2.1. Synthesis of linear precursors and oxidative folding

Peptides were assembled on Fmoc-Gln preloaded TentaGel resin (Rapp Polymere) using standard Fmoc-SPPS chemistry on a fully automated microwave CEM Liberty peptide synthesizer (see Supplementary Material¹). Cleavage from the solid support was performed using 94%(v/v) TFA, 2%(v/v) H₂O, 2%(w/v) DTT, 1%(v/v) triethylsilane, 1%(v/v) anisole. After precipitation of the crude peptide in methyl-tert-butyl ether (MTBE), the pellet was washed twice with MTBE, dried, dissolved in a mixture of acetonitrile and water [1:4(v:v)], lyophilized, and finally analyzed by RP-HPLC and LC-MS (see §2.2). Oxidative folding was performed as reported previously (Avrutina et al., 2005) followed by RP-HPLC purification.

2.2. RP-HPLC and LC-MS analysis

Analytical RP-HPLC was performed with a Varian LC 920 system equipped with a Phenomenex Synergi 4µ Hydro-RP 80 Å (250 × 4.6 mm, 4 µm) column (see Supplementary Material) at a flow rate of 1 ml min^-1. Semi-preparative RP-HPLC was performed with a Varian LC 940 system equipped with an Axia-packed Phenomenex Luna C18 (250 × 21.2 mm, 5 µm, 100 Å) column using linear acetonitrile gradients (Supplementary Material) with a flow rate of 18 ml min^-1.

LC-MS was performed with a Shimadzu LC-MS 2020 equipped with a Phenomenex Jupiter C4 (50 × 1 mm, 5 µm, 300 Å) column using linear acetonitrile gradients with a flow rate of 0.2 min^-1 (Supplementary Material).

2.3. Crystallization and data collection

Trypsin from bovine pancreas (Sigma-Aldrich) was dissolved in 1 mM HCl pH 2.0, 10 mM CaCl₂. The protein was purified on a Superdex 75 16/60 column (GE Healthcare) in 25 mM MES pH 5.5, 50 mM NaCl, 10 mM CaCl₂ at 277 K and was concentrated to 12 mg ml^-1. Prior to crystallization setup, lyophilized SOTI-III was added to 12 mg ml^-1 trypsin (in 25 mM MES pH 5.5, 50 mM NaCl, 10 mM CaCl₂) to a final concentration of 2.5 mM and kept on ice for 30 min. Initial crystallization screening was performed in Intelli-Plate 96-3 plates (Art Robbins) using a HoneyBee 961 dispensing robot (Digilab Genomic Solutions). Initial crystals were obtained with the JCSG+ screen (NeXtal; Qiagen) in condition E12 [tube 60; 0.1 M imidazole pH 8.0, 10%(w/v) PEG 8000]. Diffraction-quality crystals were obtained after several days at 292 K by mixing equal volumes of protein-complex solution with precipitant solution [0.1 M imidazole pH 7.5, 12%(w/v) PEG 8000] by hanging-drop vapour-diffusion crystallization in 15-well plates (EasyXtal; Qiagen) with a reservoir volume of 500 µl. Prior to flash-cooling in liquid nitrogen, the crystals (>0.8 × 0.1 mm) were cryoprotected with 25%(w/v) glycerol.

For SOTI-III F14A-trypsin complex crystals, lyophilized miniprotein was added to 11.5 mg ml^-1 trypsin to a final concentration of 1.5 mM and incubated for 30 min on ice. The best crystals were obtained directly in condition B3 of the initial JCSG+ screen [tube 15; 0.1 M Bicine pH 9, 20%(w/v) PEG 6000; NeXtal, Qiagen] with equal volumes of complex protein and precipitant solution. Crystals grew at 292 K after several days and were cryoprotected with 33%(w/v) glycerol prior to flash-cooling in liquid nitrogen. X-ray diffraction data were collected on BESSY synchrotron beamline MX-14.1. Data were indexed, integrated and scaled with the XDS/XSCALE package (Kabsch, 2010). Phases were obtained with Phaser (McCoy et al., 2007) using trypsin coordinates (PDB entry 2xtt ; Wahlgren et al., 2011) as a search model. The coordinates were refined with phenix.refine (Adams et al., 2004) and manually checked and corrected with Coot (Emsley & Cowtan, 2004). SOTI-III coordinates were manually built with Coot. The refined SOTI-III-trypsin complex structure was used as a molecular-replacement model for the SOTI-III F14A-trypsin complex. Figures and structure alignments were prepared using PyMOL (Schrödinger). Data and refinement statistics are summarized in Table 1.

Table 1 X-ray data-collection and refinement statistics Values in parentheses are for the highest resolution shell.   SOTI-III wild-type -trypsin complex SOTI-III F14A-trypsin complex Data collection   Space group P21 P21   Unit-cell parameters (Å, °) a = 49.4, b = 66.8, c = 108.9, = = 90.0, = 90.02 a = 48.5, b = 68.4, c = 109.8, = = 90.0, = 93.3   Resolution (Å) 32.0-1.7 (1.8-1.7) 19.9-2.0 (2.1-2.0)   Rmerge (%) 8.1 (49.1) 14.5 (45.0)   <I/(I)> 11.66 (2.85) 8.27 (2.90)   Completeness (%) 97.3 (93.7) 98.7 (98.5)   Multiplicity 3.8 (3.8) 3.2 (3.2)   Matthews coefficient VM (Å3 Da-1) 2.04 2.02   No. of molecules in unit cell (Z) 6 6 Refinement   No. of reflections 75888 48006   Rwork/Rfree (%) 17.9/22.1 16.7/22.0   B factors (Å2)     Protein 16.7 11.0     SOTI-III 21.0 16.2   R.m.s. deviations     Bond lengths (Å) 0.010 0.003     Bond angles (°) 1.30 0.83   Ramachandran plot (%)     Residues in favoured regions 97.2 97.1     Outliers 0 0   MolProbity score/percentile 1.53/90th 1.49/97th   PDB code 4aor 4aoq

2.4. Enzymatic methods

Bovine pancreatic trypsin (Sigma-Aldrich) was active-site-titrated (Chase & Shaw, 1967; Jameson et al., 1973) using 4-methylumbelliferyl guanidinobenzoate (MUGB; 1 µM; Sigma-Aldrich) in reaction buffer [50 mM Tris-HCl, 150 mM NaCl, 0.01%(v/v) Triton X-100, 0.01%(w/v) sodium azide, pH 7.6].

The concentration of inhibitory active SOTI was determined assuming a 1:1 interaction of enzyme and inhibitor (Avrutina et al., 2005; Pohlig et al., 1996). Several inhibitor concentrations (0.05-1.5 µM) were incubated with trypsin (0.2 µM) for 30 min at room temperature. Subsequently, the residual activity was quantified over 5 min by monitoring the relative fluorescence of the hydrolyzed carbobenzoxy-L-arginine-7-amino-4-methylcoumarin (75 µM; Sigma-Aldrich). The reactions were carried out in 96-well microtiter plates (Greiner Bio-One, flat bottom, black) and were monitored in a Tecan GENios reader (excitation wavelength 360 nm, emission wavelength 465 nm). The normalized reaction rates were fitted using (1), with unfixed E₀. The ratio of active sites (see above) and the fitted E₀ gives a correction factor for the active inhibitor concentration.

To determine the inhibition constants, the reaction velocity was monitored after a 30 min pre-incubation of enzyme and inhibitor. The hydrolysis of the substrate Boc-QAR-pNA (250 µM; Bachem) was analyzed over a period of 30 min at 405 nm (Tecan GENios). The apparent inhibition constant (K_i^app) was calculated by fitting (SigmaPlot 11; Marquard-Levenberg algorithm) the relative reaction velocities using the Morrison equation (1) (Morrison, 1969),

All measurements were carried out in triplicate. The substrate-independent inhibition constants (K_i) were calculated from K_i^app and K_m of the enzyme using (2),

The Michaelis-Menten constant K_m for the substrate and trypsin had been determined previously (Tischler et al., 2012).

3.1. SPPS and oxidative folding

SOTI-III and the F14A variant were synthesized by microwave-assisted solid-phase peptide synthesis (SPPS). The purity of the linear precursors was verified by HPLC (Fig. 2) and LC-ESI-MS analysis (Supplementary Fig. S1). Oxidative folding of the crude peptide was conducted in an aqueous buffer system and was monitored by analytical HPLC (Fig. 2) and LC-ESI-MS (Supplementary Fig. S1). After the final semi-preparative HPLC purification, total yields of 23.8% and 16.2% were obtained for SOTI-III and for the F14A variant, respectively.

Figure 2 Oxidative folding of SOTI-III variants and inhibition curves of bovine pancreatic trypsin. (a) HPLC traces of wild-type linear precursor (1) and folded knottin (2) at 220 nm. (b) HPLC traces of F14A linear precursor (3) and folded knottin (4) at 220 nm. (c) The inhibition curves of SOTI-III (triangles) and SOTI-III F14A (squares) were obtained by fitting data to (1), as described in §2. Error bars were calculated from three independent experiments.

3.2. Structural studies

The crystals of the complex of purified SOTI-III with bovine pancreatic trypsin had P2₁ symmetry and diffracted to 1.7 Å resolution. The structure shows additional F_o - F_c electron density for the inhibitor covering the active site of trypsin (Supplementary Fig. S2). The asymmetric unit is packed with three trypsin molecules, each blocked by the inhibitor (Fig. 3a). The SOTI-III secondary structure shows the proposed C4-C21, C11-C25 and C20-C36 disulfide linkages (Figs. 3b and 4a; Kowalska et al., 2007). While the first cystine bridge is shaped as a right-handed hook, the second and last bridges form a right-handed and a left-handed spiral, respectively (Supplementary Table S1). The overall structure of the inhibitor is mainly constructed of -turns (Fig. 4 and Supplementary Table S2). The turns are interrupted by disulfide bonds, three short antiparallel -strands (residues 9-11, 26-28 and 33-36) and a short four-amino-acid-long helix (residues 16-19). The inhibitor loop (residues 26-34), which contains the S₁ site filled by Arg32 (P₁ residue), forms a -hairpin with a -turn (Figs. 3c and 4 and Supplementary Table S2b).

Figure 3 Structure of SOTI-III in complex with pancreatic trypsin. (a) Trimeric complex of trypsin (surface representation) with SOTI-III (cartoon representation) in the asymmetric unit. (b) SOTI-III binding pocket. Cysteine residues are colored green. Disulfide bonds are labelled with residue numbers. (c) Close-up of the SOTI-III inhibitor loop (red) blocking the catalytic triad (blue). Arg32 is coordinated in the S1 pocket in a substrate-like fashion (Schechter & Berger, 1967). Dashed lines indicate hydrogen bonds. The closest hydrogen bonds of catalytic Ser200 to the inhibitor backbone are labelled in red. (d) Close-up of the subsite where Phe14 of SOTI-III is coordinated. Tyr154 of trypsin (purple) points away from residue 14.

Figure 4 Secondary structure and topology of SOTI-III. (a) Secondary-structure elements of SOTI-III. The four loops (mainly -turns) are labelled L1-L4. Green arrows indicate the three -strands. The short helix 1 is colored purple. Disulfide bonds are drawn as black lines. The cysteine residues involved are labelled I-VI. The red U-shaped line symbolizes the -hairpin with a -turn of the inhibitor loop L4. (b) Topology of SOTI-III. Residue numbers for boundaries of secondary-structure elements are shown. The coloring of secondary-structure elements is the same as in (a).

SOTI-III coordinates at the active site of each trypsin molecule without causing major dislocation of the protease loops. Here, the average r.m.s.d. of 0.227 ± 0.026 Å for all three trypsin-inhibitor complexes in the asymmetric unit is in the same range as the r.m.s.d. of each trypsin-inhibitor complex when compared with apo trypsin (r.m.s.d. = 0.259 ± 0.008 Å; Supplementary Fig. S3). SOTI-III is most flexible in the regions of loop 1, loop 3 and a three-residue helix, as observed in a B-factor putty representation of all three monomers (Supplementary Fig. S4).

On average, 15 residues of the inhibitor (750 Å²) interact with 24 residues of trypsin covering 980 Å² of the protease surface (Supplementary Table S3). A total of 12 hydrogen bonds are involved in coordination of the inhibitor. Notably, the P₁ residue Arg32 forms eight of these interacting hydrogen bonds (Supplementary Fig. S5).

Phe14 of SOTI-III binds to a cleft apart from the active site, which is formed by Ile78, Trp144 and Pro155 of trypsin (Fig. 3d). Interestingly, a detailed search in the Protein Data Bank showed that this cleft is not addressed by other trypsin inhibitors (Fig. 5; Supplementary Material).

Figure 5 Comparison of the coordination of SOTI-III with those of other peptide trypsin inhibitors. (a) Superposition of SOTI-III with other canonical inhibitors. Trypsin is shown in surface representation. SOTI-III (chain A) is colored purple, STI lime (PDB entry 1avw ; Song & Suh, 1998), SFTI-I black (PDB entry 1sfi ; Luckett et al., 1999), CPTI-II blue (PDB entry 2btc ; Helland et al., 1999) and BPTI orange (PDB entry 1f7z ; Pasternak et al., 2001). Inhibitors were superposed based on trypsin main-chain atoms. (b) Close-up of the active-site residues and the inhibitor loop with arginine or lysine P1 residues. The coloring of the inhibitors is as in (a). Active-site residues are shown as sticks, with C, N and O atoms in white, blue and red, respectively. (c) Superposition with the SOTI-III complex of 99 deposited trypsin inhibitor structures that are close to residues 12-15 of SOTI-III in chain D (<3 Å; see Supplementary Material). SOTI-III is shown in purple and trypsin is shown in white. Superposed trypsin molecules (blue lines) align well, while the superposed inhibitors (gray lines) do not align or address the Phe14 binding site of SOTI-III. Tyr45 of trypsin in the SOTI-III complex has a different rotamer conformation. (d) Close-up of the inhibitor loop region in (c) showing the P1 residue Arg32 (red box). The inhibitors superpose well with the SOTI-III inhibitor in this region.

To investigate the role of Phe14 upon trypsin binding and inhibition, a F14A variant of SOTI-III was synthesized and cocrystallized. The complex with F14A SOTI-III, which crystallized in the same P2₁ space group, shows that in monomer C of trypsin the residual space which was formerly filled by the side chain of Phe14 (Fig. 3d) is partly occupied by the protease residue Tyr154 (Supplementary Fig. S6). In one of the remaining two monomers the electron density is rather weak for the Tyr154 side chains, suggesting that the side chain might swap between those two rotamers.

The K_i values of wild-type SOTI-III and of the F14A variant were determined as described in §2. They were found to inhibit trypsin in the nanomolar range. The substrate-independent inhibition constants for SOTI-III and the F14A variant were calculated to be 60.9 ± 8.4 and 201.8 ± 27.7 nM, respectively, from the inhibition curves (Fig. 2c). This indicates that Phe14 contributes to trypsin binding. However, the main inhibitory effect is mediated by the C-terminal inhibitor loop. This is supported by the fact that more than half of the inhibitor-coordinating hydrogen bonds are formed by Arg32 within this loop region. This is not surprising, since other P₁ residues have also been reported to contribute strongly to inhibitor binding (Krowarsch et al., 2005).

3.3. Comparison studies with other trypsin inhibitors

The structure of SOTI-III was previously modelled in silico (http://knottin.cbs.cnrs.fr/ ). Although the calculated structure is similar to that of the enzyme cocrystallized with trypsin, significant conformational differences are present in the C-terminal inhibitory loop region (residues 25-35). The average r.m.s.d. of the superposed structure is 2.33 Å, with a largest r.m.s.d. of over 8 Å in the inhibitor loop region (Supplementary Fig. S7).

SOTI-III exhibits the unique, but well defined, pseudo-knot motif typical of members of the `knottin' family (Figs. 3b and 4). This structural element determines the stability and rigidity, similar to the majority of known canonical protease inhibitors (squash, Bowman-Birk, grasshopper, hirustasin, chelonianin and Ascaris inhibitors) that lack a hydrophobic core and extensive secondary structure. It contains a canonical inhibitor loop located at the C-terminus, thereby defining it as a member of the huge family of canonical serine protease inhibitors.

The C-terminal location of the inhibitor loop is a unique feature of SOTI-III. Owing to the amino-acid sequence, the additional interaction of Phe14 and the location of the inhibitor loop, we support the notion of Kowalska et al. (2007) that SOTI-III together with other Mirabilis and Spinacia serine proteinase inhibitors form a new family of inhibitors.