Additional Supporting Information may be found in the online version of this article.


Figure S1. Illustration of the geometric hashing algorithm in RosettaMatch used to design a three-residue zinc binding sites on existing protein scaffolds. Illustration derived from Clarke and Yuan, 1995 (reference 32 in the main article).

Figure S2. Ideal geometry for zinc coordination by histidine and cysteine. Coordination geometry is defined by coordination bond lengths, angles about zinc, angles about the ligating atom, and dihedral angles that put the zinc in the same plane as the histidine ring. Standard deviations are based on statistics we obtained from 1705 four-coordinated zinc binding sites in structures deposited in the Protein Data Bank. Harmonic constraints for distance, angle, and dihedral values are used to evaluate the quality of zinc-binding geometry, and the score is divided by the number of coordinating residues. Since there are four geometric features to consider, a score less than 4 would reflect a metal site that on average is within standard deviation.

Figure S3. Depictions of four experimentally tested designs of zinc-mediated ubiquitin binding proteins. Top left: 2D4X scaffold, design 1 (Spelter). Top right: 2D4X scaffold, design 2. Bottom left: 2FZ4 scaffold, design 1. Bottom right: 2ONU scaffold, design 1.

Figure S4. Comparison of interaction motifs featuring a hydrophobic helix targeting the hydrophobic surface patch of ubiquitin. Our designed interaction (white) does not resemble native interactions with ubiquitin.

Figure S5. Zinc binding data for the Spelter design. Zinc binding as seen in the 2D4X-CCH design is retained in the Spelter design with additional mutagenesis for ubiquitin binding. A) Circular dichroism thermal melts monitoring ellipticity at 222 nm wavelength. Mutation of both cysteine residues abolishes zinc binding. B) Isothermal titration calorimetry: Spelter was present in the sample chamber at 20 μM, ZnSO4 at 250 μM was injected in 2 μl increments. A fit to the titration curve gives values of Kd = 9.5 nM and N = 1.1 (left). Mutation of both cysteine residues abolishes zinc binding (right).

Figure S6. The designed metal site is specific for zinc. Addition of zinc results in a 4 °C increase in Tm, while addition of other metals results in <1 °C increase in Tm.

Figure S7. Fluorescence polarization binding curves, see Table II of main text. A) Tryptophan point mutations weaken binding affinity 2-fold in the absence of zinc. In the presence of zinc, W199A weakens affinity but W203A does not. B) Spelter binding of Ubq-H68A did not fit a one-site binding model, but the affinity is tighter than Spelter binding of Ubq.

Figure S8. The 3-D NMR experiment HNCACB used to assign our construct of ubiquitin containing an N-terminal glycine-serine extension and in buffer containing 20 mM MOPS pH 6.9, 50 mM NaCl, 0.5 mM TCEP. Residues 63 through 74 are shown for illustration purposes. Red peaks represent CB atoms, black peaks represent CA atoms. In each strip, the larger peak is from residue i (relative to the amide group defining the strip) and the smaller peak is from residue i - 1. Smaller peaks are matched with larger peaks to walk forward or backward in the protein sequence. Chemical shift values determined from this experiment are listed in Table SII.

Figure S9. Possible mechanism by which zinc improved Spelter binding to ubiquitin-H68A. A) Zinc dependence of Ubq-H68A binding is lost with the Spelter-E136A mutation. B) Zinc binding may have preoriented the loop to favor hydrogen bonding and electrostatic attraction between residue E136 on Spelter and residue K6 on ubiquitin.

Figure S10. The backside H68 delta nitrogen has a positive SASA value (solvent accessible surface area) using a 1.4 Å water probe, suggesting that it is solvent-exposed. It was not flagged as a buried-unsatisfied polar atom. However, it is partially buried by L8 in the computational model, which would result in a desolvation penalty. This may negate the energetic benefit of a His-zinc coordination bond and could explain why the H68A ubiquitin mutant bound more tightly that wild-type ubiquitin. Desolvation cost is captured in the Rosetta energy function using an implicit solvent model that penalizes burial of polar residues, but this penalty may be too lenient. Accurately capturing solvent effects remains an outstanding challenge in protein modeling.

Figure S11. Spelter binds ubiquitin with higher affinity than many naturally occurring ubiquitin-binding interactions.

Equation S1. Digital resolution of proton, nitrogen, and carbon in HSQC chemical shift perturbation studies.

Table S1. Experimentally tested zinc-mediated ubiquitin binders

Table S2. HNCACB chemical shift assignments of ubiquitin. This ubiquitin construct had an N-terminal glycine-serine extension and the buffer was 20 mM MOPS pH 6.9, 50 mM NaCl, 0.5 mM TCEP.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.