Crystal structure of a self-assembling lipopeptide detergent at 1.20 Å

Abstract

Lipopeptide detergents (LPDs) are a new class of amphiphile designed specifically for the structural study of integral membrane proteins. The LPD monomer consists of a 25-residue peptide with fatty acyl chains linked to side chains located at positions 2 and 24 of the peptide. LPDs are designed to form α-helices that self-assemble into cylindrical micelles, providing a more natural interior acyl chain packing environment relative to traditional detergents. We have determined the crystal structure of LPD-12, an LPD coupled to two dodecanoic acids, to a resolution of 1.20 Å. The LPD-12 monomers adopt the target conformation and associate into cylindrical octamers as expected. Pairs of helices are strongly associated as Alacoil-type antiparallel dimers, and four of these dimers interact through much looser contacts into assemblies with approximate D₂ symmetry. The aligned helices form a cylindrical shell with a hydrophilic exterior that protects an interior hydrophobic cavity containing the 16 LPD acyl chains. Over 90% of the methylene/methyl groups from the acylated side chains are visible in the micelle interiors, and ≈90% of these adopt trans dihedral angle conformations. Dodecylmaltoside (DDM) was required for the crystallization of LPD-12, and we find 10–24 ordered DDM molecules associated with each LPD assembly, resulting in an overall micelle molecular weight of ≈30 kDa. The structures confirm the major design objectives of the LPD framework, and reveal unexpected features that will be helpful in the engineering additional versions of lipopeptide amphiphiles.

Detergents that are able to stabilize native structures of membrane proteins in the absence of lipid bilayers are essential tools for the structural study of membrane proteins (1, 2). Despite the large number of detergents that are available, no single detergent is optimal for all purposes, and choice of the solubilizing agent is highly dependent on both the target protein and on the application. In particular, the structural study of membrane proteins by NMR or x-ray crystallography depends critically on the choice of detergent (3, 4), and many proteins cannot be studied because of problems with protein stability and aggregation in the commonly used detergents (5). Because of these issues, there is a need to expand the range of amphiphiles available for membrane proteins research.

Nondenaturing detergents act as membrane mimetics, and an ideal detergent would generate a local environment at the protein surface that is indistinguishable from that of the native lipid bilayer. A central shortcoming of many of the commonly used detergents is that the micelle interior is less ordered and less well-packed relative to the interiors of lipid bilayers (2, 6). This is a direct result of the shape of the amphiphile monomer, which has a high degree of positive intrinsic curvature, resulting in self-assembly into spherical or ellipsoidal micelles with poor internal packing. We have designed lipopeptide detergents (LPDs) as a new class of amphiphile engineered to form small cylindrical assemblies with rigid exterior surfaces that provide a natural interior packing environment for fatty acyl chains (7). The basic LPD design consists of a 25 aa α-helical peptide scaffold that supports alkyl chains covalently attached to each end of the helix. LPDs with alkyl groups ranging from 12 to 20 carbons in length (LPD-12 to LPD-20) are soluble and form relatively small micelles with low aggregation numbers and critical micelle concentrations less than 10⁻⁶ M (7). LPDs are highly effective and generally outperform traditional detergents such as N-dodecyl-β-d-maltopyranoside (DDM) in preserving the native structures of membrane proteins in solution (7). The ability of LPD to stabilize membrane proteins is most likely due to the formation of LPD-protein assemblies in which cylindrically arranged LPD α-helices support a membrane-like ring of fatty acyl chains sandwiched between the LPD peptides and the hydrophobic surface of the target protein. As a consequence, the number of LPD monomers in the protein complex depends on the cross-sectional diameter of the protein in the transmembrane region.

Stroud and coworkers introduced “peptitergents” as amphipathic α-helical peptides for the solubilization of membrane proteins (8). The peptitergent PD1 could maintain some membrane proteins in a solubilized state, but was not as effective as the more commonly used “traditional” detergents. The LPD design builds on the peptitergent concept by introducing fatty acyl chains as the apolar moiety of the peptidic amphiphiles, and the inclusion of the acyl chains has a profound effect on the behavior of the peptides. Notably, an LPD-0 control peptide consisting of the parent peptide sequence but without the attached acyl groups did not exhibit high α-helical content, did not self-associate into regular assemblies, did not stabilize solubilized membrane protein and did not have detergent properties (7). The profound difference in the behaviors of LPD-0 relative to the acyl-coupled peptides demonstrates that the self-assembly and detergent properties of LPDs are largely determined by the sequestration of the acyl tails into micelles. Thus, in contrast to the peptitergents, the α-helices in the LPDs serve mainly as a rigid scaffolding that supports the fatty acyl tails, and the key to LPD detergent behavior is the proper positioning of the alkyl chains in an amphipathic monomer, and not the amphipathicity of the α-helical peptides.

In this study, we report the crystal structure of LPD-12, a well-behaved LPD with the sequence acetyl-A(O12)AEAAEKAAKYAAEAAEKAAKA(O12)A-amide. O12 refers to modified ornithine residues generated by the formation of an amide bond between the ε-amine of the side chains and the carboxylic acid functions of dodecanoic acid. The lipopeptide was designed to have high α-helical propensity and form a mildly amphipathic helix with an alanine face opposite from a charged lysine/glutamate face. The expected length of the 25 residue α-helix is approximately equivalent to the width of a lipid bilayer. A single tyrosine was included to facilitate the quantitation of the LPD. The design places the O12 acyl chains at residue positions 2 and 24 such that they are able to lie along the alanine face of the helix. Pure LPD-12 self-assembles in solution into small, monodiperse assemblies with an aggregation number of 8 (7), a result supported by molecular dynamics modeling studies (9). The high-resolution crystal structure presented here confirms many of the elements of the target design, but also reveals unexpected features that will be helpful for the generation of new LPD variants.

Results

We determined the crystal structure of LPD-12 to a resolution of 1.20 Å (supporting information (SI) Table S1). The crystals contains two distinct octameric micelles. The “tight” octamers contains chains A-H, whereas the “loose” octamers are made up from chains I-L and the symmetry related set I′-L′ generated by a crystallographic two-fold axis. Both LPD micelles consist of eight antiparallel α-helical lipopeptide monomers that are approximately aligned about a central bundle axis (Fig. 1). The highly organized alkyl chains are contained within the hydrophobic core, and the hydrophilic peptide surfaces face the exterior. The octamers are made up from four pairs of tightly associated dimers (A/B, C/D, E/F, G/H for the tight octamer and I/J, K/L, I′/J′, K′L′ for the loose octamer) arranged with approximate D₂ symmetry. In the case of the loose octamer, the bundle axis is coincident with a true crystallographic two-fold axis. A comparison of the tight and loose octamers is presented in Fig. S1. The RMSD between the 200 Cα atoms of the tight and loose octamers is 2.9 Å, and decreases to 1.9 Å if chains C/D and K/L are excluded (Table S2). The overall dimensions of the micelles are ≈40 Å × 40 Å × 55 Å, with the longer distance corresponding to the bundle axis direction. The temperature factors of the main chain atoms are mostly in the range of 8–12 Ų, whereas some of the longer Lys, Glu and O12 side chains are in the range of 10–40 Ų (Fig. S2). For the remainder of this article, we describe only the tight octamer, except as indicated.

Fig. 1.

Orthogonal views of the “tight” LPD-12 octamer. (Center) The chain labels A-H are positioned near the C termini of the peptides, and the LPD acyl chains are in colored stick representation. The vertical black line (labeled z) represents the pseudo two-fold axis that is coincident with the bundle axis, and relates chains ABCD to chains EFGH. The horizontal dotted gray line (labeled y) represents the pseudo two-fold axis that relates chains ABCD to chains HGFE. This view is along the pseudo two-fold axis (x) that relates chains BAHG to chains CDEF. (Left) View along the bundle axis (z). Chains A, C, E and G are oriented with the C terminus pointing toward the viewer, and are labeled “+.” Chains B, D, F and H are in the opposite direction and are labeled “−.” (Right) View along the pseudo two-fold y axis.

LPD-12 Monomers.

The LPD-12 lipopeptides are α-helical as designed, with the charged lysine and glutamate face of the helix opposite from the nonpolar alanine/O12 face (Fig. 2A). The peptide chains of the twelve crystallographically independent monomers are highly similar, and superimpose with a mean Cα rmsd of less than 0.3 Å. There is more variability in the conformations of the acylated ornithine (O12) residues side chains, despite the fact that nearly 90% of the 366 observed O12 dihedral angles are in the trans conformation. Full length acylated side chains are visible in 9 of the 24 crystallographically independent O12 residues, whereas 1–4 of the terminal methyl/methylene groups of most of the other O12 side chains could not be modeled into the electron density because of disorder. Two of the O12 side chains do not extend beyond carbon C4 of the coupled dodecanoic acids. The alkyl groups of the O12 side chains at peptide residue position 2 follow a variety of paths, but the O12 conformers at position 24 are more regular and cluster into two groups according to the gauche- or trans conformation at the C1-C2-C3-C4 dihedral angle (labels “a” and “b” in Fig. 2A, respectively). The alkyl groups in the “a” cluster contact the alanine face of the peptide, as designed, whereas the all-trans“b” cluster extends away from the helix (Fig. 2B). Of the four designed (i, i + 4) intrachain salt bridges in the sequence, only the charged atoms from the E7/K11 and E18/K22 pairs interact (Fig. 2B, Fig. S3 and Table S3) and these all have temperature factors <15 Ų (Fig. S2B).

Fig. 2.

Conformation of the LPD-12 monomers. (A) Superposition of the twelve LPD-12 monomers, with the peptide backbone represented as ribbons and the side chains shown as sticks. Each lipopeptide is colored differently. The acylated ornithines at residue positions 2 and 24 are labeled O12–2 and O12–24, respectively. (B) A portion of lipopeptide chain B is shown along with Fo-Fc electron density contoured at 2.5 σ, calculated with the entire chain omitted from the model.

LPD-12 Dimers.

Pairs of LPD-12 monomers associate into tight antiparallel dimers that form the building blocks of the micelle (Fig. 3). The peptidic moieties of all of the dimers are highly similar, and superimpose with a mean Cα rmsd of 0.28 Å (Table S2). There is a two-residue offset between the peptides, resulting in a short C-terminal “overhang” in each dimer pair. The dimers form left-handed coiled-coiled superhelices with a pitch of 139 Å and crossing angle of 24.9°, features similar to those of classic coiled-coil GCN4 parallel dimer structures (10) (Table S4).

Fig. 3.

LPD-12 dimers associate as Alacoils. (A) Schematic representation of the LPD-12 antiparallel coiled-coil dimer with a-a and d-d interactions at the interface. (B) The peptide sequence of LPD-12 and its heptad repeat assignment. (C) Cartoon representation of an end-on view of a coiled-coil LPD-12 dimer. The orientation is similar to that in A. (D) Superposition of the 6 different Alacoil dimers (A/B, C/D, E/F, G/H, I/J and K/L), viewed from the exterior-facing polar surface. The gray lozenge in the center represents the local pseudo two-fold axis present in each dimer. These axes are independent of the x,y and z pseudo 2-fold axes of the octamer that are described in Fig. 1. The amino termini are labeled.

The two chains in each LPD-12 antiparallel dimer are related by a noncrystallographic pseudo two-fold axis centered between residues A10 and A14 and normal to the coiled-coil superhelix axis (Fig. 3D). These two-fold axes are unrelated to the three mutually perpendicular pseudo two-fold rotation axes of the approximate D₂ symmetry described in Fig. 1. The dimer interface exhibits knobs into holes packing between residues A3, E7, A10, A14, A17 and A21, and these can be assigned to positions a and d of a heptad repeat (Fig. 3). The small alanine side chains at the interface allow for a very close approach of two helices, with an average interhelical distance of 8.1 Å (Table S4). The dimers correspond to Alacoil structures (11), which are antiparallel coiled-coil motif with a-a and d-d interactions at the interface (Fig. 3A). The interdigitated helical backbones are offset by ∼0.25 of a heptad repead, which, along with the short ∼8 Å distance between the helical axes, is a characteristic of a ferritin-type Alacoil (11). Alacoils have been previously observed in designed protein sequences (12–15). There are no interchain salt bridges between any of the peptides in the octamers (Fig. S3), although there are extensive glutamate-lysine salt bridges between the octamers in the crystal lattice.

LPD-12 Octamers.

Four LPD-12 Alacoil dimers assemble into octameric micelles with an overall antiparallel arrangement of the monomers (Fig. 4). The individual peptide residues can be assigned to four different types of partially overlapping environments. Positions c, d and g from the heptad repeats face into the core of the assemblies and consist exclusively of alanine and O12 residues. Heptad positions b and e face outward from the bundles, and contain most of the charged lysine and glutamate residues. The a and d position residues are involved in the tight Alacoil contacts, whereas the c and f positions are on the face opposite the Alacoil, and define the much looser lateral interfaces.

Fig. 4.

The LPD-12 octamer. (A) Schematic representation of the octamer viewed along the bundle axis, with each helix represented by its helical wheel. Heptad-repeat positions c, g and d are shaded in pink, and face the octamer core. The Alacoil, type I and type II interfaces are indicated. The chain-coloring and -labeling is consistent with Fig. 1. (B–D) The three types of interfaces are shown in ribbon representation with the peptide solvent-accessible surface. All views are from the exterior surface of the micelle, and coloring is according to A. The amino termini are labeled and the Tyr-12 residues are shown in sticks. The gray lozenges in C and D correspond to the pseudo 2-fold noncrystallographic x and y axes in Fig. 1, respectively. The thin vertical line represents the bundle z axis.

There are two types of lateral interfaces involving heptad position residues c and f between adjacent Alacoil dimers (Fig. 4). The type I interfaces are characterized by a minor offset along the lengths of the helices, and are observed for helix pairs B/C and F/G in the tight octamers, and pairs J/K and J′/K′ in the loose octamers. The helices in these interfaces are 15–16 Å apart with a 35° crossing angle, and are tilted by ∼28° relative to the octamer bundle axis. Type II interfaces are observed between chains D/E and H/A in the tight octamers and L/I′ and L′/I in the loose octamers, and in these cases, the helices are 13–14 Å apart and have a crossing angle of ∼5°. Helices at type II interfaces are offset from each other by approximately half a length of helix and are approximately aligned with the bundle axis. The views in Fig. 1 Center and Right are directly into the type I and type II interfaces, respectively.

The pairs of helices involved in the different examples of the type I and type II interfaces do not superimpose very well, unlike the well defined Alacoil pairs (Table S2), indicating that these weak interfaces are structurally more variable. There are only very minor interpeptide contacts across either the type I and type II interfaces. These involve the tips of Tyr-12 across the type I interfaces, and minor van der Waals contacts between the side chains of Y12 and A5 in some of the type II interfaces (Fig. 4). There is an average buried surface area of 11 Ų and 112 Ų between the peptides in the type I and type II octamer interfaces, respectively, in contrast to an average of 478 Ų in the Alacoil interfaces. Thus, the main driving force for the self-assembly of the octamers does not come from lateral peptide-peptide interactions between the Alacoil dimers, consistent with the unremarkable properties of the LPD-0 peptide (7). Instead, the formation of LPD-12 micelles is driven by the sequestration of the O12 alkyl chains into the hydrophobic core of the LPD assemblies, similar to the behavior that determines micelle formation in traditional detergents. There have been previous examples of using peptides to form internal structures with specific properties. For example, a de novo designed four helix bundle protein was engineered to provide an internal metal binding site, but in this case, the characteristics of the internal cavity was largely determined by peptide-peptide contacts and the protein was folded even in the absence of metal ions (16).

Pairs of Alacoil dimers related by the pseudo two-fold axis aligned with the octamer bundle axis (z in Fig. 1) tilt toward each other and interact at one end through the acylated ornithine residues. Thus, dimer A/B makes contact with dimer E/F and dimer C/D makes contact with dimer G/H, forming a small patch of four aligned O12 chains at the top (Fig. 5 A and C) and bottom (Fig. 5 B and D) of the bundle. The opposite termini of these pairs of dimers are farther apart, and acyl chains attached to these ends travel toward the central axis of the bundle instead of along this axis. The reason for this becomes apparent when we consider how these tetramers assemble into octamers, because the “inverted V” shape of the AB/EF tetramers interlocks with the “upright V” shape of the CD/GH tetramers (Fig. 5E). Movie S1 presents the Alacoil, type I and type II interfaces, as well as the sequestered acyl core in the tight LPD micelle.

Fig. 5.

Assembly of the LPD-12 octamer. (A) The A/B and E/F dimers are shown with the α-helices as cylinders and the O12 residues as gray sticks. The C-termini of the four LPD chains are labeled. This view is rotated +45° about the bundle axis from the view in Fig. 1 Center, and the chain-coloring is consistent with Fig. 1. (B) Similar representation for the G/H and C/D chains. This view is rotated −45° about the bundle axis from the view in Fig. 1 Center. (C) Similar to A, but represented as the solvent accessible surface. (D) Similar to B, but as the solvent accessible surface. (E) View of the assembled octamer, as seen directly into the A/B Alacoil dimer.

Hydrophobic Core and the Role of DDM.

The sixteen O12 chains per octamer do not completely occupy the hydrophobic core of the micelles, even if we account for the 10% of the O12 atoms that we could not model into the electron density (Fig. 6A). Well-ordered DDM detergent molecules, which were required for LPD-12 crystallization, are present in both the tight and loose octamers and serve as “mortar” to fill the remaining gaps within the micelle interiors. We have located 10 DDMs in the tight octamer, and 2 × 12 DDMs in the two-fold symmetric loose octamers (Table S1). The overall volume of the interior cavity formed by the peptidic parts of the tight octamer is ≈950 ų, and is fully occupied by the O12 side chains and the DDM acyl chains if we account for the small number of atoms not located in the electron density. The interior of the octameric bundles is apparently “dry”, and we have not located any water molecules within any of the micelles.

Fig. 6.

Detergent molecules in the LPD-12 octamer. (A) View of the octamer with the peptides represented as ribbons and the O12 residues as light gray surfaces. The view is in the same orientation as in Fig. 5E, and the chain-coloring is consistent with Figs. 1 and 5. (B) Similar to A, but with the inclusion of the associated DDM molecules in space-filling representation. DDM carbons are dark gray, and oxygens are red. The DDM molecules occupy the type I and type II interfaces that are to the left and right of the yellow/red B/A Alacoil dimer, respectively. (C) Close-up of the DDM molecules flanking the Y12 residues of chains B and C in a type I interface. DDM carbons are in yellow, and |F_o − F_c| electron density calculated with the omission of the DDMs and contoured at 2.5 σ is included. (D) Similar to C, but with two DDM molecules located at the type II interface between LPD chains A and H.

The DDMs fill the type I and type II interfaces between the Alacoil dimers (Fig. 6B and Fig. S4), and thus seal the micelle interior from the external aqueous solvent. This explains the near lack of lateral peptide side chain contacts at the non-Alacoil lateral interfaces (Fig. 4 C and D). Stacks of aligned DDMs can be seen lining the openings in the bundles at the type I and type II interfaces. The maltoside head groups are found associated with either tyrosines or hydrophilic surface residues of the micelle, whereas the detergent hydrocarbon tails penetrate into the octamer interior (Fig. 6 C and D, Fig. S4, Fig. S5, and Movie S1). Interactions between DDM headgroups and aromatic side chain interactions have been described in crystals of cytochrome c oxidase (17).

The loose octamer has a higher DDM content than the tight octamer (24 per micelle versus 10 per micelle) and the interior acyl chains are slightly less organized relative to the tight octamer (Fig. S1 and Fig. S4). Consequently, the volume of the loose octamers is slightly larger than the tight octamers. DDM molecules are also present at the top and bottom of the loose octameric bundles, where the CB, CG and CD atoms of the acylated ornithines constitute a hydrophobic surface at the caps of the cylinders that cannot be protected by the LPD α-helices.

Discussion

The LPD-12 crystal structure has allowed us to visualize an LPD micelle assembly at near-atomic resolution. Previous dynamic light scattering and analytical ultracentrifugation data indicated that LPD-12 formed highly defined, monodisperse octamers (7), a result supported by molecular dynamics simulations (9). In the latter work, LPD-12 octamers with antiparallel helices were highly stable, whereas simulations with LPD-12 hexamers and decamers exhibited significant instabilities. The internal acylated ornithine residues in the octameric LPD-12 models had high trans content and long average rotation times relative to liquid alkanes, and overall the chain dynamics were reminiscent of the slow dynamics in lipid gel phases (9). The experimental structure presented here supports many of these findings, but reveals additional unexpected features. The structure confirms the following aspects of LPD micelles: (i) the LPD-12 lipopeptides form α-helices with the O12 groups aligned along the expected face of the monomers, (ii) the monomers self-associate in an antiparallel arrangement into octameric micelles, and (iii) the hydrophobic core of the octamers is tightly packed with well-organized alkyl chains. Other features, however, were not foreseen in the design: (i) the assemblies exhibit approximate D₂ symmetry, a feature presumably limited to the octameric complex, (ii) the LPD-12 helices form tight ferritin-type Alacoil dimers which serve as the building blocks for the micelles, (iii) there are only minor peptide-peptide interactions between the Alacoil dimers and there is considerable variability in the type I and type II interfaces, and (iv) well-ordered DDM molecules fill the unoccupied space in the hydrophobic core of the micelles at the type I and type II interface sites. The latter two features deserve additional comment. It is important to note that the previous characterizations of LPD by solution and computational methods did not include DDM, and pure LPD-12 forms stable, defined octamers in the absence of added detergents (7). LPD-12, however, did not crystallize without DDM, and pure LPD-12 micelles are probably more internally dynamic than the LPD/DDM mixed micelles. This suggests that added “traditional” detergents can be used to modulate the behavior of LPDs.

We have found LPDs to be effective in solubilizing both low and high molecular weight membrane proteins, and these detergents can thus mold themselves to surround a variety of membrane protein shapes and sizes. Octamers are the minimal assembly size for pure LPD, but these micelles do not have the space to accommodate a guest protein. Larger number of monomers are necessarily required to form the annular assemblies that accommodate membrane proteins, and this requires a certain plasticity in the lateral LPD-LPD contacts. It is likely that these changes arise at positions similar to the loose type I and type II interfaces seen in the octameric micelles, where there are only very minor peptide-peptide contacts. The principles for the construction of antiparallel helical bundles into tube-like assemblies with dihedral symmetries have been explored (13, 18), but it is unlikely that variations of the dihedral symmetry that we observe in the LPD-12 octameric micelles can be applied to the arrangement of the LPDs in an LPD/membrane protein complex. A symmetric arrangement would require rigidly determined detergent positions, and this would not be consistent with the diversity of membrane protein shapes. Also, LPD self-assembly is largely driven by alkyl chain sequestration, and this effect is expected to lead to more malleable micelles than the well-ordered side chain packing interactions seen in helix-helix contacts (19). Finally, our results suggest that the fine tuning of LPD acyl chain packing with traditional detergents such as DDM may be a simple way to optimize the solution properties of an LPD/protein complex. Notably, this does not compromise the fact that the solubilized LPD/protein complexes would remain relatively small as compared to traditional protein-detergent complexes and retain a relatively rigid exterior surface dominated by the charged surfaces of the LPD α-helices. Membrane proteins solubilized in this way may have highly favorable properties for study by NMR or x-ray crystallography.

Materials and Methods

Lipopeptide Synthesis.

LPD-12 was synthesized and purified according to protocols described previously (7).

Crystallization and Data Collection.

LPD-12 at 15 mg/ml (5.28 mM) was crystallized by the hanging drop method at 20°C with a precipitant solution containing 0.15 M potassium phosphate (pH 4.2) and 12% PEG 3350. Orthorhombic crystals (P2₁2₁2; a = 72.48 Å, b = 130.04 Å, c = 42.81 Å) were grown in the presence of 2:1 molar equivalents of DDM (Anatrace) to LPD-12. Crystals were cryo-protected using mother liquor supplemented with 15% ethylene glycol and flash-frozen in liquid nitrogen. An iodinated derivative was generated by the vaporizing iodine labeling method (20) and data were collected to 2.5 Å on a Bruker Proteum CCD system using CuKα radiation and reduced with SAINT-Plus software. A native dataset was collected to 1.20 Å resolution at the Advanced Photon Source, SBC-CAT 19-ID beamline, Argonne National Laboratory with 0.98 Å x-rays and were reduced with HKL3000 (21).

Structure Determination and Refinement.

Initial SAD phases were obtained from the data collected on the iodinated derivative. A total of seven iodine sites were identified by SOLVE (22). Phases were extended with density modification and autotraced with RESOLVE (22) using the 1.20 Å native dataset. The partially traced polypeptide model was subjected to initial refinement with REFMAC (23), producing an electron density map with clear peptide and acyl chains. Subsequent rounds of anisotropic refinement were carried out by with SHELXL (24) and model building in COOT (25). The final R_work and R_free were 13.6% and 17.2%, respectively. Detailed crystallographic statistics are presented in Table S1. Molecular illustrations were prepared with PyMOL (DeLano Scientific).