*2.2. NMR Spectroscopy Results and Chemical Shift Assignment Completeness*

The <sup>15</sup>N-1H HSQC spectrum of Nb23 is shown in Figure 1. The resonance spreading already appears quite satisfactory, and TROSY pulse schemes further enabled the resolution of certain overlapping peaks in the regular <sup>15</sup>N-1H HSQC. Apart from the two prolines which lack amide protons and excluding Met0 and the (His)<sup>6</sup> tag, amide connectivity assignments are missing for Gln1, Arg27, Thr28, Ser63, and Ser105, which include residues of the expectedly mobile CDR1 (Arg27 and Thr28) and CDR3 (Ser105) loops. The occurrence of conformational mobility at intermediate rate on the chemical shift scale

leading to signal broadening seems confirmed by the fact that neighboring residues in CDR1 and CDR3 (Gly26 and Gly102) exhibit below-average intensities and by the <sup>15</sup>N{1H} NOE data, where residues in conformationally rigid regions show a close-to-average ratio of peak intensity with and without hydrogen saturation (Figure 2). It is thus plausible that an unfavorable conformational exchange rate in the CDR regions could affect the detectability of some signal in <sup>15</sup>N-1H HSQC and TROSY spectra. On the other hand, the unassigned peaks other than sidechain resonances that were observed in the <sup>15</sup>N-1H HSQC or TROSY maps—namely three cross-peaks highlighted by blue boxes and letter labels in Figure 1—were addressed, but no conclusion could be achieved through the correlation patterns of the 3D triple resonance experiments acquired for backbone assignment, suggesting again that some slow conformational exchange occurring over the ms-to-µs time scale accelerates relaxation, thereby hindering the propagation of the coherence transfer pathway. The extent of population transfer from <sup>15</sup>N{1H} NOE data (Figure 2) enables, however, a tentative assignment. The negative heteronuclear NOE of boxed peak (a) is very likely to arise from Gln1. The close-to-average NOE value of boxed peak (c) could be consistent with the mobility expected at Ser63. Finally, the NOE value observed for boxed peak (b) suggests a possible attribution to Thr28, given the similar NOE value measured at Phe29. This dipolar-coupling-based assignment leaves only Arg27 (CDR1) and Ser105 (CDR3) without observable <sup>15</sup>N-1H connectivity signal that, in turn, corresponds to the signature of a conformational exchange process at the start of CDR1 and CDR3.

Typical TROSY-based 3D triple resonance spectra [14,15] (see Section 4) were used to assign the backbone and sidechain atoms. The sidechain assignment was arduous especially for residues with very long sidechains, due to the relaxation attenuation ensuing from many magnetization transfers combined with the relatively low sample concentrations, leading to noisy data with reduced intensity. The low sample concentrations were in turn due to poor protein solubility, at least for the particular sample conditions used here, and concentrations were further reduced by the subsequent protein precipitation occurring during the data acquisition.

The aromatic sidechain hydrogen atoms of Tyr, Phe, and Trp residues were assigned using the 2D experiments correlating the Hδ and Hε to the Cβ (2D CBHD and CBHE [16]) with samples in 100% D2O. The corresponding aromatic carbons were identified in the <sup>13</sup>C-1H HSQC. Due to extensive overlap of the aromatic carbon atoms in the spectra, only 32% of them could be assigned unambiguously.

The total percentages of chemical shifts assigned are reported in Table 1. Excluding Met0, the (His)<sup>6</sup> tag and two Pro residues, the backbone assignments (Cα, C', HN, N and Hα) were 95% complete, the sidechain residue assignments (including Cβ and Hβ) were 67% complete, and the aromatic residue assignments were 50% complete. Overall, the chemical shift assignment was achieved to an extent of 77%. The majority of the unassigned chemical shifts for both backbone and sidechain belong to residues of the CDR1 and CDR3 regions, which are expectedly less rigid than the remaining structure, thereby leading to inherently poor frequency spreading and/or broad line widths when unfavorable mobility rates are also involved. The completeness limits of the aromatic residue assignment could instead be totally ascribed to extensive resonance degeneracy from high mobility, for which characterization was mostly ambiguous and hence peaks unassignable, especially for carbons.

**Figure 1.** The <sup>15</sup>N−1H HSQC of Nb23 from a freshly prepared sample (247 µM in 19.5 mM bis-Tris and 21 mM NaCl). The good signal−to−noise of the spectrum allowed the application of a squared sine−bell shifted by π/6 to achieve complete resolution. Excluding Met0 and the C−terminal (His)<sup>6</sup> tag used for expression, five N−H connectivities could not be assigned (Gln1, Arg27, Thr28, Ser63, and Ser105). Only the three blue-boxed connectivities, labeled a, b, and c, out of those that were observed, could not be attributed through scalar correlation. A tentative assignment is proposed based on heteronuclear NOE (see main text). The central area highlighted with a box has been enlarged for better visualization (lower panel) to limit the assignment annotation crowding given the high density of peaks. The Asn and Gln sidechain carboxyamide pairs could be connected from the slow exchange cross−peak of 2D <sup>1</sup>H−1H NOESY, which also enabled the identification in a few cases from intra−residue NOE. The pairs are connected with blue dashed lines and the assigned ones are marked with an asterisk. The dispersion of peaks indicates a well−structured protein. The remaining peaks without labels belong to sidechain NHs, i.e., Arg, His, and Trp.

**Figure 2.** <sup>15</sup>N{ <sup>1</sup>H} NOE values, with I/I0 ratios representing the individual amide signal intensity with and without hydrogen saturation. The horizontal dotted line marks the average ratio value. Ratios below the average line indicate regions of mobility in the protein. The main regions of flexibility correspond to the supposed CDR1 (positions 26−31), a supposed loop between positions 42 and 45, and the supposed initial part of the CDR3 (positions 102−106). Residues with no bar correspond to either prolines (Pr041 and Pro88) or residues which were missing NH assignment. Based on the NOE values obtained for peaks (a), (b), and (c), that did not show scalar correlation in 3D spectra (Figure 1), a tentative assignment is proposed, respectively Gln1, Thr28, and Ser63, as indicated by the positions of the red bars.


**Table 1.** Chemical shift assignment completeness.
