4.2. NMR Assignments and Solution Structure

Protein samples were dissolved in 200 µL of aqueous buffer containing 20 mM Tris-HCl pH 7.2, 150 mM NaCl, and 0.1 mM PMSF and EDTA (5% D₂O for the lock) at a concentration of about 1 mM. Experiments were recorded at 20 °C on a Bruker AVANCE III 800 MHz (Bruker Biospin, Wissenbourg, France) equipped with a 5 mm Z-gradient TCI cryogenic probe head. ¹H chemical shifts were directly referenced to the methyl resonance of DSS, while ¹³C and ¹⁵N chemical shifts were referenced indirectly to the ¹³C/¹H and ¹⁵N/¹H absolute frequency ratios. All NMR experiments were processed with Gifa [29].

Backbone and Cβ resonance assignments were made using standard 3D triple-resonance HNCA, HNCACB, CBCA(CO)NH, HNCO, and HN(CA)CO experiments [30] and 3D [¹H,¹⁵N] NOESY-HSQC (mixing time 150 ms) and TOCSY-HSQC (isotropic mixing: 60 ms) experiments performed on the ¹⁵N,¹³C-labeled GIPC1-GH2 sample. [¹H,¹⁵N] NOESY-HSQC was used to extract the set of nOe’s restraints used for structure modeling, completed by restraints obtained from a 2D homonuclear NOESY (mixing time 200 ms) recorded on a deuterated buffer. NOE cross-peaks were assigned through automated NMR structure calculations with CYANA 3 [18]. Backbone φ, ψ, and side-chain χ¹ torsion angle constraints were obtained from a database search procedure on the basis of backbone (¹⁵N, HN, ¹³C’, ¹³Cα, Hα, ¹³Cβ) chemical shifts using TALOS-N [17]. Hydrogen bond restraints were derived from the analysis of residue (φ,ψ) values and CLEANEX-PM experiments [19,20]. When identified, the hydrogen bond was enforced using the following restraints: ranges of 1.8–2.0 Å for d(N-H,O), and 2.7–3.0 Å for d(N,O).

The final list of restraints, from which values that were redundant with the covalent geometry were eliminated, was used for structure modeling. A total of 200 three-dimensional structures were generated using the torsion angle dynamics protocol of CYANA 3 from 1219 NOEs, 88 hydrogen bonds, and 164 angular restraints. The 20 best structures (based on the final target penalty function values) were minimized with CNS 1.2 according to the RECOORD procedure [31] and analyzed with PROCHECK [32]. The rmsds were calculated with MOLMOL [33]. Models are displayed with PyMOL [34]. All statistics are given in Supplementary Materials, Table S1.