*2.3. Secondary Structure Content Assessment*

An assessment of secondary structure content was made by looking at the difference of the deviations from random conformation chemical shifts of the assigned Cα and Cβ resonances (∆δ <sup>13</sup>Cα − ∆δ <sup>13</sup>Cβ) [17]. To identify secondary structure elements using the individual carbon resonances, the chemical shifts are compared to the random coil chemical shift of the corresponding residue. A difference larger than ±0.7 ppm from the random coil chemical shift for several consecutive residues indicates the presence of secondary structure elements. Four consecutive downfield shifted Cα resonances beyond the 0.7 ppm threshold with respect to the random coil shift indicate α-helical structure, while three consecutive upfield shifted resonances in a row indicate β-strand presence. The opposite is true for Cβ resonances (downfield shift indicates β-strand, upfield shift indicates α-helix) [18]. The difference between the ∆δ <sup>13</sup>Cα and ∆δ <sup>13</sup>Cβ eliminates any possible chemical shift reference error on the individual deviations, with a positive ∆δ <sup>13</sup>Cα − ∆δ <sup>13</sup>Cβ difference indicating α-helix and a negative difference indicating β-strand. Here, a cumulative approach to identify secondary structure elements from the ∆δ <sup>13</sup>Cα − ∆δ <sup>13</sup>Cβ difference was employed by using an error threshold derived from the individual ± 0.7 ppm deviations of ∆δ <sup>13</sup>Cα and ∆δ <sup>13</sup>Cβ, i.e., p (0.7 <sup>2</sup> + 0.7 <sup>2</sup>) ∼ = 1 ppm. The results are illustrated in Figure 3, with the expected secondary structure elements highlighted in the figure. Overall, nine β-segments could be identified, a number consistent with the typical β-strand content of a canonical immunoglobulin variable domain, with a percentage of residues involved in β-strands of 49.6%. In comparison, Nb24 has a β-strand content of 50.4% when bound to antigen [11]. One possible α-helical tract was identified in the supposed CDR3 loop between residues 107 and 109.

**Figure 3.** The chemical shift indexing analysis (CSI), computed by taking the difference between the experimentally determined Cα chemical shifts and the Cα random coil chemical shift (∆δ <sup>13</sup>Cα) minus the difference between the experimentally determined Cβ chemical shift and the Cβ random coil chemical shift (∆δ <sup>13</sup>Cβ). Three negative ∆δ <sup>13</sup>Cα − ∆δ <sup>13</sup>Cβ values in a row indicate the presence of β-strand. A cumulative threshold error based on the individual ∆δ deviations of ±0.7 ppm, i.e., p (0.7<sup>2</sup> + 0.7 <sup>2</sup>) ∼ = 1 ppm, was used as a threshold to include only significantly varying consecutive negative values. Residues predicted to be in β−strands are highlighted in green in the graph. The chemical shift differences of Cys22 and Cys96 (highlighted by green hatched bars) deviate because of upfield shifts induced by aromatic sidechains. As a consequence, especially for the Cβ chemical shifts, typical values of the reduced cysteines were observed despite the presence of the disulfide bridge with the associated β structure content. Control CD spectra of oxidized and reduced Nb23 are reported in Supplementary Materials to illustrate the issue, showing that Cys22 and Cys96 form a disulfide bridge. Yellow blocks indicate the position of residues that were estimated to be in β−strands by TALOS-N.

For an alternative assessment of secondary structure content, TALOS-N [19] was also used to infer ϕ and ψ torsion angles of Nb23 sequence from its backbone and Cβ chemical shift assignments. Torsion angles are in turn characteristic for certain types of secondary structures. The secondary structure content obtained by TALOS-N assessment is also illustrated in Figure 3. Here β-strand content was also 50.4% (as for Nb24), marking a difference with the chemical shift indexing analysis.

Circular dichroism (CD) data collected for Nb23 and uploaded to the Beta Structure Selection (BeStSel) server, a CD data analysis server especially useful for identification of β structures [20], show that Nb23 is mainly composed of antiparallel β-strands with different twists. No α-helical segments were identified. The overall β-strand content of the structure was 55.2%, which is slightly exceeding the content from the chemical shift indexing and TALOS-N estimations. This is not surprising as BeStSel assessment also includes relaxed β-strands. The results from the BeStSel analysis can be found in the Supplementary Materials.
