Arginine Accelerates Sulfur Fluoride Exchange and Phosphorus Fluoride Exchange Reactions between Proteins

Abstract

Sulfur fluoride exchange (SuFEx) and phosphorus fluoride exchange (PFEx) click chemistries are advancing research across multiple disciplines. By genetically incorporating latent bioreactive unnatural amino acids (Uaas), these chemistries have been integrated into proteins, enabling precise covalent linkages with biological macromolecules and paving the way for new applications. However, their suboptimal reaction rates in proteins limit effectiveness, and traditional catalytic methods for small molecules are often incompatible with biological systems or in vivo applications. We demonstrated that introducing an arginine adjacent to the latent bioreactive Uaa significantly boosts SuFEx and PFEx reaction rates between proteins. This method is effective across various Uaas, target residues, and protein environments. Notably, it also enables efficient SuFEx reactions in acidic conditions, common in certain cellular compartments and tumor microenvironments, which typically hinder SuFEx reactions. Furthermore, we developed the first covalent cell engager that substantially enhances natural killer cell activation through improved covalent interaction facilitated by arginine. These findings provide mechanistic insights and offer a biocompatible strategy to harness these robust chemistries for advancing biological research and developing new biotherapeutics.

Graphical abstract

Introducing an arginine (Arg) residue adjacent to the latent bioreactive unnatural amino acid (Uaa) significantly accelerates both SuFEx and PFEx reaction rates between proteins. This offers a general and biocompatible strategy to harness these robust click chemistries for fundamental and applied biological research, applicable in vitro and in vivo.

Introduction

The integration of sulfur fluoride exchange (SuFEx) and phosphorus fluoride exchange (PFEx) chemistries into proteins is significantly advancing their application in biological settings.^[1,2] This is achieved through the genetic encoding of latent bioreactive unnatural amino acids (Uaas),^[3–5] which react with nearby nucleophilic functional groups via proximity-enabled SuFEx or PFEx reaction. These biospecific chemistries allow the precise construction of covalent linkages between proteins,^[5] and between proteins and other biomolecules, such as RNAs or carbohydrates,^[6,7] both in vitro and in vivo.^[8] Establishing strong, stable, and irreversible covalent connections among biomacromolecules is transforming the generation of novel protein properties,^[2,9] uncovering previously elusive biomolecular interactions,^[7,10,11] and facilitating the development of potent covalent protein-based therapeutics.^[12–14] The effectiveness of these applications crucially depends on rapid reaction rates, which are essential for enhancing yields, detection sensitivity, and therapeutic efficacy.^[8]

Despite their potential, the efficiency of SuFEx and PFEx reactions between proteins is currently suboptimal.^[15] These reactions are enabled by bringing a latent bioreactive Uaa to the close proximity of its target residue upon protein binding. The reaction efficiency is influenced by factors such as the binding affinity of the carrier protein, the length and flexibility of the Uaa’s side chain,^[1,11] the orientation and intrinsic reactivity of the Uaa’s warhead,^[7,16,17] and pH levels,^[15] which affect the nucleophilicity of target residues. Under optimal conditions, the second-order rate constant (k₂) for SuFEx reactions between proteins ranges from 5 to 50 M⁻¹s⁻¹.^[15] For comparison, k₂ for click reactions in small molecules can range from 1 to 10⁵ M⁻¹s⁻¹.^[18] While catalysts often accelerate SuFEx and PFEx reactions in small molecules,^[19–21] their effects within protein environments are yet to be explored, and most are incompatible with biomacromolecules. Additionally, the use of exogenous catalysts for in vivo applications and therapeutic interventions is impractical. Therefore, a general and biocompatible method to enhance SuFEx and PFEx reaction rates between proteins is highly desired.

In this study, we demonstrated that strategically introducing an arginine (Arg) mutation near the latent bioreactive Uaa significantly accelerated the SuFEx reaction rate between proteins. This acceleration was observed across different SuFEx-capable Uaas, various target residues, and even in acidic conditions that normally hinder SuFEx reactions. Similarly, the Arg mutation also enhanced the recently developed PFEx reaction between proteins. Additionally, we developed a covalent natural killer (NK) cell engager and showed that covalent binding and Arg-accelerated reaction markedly increased NK cell activation, suggesting a promising therapeutic application. Therefore, Arg acceleration provides a simple, biocompatible method to enhance SuFEx and PFEx reactions between proteins, expanding their potential in biological and biomedical research.

Results and Discussion

Tetramethylguanidine does not enhance SuFEx reactions between proteins

Small molecule SuFEx reactions are catalyzed by diverse catalysts, including nitrogenous Lewis bases, Lewis acids, and bifluorides salts.^[21] The SuFEx reaction used to conjugate aryl fluorosulfate small molecules to Tyr residues within peptides or proteins is facilitated by the water-soluble Lewis base tetramethylguanidine (TMG).^[22] We investigated whether TMG could similarly enhance the proximity-enabled SuFEx reaction at the protein-protein interface mediated by the genetically incorporated latent bioreactive Uaa fluorosulfate-L-tyrosine (FSY)^[2] (Figure 1a–b).

Figure 1.

TMG did not enhance proximity-enabled SuFEx reactions between proteins. (a) Schematic illustration of proximity-enabled SuFEx reaction between proteins. (b) Structures of FSY and TMG. (c) Crystal structure of the Afb–Z protein complex (PDB code: 1LP1), showing Asp36 for FSY incorporation and Asn6 for mutation to Tyr, His, or Lys. (d) SDS-PAGE analysis of cross-linking between Afb(36FSY) and MBP-Z(6H) over various incubation times with and without TMG. See Figure S10 for uncropped gels. (e) Determination of k_obs for reactions in (d). Data represent mean values, and error bars represent SEM; n = 3 independent experiments. SDS-PAGE of one experiment is shown.

We utilized the Z_spa affibody (Afb), which binds to the Z protein with a moderate affinity (K_d 6 μM). Based on the Afb-Z complex structure,^[23] we incorporated FSY into Afb at site 36 and mutated Tyr, Lys, or His at site 6 of the Z protein (Figure 1c). Upon Afb-Z binding, FSY is placed in proximity to the mutated residue for proximity-enabled SuFEx reaction. Maltose binding protein (MBP) was fused to the N-terminus of the Z protein to facilitate the separation of Afb and Z proteins with similar molecular weights. We incubated the purified Afb(36FSY) with MBP-Z(6X) in the presence or absence of 20 mM TMG at 37°C for varying durations. Samples were subsequently quenched and analyzed by SDS-PAGE. A time-dependent decrease in MBP-Z(6H) band intensity was observed, alongside a concomitant increase in the Afb(36FSY)–MBP-Z(6H) cross-linking band intensity (Figure 1d). We quantified the band intensities and plotted the decrease in normalized MBP-Z(6H) band intensity against time (Figure 1e), from which the pseudo-first-order rate constant (k_obs) was calculated. Afb(36FSY) cross-linked with MBP-Z(6H) at a similar k_obs with and without TMG. Similar results were obtained for Afb(36FSY) cross-linking with MBP-Z(6Y) or MBP-Z(6K) (Figure S1). These results indicate that TMG lacked a catalytic effect on the SuFEx reaction between FSY and Tyr, His, or Lys in the Afb–Z protein pair.

We also investigated the impact of TMG on the SuFEx reaction between nanobody 2rs15d and human epidermal growth factor receptor 2 (HER2) (Figure S2), as well as between nanobody 7D12 and human epidermal growth factor receptor (EGFR) (Figure S3). The observed SuFEx k_obs values showed no significant differences in the presence or absence of TMG. These findings collectively indicate that TMG did not enhance the proximity-enabled SuFEx reaction between these proteins.

Arg accelerates SuFEx reactions of FSY with Lys between proteins in vitro and on cells

Since the external small molecule catalyst TMG did not enhance SuFEx reactions between proteins, we investigated whether protein mutations near the SuFEx site could induce a catalytic effect. Arg, with its guanidine group, emerged as the most promising candidate to replicate the function of TMG or other nitrogenous Lewis bases. Previous studies have reported that an adjacent Arg residue facilitated the reaction between small molecule fluorosulfate probes and Tyr in target enzymes.^[24,25] However, SuFEx reactions can yield different outcomes depending on the system. For instance, SuFEx targeting Ser results in stable linkages in the small molecule-protein interface^[26,27] but unstable linkages in the protein-protein interface.^[7,28] Similarly, as shown above, TMG exhibited a catalytic effect in one system but not in the other. Thus, it is uncertain if Arg facilitation is equally effective at the protein-protein binding interface. In previous studies involving the small molecule-protein interface, Arg was always pre-seated with the target residue in the target protein. In contrast, we aimed to incorporate Arg into the same protein containing FSY, avoiding mutations in the target protein itself. This change in the protein location of Arg is crucial for in vivo applications, as target proteins are typically unamenable to genetic or chemical modifications in vivo.

The proximity-enabled SuFEx reaction between two proteins occurs in two steps (Figure 1a):^[15] first, the proteins bind to form a non-covalent complex, bringing the latent bioreactive Uaa FSY of one protein close to a target residue on the other. This proximity facilitates the SuFEx reaction between FSY and the target residue, resulting in a specific covalent bond. To expedite this reaction, introducing Arg should not significantly reduce protein binding, and the Arg side chain should be positioned near the FSY reaction site.

Based on these assumptions, we initiated our study on the effect of Arg using the 2rs15d–human HER2 protein pair. The nanobody 2rs15d specifically binds to human HER2. Guided by the crystal structure of human HER2 in complex with 2rs15d,^[29] we introduced FSY at site Asp54 to covalently target Lys150 of HER2 (Figure 2a). We identified Ser52 and Asp56 on 2rs15d as ideally situated to potentially expedite the SuFEx reaction relative to Asp54FSY. In contrast, while Gly53 and Gly55 are nearby, their mutation to Arg would orient the side chains away from the SuFEx reaction site, serving as functional controls. These sites were individually mutated to Arg, and the resulting mutant proteins were purified with yields of 1.4 to 3 mg/L of cell culture. They were then incubated with the extracellular domain (ECD) of human HER2 in PBS buffer for various durations, followed by SDS-PAGE analysis under denatured conditions (Figure 2b).

Figure 2.

Arg accelerated proximity-enabled SuFEx reaction between FSY and Lys between proteins in vitro and on cells. (a) Crystal structure of nanobody 2rs15d binding to human HER2 (PDB code: 5MY6). Residue D54 in 2rs15d was mutated to FSY to target K150 in HER2. Residues S52, G53, G55, and D56 in 2rs15d were individually mutated to Arg. (b) SDS-PAGE analysis of cross-linking between human HER2 ECD and 2rs15d(D54FSY), as well as the Arg mutants of 2rs15d(D54FSY). See Figure S11 for uncropped gels. (c) Determination of k_obs for reactions between human HER2 ECD and 2rs15d(D54FSY), and its Arg mutant, 2rs15d(D54FSY/D56R). (d) Western blot analysis of cross-linking between 2rs15d(D54FSY) or its Arg mutant, 2rs15d(D54FSY/D56R), and native HER2 on BT-474 cell surfaces. Control: cells only, without 2rs15d protein. The anti-Hisx6 antibody was used to detect the His6 tag appended to the C-terminus of the 2rs15d nanobodies. See Figure S12 for uncropped blots. (e) ELISA analysis of the impact of the Arg mutation on 2rs15d(D54Y)’s binding affinity for human HER2 ECD. n.d., not determined. (f) Crystal structure of nanobody 7D12 binding to human EGFR (PDB code: 4KRL). Residue Y109 in 7D12 was mutated to FSY to target K443 in EGFR. Residues T107, L108, and E44 were individually mutated to Arg. (g) SDS-PAGE analysis of cross-linking between human EGFR ECD and 7D12(Y109FSY), as well as the Arg mutants of 7D12(Y109FSY). See Figure S13 for uncropped gels. (h) Determination of k_obs for reactions between human EGFR ECD and 7D12(Y109FSY), and its Arg mutant, 7D12(Y109FSY/E44R). (i) Western blot analysis of cross-linking between 7D12(Y109FSY) or its Arg mutant, 7D12(Y109FSY/E44R), and native EGFR on A431 cell surfaces. Control: cells only, without 7D12 protein. The anti-Hisx6 antibody was used to detect the His6 tag appended to the C-terminus of the 7D12 nanobodies. See Figure S14 for uncropped blots. Data represent mean values, and error bars represent SEM in (c) and (h) and s.d. in (e); n = 3 independent experiments.

Without the Arg mutation, 2rs15d(D54FSY) readily crosslinked with HER2 ECD, displaying a distinct cross-linking band within 15 min. Interestingly, introducing Arg at Gly53 and Gly55 attenuated the cross-linking rate. Notably, Arg substitution at Ser52 abolished crosslinking, whereas at Asp56, it significantly expedited cross-linking, as evidenced by the clear presence of the band at just 5 min for 2rs15d(D54FSY/D56R). The k_obs for 2rs15d(D54FSY) with HER2 ECD was 0.568 h⁻¹ (Figure 2C). Remarkably, 2rs15d (D54FSY/D56R) exhibited a k_obs of 1.96 h⁻¹, indicating that the Asp56 to Arg mutation in 2rs15d(D54FSY) enhanced cross-linking efficiency by approximately 3.4-fold. The crosslinked 2rs15d(D54FSY/D56R)–HER2 ECD complex was characterized by tandem MS, confirming that FSY reacted exclusively with Lys150 of HER2 (Figure S4), indicating that the Arg mutation did not alter the reaction specificity of the nanobody.

We further assessed whether Arg could enhance the ability of 2rs15d(D54FSY) to crosslink native HER2 receptors on the surface of cancer cells. BT-474, a HER2-expressing human breast cancer cell line, was incubated with 2 μM of either 2rs15d(D54FSY/D56R), 2rs15d(D54FSY), or 2rs15d(WT). As expected, Western blot analysis of the cell lysates revealed no covalent cross-linking for 2rs15d(WT) (Figure 2d). Using a HER2-specific antibody, covalent cross-linking of 2rs15d(D54FSY) with native HER2 was detected after 1 h of incubation. In contrast, the Arg mutant, 2rs15d (D54FSY/D56R), cross-linked native HER2 within just 15 min. Detection with the anti-Hisx6 antibody showed that the Arg mutant also crosslinked faster than 2rs15d(D54FSY). Additionally, there was no difference in cross-linking specificity between 2rs15d(D54FSY) and its Arg mutant, indicating that the Arg mutation did not alter the nanobody’s crosslinking specificity. Thus, the Arg mutation accelerated the cross-linking of 2rs15d(D54FSY/D56R) with native HER2 on cancer cell surfaces.

To understand why the Arg mutation at certain sites of 2rs15d did not accelerate the reaction, we investigated its impact on the binding affinity of 2rs15d(D54Y) with human HER2 ECD. We chose 2rs15d(D54Y) because FSY is an analog of Tyr, and 2rs15d(D54FSY) was unsuitable for K_d measurement due to its covalent cross-linking. Enzyme-linked immunosorbent assay (ELISA) revealed that the Arg mutation at S52 nearly abolished the binding ability of 2rs15d(D54Y) (Figure 2e), explaining the lack of cross-linking by 2rs15d(D54FSY/S52R). Arg mutations at G53 and G55 significantly reduced binding affinity. Interestingly, although the Arg mutation at D56 decreased binding affinity by 14-fold, the corresponding mutant, 2rs15d(D54FSY/D56R), still demonstrated a 3.4-fold increase in the SuFEx reaction rate.

To investigate whether the acceleration of FSY-based SuFEx reactions by Arg could be extended to other proteins, we examined its effect on the 7D12–human EGFR protein pair. The nanobody 7D12 specifically binds to human EGFR. Guided by the crystal structure of human EGFR complexed with 7D12,^[30] we incorporated FSY at Tyr109 within 7D12 to target Lys443 on human EGFR (Figure 2f). Considering the proximity and orientation of the side chain towards Y109FSY in 7D12, we hypothesized that Arg mutations at positions E44 or L108 could expedite the SuFEx reaction. Conversely, a mutation at T107 was anticipated to serve as a control, given its side chain orientation away from the SuFEx reaction site.

These mutant proteins were purified and incubated with human EGFR ECD in PBS buffer, followed by SDS-PAGE analysis. As shown in Figure 2g, 7D12(Y109FSY) and its Arg variants all achieved time-dependent cross-linking with human EGFR ECD. As anticipated, the 7D12(Y109FSY/T107R) mutant did not enhance the reaction rate. Interestingly, 7D12(Y109FSY/L108R) showed a modest increase, whereas 7D12(Y109FSY/E44R) demonstrated a notably faster cross-linking compared to the unmutated 7D12(Y109FSY). Upon mutating Glu44 to Arg, the k_obs of 7D12(Y109FSY/E44R) was enhanced 2.5-fold compared to 7D12(Y109FSY) (Figure 2h). Tandem MS characterization of the crosslinked 7D12(Y109FSY/E44R)–EGFR ECD complex confirmed that FSY reacted exclusively with Lys443 of EGFR (Figure S5), indicating that the E44R mutation did not alter the nanobody’s reaction specificity.

We also assessed the potential of the Arg mutation to enhance the cross-linking of 7D12(Y109FSY) with native EGFR receptors on cancer cell surfaces. EGFR-expressing A431 cells, derived from epidermoid carcinoma, were incubated with 2 μM of 7D12(WT), 7D12(Y109FSY), or 7D12(Y109FSY/E44R). Western blot analysis of the cell lysates revealed that 7D12(WT) did not show covalent cross-linking (Figure 2i). In contrast, 7D12(Y109FSY) crosslinked with native EGFR after a 2-h incubation period. Remarkably, 7D12(Y109FSY/E44R) demonstrated robust cross-linking with native EGFR on A431 cells in just 15 min. The faster cross-linking by the Arg mutant and its selectivity for EGFR were further confirmed using the anti-Hisx6 antibody.

Collectively, these findings demonstrate that strategically placing Arg near FSY significantly boosted the rate of the proximity-enabled SuFEx reaction between FSY and the Lys residue at the protein-protein binding interface. Potential mechanisms for Arg-induced acceleration include the guanidine group of Arg facilitating the departure of the fluoride ion and/or lowering the pKa of the nucleophilic target residue.^[21] This effect likely requires the proper orientation of the guanidine group, meaning only specific sites in the protein, when mutated to Arg, can manifest this enhancement. In contrast, the catalyst TMG did not show an acceleration effect in proteins, likely because it cannot maintain a stable orientation relative to the Uaa or the target residue, or it fails to access tight protein binding interfaces. While the placement of the Arg mutation near the Uaa within the protein’s three-dimensional structure is crucial, a quantitative understanding of this spatial requirement remains to be established. It is of note that the Arg mutation can reduce or eliminate protein binding at certain sites. Structural information is currently used to guide the placement of the Arg mutation. Recent progress in protein structure prediction and deep learning approaches is expected to improve the prediction and optimization of Arg mutation sites in various proteins.^[31,32]

Arg accelerates SuFEx reactions between FSY and all target residues in proteins

FSY reacts with nearby Lys, His, and Tyr residues through a proximity-enabled SuFEx reaction in proteins.^[2] To explore whether an Arg mutation could enhance the FSY-based SuFEx reaction across different target residues, we employed the Afb-Z protein pair, which tolerates target residue mutation for cross-linking. Using the crystal structure of the Afb-Z complex (Figure 3a),^[23] we incorporated FSY at site 36 of Afb to covalently target residue 6 of the Z protein, where Lys, Tyr, or His was individually mutated. Additionally, Phe32, located near position 36 in Afb, was mutated to Arg to potentially accelerate the SuFEx reaction.

Figure 3.

Arg accelerated SuFEx reaction between FSY and target residues Lys, His, and Tyr in the Afb–Z protein pair. (a) Crystal structure of the Afb–Z protein complex (PDB code: 1LP1), showing site Asp36 for FSY incorporation, Asn6 for mutation to Lys, His, or Tyr, and Phe32 for mutation to Arg. (b-d) SDS-PAGE analysis of cross-linking and determination of k_obs (bottom graph) for MBP-Z(6X) variants with Afb(36FSY) or its Arg mutant, Afb(36FSY/F32R). Variants include: (b) MBP-Z(6K), (c) MBP-Z(6H), and (d) MBP-Z(6Y). See Figure S15 for uncropped gels. Data represent mean values, and error bars represent SEM; n = 3 independent experiments. SDS-PAGE of one experiment is shown. (e) Comparison of cross-linking yields at 16 h.

The resulting MBP-Z(6X) proteins, each with a different mutation at position 6, were incubated with either Afb(36FSY) or Afb(36FSY/32R) in PBS buffer. After various incubation periods, SDS-PAGE analysis was conducted, and k_obs was determined. With MBP-Z(6K), Afb(36FSY/32R) achieved a cross-linking rate approximately 3.4 times faster than Afb(36FSY) (Figure 3b), confirming the role of Arg in enhancing the SuFEx reaction with the Lys residue. Similarly, with MBP-Z(6H), Afb(36FSY/32R) demonstrated a 3.7-fold increase in cross-linking rate compared to Afb(36FSY) (Figure 3c). Additionally, the cross-linking rate of Afb(36FSY/32R) with MBP-Z(6Y) showed a substantial 10.5-fold increase over Afb(36FSY) (Figure 3d).

Beyond the reaction rate, the final cross-linking yield after 16 h of incubation was assessed (Figure 3e). Compared to Afb(36FSY), Afb(36FSY/32R) achieved 2.3-, 2.9-, and 7.9-fold increases in cross-linking yields with MBP-Z(6K), MBP-Z(6H), and MBP-Z(6Y), respectively.

The three crosslinked Afb(36FSY/32R)–MBP-Z(6X) complexes were characterized by tandem MS, confirming that FSY in the Afb specifically reacted with the target Lys, His, or Tyr in MBP-Z (Figure S6–S8). These results collectively indicate that the Arg mutation significantly enhanced the FSY-based SuFEx reaction rates with Lys, His, and Tyr residues and improved cross-linking yields with all tested residues.

Arg accelerates SuFEx reactions of mFSY between proteins

To investigate whether an Arg mutation could enhance the SuFEx reaction rates of other latent bioreactive Uaas, we incorporated meta-fluorosulfate-L-tyrosine (mFSY) into the nanobody 7D12 and examined the effect of Arg on its cross-linking efficiency with the human EGFR ECD. Unlike FSY, which has a fluorosulfate group at the para position, mFSY features this group at the meta position (Figure 4a), offering a distinct orientation.^[16] We incorporated mFSY at Tyr109 within nanobody 7D12 to target Lys443 on EGFR (Figure 2f). The variants 7D12(Y109mFSY) and 7D12(Y109mFSY/E44R) were purified, incubated with EGFR ECD, and analyzed by SDS-PAGE (Figure 4b) and k_obs determination (Figure 4c). Both mFSY-modified 7D12 mutants successfully cross-linked with EGFR ECD. The k_obs for 7D12(Y109mFSY/E44R) was 1.9-fold higher than for 7D12(Y109mFSY). Thus, the Arg mutation could accelerate the SuFEx reaction of mFSY, in addition to FSY, despite the differences in the orientation of their fluorosulfate groups.

Figure 4.

Arg accelerated SuFEx reaction of mFSY and FFY between proteins. (a) Structure of mFSY. (b-c) SDS-PAGE analysis of cross-linking (b) and determination of k_obs (c) for human EGFR with 7D12(Y109mFSY) and its Arg mutant, 7D12(Y109mFSY/E44R). (d) Structure of FFY. (e-f) SDS-PAGE analysis of cross-linking (e) and determination of k_obs (f) for human EGFR with 7D12(Y109FFY) and its Arg mutant, 7D12(Y109FFY/E44R). See Figure S16 for uncropped gels for (b) and (e). Data represent mean values, and error bars represent SEM; n = 3 independent experiments. SDS-PAGE of one experiment is shown.

Arg boosts the enhanced SuFEx reactivity of FFY

We previously developed the latent bioreactive Uaa FFY by introducing an electron-withdrawing fluorine substitution on FSY (Figure 4d), thereby enhancing the SuFEx reaction rate in proteins.^[14] Here, we evaluated the acceleration effects of FFY and Arg mutation on SuFEx reactivity, and whether Arg could further boost FFY’s reactivity. Using the 7D12-EGFR protein pair, we incorporated FFY at Tyr109 of 7D12 to target Lys443 on EGFR (Figure 2f). The previously identified E44R mutation was introduced to examine the Arg effect. Both 7D12 variants, Y109FFY and Y109FFY/E44R, were purified and incubated with human EGFR ECD, followed by SDS-PAGE analysis (Figure 4e) and k_obs determination (Figure 4f). As expected, both FFY-modified 7D12 mutants efficiently cross-linked with EGFR ECD. The k_obs for 7D12(Y109FFY) was measured at 0.176 h⁻¹, representing a 1.3-fold increase compared to 7D12(Y109FSY) (k_obs = 0.137 h⁻¹). This increase confirms FFY’s enhanced reactivity. Introducing the Arg mutation into 7D12(Y109FSY) to create 7D12(Y109FSY/E44R) resulted in a k_obs of 0.342 h⁻¹, a 2.5-fold increase over 7D12(Y109FSY), highlighting Arg’s superior acceleration compared to FFY. Remarkably, 7D12(Y109FFY/E44R) achieved a k_obs of 0.660 h⁻¹, a 3.8-fold increase over the 7D12(Y109FFY) baseline, demonstrating that the Arg mutation further elevated FFY’s cross-linking efficiency. Overall, the combined effects of FFY substitution and the E44R mutation led to a 4.8-fold increase in reaction rate compared to the original FSY variant.

We also assessed the effects of Arg mutation on the reactivity FFY within the Afb(36FFY)/MBP-Z(6X) protein pair. Interestingly, incorporation of FFY at site 36 in Afb precipitated a 3- to 4-fold reduction in k_obs compared to that of FSY. This trend was consistent across all three target residues – His, Tyr, and Lys – in the MBP-Z(6X) protein (Figure S9). However, the introduction of an Arg at site F32 in Afb markedly increased the k_obs of the Afb(36FFY/F32R) variant, equalizing it with the rate of the Afb(36FSY/F32R) variant.

In summary, the Arg mutation not only demonstrated a more potent acceleration effect on SuFEx reactivity compared to the FFY substitution but also further enhanced this acceleration beyond the level achieved by FFY alone. Moreover, the Arg mutation proved capable of augmenting the SuFEx reaction rate even when the FFY substitution was ineffective.

Arg accelerates PFEx reactions between proteins

Recent advancements have positioned PFEx reactions at the forefront of click chemistry alongside the SuFEx reaction.^[19] PFEx reactions enable the exchange of P(V)-F bonds with incoming nucleophiles, resulting in stable tetrahedral P(V)–O and P(V)–N bonds. In small molecule contexts, these reactions require a Lewis base catalyst and a silicon-based additive.^[19] At the protein level, latent bioreactive amino acids, PFY and PFK, featuring phosphoramidofluoridate groups, have been incorporated into proteins through genetic code expansion. This enables the covalent targeting of nearby His, Tyr, Lys, or Cys residues through a proximity-driven PFEx reaction (Figure 5a, 5b).^[1]

Figure 5.

Arg accelerated PFEx reaction of PFY between proteins. (a) Structure of PFY. (b) Scheme showing the proximity-enabled PFEx reaction between proteins. (c-d) SDS-PAGE analysis of cross-linking (c) and k_obs determination (d) for MBP-Z(6H) with Afb(36PFY) and its Arg mutant, Afb(36PFY/F32R). (e-f) SDS-PAGE analysis of cross-linking (e) and k_obs determination (f) for MBP-Z(6Y) with Afb(36PFY) and its Arg mutant, Afb(36PFY/F32R). See Figure S17 for uncropped gels for (c) and (e). Data represent mean values, and error bars represent SEM; n = 3 independent experiments. SDS-PAGE of one experiment is shown.

To explore whether an Arg mutation could enhance the PFEx reaction rate between proteins, we incorporated PFY into Afb to produce Afb(36PFY) and Afb(36PFY/F32R). These variants were incubated with MBP-Z(6H) or MBP-Z(6Y) for cross-linking. SDS-PAGE analysis revealed successful cross-linking between MBP-Z(6H) and the two Afb variants via the PFY-His reaction (Figure 5c). Importantly, the introduction of an Arg at position 32 in Afb(36PFY/F32R) led to a 44% increase in the PFEx reaction rate (Figure 5d). In contrast, through the PFY-Tyr reaction, cross-linking MBP-Z(6Y) with Afb(36PFY) showed no discernible results (Figure 5e). However, the Arg mutant Afb(36PFY/32R) demonstrated substantial cross-linking, with a k_obs of 0.0218 h⁻¹ (Figure 5f). These results indicate that the Arg mutation could indeed accelerate the PFEx reaction rate between proteins.

Arg boosts SuFEx reactions between proteins at acidic pH levels

An exciting frontier in SuFEx chemistry is the development of covalent protein-based therapeutics for cancer treatment.^[12,13] These drugs incorporate a latent bioreactive Uaa that selectively reacts with a natural residue on the target protein upon drug-target interaction, facilitating irreversible binding and therapeutic action. However, the efficacy of these therapeutics may be challenged by the acidic extracellular pH prevalent in tumor tissues, typically between 6.5 to 6.8,^[33] as SuFEx reactions are known to be less reactive in acidic environments.^[15] Moreover, the acidic lumen of various intracellular organelles and vesicles — with pH ranging from 4.7 to 6.7 — is common in eukaryotic cells.^[34] Targeting proteins within these acidic contexts demands SuFEx reactions that are efficient even at lower pH levels.

To address this challenge, we investigated whether an Arg mutation could enhance SuFEx reactivity under acidic conditions. We first used the 2rs15d-HER2 pair to evaluate the effect of Arg acceleration at acidic pH levels. The purified 2rs15d(D54FSY) and its Arg mutant, 2rs15d(D54FSY/D56R), were incubated with the human HER2 ECD protein in PBS buffer at pH 6.0 and 6.5, followed by SDS-PAGE analysis and k_obs determination (Figure 6a, 6b). At pH 6.5, the k_obs for 2rs15d(D54FSY) was 4.2-fold slower than at pH 7.4, confirming that acidic pH decreases SuFEx reactivity in proteins. In contrast, the Arg mutant, 2rs15d(D54FSY/D56R), exhibited a k_obs of 0.888 h⁻¹ at pH 6.5, marking a 6.6-fold increase over 2rs15d(D54FSY) at this pH (Figure 6b). Furthermore, at pH 6.0, the k_obs for 2rs15d(D54FSY) decreased further with increased acidity. However, the Arg mutant, 2rs15d(D54FSY/D56R), achieved a k_obs of 0.705 h⁻¹, demonstrating a remarkable 10.7-fold improvement over 2rs15d(D54FSY).

Figure 6.

Arg accelerated SuFEx reaction of FSY at acidic pH levels. (a-b) SDS-PAGE analysis of cross-linking of human HER2 ECD with nanobody 2rs15d(D54FSY) or its Arg mutant, 2rs15d(D54FSY/D56R), at acidic pH 6.5 and 6.0 (a), and subsequent determination of k_obs (b). (c-d) SDS-PAGE analysis of cross-linking of human EGFR ECD with nanobody 7D12(Y109FSY) or its Arg mutant, 7D12(Y109FSY/E44R), at acidic pH 6.5 and 6.0 (c), and subsequent determination of k_obs (d). See Figure S18 for uncropped gels for (a) and (c). Data represent mean values, and error bars represent SEM; n = 3 independent experiments. SDS-PAGE of one experiment is shown.

The 7D12-EGFR protein pair was similarly employed to corroborate these observations. The purified 7D12(Y109FSY) and its Arg mutant, 7D12(Y109FSY/E44R), were incubated with the human EGFR ECD protein in PBS buffer at pH 6.0 and 6.5, followed by SDS-PAGE analysis (Figure 6c). Compared to the reaction at pH 7.4, the cross-linking rate of 7D12(Y109FSY) was 5.2-fold slower at pH 6.5 and 6.5-fold slower at pH 6.0 (Figure 6d). However, the Arg mutant, 7D12(Y109FSY/E44R), increased the k_obs 3.7-fold at pH 6.5 and 2.5-fold at pH 6.0 compared to 7D12(Y109FSY).

In summary, the Arg mutation markedly increased SuFEx reaction rates between proteins at pH 6.5 and 6.0. This ability to accelerate SuFEx reactions in acidic environments is both unique and highly beneficial. As pH decreases, nucleophiles become more protonated, reducing SuFEx reaction rates. However, various intracellular organelles and vesicles – such as the Golgi apparatus, endosomes, lysosomes, autophagosomes, trans-Golgi networks, and synaptic vesicles – maintain acidic pH levels and are involved in fundamental cellular processes. Notably, the interstitial spaces in tumor microenvironment often have an acidic extracellular pH.^[33] While standard proximity-enabled SuFEx reactions can be inefficient under these conditions, the Arg mutation strategy enhanced SuFEx reactivity, making it valuable for covalent protein targeting in such acidic environments.

Arg boosts covalent engager activation of NK cells

The efficacy of NK cell-based cancer immunotherapy significantly relies on the targeted accumulation and activation of NK cells within the tumor microenvironment.^[35] To enhance this efficacy, we developed a covalent bispecific NK engager, exploring the potential of covalent binding and Arg acceleration to boost NK activation against cancer cells. This engager comprised a CD16-specific nanobody, C21,^[36] fused to an EGFR-specific nanobody, 7D12, via a (GGGGS)₄ linker (Figure 7a). C21 binds CD16 (FcRγIII) on NK cells, while 7D12 is engineered with a latent bioreactive Uaa FSY at Tyr109 to facilitate covalent binding to EGFR on tumor cells. Additionally, an Arg mutation at Glu44 was introduced to enhance cross-linking efficiency.

Figure 7.

Arg acceleration enhanced the activation of NK cells by a covalent NK engager. (a) Schematic illustration of the covalent NK engager and its mechanism for NK cell activation. (b-c) SDS-PAGE analysis of cross-linking (b) and subsequent k_obs determination (c) for human EGFR ECD with the covalent NK engager 7D12(Y109FSY)–C21, and its Arg mutant 7D12(Y109FSY/E44R)–C21. Data represent mean values, and error bars represent SEM; n = 3 independent experiments. See Figure S19 for uncropped gels. (d) Western blot analysis of cross-linking of native EGFR on A431 cancer cells by the covalent NK engager and its Arg mutant. Increased incubation periods resulted in more engager-EGFR cross-linked complexes, detected using anti-EGFR and anti-Hisx6 antibodies. The anti-Hisx6 antibody detected the His6 tag appended at the C-terminus of the 7D12-C21 engagers. Equal concentrations of engager were used for all incubation periods, and unbound engagers were washed away before cell lysis for Western blot analysis. Longer incubation periods led to more engager binding and internalization by the cells. Bound but non-cross-linked engagers were detected as free engagers in the denatured Western blot (middle panel) since the whole cell lysate was analyzed. See Figure S20 for uncropped blots. (e) Perforin release by NK-92 cells in response to three different NK engagers. A431 cancer cells were treated with the NK engagers separately and then co-cultured with NK-92 cells. Dots represent the values from independent biological replicates, center bars represent mean values, and error bars represent SEM; n = 3 independent experiments. ns, not significant; * p < 0.05; ** p < 0.01; *** p < 0.001; unpaired t test (At 400 nM, for WT versus Y109FSY, p = 0.0167; for WT versus Y109FSY/E44R), p = 0.0001; for Y109FSY versus Y109FSY/E44R, p = 0.0251. At 200 nM, for WT versus Y109FSY, p = 0.0282; for WT versus Y109FSY/E44R), p = 0.0050; for Y109FSY versus Y109FSY/E44R, p = 0.0005. At 100 nM, for WT versus Y109FSY, p = 0.0061; for WT versus Y109FSY/E44R, p = 0.0080).

Initial experiments focused on the engager’s ability to form covalent linkages with EGFR and the impact of Arg acceleration. Both 7D12(Y109FSY)-C21 and its Arg-mutated counterpart, 7D12(Y109FSY/E44R)-C21, were incubated with human EGFR ECD, demonstrating in vitro covalent cross-linking with increasing intensity over time (Figure 7b). The Arg mutant showed 2.5-fold increase in cross-linking k_obs compared to the non-mutated version (Figure 7c). The NK engagers were further assessed to crosslink the full-length native EGFR receptor on A431 cancer cell surfaces. Following various treatment durations, Western blot analysis of cell lysates revealed that both covalent engagers, unlike the noncovalent 7D12(WT)-C21, effected robust, time-dependent cross-linking of EGFR (Figure 7d). Notably, the Arg mutant showcased a significantly faster cross-linking rate compared to its non-mutated counterpart.

We then assessed the capability of the covalent engager and its Arg mutant to enhance NK cell activation against cancer cells. A431 cells were treated with various concentrations of NK engagers for one hour before the introduction of human NK-92 cells — a cytotoxic NK cell line currently in clinical trials for cancer therapy.^[37] NK-92 cell activity was gauged by measuring the release of perforin, a marker of NK cell-mediated killing. All three engagers triggered a concentration-dependent increase in NK cell activation (Figure 7e). Notably, at concentrations of 100 nM and beyond, the covalent engager 7D12(Y109FSY)-C21 markedly enhanced NK cell activation relative to its noncovalent counterpart, 7D12(WT)-C21, underscoring the efficacy of covalent binding. More critically, the Arg mutant, 7D12(Y109FSY/E44R)-C21, exhibited further increases in NK cell activation beyond the covalent engager 7D12(Y109FSY)-C21, particularly at 200 and 400 nM concentrations. Specifically, the Arg-accelerated covalent engager showed increases in NK activity by 2.4-, 2.4-, and 2.6-fold over the noncovalent engager at 100, 200, and 400 nM concentrations, respectively. These results underscore the significant role of Arg acceleration in SuFEx reactivity, highlighting its potential to advance the development of potent covalent protein therapeutics for cancer immunotherapy.

Conclusion

In conclusion, we developed a general and biocompatible method to accelerate SuFEx and PFEx reaction rates between proteins by strategically introducing an Arg mutation near the latent bioreactive Uaa. This approach proved effective across all target residues, various SuFEx-capable Uaas, and different protein systems with binding affinities ranging from μM to nM. The acceleration was observed both in vitro and on cell surfaces. The Arg mutation enhanced SuFEx reactions more effectively than FFY, which is known to enhance these reactions, and combining Arg with FFY further increased the reaction rate. Additionally, the Arg mutation was effective in conditions where FFY alone was not. This strategy also successfully enhanced PFEx reactions, a forefront in click chemistry.^[1,19] Remarkably, the Arg mutation boosted SuFEx reaction rates even at acidic pH levels, which typically hinder these reactions. Enabling SuFEx reactions at acidic pH holds particular promise for exploring acidic biological compartments and developing covalent protein therapeutics for cancer treatment using SuFEx chemistry. We also developed the first covalent NK cell engager, demonstrating that covalent binding and Arg acceleration significantly increased NK cell activation by the engager, showcasing the therapeutic potential of Arg-accelerated SuFEx reactions. Lastly, this method is straightforward and well-suited for both in vitro and in vivo applications. Introducing the Arg mutation in the same protein with the warhead-bearing Uaa eliminates the need to modify the target protein, which is crucial for in vivo and therapeutic applications where the target protein cannot be modified. In contrast, catalysts required for SuFEx and PFEx reactions in small molecules are generally unsuitable for in vivo use. Consequently, our method offers a straightforward, biocompatible strategy to leverage the potent SuFEx and PFEx chemistries within biological contexts, facilitating biological research and biotherapeutic applications.