|Home | About | Journals | Submit | Contact Us | Français|
CD1 molecules function to present lipid-based antigens to T cells. Here we present the crystal structure of CD1c at 2.5 Å resolution, in complex with the pathogenic Mycobacterium tuberculosis antigen mannosyl-β1-phosphomycoketide (MPM). CD1c accommodated MPM’s methylated alkyl chain exclusively in the A′ pocket, aided by a unique exit portal underneath the α1 helix. Most striking was an open F′ pocket architecture lacking the closed cavity structure of other CD1 molecules, reminiscent of peptide binding grooves of classical Major Histocompatibility Complex molecules. This feature, combined with tryptophan-fluorescence quenching during loading of a dodecameric lipopeptide antigen, provides a compelling model by which both the lipid and peptide moieties of the lipopeptide are involved in CD1c presentation of lipopeptides.
CD1 molecules are Major Histocompatibility Complex (MHC)-like proteins that are broadly expressed on antigen-presenting cells (APCs) and serve to present lipid-based antigens to T cells in the vertebrate immune response. Unlike the polymorphic, peptide-presenting, classical class I and II MHC molecules, CD1 proteins exhibit essentially no allelic variation, and in humans are located outside of the Chromosome 6 MHC region on Chromosome 1 at 1q22–23 (Albertson et al., 1988). There exist five human isoforms of CD1 that are grouped according to their amino acid homology, genomic organization and immunological function; CD1a, CD1b, and CD1c form Group 1 (Calabi et al., 1989), CD1d forms Group 2 and CD1e has been proposed to form a third group due to its intracellular chaperone function (Angenieux et al., 2000; de la Salle et al., 2005). Tyrosine-containing motifs present in the cytoplasmic tail of CD1b, CD1c and CD1d molecules determine where they are trafficked within the cell, and thus to what repertoire of lipid-based antigens they may be exposed. The trafficking route as well as structural features unique to each CD1 molecule endow them with the ability to bind a structurally diverse array of lipid antigens from bacterial (Willcox et al., 2007) and endogenous sources (De Libero and Mori, 2007) and present these to T cells at the cell surface (Sullivan and Kronenberg, 2007).
CD1 molecules have evolved a deep and narrow binding cavity that is well suited to anchoring the hydrophobic alkyl chains of lipid molecules. Based on the structures of CD1a (Zajonc et al., 2005b; Zajonc et al., 2003), CD1b (Batuwangala et al., 2004; Gadola et al., 2002) and mouse (Wu et al., 2006; Zajonc et al., 2005a; Zajonc et al., 2005c; Zajonc et al., 2008; Zeng et al., 1997) and human CD1d (Koch et al., 2005), the binding cavity contains two major pockets, A′ and F′. In each of these structures the A′ pocket extends deep into the CD1 molecule and is enclosed by the A′ roof. Variation in the A′ pocket size, the shape of the F′ pocket along with the presence of an exit portal (C′) and T′ tunnel unique to CD1b explain each isoform’s ability to bind different classes of lipids, varying in the length and saturation of their hydrocarbon tails and the chemical nature of their head group.
Of the CD1 isoforms, CD1c is unique in its ability to present mycobacterial phosphoketides (de Jong et al., 2007; Matsunaga et al., 2004) and polyisoprenoids (Moody et al., 2000) containing branched alkyl chains and has also been shown to present lipopeptides with a long peptide moiety similar in length to those presented by MHC class II molecules (Moody et al., 2004; Van Rhijn et al., 2009). CD1c is expressed at high amounts on B cells, myeloid dendritic cells (DCs) and thymocytes (Dougan et al., 2007) and intracellularly it broadly surveys the endocytic system (Sugita et al., 2000), particularly early and late endosomes (Briken et al., 2002; Briken et al., 2000) similar to the path of MHC class II proteins. CD1c is the only CD1 isoform that has been shown to interact with both αβ (Beckman et al., 1996; Porcelli et al., 1989; Van Rhijn et al., 2009) and γδ (Spada et al., 2000) T cells to date.
Functionally, CD1c has a clear role in mediating T cell responses to infectious pathogens, in particular Mycobacterium tuberculosis, the causative agent of tuberculosis (Beckman et al., 1996; Moody et al., 2000; Ulrichs et al., 2003). This was further elucidated through elegant studies of CD1c presentation of synthetic forms of mannosyl phosphodolichols (MPDs) and mannosyl-β1-phosphomycoketides (MPMs) identical to those present in minute quantities in the cell walls of M. tuberculosis and M. avium (de Jong et al., 2007; Moody et al., 2000). T cell-specific, CD1c-presented MPMs contain structurally distinct single alkyl chains ranging from C30–34 that have stereo-specific (S) methyl branches starting at C4, repeated on every fourth carbon thereafter (de Jong et al., 2007). Made by polyketide synthase 12 (pks12) (Matsunaga et al., 2004), an enzyme that appears to be unique to mycobacteria, the methyl branches of MPM represent an unusual pathogen-associated molecular pattern that, when presented at even minute amounts by CD1c, can potently activate a T cell response.
The molecular features of CD1c that enable it to bind and present the structurally divergent MPM and 12-mer lipopeptide antigens have remained unclear. Computational modeling and molecular dynamics simulations have provided hints of divergent features of CD1c (Garzon et al., 2009); however an experimentally derived three-dimensional structure has evaded determination. Our 2.5 Å crystal structure of CD1c in complex with a synthetic, all-S, C32 mannosyl-β1-phosphomycoketide (MPM) demonstrates how CD1c accommodates the unique methyl-branched structure of mycoketides, and reveals an open F′ groove feature that may explain CD1c’s ability to bind and present dodecameric lipopeptides (lipo-12).
Recombinant expression of CD1c was performed using the baculovirus insect expression system. An initial, expression-optimized construct was designed by fusing the β2microglobulin (β2m) subunit covalently to the N-terminus of CD1c via a 15 amino acid glycine-serine linker. Repeated attempts to crystallize this construct failed, so a crystal contact-optimized construct was further generated by swapping the CD1c α3 domain for that of the α3 domain of CD1b. Additional mutations were made to stabilize the interface between the platform and α3 domain, reduce the heterogeneity of glycosylation (four of the five putative N-linked glycosylation sites were eliminated), and to facilitate crystal packing (see Supplementary Methods). Several methods were used to establish that the CD1c α1 and α2 platform domain in this optimized mutant was functionally and structurally conserved in relation to wild-type CD1c. CD spectra of wild-type and optimized CD1c were super-imposable (Figure 1A) demonstrating overall secondary structure similarity between the two proteins. Furthermore, size-exclusion chromatography and native polyacrylamide gel shift demonstrated similar binding of the CD1c-specific F10/21A3 and L161 antibodies to both wild-type and optimized CD1c (Figures 1B and 1C), confirming that the three-dimensional epitopes of these antibodies on these two CD1c constructs was conserved. To assess the functional conservation of our optimized CD1c construct, we performed lipid loading assays based on isoelectric focusing (IEF) gel-shift similar to those described by Zhou et al. (Zhou et al., 2004) and confirmed that lipid binding was not compromised by the modifications made to CD1c to facilitate crystallization (Figure 1D). Furthermore, optimized CD1c loaded with our synthetic MPM was able to stimulate the CD8-1 T cell line, which is specific for CD1c presenting MPM, at equivalent amounts to that of native CD1c presenting MPM (Figure 1E), confirming the functional conservation of these two constructs.
Chemically synthesized mannosyl-β1-phosphomycoketide (MPM) was passively loaded into recombinant, optimized, CD1c and the resulting complex was purified by gel filtration chromatography. The complex of CD1c and MPM was crystallized and the structure was determined using molecular replacement with CD1b as the initial model. There was one complex of CD1c and MPM in the asymmetric unit. The structure was refined to 2.5 Å with a final Rcryst and Rfree of 24.0% and 27.3%, respectively, and 93.3% of the residues in the most favored region of the Ramachandran plot. (Figure 2 and Table 1). A loop connecting the α1 helix to the β-sheet was poorly resolved in the electron density and residues Q87–Y89 were excluded from the final model as a result. Furthermore, only the first four amino acids of the glycine and serine linker engineered to fuse β2m and the CD1c heavy chain were resolved in the electron density. Amino acid side chains with weak or absent electron density had their occupancies set to zero.
Examination of crystallographic symmetry mates revealed that the K108G mutation engineered into the CD1copt construct was critical to the establishment of a key crystal contact with a loop (residues 261 to 276) in the α3 domain of a symmetry mate. The presence of a large charged side chain at position 108 would create a steric clash and disrupt a crystal contact at this location.
The CD1c heavy chain is composed of α1, α2 and α3 domains that assemble non-covalently with the β2m subunit, similar to other CD1 family members. The α1 and α2 domains constituted the ligand-binding platform domain and have the canonical MHC structure of two α-helices supported by a six-stranded β-sheet (Figure 2A, 2B). This architecture supported the formation of two distinct ligand-binding cavities, lined with hydrophobic residues that are compatible with the binding of hydrophobic alkyl chains characteristic of lipid molecules. These two cavities have been named A′ and F′ based on homology with the peptide-binding pockets of MHC class I molecules and consistent with the pocket nomenclature for other CD1 family members (Zeng et al., 1997). The platform domain (α1 and α2 domains) of CD1c was most structurally related to that of CD1b based on a backbone RMSD of 1.16 Å (in comparison to CD1a: 1.50 A and CD1d: 1.60 A) consistent with CD1c’s close (59%) amino acid identity to CD1b (Figure S1).
CD1c’s A′ pocket had the shape of a wide circular tunnel with one area of constriction, located slightly underneath the A′ portal, where the A′ pocket opened to solvent (Figure 3, top panel). This constriction selected for a clockwise directionality for lipid binding in the A′ pocket, specifically for those lipids with methylated branches, as seen in MPM. CD1c’s circular A′ pocket was centered around the A′ pole formed by the van der Waals contact of V12 and F70. The A′ pole as well as the A′ roof, composed of residues L66, L162 and T166, made up the conserved features of the A′ pocket found in other CD1 isoforms (Figure 3). The A′ portal is rimmed with the hydrophobic residues L69, F72, Y73 and V155, and the polar T158; it is here where the phosphomannose head group of MPM was situated, in a conformation consistent with presentation to a T cell receptor (TCR) (Figure 2). The A′ portal was continuous with the F′ pocket, lacking a dividing wall between these two cavities (Figure 2A, lower panel). In comparison to the other human CD1 isoforms, CD1c had the second largest cavity volume, 1780 Å3, compared with CD1b (2200 Å3), CD1d (1650 Å3) and CD1a (1280 Å3) as calculated by the program Pocket-Finder (Hendlich et al., 1997). Therefore the overall structure of CD1c in complex with MPM was reminiscent of other CD1 isoforms in the location and chemical nature of their A′ and F′ pockets.
Unique to CD1c was the presence of a portal to the exterior located underneath the α1 helix, called the D′ portal due to its similarity in location to the D pocket in MHC class I molecules. Two amino acid substitutions only found in CD1c (G26 and L74) are involved in the architecture of this portal, with contributions from G28 (found in other CD1 isoforms) (Figure 4A, upper panel, Figure S1). In CD1b, most closely related to CD1c based on structural and amino acid homology, T26 and I74 closed off this portal to the exterior, despite sharing the glycine at position 28 (Figure 4A, lower panel, Figure S1). An additional opening was noted in the CD1c structure, called the E′ portal, which connects the F′ pocket to the exterior through a narrow tunnel under the α1 helix. This opening was formed by several amino acids that are either unique to CD1c (F16, L77 and V96) or are shared with other isoforms (F18, T78, I81) (Figure 4A, upper panel, Figure S1). Key to the formation of both portals was the increased distance between the α1 helix and the S2 β-strand of the α1 domain β-sheet, due in part to the presence of the large aromatic side chain at position 16 (F16) (Figure 4B, Figure S1). This caused a ~2.5 Å increase in main chain-main chain distance between the α1 helix and β-sheet at the location of these portals (Figure 4B).
The most striking aspect of CD1c was the open nature of the F′ pocket, here termed the F′ groove, a feature not found in other CD1 molecules (Figure 5). CD1c’s F′ groove is open to solvent, contrasting with the closed nature of the F′ pockets of other CD1 isoforms. This open groove was created by both smaller side chains in CD1c at positions 84, 140 and 144 (H, L and V in contrast to F, R and F in CD1b) which failed to reach across the α1 and α2 helical divide, and the extended conformation of the connecting loop between the α1 helix and the S1 strand of the α2 domain β-sheet (Figure 5, Figure S1). In all other CD1 molecules this loop is tucked back into the architecture of the F′ pocket, closing off this region to solvent (Figure 5, Figure S1). The residues lining the F′ groove in CD1c were almost entirely hydrophobic in nature; of the 21 side chain residues contributing to this surface only H84 is potentially charged (this side chain was also disordered in the electron density) whereas the remainder were composed of eight aromatic and twelve nonpolar residues (V, L or I). Therefore, ligands bound in this groove would preferentially be hydrophobic and/or aromatic in nature. It was also notable that the atomic displacement factors (or B-factors) were highest in the α-helical regions of the F′ groove (Figure S2), suggestive of a higher intrinsic disorder or flexibility to this region. This putative flexibility may be a contributing factor to CD1c’s ability to accommodate alternative ligands in this groove, such as the lipopeptide ligands containing larger peptide fragments.
Liquid chromatography-mass spectrometry (LC-MS) of our CD1c loaded with MPM showed a significant enrichment of MPM, confirming that MPM was bound in our purified complex (Figure S3). The LC-MS analysis also revealed bound hydrocarbon chains ranging in length from C8 to C18 (Figure S3). Omit map electron density showed a clear clockwise traverse of the methylated hydrocarbon chain of MPM through the A′pocket, exiting through the D′ portal (Figure 2 and Figure 4A). The C32 MPM (equivalent to the natural M. tuberculosis MPM) lipid tail was accommodated perfectly in the A′ pocket, demonstrating the optimal length of hydrocarbon chains for CD1c is approximately 27 carbons (the remaining five are methyl groups). The exact placement of the phosphomannose head group was unclear in initial rounds of refinement, so we used molecular dynamics simulations and energy minimization based on omit map density to estimate the optimal positioning of the MPM lipid ligand within the A′ pocket. Two positions were determined to be most likely based on these simulations, one designated A1 features optimal hydration and placement of the negatively charged phosphate group; the second, designated A2, was positioned according to electron density that extends out of the D′ portal, positioning the head group lower in the A′ portal in a less optimal hydrophobic environment (Figure S4). Further rounds of refinement with the lipid in the A1 position revealed omit map density for two water molecules consistent with a water-mediated stabilization of the head group (Figure 6). Five highly probable hydrogen bonds (distance < 3.3 Å, Table S1) coordinating the phosphomannose head group were evident and shown in Figure 6 as large dashes, four of which are water mediated. The fifth hydrogen bond was between the main chain carbonyl of F72 and the ring oxygen of the mannose. Additional potential hydrogen bonds were present (Figure 6 and Table S1, some, but not all, of these weak hydrogen bonds are indicated as small dashes) indicating that positioning of the phosphomannose head group in A1 was optimal both in regards to our electron density data and molecular dynamics simulations.
Density in the F′ groove of our CD1c structure was consistent with a C12 hydrocarbon chain, supported by our LC-MS analysis of lipids bound to CD1c. This LC-MS data showed that CD1c could also bind free fatty acids (FFAs) ranging from C8 to C18, with enrichment in C12 (Figure S3B). However, the presence of other hydrocarbon chains from these FFAs made it unclear whether the density in the structure represented a C12 hydrocarbon chain, an average of a population of variable hydrocarbon chain lengths (similar to the averaged density present in the first “unloaded” CD1 structure of murine CD1d (Zeng et al., 1997)), and/or an alternatively placed MPM. Therefore we have conservatively placed a C12 hydrocarbon chain in the density to represent these possibilities. Molecular dynamics simulations suggested that a portion of MPM lipid tail could be stably bound in the F′ groove (Figure S4). However, the lack of sufficient electron density for the MPM lipid tail in the F′ groove suggested the most stable conformation for MPM is in the A′ pocket, as observed (Figures 2 and and33).
The unique F′ groove structure led us to speculate that this feature may allow CD1c to bind diverse antigens for presentation to T cells. Since previous work has demonstrated a clear role for CD1c in lipopeptide presentation through characterization of a stimulatory 12-mer lipopeptide similar in sequence to the myristoylated NEF (Van Rhijn et al., 2009), we sought to test whether the F′ groove played a role in binding of the peptide portion of this antigen. The defined sequence of the peptide motif: GGKWSKXSKWSK with acylation at the N-terminal glycine, contained two tryptophan residues and a kynurenine (designated as X ), an oxidized form of tryptophan. Titration of this synthetic lipopeptide into a solution containing CD1c resulted in clear quenching of tryptophan fluorescence (Figure 7). Based on these titration data, the equilibrium binding constant (KD) of this lipopeptide for CD1c was calculated to be ~66 nM. We therefore propose a model whereby the peptide moiety is associated with the F′ groove (either buried or associated with its surface) of CD1c while the acyl chain is accommodated in the A′ pocket, similar to what we observe for MPM.
CD1 molecules, while non-polymorphic, have evolved specific structural features enabling them to bind and present unique repertoires of lipid-based ligands to T cells. All structural information to date suggests a model whereby CD1 molecules anchor their antigens through binding of their alkyl chains in pockets or tunnels, positioning the polar lipid head groups in their A′ or F′ portals, accessible to recognition by a T cell receptor. The size and complexity of these pockets dictates the lipid-based antigens that can be bound by each isoform. CD1a possesses a “molecular ruler” in the form of a restricted sized A′ pocket, limiting bound alkyl chains to a particular length. The F′ pocket of CD1a, however, can accommodate both the N-aryl branch of a siderophore lipopeptide (Zajonc et al., 2005b) or the fatty acid chain of a sulfatide (Zajonc et al., 2003). On the other hand, CD1b has evolved a connected network of A′, F′, C′ and T′ channels (Gadola et al., 2002), creating a “maze” that has the capacity to fit alkyl chains up to C80 in length. A unique C′ portal also allows variation in size of the bound lipid moiety, acting as an escape hatch for lipids of varying length. Mouse and human CD1d appear well suited to bind diacylated lipids (Koch et al., 2005; Wu et al., 2006; Zajonc et al., 2005a; Zajonc et al., 2005c; Zajonc et al., 2008; Zeng et al., 1997) such as the canonical α-galactosylceramide (α-GalCer), a potent agonist for the CD1d-restricted invariant Natural Killer T cell population. The acyl chain of CD1d-bound lipids is bound in the A′ pocket whereas the sphingosine or phytosphingosine chains are found in the F′ pocket. We have shown here that CD1c, the last of the lipid-presenting human isoforms to be structurally characterized, has structural features unlike those present in the other human CD1 isoforms. These features endow CD1c with the unique ability to present divergent antigens such as the branched mycoketide, MPM, from M. tuberculosis and lipopeptides with long peptide moieties.
In our structure, MPM appears bound exclusively in the A′ pocket, exiting through the D′ portal underneath the α1 helix. It is this portal, unique to CD1c, that appears to allow the preferential binding of the branched alkyl chain of MPM, and enables the observed variation in length (C30–C34) of MPM that can activate T cells when presented in CD1c (de Jong et al., 2007). Similar to other complexes of CD1 molecules with lipids, this binding mode situates the polar phosphomannose head group ideally for recognition by a T cell receptor. The electron density of MPM in our structure was consistent with two alternative conformations; molecular dynamics simulations with energy minimization favored the A1 conformation consistent with optimal solvation and hydrogen bonding of the phosphomannose head group. However, the electron density for the alkyl chain extended out the D′ portal, suggesting that the computationally less preferred A2 conformation may also exist in our structure.
Our structure also provides a rationale for CD1c’s specificity for MPM’s lipid length and stereochemistry of methylation. Longer length mannosyl-β2-phosphodolichol (MPD) analogs of the length C55–95 are essentially non-stimulatory to CD1c-reactive T cells (Moody et al., 2000); alkyl chains of this length would have to extend either out of the D′ portal or out of the A′ portal into solvent, both highly energetically unfavorable conformations and therefore not preferred for binding into CD1c. Furthermore, due to the constriction in the circular A′ pocket, it is unlikely that longer branched alkyl chains could wrap around this pocket in the fashion seen in CD1b (Gadola et al., 2002). Our structure also reveals why the natural, all-S conformation of M. tuberculosis C32 MPM is preferred over synthetic, stereorandom C30 MPM which produces a 20 to 40-fold decrease in stimulation of the CD8–1 T cell clone which recognizes the complex of CD1c with MPM (de Jong et al., 2007). Based on the turn that is present in the all-S C32 MPM in our structure, alternative positioning of the last four methyl groups of MPM would force them to point inward in the turn, clashing with the natural turn of the A′ pocket. The all-S form has these last four methyl groups splayed to the outside of the turn, allowing for a tighter bend to the alkyl chain. Because 1/16th (2−4) of the stereorandom MPM mixture would be composed of all-S methyl groups in these positions, this would explain the decrease, yet not abolition, of the T cell stimulation activity of this mixture.
We and others have shown that CD1c can also bind diacylated lipids similar to those found in other CD1 isoforms (Gilleron et al., 2004; Guiard et al., 2009; Shamshiev et al., 2002), so it is likely that CD1c can stably bind two alkyl chains in its A′ and F′ pockets. However, accommodation of a single alkyl chain such as that seen in MPM likely requires the presence of the methyl branches to stabilize it in the CD1c A′ pocket. Removal of these methyl branches results in a drastic decrease in T cell simulation (de Jong et al., 2007), likely due to the weakened stability of these MPM variants in CD1c. The electron density apparent in the F′ groove of CD1c suggests that this structure, while open to solvent, maintains the ability to bind alkyl chains. Indeed, a variety of free fatty acids, enriched in the hydrocarbon chain C12, were characterized by LC-MS from our MPM loaded CD1c. Furthermore, molecular dynamics simulations suggest that MPM can also be bound in the F′ groove, although our structure suggests that it is not accommodated with the same occupancy or stability as that seen in the A′ pocket. Thus, the electron density we see is likely due to averaging over the various lipid species that may be bound in the F′ groove.
The presence of the F′ groove provides a compelling explanation for how CD1c can bind, at high affinity, divergent ligands such as the dodecameric lipopeptide (lipo-12) that mimics myristoylated proteins (Van Rhijn et al., 2009). Myristoylation is the process by which myristate (saturated C14) is covalently attached via an acyl linkage to an N-terminal glycine residue. Vital to eukaryotic cellular processes and required for the functioning of certain viral proteins (such as the myristoylated Nef (Guy et al., 1987) and Gag (Veronese et al., 1988) proteins of the Human Immunodeficiency Virus (HIV), myristolation is predicted to occur on ~0.8% of mammalian proteins and ~3.7% of viral proteins (http://mendel.imp.ac.at/myristate/myrbase/). The affinity calculated for lipo-12 (~66 nM) is similar in magnitude to that of lipoarabinomannan (LAM) for CD1b (32 nM-53 nM) (Ernst et al., 1998) and to those measured for high-affinity peptides for MHC molecules. Both are higher than the measured affinities for CD1d and αGalCer (ranging from 340 nM (Naidenko et al., 1999) to ~1–10 μM (Cantu et al., 2003)). It is important to note that in all cases, these are relative measurements as CD1 molecules are likely loaded with endogenous lipids during eukaryotic synthesis or detergent molecules during refolding. In addition, lipids most likely exist in a micellar form depending on their critical micellar concentration (CMC). Thus these measurements reflect a combination of parameters, including removal of lipids from the micellar state and competition with CD1-bound lipid. However, our affinity measurements for lipo-12 and CD1c suggest involvement of the peptidic portion of the antigen, considering that unbranched, saturated synthetic versions of MPM are not stimulatory and therefore unlikely to be stably bound (de Jong et al., 2007).
Supported by our tryptophan fluorescence quenching results and affinity determination, we propose a model where the acyl chain of lipo-12 is bound in the A′ pocket, similar to the branched alkyl chain of MPM, while the peptide portion, anchored by the two tryptophans and one kynurenine, is bound in the F′ groove. It is important to note, however, that tryptophan fluorescence quenching can be caused by a variety of factors including small structural changes in the protein and therefore does not provide direct evidence of the peptide portion being bound in the F′ groove. In the absence of a crystal structure of the complex of CD1c and lipo12, other models are also plausible, such as the peptide portion associating with the surface of CD1c, outside of the F′ groove. Shorter lipopeptides, or alteration of the amino acid sequence of the peptide portion, can affect T cell reactivity (Van Rhijn et al., 2009), however it awaits further analysis to determine whether alteration of these residues affects lipopeptide loading and stability in CD1c. This model of lipopeptide presentation by CD1c provides insight into how acylated proteins can be recognized as pathogenic patterns, raising questions about how CD1c-reactive T cells may discriminate between endogenous and foreign lipopeptides.
The synthesis of mannosyl-β1-phosphomycoketide (orβ-D-mannosyl phosphomycoketide) (MPM) (structure #29) is outlined in Supplementary Methods: Schemes S1-S5. Starting from commercially available methyl (2S)-3-hydroxy-2-methylpropionate (1), compounds 11 (Scheme S1), 17 (Scheme S2) and 22 (Scheme S3) were prepared. Using a sequence of reactions that included Julia-Kocienski coupling, platinum- and palladium-catalyzed hydrogenations and the Dess-Martin oxidation, these fragments were connected synthetically (Scheme S4) in good yield to give the precursor mycoketide C7 (n-heptyl at the tail end of the lipid). The synthesized mycoketide was used to prepare mannosyl-β1-phosphomycoketide 29 according to a literature procedure (Scheme S5) (Summeren et al., 2006). The mass spectrum and other analytical data confirmed the structure of 29. Negative ion MS Calcd for C38H79O9P− (M-Na+): 707.5, found: 707.4. Synthetic details will be reported elsewhere.
The CD1copt construct was subcloned into the baculovirus transfer vector pAcGP67A (BD Biosciences) with a deca-His tag and expressed in High Five cells using the baculovirus expression system (Saphire Baculovirus DNA-Orbigen). The protein was purified with Ni-NTA resin, eluted with buffer containing 200 mM imidazole, and treated with carboxypeptidase A (Sigma) at 25°C overnight to remove the C-terminal His-tag. The protein was further purified by gel filtration using a Superdex 200 column (GE Healthcare) equilibrated with Hepes Buffered Saline (HBS: 10mM Hepes, pH 7.2, 150mM NaCl, 0.02% azide).
To load MPM into CD1c, 0.2 mg of MPM per 1 mg of CD1c were incubated at 37°C for 30 min in HBS containing 0.05% Tween 20. The CD1c-MPM complex was repurified by gel filtration using a Superdex 200 column equilibrated in HBS containing 50mM NaCl to remove excess lipid and detergent. For crystallization, peak fractions were pooled and concentrated to 10 mg/mL. Crystallization screens were carried out using the sitting drop method and initial hits were optimized in both sitting and hanging drops by varying pH, precipitant concentration and the use of additives. The best crystals were obtained using a hanging-drop setup at 18°C by combining 1 μL protein solution and 1 μL mother liquor containing 1.05 M sodium citrate, 100 mM CHES pH 9.4 and 25 mM triglycine.
X-ray data sets were measured on a MAR300 CCD at beamline 23 ID-D at the Advanced Photon Source (APS) at Argonne National Laboratory. Crystals were cryoprotected with 50% sodium malonate before cooling to 100°K. A complete 2.5 Å data set was collected using 1° oscillations over 150 frames and a 1.033×10−10 m wavelength. The crystals indexed in space group P212121 with unit cell dimensions a=53.72 Å, b=87.09 Å, c=88.81 Å. HKL2000 was used to index, integrate and scale the data (Otwinowski and Minor, 1997). Data statistics are shown in Table 1.
Using a CD1b heavy chain and β2m (Garcia-Alles L.F., 2006) as a search model, stepwise molecular replacement with Phaser (A. J. McCoy, 2007) located the single CD1c molecule per asymmetric unit. Rigid body refinement was performed using Phenix (P. D. Adams, 2010), followed by manual building using Coot (Emsley and Cowtan, 2004) and individual site and B-factor refinement using Phenix (P. D. Adams, 2010). The stereochemistry of the model was improved using a backbone optimization procedure incorporating a torsional statistical potential contingent on amino acid type, its secondary structure as well as the identity of its two neighbors (DeBartolo et al., 2009), followed by real space refinement using a two step algorithm developed by E.J. Haddadian, K.F. Freed and T.R.Sosnick (manuscript in preparation). Both composite (Phenix (P. D. Adams, 2010)) and simple (CCP4 Sfcheck (C.C.P.4., 1994)) omit maps were calculated at each stage to locate density for ligands that were not present in the model. Hydrogen bonding contacts between MPM and CD1c were calculated using the program Contacts (CCP4 program package (C.C.P.4., 1994)). Main-chain RMSD values between CD1c and CD1a, CD1b and CD1d were calculated using PDBeFold (Krissinel and Henrick). Total ligand binding groove volumes of CD1a, CD1b, CD1c and CD1d were computed using Pocket-Finder (an algorithm based on Ligsite (Hendlich et al., 1997)). All structural figures were generated using the program Pymol (DeLano Scientific).
To measure binding of lipo-12 to CD1c, 1 mL samples containing 10 nM CD1c in HBS and various concentrations of lipo-12 ranging from 5 nM-10 μM were incubated at 37°C for 1 hour. Tryptophan fluorescence of the sample was measured using a PTI fluorimeter (model 810) with an excitation wavelength of 280 nm and the maximum emission signal at 340 nm was observed. The quenching of tryptophan fluorescence with increasing amounts of lipo-12 was expressed as a ratio (E-Eo)/Eo, the change in fluorescence divided by the control fluorescence in the absence of added lipo-12. The signal ratio was plotted against lipo-12 concentration and fit using non-linear regression in GraphPad Prism to a one-site binding model.
We thank the staff of the Advanced Proton Source at GM/CA-CAT (23ID) and LS-CAT (21ID) for their use and assistance with X-ray beamlines; Ruslan Sanishvili and Joseph Brunzelle in particular for help and advice during data collection; Esmael Haddadian, Karl F. Freed and Tobin R. Sosnick for use of their torsion angle plus real space refinement algorithm; Michael Brenner for the CD1c cDNA and Albert Bendelac for helpful discussions. This study was supported by National Institutes of Health grants R01AI073922 and R01GM081642 and the Kinship Foundation Searle Scholars Award to Erin J. Adams.
Accession Numbers: Coordinates and structure factors for the complex of CD1c with MPM have been deposited in the Protein Data Bank under the accession code 3OV6.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.