High‐Resolution Crystal Structure of a Lasso Peptide

Have I got noose for you! Lasso peptides are a growing class of bioactive peptides of microbial origin. The first crystal structure of a member of this family, the glucagon receptor antagonist BI-32169, shows that the fold is built predominantly by regular secondary structural elements and a tight network of hydrogen bonds that are partially shielded from solvent by hydrophobic amino acid side chains. This results in an extraordinarily stable structure that is resistant to thermal unfolding or proteolytic digestion, which facilitates its biological function. A growing class of bioactive peptides of microbial origin has a common, unique structural feature that has been termed “lassoed tail” or “lariat protoknot”.1–3 These peptides share a macrolactam ring at the N terminus of the peptide sequence, which is formed by condensation of a Asp 8/9 or Glu 8 side chain with the free N terminus of a Gly 1 or Cys 1 residue. Through this ring, the C-terminal tail of the peptides is threaded and trapped, either by steric hindrance or by covalent modification, by disulfide formation between a cysteine within the C-terminal sequence and a ring cysteine residue. This preserves a compact structure that entropically favors receptor interaction and is extremely resistant to unfolding or proteolytic digestion. For two members of this family, the genetic origin and the route of biosynthesis have been elucidated. Peptide precursors are ribosomally synthesized and post-translationally modified by proteolytic cleavage and amide bond linkage.4–7 Some of the known lasso peptides are secreted into the surrounding media.8, 9 Their proposed biological role is to act as antibacterial agents, presumably to give advantage over competing bacterial species or strains within the same habitat. The peptides were isolated from various microbial sources, including Streptomyces strains, Rhodococcus, Microbispora and Escherichia coli and identified as potent inhibitors of various pharmaceutically relevant receptors.2, 10–12 Weber et al. discovered anantin, the first peptide of this class to be identified,11 and described the principle of macrolactam formation. In 1994, the three-dimensional structure of RP-71955, solved by NMR spectroscopic methods, provided the first demonstration of a protoknot structure.13 Since then, solution NMR structures of eight lasso peptides (Figure 1) have been determined.1, 7, 9, 13–16 a) Sequences of lasso peptides with known three-dimensional structure. The first and the 8th or 9th amino acid (grey shade) form the macrolactam ring. In RP-71955/MS-271 and BI-32169 there are two and one disulfide bonds, respectively, formed by cysteine residues highlighted in blue. The ring-penetrating residue pair in each C-terminal sequence is highlighted in green. b) Schematic representation of the structure of RP-71955 as a representative for the lasso fold. The macrolactam ring (red ribbon) is formed via condensation of the free N terminus and the glutamic acid or aspartic acid side chain of residue 8 or 9 (yellow carbon atoms). The C-terminal tail (colorless ribbon) is folded such that the peptide runs through the middle of the ring. Disulfide bonds are shown as blue spheres. From this sequence analysis, BI-32169 is a member of the lasso peptide family that exhibits hallmarks of previously characterized classes, structural cross-linking via disulfide bridges and a glycine residue at the N terminus. It differs by the sequential position of its disulfide bridge, as well as its ring penetrating residue pair located four amino acids C-terminal to the ring closing aspartic acid residue. The structure of a 19-mer peptide, BI-32169, was established by amino acid analysis, mass spectrometry, and two-dimensional NMR analysis. BI-32169 was shown to take on a bicyclic structure with the hallmarks of a lasso peptide and a disulfide bond between Cys 6 and Cys 19.10 BI-32169 and its methyl ester derivative show potent inhibitory activity against the human glucagon receptor (IC50 values of 440 and 320 nM, respectively) in a functional cell-based assay. Thus, it is a promising lead compound for the development of new agents against type II diabetes.17 Here, we present the crystal structure of BI-32169 determined at atomic resolution. The crystal structure confirms the basic principles of the lasso fold as shown by the NMR solution structures. Furthermore, unprecedented details of the peptide tertiary arrangement are revealed by the high resolution of the crystal structure. The BI-32169 sequence is dominated by hydrophobic and aromatic amino acids and therefore has limited aqueous solubility. The compound crystallizes from a 20 mg mL−1 solution in 50 % dimethyl sulfoxide (DMSO) in a buffer containing 2-methyl-2,4-pentanediol (MPD) as the precipitant. The mother liquor is thus mainly aqueous. The crystals are extraordinarily well-diffracting plates that allow for the collection of atomic resolution data on an in-house X-ray source. Refinement of the final model (Figure 2) was done without imposing any stereochemical restraints. a) 2Fo−Fc electron density map of BI-32169 contoured at 3.0 σ. The carbon atoms of residues forming the ring are colored blue, the carbon atoms forming the tail are colored white. b) Stereoplot of the molecular structure of BI-32169 (rotated in plane by 180° around the perpendicular axis relative to panel a) as an ORTEP representation showing 30 % probability thermal ellipsoids of all C (black), N (blue), O (red) and S (yellow) atoms. A stereochemical analysis of the final peptide structure18, 19 shows that it generally obeys empirically derived parameter ranges for peptide structures20 with regard to bond lengths, bond and dihedral angles, as well as being consistent with usually observed peptide conformations. All residues fall within most favorable regions of the Ramachandran plot except Trp 13, which falls just outside allowed regions with (ϕ,ψ)=(−85°,−147°). The few observed deviations from ideal stereochemistry are associated with the Asp 9/Gly 1 side chain–main chain amide bond, a cis-peptide bond between Pro 7 and Ser 8 or Trp 13. The former may be due to strain introduced by the Asp 9/Gly 1 bond formation. Crystal packing is facilitated predominantly by burying hydrophobic surface patches on adjacent chains. Four direct hydrogen bonds are formed with neighboring molecules including the C-terminal carboxylate that forms a hydrogen bond with Thr 15 OγH. There is one major void in the crystal lattice, which is lined on one side by a polar face of one molecule and filled with solvent. The peptide adopts regular secondary structure (Figure 3). The C-terminal sequence Pro 16/Trp/Ala/Cys 19 forms a 310-helix.21 A classical type I β-turn consists of residues Asp 9/Ile/Pro/Gly 12. Otherwise, the residues are in extended or coil regions. There is one cis-peptide bond between residues Pro 7 and Ser 8. This cis-peptide bond prevents formation of a continuous hydrogen-bonding ladder between ring and ring-penetrating tail residues as observed for the other lasso peptide structures.7 Here, the hydrogen-bonding pattern between the Cys 6/Asp 9 and the Trp 13/Thr 15 segments can be described as an antiparallel β-sheet with a bulge at Ser 8. On the opposite side, another antiparallel, isolated β-bridge is formed by residues Leu 2 in the ring and Asn 14 in the tail (Figure 3 c). In total, 13 intramolecular, partially solvent-shielded hydrogen bonds generate an extraordinarily stable three-dimensional arrangement. a) Structure of BI-32169. The carbon atoms of residues forming the ring are colored blue, the carbon atoms forming the tail are colored grey and the macrolactam bond is shown in yellow. b) Surface representation showing the molecular shape and surface polarity. c) Stereo representation of backbone hydrogen-bonding interactions of ring-forming residues with the peptide segment penetrating the ring. Structural information on eight lasso peptides (Figure 1) is currently available.1, 7, 9, 13–16 They have been classified as class I and class II lasso peptides; the latter is characterized by the absence of disulfide linkages.22 From this perspective, BI-32169 can be either considered as being of class I or as constituting a new class III, which exhibits a single disulfide linking a ring residue with a tail residue.16 All structures so far have been determined by NMR methods with peptides dissolved predominantly in organic solvents, such as DMSO or methanol, because the peptides are typically only sparingly soluble in aqueous solutions. The NMR solution structure of BI-32169 was determined by homonuclear two-dimensional NMR spectroscopic methods with the peptide dissolved in 100 % DMSO.16 Interestingly, the crystal and solution structures differ substantially in various aspects. Most importantly, while in the crystal structure the type I β-turn is formed by residues Asp 9/Ile/Pro/Gly 12; in the NMR structure the turn is made by residues Ile 10/Pro/Gly/Trp 13. Furthermore, the NMR structure reveals neither the 310-helix at the C terminus nor the unusual cis-peptide at Ser 8. The crystal structure provides strong evidence that the BI-32169 most probably adopts a very stable structure with little propensity for larger structural fluctuations of the peptide backbone. This is due to the presence of regular secondary structural elements stabilized by a tight network of solvent-shielded hydrogen bonds. Furthermore, in comparison to other cyclic peptides that display distinct conformations in the crystalline state, in solution and in receptor complexes, BI-32169 has additional conformational restrictions that are due to the presence of the disulfide link and the particular lasso arrangement that holds the tail threaded through the ring even for the reduced form of the peptide at high temperatures.16 Thus, the observed differences between the NMR and crystal structures are difficult to explain. Theoretically, they may be caused by a number of factors. Experimental conditions such as temperature of the structure determination or different solvents used, as well as crystal packing effects, may lead to the stabilization of distinct conformers in solution and in the crystalline state. A clearer picture may emerge once the NMR data are reviewed in the light of the crystal structure. The issue of structural integrity of the threaded tail in lasso peptides has been discussed intensely. For MccJ25, the tail is trapped within the ring by the presence of two aromatic residues above and below the macrolactam ring that keep the tail threaded even after thermolysin cleavage of the Phe 10/Val 11 peptide bond.23 In BI-32169, the rigid 310-helix at the C terminus, the disulfide bond between Cys 6 and Cys 19, the sterically demanding Trp 13 and Trp 17, and the highly stabilizing hydrogen-bonding structure are responsible for the stability of the lasso structure. BI-32169 takes on a saddle-shaped form with residues Ile 10 and Trp 17 sticking out of an otherwise extremely compact structure. The molecule exhibits an amphiphilic character with a cluster of hydrophobic residues dominating one face of the molecule, whereas the opposite face has polar character mainly due to the fact that the smaller side chains on this part of the molecular surface do not shield the backbone amide groups from solvent (Figure 3 b). The complex of incretin hormone peptides, which have a simple helical structure, with the extracellular domains (ECDs) of their cognate receptors,24, 25 most probably also including the glucagon–glucagon receptor complex, are formed by predominantly hydrophobic interactions. The amphiphilic character of both glucagon and BI-32169 suggest that the latter may act as a competitive antagonist by binding to an overlapping surface patch on the receptor ECD. In contrast to linear or regular cyclic peptide, lasso peptides identified to date are not synthetically accessible by solid-phase peptide synthesis. An interesting alternative avenue for derivatization of lasso peptides by an in vivo biosynthetic approach has been described for capistruin.26 The same technique may be applicable to other family members, such as BI-32169, and would allow the improvement of their functional and pharmacological properties. The first crystal structure of a lasso peptide described in this work unveils the atomic details of the lasso or protoknot fold and sheds light on the extraordinary tight network of interactions that render this structure so stable. It is conceivable that the other known, structurally stable lasso peptides are equally amenable to crystallization. With more high-resolution crystal structures at hand, it will be possible to gain deeper insight into the structural determinants of the lasso fold. BI-32169 was isolated from the culture broth of Streptomyces sp. (DSM 14996) as previously described.10 The compound was dissolved at 20 mg mL−1 in 50 % DMSO and submitted to crystallization trials using commercially available protein crystallization kits (Hampton Research; Aliso Viejo, CA, USA). Crystals were obtained by vapor diffusion by mixing compound solution with 20–50 % MPD, 0.1–0.5 M sodium phosphate or ammonium sulfate, 0.1 M Tris-HCl, pH 8.5. Crystals of a plate-like shape appeared overnight and grew to a size of ∼0.1 mm3. The crystals are colorless plates that exhibit extraordinary diffraction properties. Data were collected at 100 K from a single crystal on a Saturn 944 CCD (Rigaku, USA) mounted on a RU200 rotating anode with monochromatized CuKα radiation. BI-32169 crystallizes in space group P212121 with unit-cell dimensions of 19.47, 21.43 and 29.52 Å. Data were collected to 0.86 Å resolution, and scaled and merged with D*Trek27 (Table 1). Structure solution with SHELXD28 was straightforward. Due to the favorable data/parameter ratio that comes with the high resolution, structure refinement could be performed without any stereochemical restraints by full-matrix least squares optimization using SHELXL. The final model contains a very well-defined peptide (average thermal displacement factor Bav=5.5 Å2) and 18 water molecules (of the estimated 28 water molecules expected from packing density considerations). A residual cluster of difference electron-density peaks could represent a lattice site at which an MPD molecule is located; however, it was not modeled because it is probably mostly disordered. Space group P212121 Unit cell dimensions 19.47, 21.43, 29.52 Å Solvent content 18.7 % Resolution range 8.94–0.86 Å (0.90–0.86 Å) Total number of reflections 140 346 Number of unique reflections 19 602 Average redundancy 7.12 (2.52) Percent completeness 98.7 (87.2) Rmerge 0.076 (0.408) Output 16.0 (2.1) Number of non-H atoms 143 Number of solvent atoms 18 Total number of l.s. parameters 1443 R1 for 16 628 Fo>4 σ(Fo) 0.0938 R1 for all 19 602 data 0.1055 wR2 0.2729 GooF 1.027 Flack parameter 0.044 (0.03)