Covalent bicyclization of protein complexes yields durable quaternary structures

Summary

Proteins are essential biomolecules and central to biotechnological applications. In many cases, assembly into higher-order structures is a prerequisite for protein function. Under conditions relevant for applications, protein integrity is often challenged, resulting in disassembly, aggregation, and loss of function. The stabilization of quaternary structure has proven challenging, particularly for trimeric and higher-order complexes, given the complexity of involved inter- and intramolecular interaction networks. Here, we describe the chemical bicyclization of homotrimeric protein complexes, thereby increasing protein resistance toward thermal and chemical stress. This approach involves the structure-based selection of cross-linking sites, their variation to cysteine, and a subsequent reaction with a triselectrophilic agent to form a protein assembly with bicyclic topology. Besides overall increased stability, we observe resistance toward aggregation and greatly prolonged shelf life. This bicyclization strategy gives rise to unprecedented protein chain topologies and can enable new biotechnological and biomedical applications.

Graphical abstract

Highlights

• •Triselectrophiles cross-link homotrimeric proteins, forming bicyclic topologs
• •Bicyclic enzyme exhibits activity under denaturing conditions
• •Approach is rapidly applicable to a structurally diverse range of complexes

The bigger picture

The cyclization of small molecules, peptides, and macromolecules is a fundamental strategy to design precise three-dimensional molecular shapes tailored to chemical function. Proteins, however, typically consist of linear polypeptide strands, where protein folding and assembly are governed by non-covalent interactions. Protein engineering to form precise interlinked and cyclized protein chains has the potential to expand the biotechnological applications of proteins beyond their natural environments. Here, we construct a covalent bicyclic enzyme, capable of withstanding an exceptional degree of chemical and thermal stress. Such molecules can offer new avenues for sustainable industrial chemistry and biocatalysis, functioning in chemical environments that were previously inaccessible with biological components.

Modulating the stability of protein complexes is crucial for their use in biotechnological applications, such as biocatalysis, offering more sustainable alternatives to traditional chemical transformations. Protein stabilization is typically a labor-intensive process, reliant on extensive screening. Here, we apply a structure-based approach, which utilizes three-pronged reactive cross-linkers to chemically modify and tether a diverse range of trimeric protein complexes. The resulting abiological and cyclic chain connectivity offers robust enzymatic activity and complex stability under challenging thermal and chemical conditions.

Introduction

Proteins represent a versatile molecular platform, essential to life on earth and a rich source of inspiration for the development of novel biotechnological tools.^1,2,3,4,5 Self-association of proteins into homomeric complexes is widespread, facilitating an evolutionarily relevant increase in morphological complexity.⁶ However, applications such as catalysis in abiotic environments often involve extreme conditions that challenge the structural integrity of proteins and disrupt complex assemblies. Although considerable efforts have enhanced the thermal and chemical stability of protein folds (tertiary structure),^7,8,9,10,11 the stabilization of protein complexes and their supramolecular assembly (quaternary structure) remains a challenge. Such approaches have mainly focused on homodimeric protein complexes. For example, homodimeric isologous interfaces (in which each interaction surface is identical)¹² have been stabilized or artificially engineered via systematic mutagenesis¹³ and computational approaches.^14,15 The introduction of inter-molecular disulfide bridges or non-canonical amino acids, and the design of fusion proteins, have been utilized to covalently link monomer units within dimeric complexes.^{16,17,18,19,20} In contrast, many higher-order complexes, such as homotrimers, possess heterologous interfaces (i.e., where the interacting surfaces differ), which present a bigger challenge for engineering due to the inherent necessity for more mutations and the risk of rearrangement into polymeric aggregates.²¹ General approaches toward the efficient stabilization of these complexes are currently lacking.

Bioconjugation strategies have emerged as powerful techniques to enhance and alter protein functionality, providing access to powerful labeling tools, screening platforms, and novel therapeutic entities, such as antibody-drug conjugates.^22,23,24 The use of bioconjugation approaches for protein structure stabilization is uncommon, and has focused on either tertiary structure stabilization^25,26,27 or inter-chain cross-linking to trap weak complexes and probe for unknown protein-protein interactions.^28,29,30 Herein, we aimed to employ biocompatible, multivalent modifiers for the stabilization of protein complexes. We have previously reported the in situ cyclization of proteins (INCYPRO) for the stabilization of tertiary structures,^26,27,31 which utilizes C3-symmetric triselectrophiles^32,33,34 that react with three spatially aligned cysteine side chains on the protein surface. Given the C3 symmetry of protein homotrimers, we considered whether it is possible to cross-link homotrimeric complexes with these triselectrophiles, thereby constructing protein conjugates with high tolerance toward thermal and chemical stress. Our approach aimed to fundamentally alter the chain topology of trimeric complexes, converting three protein monomers into a single, covalently interlinked chain with bicyclic topology. This enabled the construction of a stress-resistant enzyme capable of catalysis under denaturing conditions. To further challenge the concept, we explored a suite of structurally diverse trimeric complexes, successfully generating a range of stabilized trimeric complexes and obtaining four additional precisely bicyclized protein trimers with substantially increased thermal stability.

Results and discussion

Construction of covalently cross-linked protein trimers

We initially selected Pseudomonas fluorescens esterase (PFE) as a target for complex stabilization. PFE is a well-characterized homotrimeric hydrolase, which has served previously as a platform for protein engineering efforts.^35,36 Based on an earlier reported crystal structure of PFE (Protein Data Bank; PDB: 1va4),³⁷ we searched for cross-linking sites on the two faces of the trimer (Figure 1A) to enable covalent linkage by a triselectrophile (Figure 1B). Given the C3-symmetry of the PFE complex, the variation of a single residue results in three equivalent modified sites per trimer. We considered suitable Cα–Cα distances of 8–15 Å (Figure S1) between modified residues and identified two solvent exposed residues (T3 and Q174) close to the central C3 axis as potential cross-linking sites (Figure 1A). Cysteine residues were introduced to the wild-type enzyme (denoted p1) at these two positions, furnishing two variants for mono-cross-linking (p2, T3C; p3: Q174C) and one for bis-cross-linking, combining both variations (p4, Figure 1A). Notably, the latter generates an interlinked bicyclic topology upon modification. The selected variation sites are distant from the catalytic pocket and therefore not expected to directly interfere with enzymatic activity (Figure S1). Each PFE monomer bears one native cysteine which is, however, buried and not accessible for electrophilic modification.

Figure 1.

Covalent multimerization and bicyclization

(A) Crystal structure of Pseudomonas fluorescens esterase (p1, PDB: 1va4), including modification sites (red, T3, and Q174) with amino acid variations and corresponding protein (and complex) names. For sequences and design details see Figure S1.

(B) Cysteine-selective cross-linking using chloroacetamide or iodoacetamide-based cross-linkers (Ta-Cl₃, Ta-I₃).

(C) Scheme of trimer chain topology. Mass spectra (ESI-TOF) of cross-linked protein trimers including the charge state of the most abundant signal (see Tables S2–S5 for complete lists of MS signals).

The chemical cross-linking of the PFE variants was initially pursued with the chloroacetamide bearing triselectrophile N,N′,N″-((1,3,5-triazinane-1,3,5-triyl)tris(2-oxoethane-2,1-diyl))tris(2-chloroacetamide) (Ta-Cl₃) (Figure 1B) and analyzed using sodium dodecyl sulfate (SDS) polyacrylamide gel electrophoresis (PAGE). We observed only limited reactivity, however (Figure S2), particularly for variant p2 (T3C), failing to form detectable levels of product. To address this issue, we envisioned the use of a more reactive triselectrophilic cross-linker, retaining the triazinane core but replacing chloroacetamide with the more reactive and cysteine-specific iodoacetamide. The corresponding cross-linker N,N′,N″-((1,3,5-triazinane-1,3,5-triyl)tris(2-oxoethane-2,1-diyl))tris(2-iodoacetamide) (Ta-I₃) (Figure 1B), was obtained via the Finkelstein reaction, employing Ta-Cl₃ and NaI (see experimental procedures).

SDS-PAGE indicated the formation of a defined covalently linked multimeric species for each of the variants (p2, p3, and p4) upon incubation with Ta-I₃ (Figure S3). Interestingly, all cross-linked variants showed a slower mobility than expected for their molecular weight (MW = 90 kDa, apparent MW = 100–150 kDa, Figure S3), which we suspected to be due to their unusual chain topology (Figure 1C). The terminally linked p2₃Ta exhibits an extended star-like topology, whereas p3₃Ta is cross-linked via a central cysteine (C174), resulting in a branched topology. Trimer p4₃Ta₂ is cross-linked by two entities and therefore forms a compact bicyclic structure. To further analyze the reaction products, high-resolution mass spectrometry (MS) was performed confirming the formation of the expected covalently linked trimers: mono-cross-linked p2₃Ta and p3₃Ta, as well as bis-cross-linked p4₃Ta₂ (Table S1).

We subsequently performed time-of-flight (TOF) MS under denaturing conditions. The resulting spectra revealed a distinct distribution of charged states for each covalent protein trimer (Figure 1C; Tables S2–S4). The N-terminally linked, star-shaped p2₃Ta experiences extensive protonation (z(max) = 91, ca. 30 charges per monomer), with an average protonation per monomer comparable to the three unmodified variants (ca. 37 charges, Table S5). Branched p3₃Ta exhibits reduced protonation levels (z(max) = 61, ca. 20 charges per monomer). This trend continues for the presumably most compact bicyclized trimer p4₃Ta₂ (Figure S4), resulting in a considerably lower number of charges (z(max) = 42, ca. 14 charges per monomer). Analogous behavior is observed in SDS-PAGE (Figure S3) where the three covalent trimers show subtle differences in mobility (migration p4₃Ta₂ > p3₃Ta > p2₃Ta), again suggesting the highest degree of compactness for bicyclic p4₃Ta₂ in the unfolded state.

Robust catalysis by triselectrophile-linked enzyme complexes

We next investigated the impact of cross-linking on thermal stability using differential scanning fluorimetry (DSF), revealing enhanced stability for all cross-linked variants when compared with their unmodified precursors (Figure 2A). The most pronounced improvement was observed for bicyclic p4₃Ta₂ (ΔT_i = 8.4°C). In addition, we tested protein resistance toward chemical stress, employing guanidine hydrochloride (GuHCl) as a chaotropic denaturant. For all proteins, we observed decreasing T_i values with increasing GuHCl concentrations in line with the expected increased unfolding propensity under increasingly chaotropic conditions (Figure S5; Table S6). The sensitivity toward GuHCl, however, varies considerably between the different topologies. The wild-type protein p1 and the non-cross-linked variants p2, p3, and p4 exhibit strong sensitivity toward GuHCl with the absence of a defined melting curve at 1.5 M GuHCl (Figures 2B and S5). Cross-linking increases thermal stability in presence of GuHCl in all three cases (Figures 2C and S5), with the most stable cross-linked trimer p4₃Ta₂ retaining a cooperative unfolding profile up to 1.5 M GuHCl.

Figure 2.

Cross-linking increases protein stability and enzymatic activity under stress

(A) Melting temperatures (T_i) of unreacted and cross-linked variants.

(B and C) Thermal denaturation profiles of p1 and p4₃Ta₂, respectively, in the presence of varying concentrations of guanidine hydrochloride (GuHCl). For thermal denaturation curves of p2, p3, p4, p2₃Ta, and p3₃Ta see Figure S5.

(D) Relative initial rates of enzymatic reaction for p1 and the three covalent trimers (relative to initial rate of p4₃Ta₂ at the given GuHCl concentration). The absolute value for each initial rate is provided (for details see Table S7). The errors account for 1σ (n = 3).

(E and F) Reaction course of substrate conversion by p1 and p4₃Ta₂ in the presence of varying concentrations of GuHCl. For activity curves of p2, p4, p4, p2₃Ta, and p3₃Ta see Figure S6. All activity measurements were performed with 2 nM PFE trimer (50 mM HEPES, 50 mM NaCl, pH 7.5). The errors account for 1σ (n = 3).

Given the resistance of the covalently linked trimers to chemical denaturation, we tested if this translates into increased enzymatic activity under these conditions. PFE esterase activity was analyzed by monitoring the conversion of p-nitrophenyl acetate to p-nitrophenolate (p-NP, Figure S6).³⁸ In the absence of denaturant, all proteins show comparable activity (Figure 2D) indicating that enzyme activity has not been affected by cysteine variation and cross-linking. As expected, all enzymes experience a reduction in activity with increasing concentrations of GuHCl (Figure 2D; Table S7). However, cross-linking considerably enhanced resistance to the denaturant, with highest residual activities observed for bicyclic PFE trimer p4₃Ta₂ (Figure 2D). Although unmodified p1 only facilitates product formation up to 1 M GuHCl (Figure 2E), p4₃Ta₂ remains active in the presence of up to 2 M GuHCl (Figure 2F).

Bicyclic p4₃Ta₂ resists aggregation

Bicyclic trimer p4₃Ta₂ experienced the largest stabilizing effect, surpassing both single-cross-linked trimers (p2₃Ta and p3₃Ta). To further investigate this pronounced stabilization, p4₃Ta₂ was studied in more detail. Circular dichroism (CD) spectra under native conditions revealed a spectrum analogous to the wild-type enzyme p1, suggesting an unchanged overall fold (Figure 3A). Assessing the thermal denaturation behavior via CD (λ = 220 nm, Figure 3B), melting temperatures were obtained (p1 T_m = 69.5°C; p4₃Ta₂ T_m = 77.0°C), which are in the range of above DSF results (Figure 2A). The slightly higher difference suggested a time dependence of stability, with a slower temperature ramp rate applied for CD compared with DSF measurements. We noted the complete absence of CD signal at high temperature (Figure S7) and the distinctive spike in the associated high tension (HT) values (Figure 3B, overlaid), both suggesting that the predominant inactivation pathway involves protein aggregation and subsequent precipitation.³⁹

Figure 3.

Bicyclic p4₃Ta₂ resists aggregation

(A) Circular dichroism (CD) spectra of p1 and p4₃Ta₂ measured at 20°C at a 4 μM trimer concentration (in 50 mM potassium chloride, 50 mM potassium phosphate, pH 8).

(B) Thermal denaturation CD profiles of p1 (T_m = 69.5°C) and p4₃Ta₂ (T_m = 77.0°C) measured at 220 nm. HT values suggest precipitation close to the midpoint temperature (T_m derived from the maximum of the CD first derivative curve as presented in Figure S7).

(C) Top: PFE trimer structure (PDB: 1va4) with approximate distances. Bottom: temperature-dependent size distribution derived from DLS experiment for p1 (gray) and p4₃Ta₂ (orange). Measured at a 15 μM trimer concentration (in 50 mM HEPES, 50 mM NaCl, pH 8). The errors account for 1σ (n = 6). Above 69°C, it was not possible to reliably fit size distribution data for p1. For p4₃Ta₂ data at higher temperatures, see Figure S8.

To compare the aggregation behaviors of p1 and p4₃Ta₂, dynamic light scattering (DLS) measurements were performed, confirming for both the presence of homogeneous, well-dispersed trimers at ambient temperatures (up to T = 40°C, Figure 3C). Notably, the observed radius (r ∼ 4 nm) corresponds to the size of the PFE trimer in the crystal structure (Figure 3C). For p1 (gray), temperature-dependent DLS measurements indicated the appearance of higher-order oligomeric structures around 45°C (Figure S8). Initially, species with a radius of ca. 30 nm (T = 50°C) and 60 nm (T = 60°C) were detected. At 69°C, we mainly observed the presence of aggregates >1 μm, which corresponds to the CD-derived melting temperature of p1 (T_m = 69.5°C). Above 69°C, it was not possible to reliably fit size distribution data for p1, most likely due to severe precipitation.

In comparison, bicyclic p4₃Ta₂ (orange) remained broadly monodisperse up to 65°C (Figure S8) suggesting a delay of aggregation onset by ca. 20°C. At 69°C, p4₃Ta₂ shows species with increased size around 30 nm, analogous to p1 at 50°C (Figure 3C). Eventually, p4₃Ta₂ also showed additional higher-order oligomeric states (60 nm at 70°C and >1 μm at 74°C, Figure S8). To further assess the effect of aggregation on long-term stability, we tested storage at 50°C, a temperature below the melting temperatures of both proteins (p1 T_m = 69.5°C; p4₃Ta₂ T_m = 77.0°C). After 24 h incubation at 50°C, both enzymes retained their activity (Figure S9). After longer storage (more than 1 day), however, p1 lost activity, whereas p4₃Ta₂ kept its activity for multiple weeks. At this temperature, unmodified p1 already showed the formation of oligomeric structures (Figure 3C), which could lead to aggregation and loss of enzymatic activity over longer time periods.

Crystal structure of bicyclic p4₃Ta₂

In pursuit of a crystal structure of p4₃Ta₂, conditions were screened to obtain suitable crystals for X-ray diffraction analysis. A dataset was collected, determining the crystal structure of p4₃Ta₂ at a resolution of 2.5 Å (PDB: 8pi1, space group C121, Table S8) following molecular replacement with the PFE trimer as a search model (derived from PDB: 1va4). The obtained structure harbors five protomers of p4₃Ta₂ per asymmetric unit (Figure 4A). Between both introduced cross-linking sites (C3 and C174) additional well-defined electron density was observed that matches the molecular structure of the cross-link (Figure 4B), thereby validating the expected site-specific modification and bicyclization of the p4 trimer.

Figure 4.

Crystal structure of bicyclic p4₃Ta₂

(A) The asymmetric unit contains five p4₃Ta₂ protomers (space group: C121, for details see Table S8).

(B) 2Fo−Fc electron density map around Ta cross-linked via cysteines C3 and C174, respectively, in the three monomers in p4₃Ta₂.

(C) Overlayed crystal structures of p1 (gray, PDB: 1va4) and p4₃Ta₂ (orange, backbone atom RMSD: 0.18 Å).

(D) Surface representation of p4₃Ta₂ showing protein subunits (orange) and Ta cross-links (red).

Overall, the crystal structure of p4₃Ta₂ (orange) aligns closely with the previously reported structure of wild-type PFE (p1, gray, Figure 4C, backbone atom root-mean-square deviation [RMSD]: 0.18 Å), demonstrating minimal perturbation of the macromolecular structure despite the modification of six residues within the cross-linked trimer. On one face of the trimer (top view), the cross-link connects the three N-terminal cysteines (C3) located in a short β strand. On the other side (bottom view), the cross-linked central cysteines (C174) are located at the beginning of an α helix (Figure 4C). Interestingly, at both sites, the cross-link is located within the cavity formed at the junction of the three protein monomers (Figure 4D).

Bicyclization of structurally diverse homotrimeric proteins

The bicyclization of trimeric PFE provides a unique architecture that has not been reported for protein complexes before. To assess the applicability of this approach to other homotrimeric proteins, we searched the RCSB PDB for C3-symmetric homotrimers with 50–400 residues per monomer and identified ca. 4,000 structures. To simplify protein production and characterization, this pool was reduced by selecting for proteins with a hexahistidine tag, at least one tryptophan, and confirmed expression in E. coli. In addition, we allowed a maximum of one cysteine residue per monomer to minimize off-target reactivity. These filter criteria produced an initial set of 119 trimers (Table S9), of which 18 structurally diverse protein complexes spanning 14 unique folds, as classified by their class, architecture, topology, and homologous superfamily (CATH) domain identities,⁴⁰ were selected after visual inspection. To enable bicyclization, two residues per monomer were substituted for cysteine, with distances of 8–15 Å between Cα atoms of the resulting cysteine triplet, producing a total of 18 designs (Tables S10 and S11). Protein variants were expressed in BL21(DE3) cells in an IPTG inducible pET28(+) vector, retaining the purification tags from the original PDB sequences. Expression and purification were successful for 17 of the 18 variants (>0.5 mg protein from 1.5 L culture).

Cross-linking reactions with Ta-I₃ were performed in analogy to the bicyclization of p4. Initially, stabilization effects were analyzed by thermal denaturation experiments using DSF to compare T_i values prior to and after cross-linking. Of the 17 obtained proteins, 10 showed a substantial increase in thermostability upon Ta-I₃ treatment (ΔT_i > 5°C, Figures S10 and S11) with 7 of them surpassing the stabilization effects observed for PFE (ΔT_i > 10°C). The ten proteins that showed clearly increased thermal stability upon cross-linker treatment were analyzed by high-performance liquid chromatography (HPLC) coupled with MS to assess their cross-linking status. For five, we did not obtain interpretable spectra, which may suggest inhomogeneity or incompleteness of the cross-linking reaction or could be a result of the inherent technical difficulties of MS analysis of large, interlinked protein chains. For the remaining five variants, we confirmed covalent trimer formation (Figures S12–S16). One of these, however, only reacted with one cross-linker (Figure S12). For the other four complexes (Figure 5), we confirmed the formation of the bis-crosslinked trimers (monomer₃Ta₂, Figures S13–S16), additionally validated by SDS-PAGE (Figure S17).

Figure 5.

Stabilization of different homotrimeric proteins via bicyclization

(A) Characterization of l4₃Ta₂ (Ip1913, PDB: 3fnj).

(B) Characterization of b4₃Ta₂ (BH3489, PDB: 1vmf).

(C) Characterization of a4₃Ta₂ (Au4130, PDB: 3c6v).

(D) Characterization of e4₃Ta₂ (ECHD, PDB: 5c9g). Reported crystal structures and cysteine variation sites with average distances (d_avg) between Cα atoms of cross-linking residues are shown. Calculated and found molecular weights (MWs) and CD-derived thermal denaturation curves of unmodified (black) and corresponding cross-linked trimers (orange) are provided, including the resulting difference in melting temperatures (ΔT_m). T_m values (Table S12) represent the maximum of the first derivative (corresponding HT curves and CD spectra are available in Figure S18).

The four confirmed bicyclic proteins originate from a diverse set of structural classes with MWs ranging from 34 to 91 kDa, and distances between introduced cross-linking sites ranging from 10 to 15 Å (Figure 5). To further validate their thermal stability, CD thermal denaturation experiments were performed providing increases in melting temperatures (ΔT_m = 6°C–39°C) in line with DSF-derived values. The smallest bicyclic trimer, l4₃Ta₂ (MW = 34 kDa, Figure 5A), was derived from lp_1913,⁴¹ a rhodanese-like domain from Lactobacillus plantarum, and revealed a considerably increased melting temperature (ΔT_m = 19°C). Bicyclization of the BH3498⁴² variant b4 resulted in b4₃Ta₂ (MW = 50 kDa, Figure 5B) experiencing the smallest stabilization effect (ΔT_m = 6°C) among the four tested variants. The heterologous interfaces of this highly compact trimeric structure are large relative to the size of the trimer, which may explain the moderate stability enhancement upon cross-linking. Thermal denaturation of l4₃Ta₂ and b4₃Ta₂, as well as their monomeric precursors (I4 and b4), produced an unexpected reduction in ellipticity at 220 nm; however, this wavelength remained the most effective way of assaying unfolding by CD. As the third example, bicyclic trimer a4₃Ta₂ (Figure 5C) was derived from Au4130,⁴³ an uncharacterized enzyme from Aspergillus fumigatus containing a tautomerase-3 domain. It showed the most pronounced thermo-stabilization effect upon cross-linking (ΔT_m = 39°C). Notably, the un-cross-linked protein itself was the least stable of the tested targets (T_m = 41°C). The fourth bicyclic trimer e4₃Ta₂ (Figure 5D), originated from ECHD, a putative enoyl-coenzyme A (coA) hydratase variant,⁴⁴ shares the most structural similarity with PFE among the four additional proteins. However, the sequence similarity is minimal (22% sequence identity), and their structural domains are non-homologous. Bicyclization of e4 again resulted in a substantially increased thermal stability (ΔT _m = 17°C).

Conclusions

We report the bicyclization and stabilization of trimeric proteins, which have been considered particularly difficult targets for stabilization efforts. In the case of PFE, bicyclization enhanced enzyme activity under chemical stress and reduced the formation of higher-order aggregates, as well as precipitation. This suggests that the cross-links tether the quaternary structure, thereby preventing exposure of the hydrophobic interior, which, in turn, results in a reduced tendency to nucleate initial, soluble aggregation states. This may also be supported by the polar nature of the cross-link conveying additional solubility. Most notably, the increased thermal stability and solubility of p4₃Ta₂ results in an exceptional longevity with full enzyme activity after more than 3 weeks of storage in buffer at 50°C.

Crucially, we show the bicyclization strategy to be applicable across diverse structural domains. The protein conjugates we present here constitute an abiological topology and were obtained by converting protein complexes into quasi-tertiary structures with augmented connectivity. This topology results in increased compactness in the denatured state and may thereby mimic an intermediate state in the protein folding trajectory. Such systems can therefore shed new light on protein folding and its modulation.^2,45,46,47 Considering the crucial role of proteins in emergent biotechnological and biomedical applications, the reported chemically modified complexes constitute an attractive modality for applications in sustainable biocatalysis, protein therapy, and diagnostics.

Experimental procedures

Further experimental procedures can be found in the supplemental information.

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Tom Grossmann (t.n.grossmann@vu.nl).

Materials availability

All materials generated in this study are available from the lead contact with a completed materials transfer agreement and reasonable compensation by the requestor for its production, processing, and shipping.

Synthesis of Ta-I₃

A round bottom flask was charged with 100 mg (0.206 mmol, 1 equiv) of Ta-Cl₃ (obtained as earlier reported)²⁷ and suspended in 2.1 mL acetone. Sodium iodine 559 mg (3.70 mmol, 18 equiv) was added, and the reaction mixture was stirred vigorously at room temperature for 16 h. 2 mL of water was added, and the clear yellowish solution was directly subjected to reversed-phase chromatography (water + 0.1% formic acid (FA)/acetonitrile (ACN) + 0.1% FA 90/10 to 10/90 in 15 min). Pure fractions were combined and lyophilized to give 120 mg (0.157 mmol) of a white lyophilizate with 76% yield. ¹H NMR (600 MHz, MeOD) δ 5.37 (s, 6H), 4.28 (s, 6H), 3.80 (s, 6H). ¹³C NMR (151 MHz, MeOD) δ 171.86, 169.79, 56.93, 42.59, -2.45. High-resolution MS (Agilent 6230 ESI-TOF HPLC-MS) calculated for C₁₅H₂₁I₃N₆O₆Na⁺ [M + Na]⁺ = 784.8549, found 784.8510.

Cross-linking reactions

Protein cross-linking was carried out by mixing 15 μM of protein (trimer concentration, i.e., monomer concentration of 45 μM) with 1 mM of Ta-Cl₃ or 0.3 mM of Ta-I₃ (freshly prepared in buffer at a concentration of 5 or 1 mM, respectively) in 50 mM HEPES, 50 mM NaCl buffered at pH 8, in a total reaction volume of 80–300 μL. Prior to cross-linking, protein stocks were concentrated to at least 65 μM (trimer concentration) in purification buffer (50 mM HEPES, 150 mM NaCl, at pH 8). This buffer also contained 0.5 mM TCEP to ensure reduction of cysteine side chains prior to cross-linking. The subsequent dilution of the protein upon mixing with the cross-linker resulted in dilution of the TCEP concentration to a maximum of 115 μM. Reactions were incubated at 20°C (except where indicated) overnight (Ta-Cl₃), or 2–4 h (Ta-I₃), and excess cross-linker was subsequently removed by buffer exchange with an Amicon centrifugal concentrator. Cross-linked samples for crystallography were further purified by an additional size exclusion chromatography step, as in the supplemental information section protein expression and purification. Cross-linking reactions for the determination of predominant charge states of PFE variants p2₃Ta, p3₃Ta, and p4₃Ta₂ were quenched by addition of formic acid to a final concentration of 1% (v/v).

Thermal denaturation (DSF and CD)

Thermal stability was initially assessed by DSF using a Tycho NT.6 instrument. The fluorescence ratio of 350/330 nm was measured between 35°C–95°C (with a temperature ramp rate of 30°C/min) at a protein trimer concentration of 5 μM (50 mM HEPES, 50 mM NaCl, pH 8) and a given concentration of GuHCl. The fixed temperature ramp rate of this instrument is 30°C/min, from which we determined a comparative inflection midpoint (T_i). This value is likely to overestimate absolute stability; therefore, thermal stability was further validated by subsequent CD experiments at reduced ramp rate (1°C/min). Samples for CD were measured at a trimer concentration of 3–5 μM (50 mM potassium phosphate, 50 mM potassium chloride, pH 8) in a 1 mM quartz cuvette in a JASCO J1500 instrument. Thermal melts were recorded at 220 nm from 5°C to 95°C with a ramp rate of 1°C/min to determine the equilibrium unfolding midpoint (T_m). Full spectra were also recorded at 5°C and 95°C between 200 and 260 nm with a bandwidth of 1 nm, averaging four spectra accumulations. HT voltage was maintained below 500 V at every wavelength for all measurements. CD measurements are reported as mean residual ellipticity (MRE), converted by the following equation, in which n is the number of peptide bonds in the given protein.

$$MRE\left(\mathit{deg}.{cm}^2.{dmol}^{-1}\right)=\frac{elipticity\left(mdeg\right)\times {10}^6}{pathlength\left(mm\right)\times proteinconcentration\left(\mu M\right)\times n}$$

Relative DSF and CD measurements for unmodified and cross-linked variants were in good accordance (ΔT_i vs. ΔT_m), however absolute T_m values were approximately 5°C–10°C lower, likely due to the slower temperature ramp.

Dynamic light scattering

Temperature-dependent dynamic light scattering (DLS) measurements were obtained using a Prometheus Panta (Nanotemper) instrument at a trimer concentration of 15 μM (50 mM HEPES, 50 mM NaCl, pH 8). Measurements were taken at 0.17°C intervals with a temperature ramp rate of 1°C/min. The determined autocorrelation function at each given temperature was fitted to obtain a size distribution using Panta Control (version 1.6.3) software. Six size distribution fits within a +0.95°C range of the temperature were averaged (e.g., data for T = 50°C represent an average of six measurements between 50.00°C and 50.95°C). The cumulant radii (indicating the predominant particle size) for both samples at every measured temperature between 20°C and 95°C were also calculated using Panta Control software, presented in Figure S8.

Crystallography

Initial screening of crystallization conditions was performed using the JSCG core suite I–IV at 20°C. Screens were assembled in 96-well plates using the sitting-drop vapor diffusion method, with a drop volume of 200 nL dispensed by a Mosquito crystallization robot (SPT Labtech). A protein concentration of 20 mg/mL was mixed in a 1:1 ratio with each crystallization solution. Refinement screens (drop size 1 μL) from initial hits identified two optimal crystallization solutions for growth of p4₃Ta₂ crystals: solution I (100 mM N-cyclohexyl-2-aminoethanesulfonic acid, pH 10, 25% [w/v] PEG3000) and crystallization solution II (210 mM lithium acetate, 19% [w/v] PEG3350). Obtained crystals were collected, cryo-protected by soaking in 15%–20% (v/v) glycerol, and flash frozen in liquid nitrogen. Diffraction data were obtained at the i04 beamline at the Diamond Light Source (United Kingdom). Data were integrated using xds⁴⁸ and two datasets were scaled using xscale. The crystal structure was solved by molecular replacement using Phaser within the CCP4i2 software suite⁴⁹ and PDB: 1va4 as a model. Structure building and refinement was performed using Coot, Refmac5, and PDBRedo. X-ray collection and refinement statistics are provided in Table S8. The crystal structure of bicyclic p4₃Ta₂ has been deposited to the PDB: 8pi1.

Published: November 2, 2023

Data and code availability

The crystal structure of p4₃Ta₂ (bicyclic PFE) has been deposited to the PDB with accession number PDB: 8pi1. All other data supporting this study are available in the manuscript and supplemental information.