The human MHC class I-like molecule CD1b is distinctive among CD1 alleles in that it is capable of presenting a set of glycolipid species that show a very broad range of variation in the lengths of their acyl chains. A structure of CD1b complexed with relatively short acyl chain glycolipids plus detergent suggested how an interlinked network of channels within the Ag-binding groove could accommodate acyl chain lengths of up to 80 carbons. The structure of CD1b complexed with glucose monomycolate, reported in this study, confirms this hypothesis and illustrates how the distinctive substituents of intracellular bacterial glycolipids can be accommodated. The Ag-binding groove of CD1b is, uniquely among CD1 alleles, partitioned into channels suitable for the compact accommodation of lengthy acyl chains. The current crystal structure illustrates for the first time the binding of a natural bacterial lipid Ag to CD1b and shows how its novel structural features fit this molecule for its role in the immune response to intracellular bacteria.

The CD1 proteins are a family of Ag-presenting molecules that present lipid Ags to T lymphocytes (1). Human CD1 molecules segregate into two groups according to sequence homology: group 1 contains CD1a, CD1b, CD1c, and CD1e molecules, whereas group 2 contains CD1d molecules.

Binding studies and structural studies have led to the definition of a general molecular mechanism for lipid Ag presentation by CD1 molecules. The crystal structure of mouse (m)CD1d1 suggested that CD1 molecules accommodate hydrophobic acyl chains in a hydrophobic binding groove (2). Functional studies supporting this view were conducted using a series of analogues of glucose monomycolate (GMM)5 presented by CD1b molecules varying in glycosylation, hydroxylation, and lipid length (3, 4). The combination of these structural and functional studies suggested that the alkyl components of the Ags bind within a hydrophobic groove in the CD1 protein, the hydrophilic glycan moiety protrudes from the binding groove, making contacts with the TCR (5).

This model for lipid Ag presentation has been recently confirmed by the crystal structures of human CD1b (6) and CD1a (7) in complex with specific glycolipids. In addition, the structure of CD1b suggested that, unlike other CD1 molecules, this molecule has a unique binding groove architecture characterized by a series of channels capable of accommodating a very broad range of acyl chain length.

Refolding of denatured CD1b molecules in the presence of ganglioside GM2 and phosphatidylinositol (PI) was obtained using a protocol based on the inclusion of single acyl chain detergents in the refolding to prevent CD1b precipitation (6). The three-dimensional structure of CD1b molecules refolded using this protocol showed the presence of detergent moieties occupying the channels that were not filled by the lipid ligands. To test the hypothesis that longer acyl chain lipids (such as mycolates and GMM) occupy the full set of interlinked channels, the crystal structure of CD1b-GMM complex, refolded in the absence of detergent, was determined.

The extracellular α13 domains of human CD1b (residues 1–283) and β2-microglobulin (β2m) were synthesized separately using a prokaryotic expression system (pET; Novagen, Madison, WI) (6). Both recombinant proteins were purified from Escherichia coli inclusion bodies and solubilized in 6 M guanidine-HCl solution containing 10 mM DTT. GMM was purified from Nocardia farcinica as previously described (8), solubilized at 200 μg/ml in 150 mM NaCl and 0.5% Tween 20 solution, and sonicated. CD1b and β2m proteins were refolded with the GMM by oxidative refolding chromatography using a protocol previously described (9, 10). The concentrated refolding mix was purified by size-exclusion chromatography using a HiLoad 26/60 Superdex-75 (Amersham Pharmacia, Piscataway, NJ). Crystallization conditions were screened using purified monomeric CD1b-GMM glycolipid complexes (7.5 mg/ml in 20 mM Tris-Cl (pH 6.5), 25 mM NaCl, and 2 mM EDTA). Nanoliter-scale crystallization experiments were set up using a Cartesian Technologies Microsys MIC4000 (Genomic Technologies, Huntingdon, U.K.) (11, 12). Conditions yielding crystals in nanoliter-scale sitting drops (100 nl of protein plus 100 nl of reservoir solution) were subsequently scaled up and optimized by hand pipetting. Optimally diffracting crystals were grown at 20°C from 2-μl sitting drops, using a 1:1 protein to precipitant (100 mM Na citrate (pH 5.6), 0.20–1.25 M ammonium sulfate, and 0.20–0.75 M Li2SO4) ratio and 5–15 mM urea as an additive in the drops.

Crystals were flash frozen at 100 K in mother liquor containing 20–25% glycerol. X-ray diffraction data were collected at beam lines 14.2 (Synchrotron Radiation Source (SRS), Daresbury, U.K.) and ID-14 EH4 (European Synchrotron Radiation Facility (ESRF), Grenoble, France) with 0.98- and 1.01-Å radiation, respectively, recorded on Area Detector Systems Corporation (Poway, CA) Quantum 4 charge-coupled device detectors. The data were processed using DENZO, SCALEPACK (13), and TRUNCATE (CCP4 suite). Processing statistics are given in Table I. The crystals belong to spacegroup P3121 with diffraction to 3.8 Å at SRS 14.2 (with unit cell dimensions, a = b = 97.5 Å, c = 115.2 Å) and to 3.1 Å at ID-14 EH4 (with unit cell dimensions, a = b = 97.0 Å, c = 114.8 Å). The crystallographic asymmetric unit contains one molecule of CD1b-GMM with a solvent content of 67%.

Table I.

Crystallographic statistics

ESRF ID-14 EH4SRS 14.2
Data collection   
 Spacegroup P3121 P3121 
 Unit cell (Å) a = b = 97.0, c = 114.8 a = b = 97.5, c = 115.2 
 α = β = 90°, γ = 120° α = β = 90°, γ = 120° 
 Resolution range (Å) 20–3.1 20–3.8 
 Completeness (outer) (%) 99.9 (100.0)a 100.0 (100.0)a 
 Total observations 150,075 134,566 
 Unique reflections 11,697 6,512 
 Average I/ς(I) (outer)a 34.9 (5.7)a 19.4 (4.8)a 
Rmerge (outer) (%) 7.7 (60.4)a 22.6 (93.1)a 
Model refinement   
 Maximum resolution (Å) 3.10  
 Reflections (working set/test set)c 10,887/800  
Rwork/Rfree (%)d 23.3/29.1  
 RMSD from standard stereochemistry   
  Bonds (Å) 0.009  
  Angles (°) 2.11  
 Number of atoms   
  Protein/lipid/water 3,011/76/109  
 Mean B-factors   
  Protein/lipid/water 84.6/76.8/86.0  
 Ramachandran plot   
  Most favored/additional (%) 69.9/29.8  
  Generous/disallowed (%) 0/0.3  
ESRF ID-14 EH4SRS 14.2
Data collection   
 Spacegroup P3121 P3121 
 Unit cell (Å) a = b = 97.0, c = 114.8 a = b = 97.5, c = 115.2 
 α = β = 90°, γ = 120° α = β = 90°, γ = 120° 
 Resolution range (Å) 20–3.1 20–3.8 
 Completeness (outer) (%) 99.9 (100.0)a 100.0 (100.0)a 
 Total observations 150,075 134,566 
 Unique reflections 11,697 6,512 
 Average I/ς(I) (outer)a 34.9 (5.7)a 19.4 (4.8)a 
Rmerge (outer) (%) 7.7 (60.4)a 22.6 (93.1)a 
Model refinement   
 Maximum resolution (Å) 3.10  
 Reflections (working set/test set)c 10,887/800  
Rwork/Rfree (%)d 23.3/29.1  
 RMSD from standard stereochemistry   
  Bonds (Å) 0.009  
  Angles (°) 2.11  
 Number of atoms   
  Protein/lipid/water 3,011/76/109  
 Mean B-factors   
  Protein/lipid/water 84.6/76.8/86.0  
 Ramachandran plot   
  Most favored/additional (%) 69.9/29.8  
  Generous/disallowed (%) 0/0.3  
a

Values in parentheses refer to the highest resolution shells (3.21–3.10 Å for ID-14 EH4 and 3.93–3.80 Å for SRS 14.2 datasets).

b

Rmerge = Σ(Iobs− 〈I〉)/Σ〈I〉.

c

All reflections with F > 0 were included.

d

R = Σ|FobsFcalc|/Σ Fobs, where Fobs and Fcalc are the observed and calculated structure factor amplitudes, respectively. Rfree is as for Rwork but calculated for a test set comprising reflections not used in refinement.

The CD1b-GMM structure was initially solved from data collected at SRS 14.2 by molecular replacement with the program EPMR (14) using coordinates for unliganded CD1b derived from the CD1b-PI structure (Protein Data Bank (PDB) entry 1GZQ, (6)). The solution was unambiguous with a correlation coefficient of 67% for data from 20.0- to 3.8-Å resolution. The structure was refined against the higher resolution data collected at ID-14 EH4 using rigid-body, positional, and group B-factor refinement in CNS (15) (Table I).

An initial Fo-Fc map was calculated using the molecular replacement solution (unliganded) after rigid-body, positional, and B-factor refinement. At a contour level of 2.0 ς, tubular electron density was clearly visible in the A′, T′, F′, and C′ channels. The density spanning A′-T′-F′ was continuous with the exception of missing density at three points in the T′ channel giving gaps of ∼0.5, 2.3, and 2.5 Å. Density was also visible for some hydroxyl groups of the sugar head group.

Between each further refinement round, the model was manually checked and glycolipid atoms built in using O (16), with reference to 2Fo-Fc and Fo-Fc electron density maps generated using phase information calculated from the models. Glycolipid parameter and topology files for CNS were generated using PRODRG (17). Water molecules were automatically picked using CNS and manually verified. Protein stereochemistry was validated using PROCHECK (18). Statistics for the refined structure are shown in Table I.

Atomic interactions between glycolipid and protein were calculated using LIGPLOT (19). Volume calculations for cavities within the protein were calculated using VOLUMES (R. M. Esnouf, unpublished program). Variant glycolipids were modeled into the CD1b binding groove and energy minimized in CNS. The CD1b-GMM specific TCR, LDN5 (3), was built using sequence information from the IMGT database (20) and modeled using SWISS-MODEL (21). Superimposition of coordinates for comparative analyses was conducted using SHP (22). Sequence alignments were performed using the MULTALIN web interface (23), and the output was formatted using ESPRIPT (24). Hydrophobicity calculations were performed via the PROTSCALE web interface (25) using the Eisenberg method (26). Figures were produced using BOBSCRIPT (27) and RASTER3D (28).

GMM extracted from N. farcinica was analyzed by mass spectrometry as previously described (29). The mass of the GMM bound in the CD1b-GMM crystal structure was analyzed using the following protocols. Drops containing crystals were removed by pipette from crystallization trays and dissolved in 6 M guanidium-HCl to denature the protein and dislodge the GMM from the CD1b binding groove. This treatment also cleaved off the sugar group in the GMM to afford free mycolic acids. Separation of mycolic acids from the protein was conducted by organic extraction using chloroform/methanol. The organic phase containing the liberated mycolic acid was dried under a nitrogen stream, and the resulting pellet was dissolved in chloroform/methanol. This sample was analyzed by electrospray mass spectroscopy. Measurements were conducted in positive-ion mode and performed on a triple quadrupole LCT instrument (Micromass, Altrincham, U.K.) fitted with an atmospheric pressure electrospray source. Samples were directly injected (10 μl) using a Rheodyne injector (Rheodyne, Bensheim, Germany) with methanol as the mobile phase. The machine was run at a flow rate of 200 μl/min with the cone voltage at 35 V and the spraying needle voltage set at 3 kV. The scan rate was 1 s for the mass range from 200 to 2000 Da.

The structure of CD1b-GMM was determined by molecular replacement (see Materials and Methods) and refined at 3.1-Å resolution (Table I). The final model contains CD1b residues 3–282 (extracellular α13 domains), β2m residues 1–100, and one GMM molecule (Fig. 1, A and B).

FIGURE 1.

Structure of the CD1b-GMM complex and ligand-binding groove of CD1b. A, Overall structure of CD1b-GMM complex. The main chain for the α13 domains of CD1b plus β2m is depicted schematically in ribbon representation (in blue and gray, respectively), and the nonhydrogen atoms of the glycolipid GMM are drawn as van der Waals spheres (carbon in gray; oxygen in red). Residues responsible for flexibility between the α1α2 and α3 domain are colored in gold. B, Electron density for the glycolipid in the CD1b-GMM crystal structure. GMM is drawn in ball-and-stick (carbon in black; oxygen in red). The electron density is from a Fo-Fc omit map calculated after simulated annealing with the glycolipid omitted, contoured at 2.5 ς, and represented as a red mesh. C, The ligand-binding groove of CD1b. The hydrophobic binding surface is drawn as a gray mesh with carbon from the acyl tails occupying the groove drawn as van der Waals spheres. Atoms occupying each channel and the channel label are in matched colors. D, The chemical structure of the C60 species of GMM. In A–C, the structure is viewed from a common orientation. E, Mass spectrum of the GMM sample after extraction from N. farcinica. F, Mass spectrum of mycolic acids recovered from CD1b-GMM crystallization drops. The total number of carbons in the meromycolate and α-chains in each species is indicated above each peak. Values on the mass spectra are given in daltons.

FIGURE 1.

Structure of the CD1b-GMM complex and ligand-binding groove of CD1b. A, Overall structure of CD1b-GMM complex. The main chain for the α13 domains of CD1b plus β2m is depicted schematically in ribbon representation (in blue and gray, respectively), and the nonhydrogen atoms of the glycolipid GMM are drawn as van der Waals spheres (carbon in gray; oxygen in red). Residues responsible for flexibility between the α1α2 and α3 domain are colored in gold. B, Electron density for the glycolipid in the CD1b-GMM crystal structure. GMM is drawn in ball-and-stick (carbon in black; oxygen in red). The electron density is from a Fo-Fc omit map calculated after simulated annealing with the glycolipid omitted, contoured at 2.5 ς, and represented as a red mesh. C, The ligand-binding groove of CD1b. The hydrophobic binding surface is drawn as a gray mesh with carbon from the acyl tails occupying the groove drawn as van der Waals spheres. Atoms occupying each channel and the channel label are in matched colors. D, The chemical structure of the C60 species of GMM. In A–C, the structure is viewed from a common orientation. E, Mass spectrum of the GMM sample after extraction from N. farcinica. F, Mass spectrum of mycolic acids recovered from CD1b-GMM crystallization drops. The total number of carbons in the meromycolate and α-chains in each species is indicated above each peak. Values on the mass spectra are given in daltons.

Close modal

To determine any significant differences in conformation or relative domain positioning between different CD1b-ligand structures, the CD1b-GMM structure was superimposed with the CD1b-PI and CD1b-GM2 structures (PDB entries 1GZQ and 1GZP, respectively). The CD1b-GMM crystals are of a different space group to those of CD1b-PI/GM2 (P3121 rather than C2221), and some conformational changes in peripheral regions of the protein are observed due to differences in crystal packing. There are no significant differences in main-chain conformations apart from those regions. The relative position of the α3 domain with respect to the α1α2 domain differs by 5.5° between the CD1b-GMM and CD1b-PI structures. This hinge-like effect results from a conformational flexibility of the linker region (between the α2 and α3 domains) centered on residues 175–184 (Fig. 1 A). A similar point of flexion has been noted between the α1α2 and α3 domains of classical MHC class I molecules (30).

The Ag-binding groove of CD1b is a network of hydrophobic channels in the core of the α1α2 domain (Fig. 1,C). Three channels, denoted as A′, C′, and F′, connect directly to the surface, and a fourth channel, the tunnel T′, runs through the core below the α1 and α2 helices. A′, T′, and F′ are sequentially connected, whereas C′ also connects to T′ and leads from the TCR recognition surface to a portal in the side of the molecule beneath the α2 helix (Fig. 1 A) (6).

Electron density maps calculated using phase information from the final model show clear density for GMM. At a level of 2.5 ς on a simulated annealing Fo-Fc omit map calculated after omitting the GMM, the density is continuous for the sugar head group, α-chain, and meromycolate chain (Fig. 1 B). The meromycolate chain completely occupies the A′, T′, and F′ channels, and protrudes out of the F′ channel into the TCR recognition surface. Additional, weaker electron density is visible extending beyond the TCR binding surface to the same height as the sugar head group of the GMM. This suggests the presence of a longer meromycolate chain bound at a partial occupancy; however, this minor species is not included in the final model (see next section).

The α-chain occupies the C′ channel and stops well short of the portal. There was no electron density visible at any stage of model building beyond what is seen in the final model. The electron density for the protein shows the C′ channel portal in an open state.

GMM purified from the bacterium N. farcinica comprises a range of species with molecular mass spanning from 930.0 to 1040.1 Da (C50-GMM-C58-GMM; Fig. 1,E). Recovery of GMM from CD1b-GMM crystals using organic solvents proved unsuccessful. However, under basic denaturation conditions as revealed by mass spectrometry, four peaks could be accounted for as free mycolic acids liberated from bound GMM species in the CD1b-GMM crystals (Fig. 1 F). These correspond to mycolic acids containing a total of 51, 53, 55, or 57 carbon atoms in the meromycolate chain plus the α-chain (denoted as C54, C56, C58, and C60, respectively) counting the total number of carbons up to the ester group bridging the glucose moiety, i.e., including the α, β, and carboxylate atoms.

The refined structure for CD1b-GMM contains the largest of the GMM species detected, with 8 carbon atoms in the α-chain and 49 carbons atoms in the meromycolate chain (i.e., the C60 species; Fig. 1,D). As described above, weaker electron density seen protruding further out of the F′ channel could possibly be due to an even larger species of GMM bound at low occupancy. Atoms were not modeled into this density because the presence of such species was not confirmed by mass spectrometry and the x-ray data were of limited resolution. The mass spectrum of the lipid sample directly after extraction from N. farcinica shows the predominant species to be a C54-GMM (Fig. 1 E), whereas the C58-GMM is the predominant species in the crystallization drops (F). This would suggest that the process of refolding has enriched the fraction of CD1b-GMM complexes containing larger-species GMM.

The CD1b-GMM structure was compared with the CD1b-PI/GM2 structures to analyze binding groove architectures, glycolipid positions, and protein-glycolipid interactions. In both the CD1b-PI and GM2 structures, the relatively short lipid acyl chains occupy the C′ and A′ channels, whereas the F′ and T′ channels are occupied by detergent molecules (6). An analysis of the CD1b-PI and GM2 structures when combined with the mutagenesis analysis by Niazi et al. (31) on residues affecting glycolipid Ag presentation led to the hypothesis that the meromycolate chain of a GMM ligand would occupy the network of A′, T′, and F′ channels (6). The current structural analysis confirms that the meromycolate chain of GMM is bound in the CD1b groove traversing from the A′ (via T′) to the F′ channel, whereas the α-chain occupies the C′ channel.

Comparison of side-chain conformations between the CD1b-GMM and CD1b-PI structures generally reveals relatively minor conformational changes for residues lining the channels (root-mean-square deviation (RMSD) values of <0.50 Å between equivalent side-chain atoms). However, side-chain conformations for the residues lining the connecting regions between the F′-T′ channels and for residues at the bottom of the A′ channel are significantly different in the two structures (RMSD values of 0.50–1.50 Å).

The volume in the channel around each acyl carbon atom was calculated for the CD1b-glycolipid structures (effectively sectioning the channel using atom coordinates as grid points). In terms of volume (Fig. 2), the most significant differences between CD1b-GMM and CD1b-PI/GM2 are in the connecting regions between the A′-T′ and T′-F′ channels (i.e., corresponding to the regions with high RMSD for the comparison of side chains) (Fig. 2, A and B). The connecting regions between these channels are curved or bent, with A′-T′ taking a hairpin shape. The differences in channel volume can be attributed to the fact that, in the CD1b-GMM structure, the acyl chain is continuous and tracks smoothly around the curve, whereas in CD1b-PI/GM2, the acyl chain termini and detergent termini occupy these regions and point into the outer wall of the bend. Residues forming the wall of the channels have altered conformations to accommodate these termini. The key role of the residues at the A′-T′ junction in presenting long acyl chain ligands has been previously highlighted by mutagenesis data (31).

FIGURE 2.

Comparison of the binding grooves of CD1b-GMM and CD1b-PI. The van der Waals surface of the CD1b binding groove is depicted in gray/green and gray/yellow for the CD1b-GMM and CD1b-PI structures, respectively. GMM and PI are drawn in ball-and-stick (carbon in gray; oxygen in red). In A and C, the view is rotated by 90° about the horizontal from that of Fig. 1,A, and the A′-T′ junction is indicated by an arrow. In B and D, the view is as in Fig. 1 A, and the F′-T′ junction is indicated by an arrow. In E and F, superimpositions of the two surfaces are viewed in close-up to highlight differences in the A′-T′ and F′-T′ junctions. The orientation is as in A and B, respectively, and the GMM is drawn in ball-and-stick.

FIGURE 2.

Comparison of the binding grooves of CD1b-GMM and CD1b-PI. The van der Waals surface of the CD1b binding groove is depicted in gray/green and gray/yellow for the CD1b-GMM and CD1b-PI structures, respectively. GMM and PI are drawn in ball-and-stick (carbon in gray; oxygen in red). In A and C, the view is rotated by 90° about the horizontal from that of Fig. 1,A, and the A′-T′ junction is indicated by an arrow. In B and D, the view is as in Fig. 1 A, and the F′-T′ junction is indicated by an arrow. In E and F, superimpositions of the two surfaces are viewed in close-up to highlight differences in the A′-T′ and F′-T′ junctions. The orientation is as in A and B, respectively, and the GMM is drawn in ball-and-stick.

Close modal

Mycobacterial mycolic acids that bear substitutions along the meromycolate chain have been identified (Fig. 3) (32). These substitutions can be cyclopropyl, methyl, methoxy, or keto groups, and occur at variable positions on the chain. To investigate how CD1b could accommodate such structures, the CD1b-GMM crystal structure was used as a template for the modeling of these lipids into the binding groove. Lipids were inserted into the binding groove, and the complexes were energy minimized with the protein backbone harmonically restrained. The substituent groups in the resultant models were positioned at a wide range of points in the A′ channel, A′-T′ junction, or T′ tunnel. Two representative examples are illustrated in Fig. 3. Analysis of the model structures revealed no stereochemical clashes. The CD1b binding groove is capable of accommodating these groups with only minor changes to side-chain conformations (for example, Fig. 3). Volume calculations on the channels reveal that there is no significant change in the total volume of the binding groove; however, volume calculations along sections of the binding groove show that there are local changes in volume. The substituent groups are accommodated by slight increases (typically 5–10%) in the channel volume at these points. Because these increases are achieved by changes in side-chain conformations, they tend to be counterbalanced by subtle decreases in volume extending through neighboring regions. These model-based results imply a degree of flexibility in the CD1b binding groove, consistent with this molecule’s ability to bind lipids containing substituent groups at a wide range of positions. The current structural data provide no evidence of any mechanism whereby differences in such substituent groups could influence TCR recognition as reported by Grant et al. (33).

FIGURE 3.

Conformational changes of residue side chains to accommodate additional functional groups present on modeled lipids. Modeling into the CD1b binding groove of a type 3 α-mycolic acid (A) and a type 2 methoxymycolic acid (B). i, Chemical structures of modeled lipids. ii, Close-up views of the modeled CD1b-glycolipid structures centered on the methoxy group of the methoxymycolic acid and a cyclopropyl and alkene group of the α-mycolic acid. The positions of the substituent groups depicted in the close-up view are circled in yellow on the chemical structures. For comparison, coordinates of the CD1b-GMM crystal structure are shown (dark-gray ball-and-stick). Modeled coordinates are depicted as ball-and-stick (carbon in black on protein side chains and in gray on lipid; oxygen in red; nitrogen in blue). Bonds between substituent group atoms in the modeled lipids are colored yellow. iii, Global view of the positions of the substituent groups in the CD1b binding groove. The panels show the surface of the CD1b binding groove drawn as a gray mesh with the substituent group atoms drawn as yellow van der Waals spheres. The view in A is from beneath the binding groove (rather than from above as in Fig. 2,A). The view in B is into the side of the binding groove through the α1 helix (rather than through the α2 helix as in Fig. 2 B).

FIGURE 3.

Conformational changes of residue side chains to accommodate additional functional groups present on modeled lipids. Modeling into the CD1b binding groove of a type 3 α-mycolic acid (A) and a type 2 methoxymycolic acid (B). i, Chemical structures of modeled lipids. ii, Close-up views of the modeled CD1b-glycolipid structures centered on the methoxy group of the methoxymycolic acid and a cyclopropyl and alkene group of the α-mycolic acid. The positions of the substituent groups depicted in the close-up view are circled in yellow on the chemical structures. For comparison, coordinates of the CD1b-GMM crystal structure are shown (dark-gray ball-and-stick). Modeled coordinates are depicted as ball-and-stick (carbon in black on protein side chains and in gray on lipid; oxygen in red; nitrogen in blue). Bonds between substituent group atoms in the modeled lipids are colored yellow. iii, Global view of the positions of the substituent groups in the CD1b binding groove. The panels show the surface of the CD1b binding groove drawn as a gray mesh with the substituent group atoms drawn as yellow van der Waals spheres. The view in A is from beneath the binding groove (rather than from above as in Fig. 2,A). The view in B is into the side of the binding groove through the α1 helix (rather than through the α2 helix as in Fig. 2 B).

Close modal

In an attempt to generate additional insights into the similarities and differences in the binding groove architectures of the CD1 molecules, the sequences of CD1a, CD1b, CD1c, CD1d, and mCD1d1 were analyzed in terms of hydrophobicity (Fig. 4,A). The analysis reveals a particularly complex pattern of hydrophobicity in CD1b, and in Fig. 4,B, specific regions/patches of hydrophobicity are related to distinct structural features in the binding groove. Of the other family members, CD1c shows the most similarity to CD1b in the hydrophobicity analysis. However, certain features (e.g., patch 7; Fig. 4) that map to the T′ channel are unique to CD1b. It has previously been predicted that the substitution of G98 by a valine in CD1a and CD1c would abolish the continuous T′ tunnel (Fig. 4,E) (6). This has now been confirmed for CD1a by the determination of its crystal structure in complex with sulfatide (7). The CD1a structure also reinforces the conclusion that the A′ region is the section of the binding groove that shows most similarity between group 1 family members. The presence of CD1b-like A′ channel and A′-T′ junction in the CD1a binding groove can be related to the conservation of hydrophobic patches 1, 4, 13, and 14 (Fig. 4). Because CD1c exhibits a similar pattern in its primary structure, it is likely that it too has a CD1b-like A′ channel. This conservation of A′ structure is consistent with the shared ability of CD1a, CD1b, and CD1c to bind sulfatide (34); however, Zajonc et al. (7) note that variations in the detailed architecture in this region are important in modulating the ligand-binding characteristics of CD1a vs the other CD1s.

FIGURE 4.

Comparison of CD1 molecules. A, Comparison of hydrophobic patches between the different CD1 sequences. Hydrophobicity is colored as depicted in the scale bar. B—D, Three views of the CD1b binding groove with key residues drawn in ball-and-stick; the surface of the binding groove is drawn as a solid pale-blue surface, and GMM is drawn in gray van der Waals spheres. The colors of the residues in B–D correspond to the colored numbers of the hydrophobic patches in A. In B, the view is at ∼45° to the upper surface of the binding groove. The view in C is that of Figs. 1,A and 2,B, whereas in D, it is rotated by 180° about the vertical (as in Fig. 3 B). E, Sequence alignment between CD1b, CD1a, and CD1c. F, Sequence alignment between CD1d and mCD1d1.

FIGURE 4.

Comparison of CD1 molecules. A, Comparison of hydrophobic patches between the different CD1 sequences. Hydrophobicity is colored as depicted in the scale bar. B—D, Three views of the CD1b binding groove with key residues drawn in ball-and-stick; the surface of the binding groove is drawn as a solid pale-blue surface, and GMM is drawn in gray van der Waals spheres. The colors of the residues in B–D correspond to the colored numbers of the hydrophobic patches in A. In B, the view is at ∼45° to the upper surface of the binding groove. The view in C is that of Figs. 1,A and 2,B, whereas in D, it is rotated by 180° about the vertical (as in Fig. 3 B). E, Sequence alignment between CD1b, CD1a, and CD1c. F, Sequence alignment between CD1d and mCD1d1.

Close modal

The remainder of the binding groove of CD1b is a distinctive network of channels rather than the single F′ pocket reported for CD1a (7), because it is partitioned by bulky side chains pointing into the center of internal cavities (e.g., F77 and F114). These features map to hydrophobic patches 5 and 9 (Fig. 4). Their effect is an increase in the area of the hydrophobic surface and the formation of a tubular channel (Fig. 4 B). The hydrophobicity analysis indicates that CD1c has some similarity to CD1b in these regions and thus may also differ from the single F′ cavity or pocket-type structure seen in mCD1d1 (2) and CD1a (7). In CD1b, the combination of an interconnecting T′ tunnel with a partitioned binding groove confer the unique features that allow it to bind lipids with longer acyl chains.

To gain insight into how TCRs bind to the CD1b-GMM complex, CD1b-GMM and a model of a cognate TCR, LDN5 (Ref. 3 and Materials and Methods), were superimposed onto all available crystal structures of MHC-TCR complexes. Due to the elevated position of the glycolipid head group of GMM and the structure of the α2 helix (which at its midpoint forms a more prominent apex, unlike that of the α2 helix in MHC class I or II molecules; Fig. 1 A), there are steric clashes with the backbone of the TCR in all resultant model complexes. Detailed conclusions must therefore await the structure determination of a CD1-TCR complex. However, these docking experiments imply that the complementarity-determining region 3 loops of a TCR would sample parts of the lipid.

In addition to the structure of the glycolipid head group, its position is likely to be an important determinant for recognition by TCR. Site-directed mutagenesis experiments have identified several residues in CD1b that appear to be required for TCR recognition (35). These residues fall into two groups in the context of the structure, the first group being R79, E80, and D83 on the α1 helix, and the second group being T165 and T157 on the α2 helix (35) (Fig. 5,A). Both groups of residues may contribute to TCR recognition directly by making contacts with the complementarity-determining region loops of the TCR; however, they may also contribute by stabilizing a specific orientation of the glycolipid head group by hydrogen bonding. Indeed the CD1b-GM2, PI, and GMM crystal structures show that residues from both clusters (specifically R79 and T157) can make hydrogen bonds to the glycolipid headgroups (6) (Fig. 5, B–D). Mutagenesis of other CD1b residues affects Ag-presenting function to a lesser degree (35), and in the crystal structures, these do not make any direct interactions with the glycolipid (Fig. 5 A). In common with the CD1a structure (7), the CD1b-GMM structure reveals a portion of acyl tail protruding from the F′ pocket. This is an additional component of the glycolipid that may potentially contribute to TCR recognition.

FIGURE 5.

Head group orientation of different lipids from crystal structures and positions of residues implicated in TCR recognition. A, View of the TCR binding surface of the CD1b-GMM complex with residues implicated in functional studies as critical to TCR recognition indicated as spheres at their Cα positions. Those drawn in red mark residues that, when mutated, cause a >1-log reduction in CD1b Ag presentation, whereas those in purple show a >0.5-log reduction (mutagenesis data from Ref. 35 ). Comparison of GMM (B), GM2 (C), and PI (D) structures in complex with CD1b. The main chain for the α1 and α2 helices of CD1b is represented as coil (yellow). The glycolipid and key side chains from CD1b are shown in ball-and-stick (lipid carbon atoms in gray; protein carbon atoms in black; oxygen in red; nitrogen in blue). Hydrogen bonds are depicted as dashed lines. The view in B–D is somewhat tipped about the horizontal from that of Fig. 1 A.

FIGURE 5.

Head group orientation of different lipids from crystal structures and positions of residues implicated in TCR recognition. A, View of the TCR binding surface of the CD1b-GMM complex with residues implicated in functional studies as critical to TCR recognition indicated as spheres at their Cα positions. Those drawn in red mark residues that, when mutated, cause a >1-log reduction in CD1b Ag presentation, whereas those in purple show a >0.5-log reduction (mutagenesis data from Ref. 35 ). Comparison of GMM (B), GM2 (C), and PI (D) structures in complex with CD1b. The main chain for the α1 and α2 helices of CD1b is represented as coil (yellow). The glycolipid and key side chains from CD1b are shown in ball-and-stick (lipid carbon atoms in gray; protein carbon atoms in black; oxygen in red; nitrogen in blue). Hydrogen bonds are depicted as dashed lines. The view in B–D is somewhat tipped about the horizontal from that of Fig. 1 A.

Close modal

Refolding of denatured CD1b molecules with different ligands, either in the presence of single acyl chain detergents (6) or using a protocol based on oxidative refolding chromatography (10) results in an identical three-dimensional structure of CD1b. The architecture of the CD1b binding groove is therefore independent of refolding pathway or ligand. This structural stability is particularly important for the generation of CD1b tetramers for the monitoring of glycolipid-specific T cell responses. Structural and mass-spectrometric analysis reveals that longer GMM species are preferentially selected by CD1b. Given the potential contribution to TCR recognition by acyl chains protruding out of the F′ pocket, the preferential loading of longer GMMs may narrow the population of T cells stained by CD1b-GMM tetramers. Such problems can be circumvented in tetramer production by the use of purified samples of shorter chain GMMs that are known to bind to CD1b and be recognized by CD1b-specific T cell clones (29). The ability to refold incorporating long glycolipid species provides an invaluable tool for understanding the T cell response to bacterial infection. The complexity of the CD1b binding groove architecture implies that in vivo loading of such long-chain meromycolates will require conformation changes in CD1b. The mechanisms controlling this loading process require investigation. The current CD1b-GMM structure demonstrates how a meromycolate chain of 49 carbon atoms saturates the capacity of the A′-T′-F′ superchannel. In contrast, the α-chain of 8 carbons only partially occupies the C′ channel in CD1b-GMM. The previously reported structure for CD1b-PI indicates that a GMM with a longer α-chain will extend further into this channel (6). The full capacity of this channel is 16 carbon atoms; any additional α-chain carbon atoms may exit the binding groove via the portal under the α2 helix.

The crystal structure of CD1b-GMM provides the first detailed information on the presentation of a natural bacterial glycolipid Ag. Recent functional studies have confirmed the ability of CD1b molecules to present full-length GMM to T cells (T.-Y. Cheng and B. Moody, unpublished data). Further insights into T cell recognition of CD1b-glycolipid complexes await the determination of CD1b-glycolipid-TCR complex structures.

We are grateful to the staff of the SRS (Daresbury, U.K.) and the ESRF and European Molecular Biology Laboratory outstation (Grenoble, France) for the assistance with x-ray data collection. We thank Jonathan Grimes, Erika Mancini, and Geoff Sutton for help and advice during data collection; Robert Esnouf for providing advice and scripts for running VOLUMES; Jonathan Diprose for writing the script for generating Fig. 4 A; and Peter Ashton for carrying out the mass spectrometry measurements.

1

This work was funded by United Kingdom Medical Research Council (Grants G9900061 (to E.Y.J.) and G0000895 (to G.S.B.)), the European Commission Integrated Programme (Structural Proteomics in Europe, Grant QLRT-2001-00988), Cancer Research UK (Grants C375/A2321 (to E.Y.J.) and C399/A2291 (to V.C.)), and the Cancer Research Institute. E.Y.J. is a Cancer Research UK Principal Research Fellow. G.S.B. is a Lister Jenner Research Fellow.

2

The atomic coordinates of the CD1b-GMM structure have been deposited in the Protein Data Bank (http://www.rcsb.org; PDB ID code 1UQS).

5

Abbreviations used in this paper: GMM, glucose monomycolate; PI, phosphatidylinositol; β2m, β2-microglobulin; RMSD, root-mean-square deviation; m, mouse.

1
Porcelli, S., C. T. Morita, M. B. Brenner.
1992
. CD1b restricts the response of human CD48 T lymphocytes to a microbial antigen.
Nature
360
:
593
.
2
Zeng, Z., A. R. Castano, B. W. Segelke, E. A. Stura, P. A. Peterson, I. A. Wilson.
1997
. Crystal structure of mouse CD1: an MHC-like fold with a large hydrophobic binding groove.
Science
277
:
339
.
3
Moody, D. B., B. B. Reinhold, M. R. Guy, E. M. Beckman, D. E. Frederique, S. T. Furlong, S. Ye, V. N. Reinhold, P. A. Sieling, R. L. Modlin, et al
1997
. Structural requirements for glycolipid antigen recognition by CD1b-restricted T cells.
Science
278
:
283
.
4
Ernst, W. A., J. Maher, S. Cho, K. R. Niazi, D. Chatterjee, D. B. Moody, G. S. Besra, Y. Watanabe, P. E. Jensen, S. A. Porcelli, et al
1998
. Molecular interaction of CD1b with lipoglycan antigens.
Immunity
8
:
331
.
5
Porcelli, S. A., B. W. Segelke, M. Sugita, I. A. Wilson, M. B. Brenner.
1998
. The CD1 family of lipid antigen-presenting molecules.
Immunol. Today
19
:
362
.
6
Gadola, S. D., N. R. Zaccai, K. Harlos, D. Shepherd, J. C. Castro-Palomino, G. Ritter, R. R. Schmidt, E. Y. Jones, V. Cerundolo.
2002
. Structure of human CD1b with bound ligands at 2.3 Å, a maze for alkyl chains.
Nat. Immunol.
3
:
721
.
7
Zajonc, D. M., M. A. Elsliger, L. Teyton, I. A. Wilson.
2003
. Crystal structure of CD1a in complex with a sulfatide self antigen at a resolution of 2.15 Å.
Nat Immunol.
:
8
Moody, D. B., M. R. Guy, E. Grant, T. Y. Cheng, M. B. Brenner, G. S. Besra, S. A. Porcelli.
2000
. CD1b-mediated T cell recognition of a glycolipid antigen generated from mycobacterial lipid and host carbohydrate during infection.
J. Exp. Med.
192
:
965
.
9
Altamirano, M. M., R. Golbik, R. Zahn, A. M. Buckle, A. R. Fersht.
1997
. Refolding chromatography with immobilized mini-chaperones.
Proc. Natl. Acad. Sci. USA
94
:
3576
.
10
Karadimitris, A., S. Gadola, M. Altamirano, D. Brown, A. Woolfson, P. Klenerman, J. L. Chen, Y. Koezuka, I. A. Roberts, D. A. Price, et al
2001
. Human CD1d-glycolipid tetramers generated by in vitro oxidative refolding chromatography.
Proc. Natl. Acad. Sci. USA
98
:
3294
.
11
Brown, J., T. S. Walter, L. Carter, N. G. A. Abrescia, A. R. Aricescu, T. D. Batuwangala, L. E. Bird, N. Brown, P. P. Chamberlain, S. J. Davis, et al
2003
. A procedure for setting up high-throughput nanolitre crystallization experiments. II. Crystallization results.
J. Appl. Crystallogr.
36
:
315
.
12
Walter, T. S., J. Diprose, J. Brown, M. Pickford, R. J. Owens, D. I. Stuart, K. Harlos.
2003
. A procedure for setting up high-throughput nanolitre crystallization experiments. I. Protocol design and validation.
J. Appl. Crystallogr.
36
:
308
.
13
Otwinowski, Z., W. Minor.
1997
. Processing of x-ray diffraction data collected in oscillation mode. C. W. Carter, Jr, and R. M. Sweet, Jr, eds. In
Methods in Enzymology: Macromolecular Crystallography
Vol. 276, Pt. A.
:
307
. Academic, New York.
14
Kissinger, C. R., D. K. Gehlhaar, D. B. Fogel.
1999
. Rapid automated molecular replacement by evolutionary search.
Acta Crystallogr. D
55
:
484
.
15
Brunger, A. T., P. D. Adams, G. M. Clore, W. L. DeLano, P. Gros, R. W. Grosse-Kunstleve, J. S. Jiang, J. Kuszewski, M. Nilges, N. S. Pannu, et al
1998
. Crystallography & NMR system: a new software suite for macromolecular structure determination.
Acta Crystallogr. D
54
:
905
.
16
Jones, T. A., J. Y. Zou, S. W. Cowan, M. Kjeldgaard.
1991
. Improved methods for building protein models in electron-density maps and the location of errors in these models.
Acta Crystallogr. A
47
:
110
.
17
van Aalten, D. M., R. Bywater, J. B. Findlay, M. Hendlich, R. W. Hooft, G. Vriend.
1996
. PRODRG, a program for generating molecular topologies and unique molecular descriptors from coordinates of small molecules.
J. Comput. Aided Mol. Des.
10
:
255
.
18
Laskowski, R. A., D. S. Moss, J. M. Thornton.
1993
. Main-chain bond lengths and bond angles in protein structures.
J. Mol. Biol.
231
:
1049
.
19
Wallace, A. C., R. A. Laskowski, J. M. Thornton.
1995
. LIGPLOT: a program to generate schematic diagrams of protein-ligand interactions.
Protein Eng.
8
:
127
.
20
Lefranc, M. P..
2003
. IMGT databases, web resources and tools for immunoglobulin and T cell receptor sequence analysis, http://imgt.cines.fr.
Leukemia
17
:
260
.
21
Guex, N., M. C. Peitsch.
1997
. SWISS-MODEL and the Swiss-PdbViewer: an environment for comparative protein modeling.
Electrophoresis
18
:
2714
.
22
Stuart, D. I., M. Levine, H. Muirhead, D. K. Stammers.
1979
. Crystal structure of cat muscle pyruvate kinase at a resolution of 2.6 Å.
J. Mol. Biol.
134
:
109
.
23
Corpet, F..
1988
. Multiple sequence alignment with hierarchical clustering.
Nucleic Acids Res.
16
:
10881
.
24
Gouet, P., E. Courcelle, D. I. Stuart, F. Metoz.
1999
. ESPript: analysis of multiple sequence alignments in PostScript.
Bioinformatics
15
:
305
.
25
Appel, R. D., A. Bairoch, D. F. Hochstrasser.
1994
. A new generation of information retrieval tools for biologists: the example of the ExPASy WWW server.
Trends Biochem. Sci.
19
:
258
.
26
Eisenberg, D., E. Schwarz, M. Komaromy, R. Wall.
1984
. Analysis of membrane and surface protein sequences with the hydrophobic moment plot.
J. Mol. Biol.
179
:
125
.
27
Esnouf, R. M..
1999
. Further additions to MolScript version 1.4, including reading and contouring of electron-density maps.
Acta Crystallogr. D
55
:
938
.
28
Merritt, E. A., D. J. Bacon.
1997
. Raster3D: photorealistic molecular graphics. In
Macromolecular Crystallography, Part B
Vol. 277
:
505
. Academic, San Diego.
29
Moody, D. B., V. Briken, T. Y. Cheng, C. Roura-Mir, M. R. Guy, D. H. Geho, M. L. Tykocinski, G. S. Besra, S. A. Porcelli.
2002
. Lipid length controls antigen entry into endosomal and nonendosomal pathways for CD1b presentation.
Nat. Immunol.
3
:
435
.
30
Madden, D. R..
1995
. The three-dimensional structure of peptide-MHC complexes.
Annu. Rev. Immunol.
13
:
587
.
31
Niazi, K., M. Chiu, R. Mendoza, M. Degano, S. Khurana, D. Moody, A. Melian, I. Wilson, M. Kronenberg, S. Porcelli, R. Modlin.
2001
. The A′ and F′ pockets of human CD1b are both required for optimal presentation of lipid antigens to T cells.
J. Immunol.
166
:
2562
.
32
Watanabe, M., Y. Aoyagi, H. Mitome, T. Fujita, H. Naoki, M. Ridell, D. E. Minnikin.
2002
. Location of functional groups in mycobacterial meromycolate chains; the recognition of new structural principles in mycolic acids.
Microbiology
148
:
1881
.
33
Grant, E. P., E. M. Beckman, S. M. Behar, M. Degano, D. Frederique, G. S. Besra, I. A. Wilson, S. A. Porcelli, S. T. Furlong, M. B. Brenner.
2002
. Fine specificity of TCR complementarity-determining region residues and lipid antigen hydrophilic moieties in the recognition of a CD1-lipid complex.
J. Immunol.
168
:
3933
.
34
Shamshiev, A., H. J. Gober, A. Donda, Z. Mazorra, L. Mori, G. De Libero.
2002
. Presentation of the same glycolipid by different CD1 molecules.
J. Exp. Med.
195
:
1013
.
35
Melian, A., G. F. Watts, A. Shamshiev, G. De Libero, A. Clatworthy, M. Vincent, M. B. Brenner, S. Behar, K. Niazi, R. L. Modlin, et al
2000
. Molecular recognition of human CD1b antigen complexes: evidence for a common pattern of interaction with αβ TCRs.
J. Immunol.
165
:
4494
.