The identification of insertions or deletions of secondary structure within a protein architecture depends upon the reference protein used for such comparison. The reference protein should ideally comprise the essential structural architecture, with no extraneous insertions or deletions beyond the basic folding and stability requirements. In the case of the β-trefoil, where extant naturally evolved proteins exhibit varying degree of C₃ rotational symmetry, the reference protein would ideally constitute a purely-symmetric architecture so that any asymmetric features in an evaluated protein can readily be identified. There are several de novo designed β-trefoil proteins having an exact threefold symmetric primary structure; including Threefoil (Broom et al., 2015), Mitsuba-1 (Terada et al., 2017), Phifoil (Longo et al., 2014) and the Symfoil family of proteins (Lee and Blaber, 2011; Lee et al., 2011). Threefoil was designed to have carbohydrate binding function and contains specific turn/loop secondary structure for this purpose. Similarly, Mitsuba-1 was designed to have a galactose binding site afforded by specific surface turn/loop secondary structure. In contrast, Symfoil was designed exclusively from the standpoint of optimized folding kinetics and thermodynamics, and is notably devoid of any specific functionality. Symfoil (using the Symfoil-4T variant) as a reference structure identifies five residue insertions within turns T2, T6 and T10 in Threefoil, and seven residue insertions of the same turns in Mitsuba-1 (Figure 2). Thus, the Symfoil protein was considered as the most appropriate reference protein with which to quantify secondary structure heterogeneity among β-trefoil proteins.

The RCSB structural databank (www.rcsb.org) was queried for β-trefoil proteins solved to better than 2.5 Å resolution. A total of 45 proteins were identified, representing 10 superfamilies, 17 families, and 45 domains, and with an average resolution of 1.81 ± 0.31 Å (Table 1). Only the de novo designed β-trefoil proteins exhibit an exact threefold rotational symmetry; all naturally-evolved β-trefoil proteins exhibit varying degrees of primary, secondary and tertiary structure symmetry.

Structural overlays of individual β-trefoil proteins onto the Symfoil protein coordinates (using the Symfoil-4T variant, RCSB 3O4B) were performed using the Swiss PDB Viewer software (Guex and Peitsch, 1997) and selecting for Cα atoms. An iterative fitting process was used to optimize the overlay. The number of matching Cα atoms was noted, as well as the rmsd for the fit (Table 1). This overlay was then examined for insertions or deletions in specific secondary structure elements as defined in the Symfoil structure (Figure 1). The percent of Cα matches per secondary structure element was also determined.

Sequence logo plots are a graphical representation of an amino acid (or nucleic acid) multiple sequence alignment (Schneider and Stephens, 1990; Crooks et al., 2004). Each logo consists of stacks of symbols, one stack for each position in the sequence. The height of symbols within a stack indicates the relative frequency of each amino at that position. A sequence logo plot was generated for β-strands S1, S5, and S9 as a group; similarly, S2, S6, and S10 as a group; S3, S7, and S11 as a group; and S4, S8, and S12 as a group (i.e., all sets of C₃ symmetry related strands, n = 126), for all representative β-trefoil proteins in Table 1 and using structural overlays as described above. Image generation utilized the web logo server at https://weblogo.berkeley.edu/ with colors based on chemical properties: polar amino acids (G,S,T,Y,C,Q,N) are green, basic (K,R,H) blue, acidic (D,E) red and hydrophobic (A,V,L,I,P,W,F,M) amino acids are black.

An analysis of the secondary structure length heterogeneity for the β-trefoil superfamily of proteins, compared to the Symfoil reference, shows that the heterogeneity is localized almost exclusively to turn secondary structure; indeed, all β-strands show a remarkable absence of relative insertion or deletion (i.e., all β-strands show a marked conservation of length (Figure 3). Furthermore, the heterogeneity in the turn regions principally involves insertions, as opposed to deletions, compared to the Symfoil reference protein. However, there are two notable exceptions to this general rule at turns T4 and T8, where some β-trefoils have limited deletions of up to three amino acids.

An analysis of the Cα structural conservation for regions of secondary structure in β-trefoil proteins, compared to the Symfoil-4T reference, shows that not only do β-strand regions show highly-conserved lengths, but that their overall conformation as β-strands is also highly-conserved (Figure 4). It can be seen that for the entire superfamily of β-trefoils a >90% structural conservation (i.e., <1.5 Å rmsd) is present with the symmetry-related sets of β-strands S1/S5/S9, S3/S7/S11, and S4/S8/S12. The S2/S6/S10 set exhibits 76–84% Cα structural conservation. Among turn secondary structure, turns T4 and T8 (which are symmetry-related) exhibit the least Cα structural conservation.

The Ricin B-like, Cytokine, and 30 K Lipoprotein superfamilies have the greatest number of members, with 22, 10, and 3 members, respectively (Table 1). The secondary structure length heterogeneity for these individual families is shown in Figure 5. This graph suggests that the general turn heterogeneity observed in the overall superfamily graph (Figure 3) is a composite of patterns of turn heterogeneity unique to the individual superfamilies or families. Thus, the Ricin B-like lectin superfamily exhibits the greatest turn heterogeneity (i.e., extensions) at T2, T3, T4, T6, and T10; while the Cytokine superfamily exhibits turn extensions principally at T3, T4, T7, T9, and T11; and the 30 K Lipoprotein superfamily exhibits turn extensions principally at T2, T6, and T10. Thus, each different superfamily exhibits characteristically different turn heterogeneity (i.e., extensions).

The sequence logo plots for the β-strand secondary structure exhibit characteristic patterns of hydrophobic residues (Figure 6). In β-strands S1/S5/S9 position #4 is principally hydrophobic: Ile and Leu account for 80% of all amino acids at this position, with the other residues being Phe, Tyr, Val and Met. There is some indication of hydrophobic preference at position #2, with Val and Phe accounting for approximately 40% of positions (and if Y is considered hydrophobic, then ∼50% of residues at position #2 are hydrophobic). In β-strands S2/S6/S10 positions #3 and #5 show a clear hydrophobic preference. Leu accounts for ∼50% of residues at position #3, with the majority of other residues being either Val, Ile, Phe or Trp. At position #5 Leu, Val, Ile, Ala and Met account for ∼66% of residues. In β-strands S3/S7/S11 Val, Leu and Ile account for ∼75% of residues at position #2. Ala, Leu, Val and Ile account for ∼50% of residues at position #4, with Gly another major residue at this position. In β-strands S4/S8/S12 there is a remarkable ∼70% preference of aromatic residues W or F at position #2 (with Leu, Val and Ile comprising the majority of the remainder). Hydrophobic residues are also preferred at position #4, with Ile, Leu, Phe, and Val comprising ∼60% of residues. Thus, in all β-strands there is a hydrophobic (P)/hydrophilic (H) pattern of H-P-H-P-H. Binary patterning of hydrophobic/hydrophilic amino acids is a key determinant of protein secondary structure, with an alternating hydrophobic/hydrophilic pattern favoring the formation of amphipathic β-strand secondary structure (West and Hecht, 1995; Xiong et al., 1995). These hydrophobic residues within the H-P-H-P-H patterning of the β-trefoil β-strands contribute to a highly-cooperative core packing group in the β-trefoil structure (Blaber, 2021).

Among the de novo designed symmetric β-trefoil proteins Symfoil is the most compact, principally due to the absence of specific functional surface turns/loops. Analyses of structural variations (i.e., insertions or deletions) of other β-trefoil proteins indicate that the vast majority of structural heterogeneity is associated with insertions in surface turn/loop regions in comparison to Symfoil. However, there is evidence of some β-trefoil proteins having relative truncations in the T4 and T8 regions (Figures 3, 5A). Specifically, 1FWV, 1ABR, 1KNL, 2ZQO, 1SZ6, 1XYF, and 1XEZ (all members of the Ricin B-like lectin superfamily, Table 1) have three amino acid deletions in both the T4 and T8 regions. These deletions effectively eliminate the hydrophobic residue at the #2 position in the S5 and S9 β-strands (which participate in the cooperative central core); thus, these truncations of the T4 and T8 turns may result in a less stable, or less cooperatively-folding, protein. The Symfoil protein therefore represents a “minimal” or “essential” β-trefoil architecture—one that is highly-conserved in the family of β-trefoil proteins—and is therefore a useful reference structure by which to characterize secondary structure heterogeneity in β-trefoil proteins.

The highly-conserved β-strands, and highly-divergent turn/loop regions, when comparing members of the β-trefoil superfamily, strongly suggests that functionality has its principle basis in turn/loop structure. For example, the specific heparin-binding functionality of FGF-1 (Cytokine superfamily) has been localized principally to an extension within the T11 region (Brych et al., 2004) while interaction with FGF receptor involves the T1, T4, and T8 regions (Olsen et al., 2004). Lectin functionality in the shellfish lectin MytiLec-1 and M. oreades mushroom lectin is localized to regions T2, T6, and T10 (Broom et al., 2015; Terada et al., 2017). The inhibitory function of Kunitz (STI) protease inhibitors is due to active site binding of an extended T4 loop region (Song and Suh, 1998). Ricin B-like lectin interactions involve the T2/T3 and T10/T11 regions (Suzuki et al., 2009). The Pmt2-MIR domain (superfamily MIR domain) interaction with tetraethylene glycol ligand involves regions T4 and T7 (Chiapparino et al., 2020). The interaction between LAG-1 (CSL) DNA-binding protein and DNA ligand principally involves the T1 region (Friedmann et al., 2008). The interaction between Agglutinin and T-disaccharide involves the T6 and T10 region (Transue et al., 1997). The interaction between clitocypin and cathepsin V involves the T1 and T3 regions (Renko et al., 2010). This representative summary of binding interactions provides strong support for a primary assignment of functionality to specific and structurally-heterogenous turn/loop regions in β-trefoil proteins.

Symmetric relationships among turn/loop structures in β-trefoils appears most apparent within the symmetry-related set of T2/T6/T10 turn positions. There are β-trefoil proteins having relative insertions of n = +1 (1UPS), n = +5 (3PG0), n = +6 (4IY9), n = +7 (5XG5), and n = +8 (1T9F) amino acids, relative to the Symfoil (i.e., 3O4B) reference structure. Additionally, similar examples exist having no relative insertions (i.e., n = 0; 1Q1U/1IHK/2FDB/1IJT/1BFG/1RG8) as well as n = −1 deletions (1WD3) (Figure 7). The most parsimonious explanation for such structural conservation of these symmetry-related turns is for duplication/fusion events to occur subsequent to trefoil motif structural evolution. This implies the likelihood of multiple independent instances of the evolution of β-trefoil proteins from simpler (i.e., trefoil-fold) motifs, and supports the evolutionary hypothesis put forth by Meiering (Broom et al., 2012) that the emergence of β-trefoil proteins is a recurring and ongoing evolutionary mechanism.

In the simplest example of duplication and fusion of individual trefoil-motifs leading ultimately to formation of a β-trefoil protein, the junction of gene fusion is the T4 turn region (Ponting and Russell, 2000; Lee and Blaber, 2011; Lee et al., 2011). Thus, the β-trefoil architecture contains two symmetry-related turns T4 and T8, with the “third” member of this symmetrically-related set defined by the adjacent (but discontinuous) N- and C- termini (see Figure 1). As with the T2/T6/T10 turns, a number of β-trefoil proteins exhibit a unique structural symmetry when comparing the T4 and T8 turns (e.g., 1FWV, 1ABR, 1KNL, 2ZQO, 1SZ6, 1XYF, 1XEZ; as described above). This implies that this turn formed prior to the duplication and fusion event that yielded the mature β-trefoil architecture. However, this results in a structural conundrum. The existence of a T4 region results from the fusion of two trefoil motifs. Two such turns (i.e., T4 and T8) would be generated by a subsequent tandem duplication of such a construct; however, this would yield a total of four sequential trefoil motifs. The apparent solution to the presence of an “extra” trefoil motif is for the latter fusion to include a truncation event affecting one trefoil motif (Jeltsch, 1999; Peisajovich et al., 2006; Longo et al., 2013).

In addition to providing a potential functional role, turns also serve to connect adjacent β-strand secondary structure (forming a β-hairpin), minimizing the entropic penalty of association, and thereby influencing stability and folding (Nagi et al., 1999; Thompson and Eisenberg, 1999; Lindberg et al., 2006). The reaction coordinate of cooperative protein folding typically describes a highly-polarized transition state or folding nucleus (Abkevich et al., 1994; Went and Jackson, 2005; Faísca, 2009). Establishment of this folding nucleus is the rate limiting step in folding, and once formed, serves to rapidly condense formation of the overall native structure. An isolated 42-mer trefoil motif (i.e., “Monofoil”) derived from the Symfoil protein spontaneously oligomerizes to yield an intact β-trefoil architecture (Lee and Blaber, 2011; Lee et al., 2011); thus, a serviceable folding nucleus resides within each repeating motif in the Symfoil protein (Blaber, 2020; Parker et al., 2021). However, phi-value analysis (Fersht and Sato, 2004) indicates that the effective folding nucleus in the Symfoil protein, and the related fibroblast-growth factor-1 β-trefoil protein, while not identical, are both centrally-located and more expansive than an individual trefoil motif (Longo et al., 2014; Xia et al., 2016). This more expansive central definition includes turns T4 and T8, which are novel turn structures generated by the fusion of trefoil motif repeats. These novel turns are postulated to promote local β-hairpin interactions, thereby generating a more efficient folding nucleus compared to an isolated trefoil motif. However, destabilizing mutations targeting the folding nucleus region of Symfoil indicate that the C₃ symmetry provides for alternative folding nuclei in other regions of the protein able to salvage foldability (Longo et al., 2013; Tenorio et al., 2020). The survey of turn region lengths in the β-trefoil superfamily indicates that the central region comprises turns having generally the shortest lengths (Figure 3). Thus, central turns may be somewhat “privileged” regions of secondary structure where considerations of efficient folding nucleus formation impact the optimal turn length and sequence design.

The secondary structure elements of the fundamental β-trefoil are limited to β-strand and reverse turn, and thus describe a comparatively simple protein architecture. Knowledge essential for the de novo design of β-trefoil proteins is extensive: 1) The β-strand secondary structure is the key determinant of the conserved basic architecture for this protein superfamily; 2) Conserved β-strand characteristics have been elucidated as regards length and hydrophobic patterning; and 3) The role of β-strand hydrophobic residues in cooperative core-packing interactions has been well-characterized. In this regard, it is interesting to note the different independent solutions for the set of hydrophobic core-packing residues (referencing Figure 6) utilized by the de novo designed symmetric β-trefoil proteins Symfoil [3O4B; generated through top-down symmetric deconstruction of FGF-1 (Lee and Blaber, 2011; Lee et al., 2011)], Phifoil [4O4W; generated by folding nucleus symmetric expansion of FGF-1 (Longo et al., 2014)], Threefoil [3PG0; generated by consensus sequence of a carbohydrate-binding ricin sequence (Broom et al., 2012)], and Mitsuba-1 [5XG5; generated by computational sequence constraint of the shellfish lectin MytiLec-1 (Terada et al., 2017)]. The sequence logo plot for this set of core-packing residues (Figure 8) suggests that, as long as the appropriate hydrophobic patterning and compatible van der Waals interactions are satisfied, a variety of alternative core-packing arrangements are permissible, thereby indicating a lowered threshold for successful design.

The general attributes of the folding nucleus for Symfoil have been identified, and the potential for redundant folding nuclei demonstrated. Evolutionary considerations indicate highly-permissive design pathways of foldability involving diverse fusion/truncation of trefoil motifs. Turn regions have been identified as the key regions of structural variability, and are the principle determinants of ligand functionality characteristic of this superfamily. As connectors of adjacent β-strand secondary structure, turn regions also influence the entropic penalty for the assembly of local β-hairpin structure, and this plays an important role in the formation of the folding nucleus.

Protein design must simultaneously solve at least three different problems: 1) protein foldability (i.e., folding kinetics requirements), 2) protein stability (i.e., thermodynamic requirements), and 3) the accommodation of specific function (with potential structural dynamics requirements). Analysis of the β-trefoil architecture suggests that it is readily amenable to a two-step design process, with the initial step focusing upon the design of a foldable, stable “scaffold” (and many avenues appear possible); subsequently followed by a second step of functional mutation. The present analysis indicates that the first step involves β-strand secondary structure and key hydrophobic patterning design (building upon current extensive knowledge in this area). The C₃ symmetry substantially reduces the combinatorial search of appropriate primary structure solutions. The second step focuses upon turn/loop regions and their mutation to generate desired functionality (the β-trefoil architecture perhaps best suited to ligand functionality). This second step is less-well characterized and therefore open to expansive and novel opportunities. The C₃ symmetry provides for monovalent or multivalent ligand binding opportunities. In an alternative approach, if specific loop regions are associated with unique functional properties, and the β-strands as structural elements, then diverse chimeras with novel combined structure/function attributes might be constructed using computational approaches (Ferruz et al., 2021). Overall, the β-trefoil architecture has many attractive features for de novo protein design, applied especially to ligand functionality. The adoption of heparin-binding functionality into a benign β-trefoil scaffold using the principles described herein has recently been demonstrated (Tenorio et al., Forthcoming 2022).