Synthesis of Multi-Protein Complexes through Charge-Directed Sequential Activation of Tyrosine Residues

INTRODUCTION

Site-selective post-translational modification of proteins has long been an object of study for both chemists and biologists. The augmentation of the native properties of these biomolecules to contain fluorophores,^1,2 drug molecules,³ and other signaling moieties has been of tremendous value to both our understanding of biology and the creation of new therapeutics.^3-6 While much work has been done in the attachment of small molecule cargoes, the facile attachment of additional whole proteins has proven more difficult. The primary challenge of protein–protein coupling lies in the steric inhibition that must be overcome in the joining of large molecules as well as the difficulty in the generation of suitable selective reaction handles due to the frequent occurrence of most native amino acids. Current methods for the site-selective synthesis of protein conjugates involve the introduction of uniquely reactive unnatural amino acids (UAAs) into the backbone sequence,^7,8 maleimide-cysteine linkages,⁹ intein-based strategies,¹⁰ and enzymatic coupling reactions.^6,11-14

Although these methods have addressed many important challenges in chemical biology, limitations still exist. The site selective incorporation of unnatural amino acids into proteins relies on stop codon repurposing, which requires special growth media and often lowers protein yield.^7,8 Thiol-reactive cross-linkers, such as maleimides, are specific to cysteine residues on the protein surface that are rare enough to afford selectivity and can be readily inserted via single base mutagenesis. However, the maleimide reactive handle is challenging to install on the surface of full proteins.⁹ For many enzymatic coupling techniques, such as sortase, butelase, and farnesyltransferase, coupling can only occur at the N- and C-termini.^6,11-14 More recent work has demonstrated that a combination of sortase and butelase can be used to sequentially conjugate multiple coupling partners onto a single protein.¹⁵ A similar method for synthesizing multiprotein conjugates was also developed that uses an asparaginyl endopeptidase instead of the butelase and sortase combination.¹⁶ Yet another approach to terminal protein ligation is the use of intein-based splicing which enables the facile combination of peptides and proteins through thioester intermediates.^10,17 While powerful, these approaches are still limited to conjugation via N or C-termini, which limits the relative orientation of the conjugate. Most other coupling approaches such as SpyCatcher/SpyTag require the insertion of bulky recognition domains and pose additional risks of immunological rejection.¹⁸

Tyrosinases, a broadly distributed class of enzymes involved in the melanin synthesis pathway in every kingdom of life, provide promising new alternatives. Tyrosinase enzymes normally catalyze the oxidation of tyrosine to a dopaquinone intermediate, which then cyclizes to form dopachrome.^19-23 Recent studies have repurposed the commercially available mushroom tyrosinase (abTYR) to catalyze site specific oxidative coupling reactions at N- and C-terminal tyrosine residues using different types of coupling partners. Strained alkynes²⁴ and nucleophilic coupling partners, such as secondary amines,²⁵ anilines,²⁶ and surface exposed cysteine residues,²⁷ have expanded the possible sites of enzymatic protein bioconjugation past the N- and C-termini. This methodology is versatile and can take place at temperatures ranging from 4 °C to room temperature and near-neutral pH to preserve the function of sensitive proteins.²⁷

In spite of the recent success and convenience of abTYR, it is not without limitations. In a recent investigation, abTYR was unable to activate tyrosine residues at the C terminus of protein L, regardless of linker length.²⁶ Both abTYR and protein L are over 100 kDa in mass and are negatively charged, suggesting that electrostatic repulsion and steric hindrance may be the factors preventing the activation of this particular substrate by abTYR.^28,29 A smaller, 35.5 kDa tyrosinase from Bacillus megaterium (megaTYR, referred to as bmTYR in ref 26) was found to activate C-terminal tyrosines on protein L regardless of linker length.^21,26 In this study, we characterize the substrate preferences of these tyrosinases and show that abTYR is unable to oxidize tyrosines in negatively charged substrates, whereas megaTYR and the recently reported tyrosinase from the archaeon, Candidatus Nitrosopumilus koreensis (cnkTYR),³⁰ are much more promiscuous. In this report, we generate new versions of megaTYR selective for tyrosine residues in specific charge contexts. Moreover, we leverage this new tool to enable the construction of protein trimers via charge-directed sequential activation of tyrosine residues (CDSAT).

RESULTS AND DISCUSSION

In our search for more sequence permissive enzymes than abTYR (Figure 1A) we identified tyrosinases from Bacillus megaterium (megaTYR) and Candidatus Nitrosopumilus koreensis (cnkTYR) as optimal candidates for recombinant expression. These particular tyrosinases were chosen as they are both monomeric tyrosinases that have been shown to be easily expressible in E. coli.^30,31 Moreover, megaTYR was selected because several previous studies showed that the enzyme tolerated point mutations while retaining high activity to free tyrosine.^21,31,32 The published crystal structures for megaTYR were beneficial for protein engineering, as they allowed rational design of point mutants and facilitated computational analysis of the protein (Figure 1B).³¹ In contrast, cnkTYR was recently shown to have excellent monophenol monooxygenase activity and to tolerate low temperatures without significant loss in catalytic efficiency which posed an attractive candidate despite the lack of published crystal structure (homology model presented in Figure 1C).^23,30

In addition to the previously published work on protein L,²⁶ the motivation for these explorations was the realization that abTYR was incapable of activating phenols attached to a single stranded DNA (ssDNA) substrate. In contrast, our first experiment with megaTYR showed significant levels of coupling between GFP and the phenol-DNA (Figure S1). This reaction occurs as two distinct steps: the initial activation of the phenol to the ortho-quinone followed by addition of the nucleophile. These results suggested that some aspect of the DNA substrate was preventing activation by abTYR but was tolerated by megaTYR (Figure S1). We reasoned that this was likely the high degree of negative charge on the nucleotide substrate.

We next examined the substrate preference of the new tyrosinases, specifically with respect to their charge and steric tolerance. A peptide library with tyrosine residues at their C-termini was obtained, shown in Supporting Information Table 1, which contained a variety of charged and neutral sequence elements. C-terminal tyrosine peptides were chosen as they most closely resemble the protein constructs previously shown to be effective substrates for the tyrosinase oxidation. Future studies will be needed to examine the substrate activation on internal tyrosine residues as previous studies have shown tyrosinase activity in unstructured peptide substrates.^33,34 However, our previous work has shown that even highly exposed tyrosine residues in folded proteins are not readily targeted unless they are adjacent to the N- or C-terminal position.^26,27 The peptides from the C-terminal library were coupled to Y200C GFP in 30 min reactions and analyzed via ESI-TOF MS to determine conversion to the modified peptide–protein conjugate. To ensure that peptide coupling was not biased by electrostatic interactions between GFP (pI = 6.2) and the peptides, control reactions were run on a p-aminophenylalanine modified MS2 capsid, which has a pI of 8.0. These reactions reached similar levels of modification (SI Figure S2). In a comparison of abTYR to the non-eukaryotic tyrosinases, it is readily apparent that abTYR is significantly more sensitive to negative charge (Figure 1D). In fact, abTYR shows no activity toward any peptide with 3 or more acidic residues adjacent to the terminal tyrosine. Also of note is the fact that megaTYR shows high activity toward all peptide substrates, with a drop in activity only in the case of the RRRRY peptide. These results supported the earlier challenge activating ssDNA substrates using abTYR and suggest that megaTYR may serve as a highly permissive tyrosine activator. One additional observation of this study was that cnkTYR samples showed no discernible pattern of substrate specificity. The structures of the three tyrosinases were compared to determine the origin of their selectivities, with cnkTyr represented by a homology model created using the I-TASSER server.^29,30,35-38 abTYR is found as a protein complex and shows significant steric occlusion of the active sites (Figure 1A),²⁹ while megaTYR and cnkTYR are monomeric and show a much more open active site (Figure 1B,C).^30,35 Due to their more predictable behaviors and established crystal structures, we focused our work on megaTYR and abTYR in subsequent experiments. However, further exploration or engineering of cnkTyr could yield a more specific phenol monooxygenase capable of acting with higher precision.

In the interest of improving megaTYR for even broader activity, we performed an evolutionary conservation analysis of the megaTYR backbone in order to determine target positions for engineering. Using the ConSurf web server, the 100 closest nonredundant homologues of megaTYR were clustered and a multiple sequence alignment was prepared (Figure 2A).^39-42 A normalized conservation score was calculated for each position using an empirical Bayesian methodology, which accounts for phylogenetic relationships between homologues in its construction of an evolutionary model.⁴³ The evolutionary conservation was mapped onto the megaTYR structure and used to identify mutable anionic residues with low conservation scores (Figure 2B, Table S2). Position R209 was also chosen as a position of interest, despite its mild conservation, due to its established role in potentiating diphenolase activity and its proximity to the active site (Figure 2C).³¹ We also chose to study R280 as a more distal mutable cationic residue with a low conservation score (Table S2). Previously studied mutants at positions V218 and F197 were not further investigated, as their impact on megaTYR activity was not beneficial to our purpose of producing a more electrostatically targeted tyrosinase.^21,32

The utility of modulating tyrosinase charge was explored in abTYR and megaTYR. Mutants of megaTYR identified from literature and the conservation screen were modeled in Maestro using the charge mapping tool from the Epik and MacroModel modules to calculate the surface potential for each of the proteins compared to a point charge placed 14 Å from the active site copper atoms (Figure 3A, Table S2). We reasoned that surface-exposed polar and ionic residues proximal to the active site may be attractive targets, as they would have larger effects on active site potential than more distal residues, and modeled them in a similar manner (Table S3). We found that abTYR monomers had a net potential of −4.4 V due to high numbers of acidic residues on their surfaces, corresponding to an isoelectric point (pI) of only 4.41. When the full active complex of abTYR is considered the net potential drops to −15.7 V.²⁹ In contrast, megaTYR was shown to have a net potential of +4.4 V and a pI of 8.66, correlating well with the observed charge preference.³⁵

In accordance with our assumption that more proximal residues would have a larger impact on active site potential, we found that R209H had a potential of +3.0 V, a significant reduction compared to the wild type, whereas R280H has a much smaller impact on the electrostatics near the active site (Figure 3B, Table S2). Similarly, D55K had the largest impact on the active site potential among the D/E to K mutants, corresponding to a potential of +7.33 V. Because of this, we also chose to model D55R to compare the effects of mutating anionic residues to lysine or arginine. While D55K had a potential of +7.33 V, D55R showed a slightly more positive net potential of +7.5 V, due to the higher pK_a of arginine (Figure 3B, Table S2).

On the basis of data from the conservation screen and potential modeling, we generated megaTYR with R209H, D55K, and D55R mutations (structures with residues highlighted in Figure 4A). After initial screening at a tyrosinase concentration of 5 U/mL, it was observed that R209H decreased the charge preference of megaTYR, while D55K and D55R were found to increase the preference for anionic substrates as shown by the higher level of modification with the EEEEY peptide (Figure 4B). To assess the degree of substrate preference, we ran competition assays containing the EEEEY and RRRRY peptides in equal concentrations. In these assays, abTYR only showed activity toward RRRRY, while R209H showed equal preference for the two (Figure 4C). Interestingly, wild type megaTYR showed a 90% preference for EEEEY, while the D55K mutant showed a 95% preference and D55R a surprising >99% preference for the anionic peptide (Figure 4C). As D55K experienced expression challenges and was not capable of complete substrate differentiation, we chose to focus on D55R as the higher performing mutant.

We then performed a full substrate screen on each megaTYR variant (wild type, R209H, and D55R) at the same concentration as the initial substrate screen (12 U/mL) and found that these mutations only had a significant effect on the enzymatic activity toward highly charged peptide substrates. Interestingly, D55R megaTYR exhibited low conversion for the RRRRY peptide but retained high levels of activity toward the less highly charged cationic peptides (Figure 4D). Generation of sequence-orthogonal tyrosinases could enable the facile creation of complex multidomain proteins by coupling anionic and cationic tagged subunits sequentially. This could be done in a manner similar to peptide synthesis but using entire proteins as building blocks. The controlled production of multiprotein polymers would be an enabling technology for protein based materials, biomedical technologies, and enzyme engineering.

Although the megaTYR mutants to date were not fully prevented from acting on cationic substrates, the range in selectivity between abTYR and D55R megaTYR proved a tantalizing target. To test this, the cysteine-containing Y200C sfGFP was modified with EEEEY and RRRRY C-terminal tags, which should be unreactive with abTYR and D55R megaTYR, respectively (Figures S3, S4). After purification, the Y200C sfGFP-EEEEY (Y200C GFP-E4Y) was exposed to abTYR for 30 min in a variety of conditions. Notably, the mushroom tyrosinase was able to append GGGGY and RRRRY peptides to the Y200C GFP-E4Y without oxidizing the EEEEY tag (Figures S3, S4). In addition, exposing the Y200C GFP-E4Y to abTYR resulted in no discernible oxidation by ESI-MS. This aligns with our previous peptide screens, and suggests that the EEEEY tag is not succeptible to abTYR activation (Figure S3). Similarly, Y200C sfGFP-RRRRY (Y200C GFP-R4Y) was exposed to D55R megaTYR under similar conditions, with the EEEEY peptide used in place of the RRRRY peptide. Similarly to abTYR, D55R megaTYR was able to selectively oxidize the added peptides without modifying the tyrosine tag on the Y200C GFP-R4Y. Reaction selectivity was further confirmed when exposure of Y200C GFP-R4Y to D55R megaTYR alone did not result in any oxidation (Figure S4). Together, these data suggest that abTYR and D55R megaTYR can be used for CDSAT on proteins and peptides.

To validate reaction orthogonality further, we next compared abTYR and maleimide mediated biotinylation of Y200C GFP-E4Y (Figure 5A-C). As a positive control, Y200C GFP-E4Y was selectively modified with maleimide biotin to generate biotinylated Y200C GFP-E4Y (Figure 5D). In parallel, we modified Y200C GFP-E4Y with a phenol-biotin in the presence of abTYR and showed selective oxidation of the small molecule to afford the Y200C GFP-E4Y-biotin conjugate (Figure 5E). After purification using a 10 kDa MWCO spin concentrator, both the maleimide and phenol GFP-E4Y-biotin were combined with Y200C sfGFP and 12 U/mL D55R megaTYR for 30 min and analyzed via ESI-TOFMS. We observed over 70% conversion to the trimer composed of two GFP proteins and a biotin molecule, making this one of the first techniques capable of creating 3-component fusions without the need for solid support tethering or removable protective groups (Figure 5F,G). In addition, some higher order modifications appeared, arising from residual disulfide formation between unreacted thiols in the first step. To combat the undesired formation of disulfide bonds between thiols in the protein coupling partner, we suggest pretreating these proteins with a 2x molar excess of TCEP in pH 7.0 Tris-HCl buffer, followed by an exchange into the reaction buffer via spin filtration. This pretreatment step was used for all subsequent reactions.

To evaluate the potential of an iterative tyrosinase approach for the creation of protein trimers, Y200C GFP-E4Y was combined with a MYGGS-tagged nanoluciferase and 20 U/mL abTYR to activate the neutral luciferase tag selectively (Figure 6A-C). Clean conversion to the Y200C GFP-E4Y-nanoluciferase dimer was observed, with no excess oxidation. To remove any unreacted starting material, the solution was purified using a 50 kDa MWCO spin concentrator (Figure 6C). As a third component, the thiol-containing S152CmCherry was added to the dimer solution with 20 U/mL D55R megaTYR to activate the EEEEY tag on the GFP. This strategy led to smooth conversion to the mCherry-GFPluciferase trimer (Figure 6D). We saw 100% conversion from the dimer to the final product, indicating that megaTYR is capable of catalyzing tyrosinase oxidation on anionic tags even in the presence of multiprotein complexes. The trifunctional nature of the construct identity was further validated by luminescence and fluorescence measurements. Upon introducing the nanoLuc substrate, FRET was observed between the luciferase, GFP, and mCherry subunits. In addition, direct excitation of the GFP fluorophore led to emission from the mCherry segment via energy transfer (Figure S5). This new capability to oxidize tyrosine residues in different charge contexts iteratively highlights the potential of this system to generate complex protein products for future applications.

Supplementary Material