Design of a gold clustering site in an engineered apo-ferritin cage

The study addresses the challenge of atomically precise design and regulation of protein-protected gold nanoclusters. The authors engineered a specific gold clustering site within the 4-fold symmetric axis channel of an apo-ferritin (horse liver L-chain) cage by introducing mutations (R168H/L169C). The goal was to control Au(I) accumulation and cluster formation inside a biocompatible protein scaffold and to elucidate the sequence of metal uptake, the roles of key residues, and how precursor concentration tunes cluster nuclearity and structure. The work is significant for designing atomically precise metal clusters in proteins for catalysis and bio-nanotechnology, providing structural insight into controlled multinuclear Au(I) cluster formation within a confined, symmetric protein channel.

Prior work has demonstrated the importance of atomically precise gold nanoclusters in catalysis and biomedicine, and protein-directed synthesis often yields emissive clusters, typically containing Au(0). Ferritin is a versatile protein nanocage with well-characterized ion channels (notably the 3-fold pores) that facilitate metal ion entry and transport. Earlier studies observed Au binding at ferritin interior sites (e.g., Cys/Met/His motifs) and even nucleation of gold sub-nanoclusters within ferritin, as well as engineered metal coordination (e.g., cadmium) at the 4-fold channel. The 3-fold channel has been implicated experimentally and theoretically in metal entry and transfer to interior binding/catalytic sites. Typical Au coordination in proteins involves soft S donors (Cys/Met) supported by His, consistent with HSAB principles, and aurophilic Au–Au interactions often stabilize multinuclear Au(I) assemblies. These foundations motivate the present design of an Au clustering site at the 4-fold channel to achieve tunable, symmetric multinuclear Au(I) clusters.

• Protein engineering and purification: Constructed apo-R168H/L169C mutant of recombinant horse liver ferritin L-chain (rHLFr) via inverse PCR; expressed in NovaBlue cells; purification by heat treatment followed by ion-exchange (Q Sepharose HP) and size-exclusion (S-300) chromatography.
• Preparation of Au–ferritin composites: Mixed 5 mL of 5 µM purified protein with specified equivalents (10–400 equiv.) of chloro(dimethylsulfide)gold(I) (Me2SAuCl) from 10 mM ACN stock; stirred at 20 °C for 2 h; dialyzed against 50 mM Tris-HCl pH 8.0 overnight; further purified by SEC (G-200/G-25); stored at 4 °C.
• Stability/optical measurements: UV–vis spectroscopy; EEM fluorescence spectroscopy to assess photoluminescence.
• Quantification of Au loading: ICP-MS for Au and BCA assay for protein to determine Au atoms per ferritin cage at varying equivalents.
• XPS: High-resolution XPS (Thermo K-Alpha, Al Kα) to determine Au oxidation state and analyze C 1s components; C 1s (284.8 eV) used for charge correction.
• Crystallization and X-ray diffraction: Hanging-drop vapor diffusion with 0.5–1 M (NH4)2SO4 and 10–20 mM CdSO4; data collected at 100 K at SPring-8 BL45XU or Tsinghua (Rigaku XtaLAB Synergy) with HyPix-6000; data processed with HKL2000/CrysAlisPro; space group F432.
• Structure solution and refinement: Molecular replacement (MOLREP) using apo-rHLFr WT (PDB 1DAT); scaling (AIMLESS), refinement (REFMAC5) and model building (COOT). Metals assigned from anomalous difference maps (cutoff ~4σ) at multiple wavelengths (1.15 Å, 1.00 Å) to distinguish Au from Cd. Water placement and occupancy/B-factor refinement performed with attention to Fo–Fc features; side chains with insufficient density truncated/omitted as noted.
• Electron-density analyses: Anomalous and 2Fo–Fc maps examined to assign Au positions; occupancies and B-factors tabulated; residual Fo–Fc densities in 4-fold channel noted but unassigned due to lack of anomalous signal and unusual shapes (possible water or weakly bound Au).
• Computational chemistry: DFT calculations (Gaussian 16) using PBE0 with D3-BJ dispersion; geometry optimizations with 6-311G(d) for C,N,S,H and SDD ECP for Au; Au positions and Cα atoms of coordinating residues fixed; frequency calculations for thermal corrections to Gibbs free energy; single-point energies with def2-TZVP and SMD solvation plus D3-BJ. Wavefunction analyses (Multiwfn): ADCH charges, multi-center bond order (MCBO), localized orbital locator (LOL), and NBO/NAO orbital composition for selected coordination sites. Boltzmann analysis of conformer populations for four-Au models.

• Successful engineering of a gold clustering site at the 4-fold symmetric axis channel of apo-ferritin (R168H/L169C), yielding up to 12 Au atom positions per channel.
• Quantitative Au loading (ICP-MS/BCA, Au atoms per cage): 10 equiv: 4±2; 50 equiv: 46±10; 100 equiv: 87±8; 200 equiv: 158±25; 400 equiv: 203±38.
• XPS confirms Au remains in +1 oxidation state across all loadings; Au 4f7/2 ~84.30–84.72 eV and 4f5/2 ~88.12–88.39 eV with slight right shift attributed to Au(I)–His/Cys coordination.
• Crystallography: Apo-R168H/L169C resolved at 1.5 Å; Au composites at 1.9 Å; cages remain structurally conserved (Cα RMSD 0.203–0.314 Å vs. mutant apo). Total assigned Au atoms from occupancies: 50 equiv 51.6; 100 equiv 98.4; 200 equiv 102.0; 400 equiv 140.0 (lower than ICP-MS at higher equivalents due to low-occupancy, non-specific Au not visible crystallographically).
• Four Au binding sites identified: 3-fold site (Cys126/His114/Glu130, Cl−), Cys48 site (Cys48/His49/Glu45/Arg52), Met96 site (Met96/His147; only at 400 equiv), and designed 4-fold site (R168H/L169C). Binding motif commonly (Cys/Met)–Au–His.
• Binding order with increasing precursor: 3-fold site first (10 equiv), then 4-fold and Cys48 (by 50 equiv), finally Met96 (400 equiv). Order: 3-fold > 4-fold = Cys48 » Met96.
• Stepwise accumulation and residue dynamics: At Cys48, progression from mono- to Cys-bridged dinuclear Au with His49 involvement; Glu45 and Arg52 show conformational changes aiding accumulation. At 3-fold, transient low-occupancy Au observed at 50–100 equiv suggest intermediates stabilized by aurophilic interactions; stable Cys126-bridged dinuclear Au forms by 200–400 equiv, assisted by His114 and Cl−.
• 4-fold channel Au cluster: 8 Au positions at 50 equiv; 12 positions at 100–400 equiv; average Au–Au 2.6–3.4 Å indicating aurophilic interactions. R168H exhibits concentration-dependent coordination/dynamics, stabilizing Au at higher loadings and driving cluster rearrangement at 400 equiv. Extremely short modeled Au–Au distances (e.g., 1.81 Å between Au2 and Au3) indicate mutually exclusive occupancies (not coexisting simultaneously).
• Occupancy trends indicate binding affinity at 4-fold site: Au1 > Au2 > Au3. At ≥200 equiv, rearrangement shifts electron density from center to edge; R168H reorients to coordinate Au.
• QC wavefunction analysis (Au12 model from 200 equiv structure): ADCH charges positive on Au1/Au3 and near neutral/negative on central Au2; strong multi-center interactions identified among Au2 quartet and between Au1/Au2 and L169C sulfur (normalized MCBO > 0.4); LOL maps support multi-center delocalization in cluster and coordination to R168H/L169C.
• NBO/NAO analyses indicate Au–N (His) bonding mainly from Au 5d and N 2p; Au–S bonding from Au 5d/6s and S 3p, consistent with soft-soft interactions and His as supportive donor.
• No photoluminescence observed for Au(I) clusters (UV–vis/EEM), consistent with non-emissive Au(I) states lacking Au(0).

The engineered R168H/L169C mutations create a confined, symmetric pocket at the 4-fold channel that promotes accumulation and organization of Au(I) into multinuclear clusters stabilized by S (Cys) and N (His) donors and aurophilic interactions. Structural snapshots across precursor concentrations reveal the sequence of Au uptake and redistribution: Au enters via the 3-fold channel, initially binds at the 3-fold site, then migrates to interior sites including the designed 4-fold channel and Cys48; only at high loading does the Met96 site bind. Within the 4-fold channel, Au nuclearity and topology are tunable by precursor concentration, with 8 positions at low loading and 12 at higher loadings, and a distinct rearrangement at 400 equiv driven by changes in metal–ligand equilibria and the dynamic side-chain behavior of R168H. Fractional occupancies and short Au–Au distances show that certain positions are mutually exclusive, consistent with a dynamic cluster rather than a single static configuration. QC calculations corroborate multi-center bonding and electron delocalization among Au atoms and between Au and protein ligands, explaining the stability of the observed clusters and the residual electron density features. Overall, the findings validate the design strategy: a symmetric protein channel can be engineered to template and regulate multinuclear Au(I) cluster formation with atomic-level control via precursor dosing and residue coordination dynamics.

The authors engineered a gold clustering site at the 4-fold symmetric channel of apo-ferritin (R168H/L169C), revealing by crystallography and quantum chemical analysis a unique, tunable multinuclear Au(I) cluster with up to 12 Au positions. They elucidated a stepwise Au accumulation pathway into the cage, identified key residues (e.g., His49, Glu45, Arg52, Glu130, R168H, L169C) and their conformational dynamics, and determined the binding order of four Au sites (3-fold > 4-fold = Cys48 » Met96). Increasing Au precursor concentration controls cluster nuclearity and induces structural rearrangement mediated by R168H dynamics. This work demonstrates a generalizable strategy for designing precise metal clusters within symmetric protein scaffolds and informs the design of N,S-donor ligands for controlled Au cluster assembly. Future work could target catalytic function, redox-state modulation (e.g., controlled Au(0) incorporation), kinetic control of cluster interconversion, and extension to other metals or mixed-metal clusters in engineered protein channels.

• Exact counts of Au atoms in the 4-fold channel cannot be unambiguously assigned due to fractional occupancies and mutually exclusive positions; some very low-occupancy, non-specific Au ions are likely invisible to crystallography.
• Residual Fo–Fc densities near the 4-fold channel remain unassigned (possible water ligands or weakly bound Au), introducing uncertainty in complete cluster mapping.
• Certain side chains (e.g., Cys169 at 400 equiv, Gln158, Lys172/C-termini in some structures) were not modeled due to insufficient density, limiting detailed coordination assignments.
• Crystallographic totals of Au are lower than ICP-MS/BCA at higher equivalents, reflecting detection limits for low-occupancy sites.
• The clusters are dynamic; observed structures represent ensemble averages under crystallization conditions and may shift with environment or solution conditions.
• No photoluminescence was observed; while consistent with Au(I), it limits immediate optical applications without further redox manipulation.