Sup. Figs. 1A and 1B).

2.3. Crystallization and Optimization

Human CD14 (aa 1-336; C306S) was crystallized using the hanging-drop vapor-diffusion method. Crystallization conditions were initially screened using commercially available sparse matrix kit conditions including Hampton Research Crystal Screens I and II, Emerald Biosystems Wizard I and II, Hampton Research Natrix, and Hampton Research Salt Reaction sparse matrix screening kits, and a handmade crystal screen targeting the previously published mouse CD14 crystallization condition (1 μl protein solution and 1 μl of crystallization buffer containing 100 mM sodium HEPES (pH 7.5), 1.9 M Li₂SO₄, and 5 mM NiCl₂) (52). Each condition was tested on a 12 mm x 0.22 mm siliconized glass cover slide (Hampton Research) over a VDX48 plate with sealant (Hampton Research) by mixing 1 μl of each screening solution with 1 μl protein solution (10 mg ml⁻¹ protein in 0.02 M Tris-HCl, pH 8.5 and 0.1 M NaCl) equilibrated against 300 μl screening solution in the reservoir. Sparse matrix screening in hanging drop vaporization trays at 22°C, 18°C, and 4°C lead to three conditions with single crystals: Emerald Biosystems Wizard I #28 (20% PEG 3,000, 0.1 M HEPES, pH 7.5, and 0.2 M NaCl) after 2 - 4 days at 22°C; Hampton Research Crystal Screen I #18 (20% PEG 8,000, 0.1 M Na cacodylate*3H₂0, pH 6.5, and 0.2 M Mg(OAc)*4H₂0) after 6 days at 18°C; and Emerald Biosystems Wizard II #28 (20% PEG 8,000, 0.1 M MES, pH 6.0, and 0.2 M Ca(OAc)₂) after 3 weeks at 22 °C (56-58). The best initial diffraction resolution (6.2 Å) was obtained using Emerald Biosystems Wizard II #28.

Extensive work was conducted to improve diffraction resolution. During crystal optimization, various additives, detergents, proteases, and cryoprotectants were screened, utilizing multiple crystal harvesting time points (59-61). Further post-crystallization optimization work included in situ proteolysis, dehydration via serial transfer on cover slips or over the reservoir for different durations, macromolecular crystal annealing, flash annealing, and reductive methylation of surface lysine residues using dimethylamine-borane complex (62, 63). Treatment with each of these additional techniques failed to improve the diffraction resolution (data not shown).

Single crystals of purified deglycosylated CD14 were grown at 22°C to a maximum dimension over ~40 days on a 22 mm x 0.22 mm siliconized glass cover slide (Hampton Research) over a VDX24 plate with sealant (Hampton Research) by mixing 1.5 μl 30% PEG 6,000, 0.1 M MES, pH 6.0, and 0.2 M Ca(OAc)₂ crystallization solution with 1.5 μl protein solution (10 mg ml⁻¹ protein in 0.02 M Tris-HCl, pH 8.5 and 0.1 M NaCl) equilibrated against 1 mL screening solution in the reservoir. Crystals were harvested from the drop and plunged directly into liquid N₂ without additional cryoprotection.

2.4. Data collection and structure determination

A native diffraction data set was collected at 4.0 Å resolution from a single vitrified crystal in the optimized 20% PEG 6,000, 0.1 M MES, pH 6.0, and 0.2 M Ca(OAc)₂ crystallization solution on the LS-CAT beamline, Sector 21 ID-F (Advanced Photon Source, Argonne, IL). Diffraction data were indexed and integrated using XDS followed by scaling and merging with XSCALE. The structure solution was determined via molecular replacement with Phaser in the CCP4 suite, using the molecular coordinates of mouse CD14 (PDB ID 1WWL), modified with CHAINSAW. Iterative cycles of model fitting were completed in COOT, followed by refinement in REFMAC. The final structure solution of human CD14 was deposited (PDB ID 4GLP; http://www.pdb.org/pdb/search/structidSearch.do?structureId=4GLP) and verified with Molprobity in PHENIX. Data collection and refinement statistics are summarized in Table 1.

Table I.

Data collection and refinement statistics

	Human CD14
Data collection
Space group	P 32 2 1
No. of mol. in asym. unit	1
Cell dimensions
a, b, c (Å)	a = 147.52, b = 147.52, c = 44.07
β (°)	β = 90
Resolution (Å)	4.0
Rsyma,b (%)	8.6 (84.2)
I/σa	17.20 (3.19)
Completenessa (%)	99.6 (100.0)
Refinement
Resolution range (Å)a	27.9-4.0 (4.1-4.0)
Total reflectionsa	53002 (3887)
Unique reflectionsa	4780 (341)
R factor (%)c	28.5
Rfree (%)c	32.1
No. atoms in protein	2346
Average B factor	10.9
Wilson B factor	39.2
r.m.s. deviation
Bonds (Å)	0.010
Angles (°)	1.640

^aNumbers in parentheses correspond to the highest resolution shell. r.m.s.d., Root-mean-square deviations from ideal geometry.

^bRmerge = ∑h∑I ∣Iih-<Ih>∣/∑h∑I <Ih>, where <Ih> is the mean intensity of the observations Iih of reflection h.

^cR factor = ∑ (∥Fobs∣-Fcalc∥)/∑∣Fobs∣; R_free is the R factor for a subset (5%) of reflections that was selected prior to refinement calculations and not included in the refinement.

2.5. LPS Binding Assay

The binding of LPS by PNGaseF deglycoslated human CD14 (34 kDa) was detected by a native PAGE gel shift method (64). 10 μg human CD14 was mixed with a 2 molar ratio of the Ra chemotype of E. coli LPS (3.79 kDa; Sigma-Aldrich), with or without a 1:100 molar ratio of CD14:human LBP (60 kDa; kindly provided by Dr. Jerrold Weiss, University of Iowa). The reactions were incubated for 1 h at 37°C in a total volume of 5 μL in PBS buffer containing 0.01M EDTA. 5 μL bromophenol blue with glycine dye was added and each sample was loaded in a 4-20% Mini-PROTEAN TGX precast gel, pH 6.8 (Bio-rad) run at 100V in the recommended running buffer without SDS at 4°C for 3 hours (Sup. Fig. 2).

2.6. Bioactivity Assay

Various concentrations of the Ra chemotype of E. coli LPS (Sigma) were pre-incubated for 1 h at 37 °C at a 2:1 molar ratio LPS:CD14 in RPMI media containing 1% human serum albumin. For some reactions, LBP was added as a catalyst at a 1:100 molar ratio of LBP:CD14. The bioactivity of CD14 in these reactions was assessed using SW620 human epithelial cells which express TLR4/MD-2 and lack detectable levels of membrane CD14 (42). SW620 cells were cultured in RPMI complete media containing 1% bovine serum albumin and seeded in a 96-well plate at a density of 1×10⁶ cells/well and allowed to adhere for 1 h at 37°C, 5% CO₂. To remove serum, SW620 cells were carefully washed 4 times with PBS, 2 times with Gibco FreeStyle 293 Expression Media (Invitrogen), and 2 times with RPMI containing 1% human serum albumin. Preincubated LPS samples were added to the washed SW620 epithelial cells. Supernatant was harvested 6 hours later and IL-8 production was measured using a human IL-8 Cytoset sandwich ELISA kit (Invitrogen).

3. Results

3.1 Generation of Bioactive Human CD14 Suitable for Crystallization

A number of important parameters were considered in generating a soluble form of human CD14 suitable for crystallization. Since soluble forms of CD14 are naturally heterogeneous at the C-terminus, we generated an expression construct which lacks the C-terminal protein sequence required for GPI anchoring. This was done with the knowledge that C-terminal truncation of CD14 does not affect LPS associated bioactivity, which resides entirely within the amino-terminal half of the protein (65, 66). Additionally, since our CD14 construct is truncated between C306 and C352, which are predicted to form the final disulfide bond, we mutated the unpaired C306 residue to serine. CD14 was expressed as an Fc tagged fusion protein and purified over protein A. An engineered thrombin cleavage site enabled removal of the Fc tag, and further purification steps resulted in highly pure soluble CD14 protein suitable for crystallization (Sup. Fig. 1A and 1B). To further ensure improved protein homogeneity for crystallization, we removed potentially heterogeneous N-linked glycans using PNGase F. This enzymatic deglycosylation was preferred because mutation of the N-linked glycosylation sites in CD14 prevents cellular secretion (52). As observed by others, PNGase F treatment reduced the molecular weight of CD14 by approximately 10kD ((67, 68) (Sup Fig 1C).

To ensure that our soluble human CD14 is biologically active, we measured the ability of the purified protein to bind LPS and to mediate MD-2/TLR-4 dependent chemokine production. Binding was assessed by mixing a 2:1 molar excess of the Ra chemotype of E. coli LPS with CD14 in the absence and presence of a catalytic amount of LBP. CD14 formed a complex with LPS as evidenced by altered mobility on a gradient native PAGE gel relative to CD14 alone (Sup. Fig. 2). We also assessed the ability of our purified soluble CD14 to induce LPS-mediated activation of human epithelial SW620 cells. These cells naturally express the TLR4/MD-2 signaling complex, but lack endogenous CD14 such that LPS induced activation is greatly enhanced by addition of active forms of soluble CD14 (41, 42). To this end, purified soluble CD14 was pre-incubated with increasing concentrations of LPS in the presence of LBP to facilitate CD14-LPS complex formation. Preincubated LPS, LBP and CD14 potently stimulated IL-8 production of SW620 cells and enabled detection of 100 pg/mL LPS, with maximal cell stimulation observed at a 10 ng/mL LPS (Sup. Fig. 3). These LPS-induced IL-8 production levels were comparable to that mediated by human serum which naturally contains LBP and CD14 catalytic activity. In contrast, either LPS alone or LPS preincubated with LBP poorly activated the cells. Taken together, these results show that the soluble human CD14 (aa 1-336, C306S) used herein binds LPS and is biologically active.

3.2 Crystal Structure of Human CD14

Human CD14 crystallized in the trigonal space group P3₂21 (Table I). The crystal structure begins with aa 26, indicating complete removal of the leader sequence by HEK 293F cells during secretion along with low electron density for the first six amino acids, aa 20-25. The overall X-ray crystal structure of human CD14 reveals a bent solenoid that is formed by 10 leucine rich repeats, each creating a single turn (Fig. 1A). This is consistent with the leucine rich repeat sequence motifs of human CD14 (69), and the expected bent solenoid fold generally characteristic of leucine rich repeat containing proteins (70). The structure also reveals five alpha helices on the convex side, as well as eleven beta strands that are coordinated in a parallel beta sheet on the concave side of the bent solenoid. The pairwise proximity between cysteine residues in the crystal structure at positions 34 and 51, 187 and 217, as well as 241 and 272 are consistent with the expected disulfide bonds of human CD14, which have been shown to be important for proper protein folding (49). Interestingly, recombinant purified human CD14 lacking the C306S mutation, a mutation originally designed to avoid unnatural disulfide binding, did not crystallize and formed aggregates in the crystallization drop (data not shown).

Fig. 1. Human CD14 X-ray crystal structure.

A, The X-ray crystal structure of human CD14 aa 26-335 is shown in ribbon (red). Alpha helices (α1, α4, α5, α6, α7), beta strands (β3-β13), and the positions of cysteine residues (orange) are indicated. The site directed mutagenesis of C306S is shown in magenta. B, Secondary structure alignment of human CD14 (red) and mouse CD14 (gray) was created using the default parameters of MultiSeq in VMD. Disulfide bonds in human CD14 (orange) and mouse CD14 (yellow) are indicated. The site directed mutagenesis of C306S is shown in human CD14 (magenta). C, A structure based sequence alignment of human and mouse CD14 was created using the default parameters of MultiSeq in VMD. MultiSeq output was used to assign human CD14 secondary structure elements upon visualization in VMD, and mouse CD14 secondary structure positions were derived from the header notation of 1WWL.pdb. Numbered leucine rich repeat regions, which are conserved in location between both species, are shown in brackets ([LRR]). Alpha helices (red rectangles), beta strands (yellow rectangles), turns and loops (black dashed lines) are indicated. Cysteine residues (orange) and residues located in the rim of the N-terminal pocket are boxed for human CD14 (maroon) and mouse CD14 (blue). Since amino acid numbering is not standardized in previous publications, we have chosen to compare mouse and human CD14 crystal structures by starting amino acid numbering at the first methionine residue.

Another key structural feature present in the crystal structure of human CD14 is an N-terminal pocket or cavity (Fig. 2A). Although we cannot know with certainty the position of each side chain at this resolution, the overall cavity of human CD14 is comprised largely of hydrophobic residues present in α1 through α5 and β1 through β6 along with their connecting loops. It is reasonable to assume that this hydrophobic pocket serves as the binding site for the lipid chains of LPS and other acylated CD14 ligands (Fig. 2B), especially as the regions making up the pocket are entirely contained within the N-terminal region (aa 20-171) of CD14 that has been shown to be sufficient for bioactivity (65, 66).

Fig. 2. Comparison of mouse and human CD14 N-terminal ligand binding pocket.

A, Mouse (gray) and human (red) CD14 ribbon structures are overlaid over mouse (translucent gray) and human (translucent red) CD14 space filling structures drawn in VMD. CD14 structures are rotated 90° relative to Fig. 1B. The N-terminal hydrophobic binding pocket is indicated (black arrow). B, Space filling hydrophobicity structures were drawn in Chimera for human CD14 (left) and mouse CD14 (right). The Kyte-Doolittle scale was used to compare hydrophobicity by showing hydrophobic (orange), neutral (white), and hydrophilic (blue) residues as well as glycosylation sites (red).

It is instructive to examine the rim residues of the CD14 hydrophobic pocket as these residues likely engage portions of LPS, or other acylated CD14 ligands, that are not accommodated by the pocket itself. Positively charged residues are located in the rim at K71 and R72 and just outside the rim at R80, K87, and R92 (Fig. 3A). Hydrophobic residues at W45, F49, V52, F69, Y82, and L89 encircle the rim and their side chains overlay the entrance to the pocket. When measured using the program Chimera, the pocket diameter of human CD14 from W45 to L89 is 15.7 Å wide. Interestingly, the hydrophobic pocket entrance is subdivided by an interaction between the α1 residue F49 and the α3 residue Y82 (Fig. 3B).

Fig. 3. Comparison of mouse and human CD14 pocket rim residues.

A, The structure of the rim of the predicted N-terminal ligand binding pocket of both mouse and human CD14 are overlaid. Rim residues of human CD14 are denoted in red and those of mouse CD14 are denoted in blue. Regions important for LPS binding, as determined from previous mutagenesis, blocking antibody, and epitope mapping studies are indicated in black with gray text. B, A space filling representation of the residues located on the rim of the predicted N-terminal ligand binding pocket are human and mouse CD14 are displayed in red and blue, respectively. Regions important for LPS binding are displayed in black ribbon.

3.3 Comparison of Human and Mouse CD14 Structure

Human soluble CD14 purifies and crystallizes as a monomer (Sup. Figs. 1A and 1B). In contrast, the murine protein crystalized as a dimer with the dimer interface comprised of C-terminal beta strand interactions (52). The mouse and human CD14 structures are comparable at the C-terminus because both were truncated at similar positions to make them more amenable to crystallization (52). Different crystallization conditions, constructs, and cellular expression systems may account for the differences observed in crystal packing between the two proteins. Nevertheless, it is unlikely that the C-terminal dimerization of truncated mouse CD14 is biologically relevant (52), as numerous biophysical studies support the idea that bioactive soluble CD14 is a monomeric protein (4, 12, 13, 21-24).

Human and mouse CD14 crystal structures contain a bent solenoid fold and asparagine ladder, which are structural features typical of leucine rich repeat proteins (70). In fact, both proteins are highly superimposable with a r.m.s.d. of 1.089 Å (Fig. 1B). The secondary structure of human CD14 includes alpha helices and beta strands that are equivalent to α1, α4, α5, α6, α7, and β3-β13 of the published mouse CD14 structure (Figs. 1B and 1C) (52). Although helical coils exist in human CD14 at locations that are synonymous to helices α2 and α3 of mouse CD14, the structure resolution falls just outside the range needed to determine helical packing with certainty (Figs. 1A and 1C; (71, 72). Human and mouse CD14 both contain a large N-terminal hydrophobic pocket consistent with their similar roles in the binding and delivery of various lipidated molecules including LPS (Figs. 2A and 2B). Similar to human CD14, the main pocket and sub-pockets present in mouse CD14 are comprised of hydrophobic residues within α1-5, β1-6, and their connecting loops (52). Using a structure based sequence alignment, created using MultiSeq in VMD, we compared the general amino acid characteristics present in the residues at the rim of the proposed binding pocket in both structures (73). The authors of the mouse CD14 crystal structure identified hydrophilic and hydrophobic amino acids in rim residue positions indicated in blue ((52); Fig. 3A). Many of these residues, including P39, W41, F45, and Q77 are conserved with rim residues at the pocket entrance of human CD14. Thus, the N-terminal hydrophobic pocket found in our crystal structure revealed a conservation of pocket size and hydrophobicity with that of mouse CD14.

Despite the wide ranging structural similarities between mouse and human CD14, differences do exist. For example, although the amino acids found at equivalent positions of mouse and human CD14 generally have similar physical properties, there is variation in the charge of certain rim residues. Namely, the positively charged K38 residue present in the mouse protein is not retained in the human protein. Likewise, the negatively charged D44 residue in human CD14 is not found in mouse CD14. Finally, the diameter and subdivision of the entrance to the pocket differs between species. The mouse crystal structure reveals an 8.0 Å wide entrance to the pocket that is bounded on one side by a hydrophobic interaction between F45 and F78 ((52); right side of Fig. 3A). Residues F45, L49, and I81 in the mouse CD14 structure create additional hydrophobic interactions which shield this side of the pocket. These hydrophobic interactions of the mouse CD14 pocket are not present in the human CD14 structure whose pocket extends to include the rim residues T85 and L89 (Fig. 3A). The biological significance of these differences between the pocket and associated rim of mouse and human CD14 is difficult to define without structural information from ligand bound proteins.

4. Discussion

CD14 sensitizes cells to LPS by delivering this bioactive lipid to MD-2; an essential component of the TLR4 signaling complex. All members of the MD-2 related lipid-recognition protein superfamily are characterized by two β sheets organized in an αβ cup fold to create a centralized hydrophobic LPS binding cavity (74). In the crystal structures of human and mouse MD-2 bound to lipid IVa, all four acyl chains of the ligand are buried inside and occupy the majority of the hydrophobic pocket volume (75, 76). Crystal structures of MD-2 in the context of the entire mouse and human TLR4/MD-2/LPS homodimeric signaling complex reveals a narrow and deep (1720 ų volume) hydrophobic binding pocket that completely sequesters five of the six fatty acid chains (77, 78). The remaining unbound acyl chain lies along the surface of MD-2 and, together with the F126 loop of MD-2, creates a new hydrophobic patch that promotes homodimer formation by association with TLR4 from an adjacent TLR4/MD-2/LPS complex (19, 77).

Compared to MD-2, CD14 has been reported to afford less protection against enzymatic removal of secondary fatty acids of bound LPS (79). This probably reflects that fact that at about 820 ų, the hydrophobic pockets of both human and mouse CD14 are considerably smaller than that of MD-2 (52) (Figs. 2A and 2B). Similar to MD-2, previous circular dichroism, tryptophan fluorescence, and NMR studies suggest that LPS binding does not induce large structural changes in human CD14 (50, 80). Given these volume constraints and pocket rigidity, it is unlikely that CD14 can accommodate all the acyl chains of LPS. This may contribute to the reduced affinity of CD14 for LPS compared to MD-2 (22, 23).

Given the physical limitations of the hydrophobic pocket, we hypothesize that CD14 stabilizes additional fatty acid chains on a hydrophobic cluster located near α1 just outside of the pocket entrance. A similar hydrophobic cluster is present in a comparable location in the crystal structure of mouse CD14 ((52); Fig. 2B). In human CD14, the hydrophobic cluster is comprised of the amino acids F32, F49, V52, S53, A54, V55, and L89 (Fig. 2B). This cluster and associated acyl chains of LPS could perhaps create a new protein-protein interaction surface that may help facilitate the transfer of LPS to another molecule such as MD-2.

Our crystal structure confirms that the hydrophobic pocket encompasses the N-terminal half (aa 20-171) of CD14 that has been shown to be the bioactive portion of the protein (28). Extensive deletion mutagenesis, alanine scanning mutagenesis and epitope mapping of inhibitory anti-CD14 antibodies have revealed four separate regions in the N-terminal half of CD14 that are critical for LPS binding and cell activation (Fig. 3B). These four functionally important regions are delineated by amino acids 26-32, 41-44, 56-64 and 78-83 which are physically located near β1, within the loop preceding α1, across β3 and within α3, respectively (28, 81-83),(84). Additionally, LPS binding has been shown to protect aa 57-64 from endoprotease AspN digestion (80). The residues in the first three regions near α1, β1 and β3 appear to fulfill critical roles in forming the pocket entrance, capping the leucine rich repeat beta sheet, and/or stabilizing proper folding of the solenoid.

Residues in the fourth region encompass a Y82 loop located on the opposite side of the pocket entrance. Although Y82 appears to weakly block the entrance of the pocket through interaction with F49, we believe that upon binding LPS, these hydrophobic residues may change position to accommodate the acyl chains within the pocket. Although the corresponding interaction is represented by F45 and F78 in mouse CD14, additional hydrophobic residues present in mouse CD14, but absent in human CD14, close off the pocket entrance (Figs. 3A and 3B). Overall, our human CD14 crystal structure allows for the resolution of an N-terminal hydrophobic pocket predicted by multiple studies to be the LPS binding site. Compared to mouse CD14, our work provides an alternative definition of rim residues, reveals an expanded pocket entrance, and identifies a unique hydrophobic bridge feature that incorporates residues shown to be important in LPS binding and cell activation (Figs. 3A and 3B).

Hydrophilic residues at the opening of MD-2 have been shown to be important for LPS binding by properly aligning the negatively charged 1 and 4′ phosphates of LPS (42, 77). Similar to MD-2, charged rim residues in CD14 may be important for orienting or binding LPS. For example, previous work has shown that mutation of E47, which falls near the front of the rim in α1, to either lysine or arginine can block binding of P. gingivalis LPS (85). Solution NMR spectroscopy analyzing soluble CD14 aa 20-171 bound to the Re chemotype of LPS did not have structural assignments available upon publication, but does show high average temperature factor evidence for hydrophilic residues at the rim suggesting local flexibility to accommodate LPS binding (50). Further work to develop a ligand bound CD14 structure would help to clarify the roles of both the hydrophobic binding pocket and the hydrophilic rim residues required for ligand binding.

A number of natural and synthetic acylated agonists and antagonists are known to be shuttled by CD14. For example, CD14 delivers tri- and di-acylated agonists to complexes formed by members of the TLR2 subfamily (33-40, 86). Natural antagonists of LPS, including R. sphaeroides pentacylated lipid A, tetracylated lipid IVa and monophosphoryl lipid A, an approved vaccine adjuvant in Europe, are shuttled by CD14 to MD-2/TLR4 (78, 87-90). Eritoran, a synthetic LPS antagonist, that is structurally similar to lipid IVa, is also delivered to MD-2 by CD14 (91, 92). More recently, synthetic LPS inhibitors derived from either diacylated sugars or tetraacylated sulfate containing compounds have been identified and shown to competitively block LPS binding to CD14 (93-95). It is highly likely that the hydrophobic binding pocket in human CD14 accommodates these natural and synthetic agonists and antagonists especially as they contain fewer acyl chains than hexa-acylated enteric bacterial LPS (89, 96). Since TLR4 inhibitors are largely LPS mimics that are either delivered by CD14 or directly compete with LPS for interaction with CD14, the structural information presented here may further the design of drugs for the treatment of sepsis and other inflammatory diseases.