Advancing Protein Therapeutics through Proximity-Induced Chemistry

SUMMARY

Recent years have seen a remarkable growth in the field of protein-based medical treatments. Nevertheless, concerns have arisen regarding the cytotoxicity limitations, low affinity, potential immunogenicity, low stability, and challenges to modify these proteins. To overcome these obstacles, proximity-induced chemistry has emerged as a next-generation strategy for advancing protein therapeutics. This method allows site-specific modification of proteins with therapeutic agents, improving their effectiveness without extensive engineering. In addition, this innovative approach enables spatial control of the reaction based on proximity, facilitating the formation of irreversible covalent bonds between therapeutic proteins and their targets. This capability becomes particularly valuable in addressing challenges such as the low affinity frequently encountered between therapeutic proteins and their targets, as well as the limited availability of small molecules for specific protein targets. As a result, proximity-induced chemistry is reshaping the field of protein drug preparation and propelling the revolution in novel protein therapeutics.

Graphical Abstract

eTOC Blurb

In this review, Cheng et al. provide an overview of the chemical mechanisms, functional groups, and introduction methods associated with proximity-induced chemistry. Additionally, they introduce innovative protein drugs developed via proximity-induced chemistry and discuss their therapeutic effectiveness across pre-clinical and clinical investigations.

Introduction

Protein therapeutics have emerged as a powerful approach in medicine, harnessing the diverse and dynamic roles of proteins in the body to address various diseases. Proteins play crucial roles in metabolic reactions, cell signaling, and immune responses, offering a tremendous opportunity for disease intervention^1–4. The development of protein therapeutics dates back to 1982 with the first United States Food and Drug Administration (FDA)-approved recombinant protein humulin, and since then, the scientific community has recognized their immense potential^1,5,6. With advancements in biotechnology, protein therapeutics have experienced significant growth, leading to the approval of over 200 therapeutic proteins by the FDA^7,8. These therapeutics encompass a range of categories, including monoclonal antibodies, enzyme replacement therapy, growth factors, interferons, hormones, and fusion proteins. Protein therapeutics offer several advantages over traditional small molecule drugs, such as high target specificity, multifunctionality, reduced adverse effects, ability to target undruggable surfaces, and effectiveness in treating genetic mutations^1,9,10. Despite revolutionizing the field of medicine, protein therapeutics still face limitations in terms of therapeutic efficacy. These limitations arise from their lack of inherent cytotoxic activity, relatively low affinity for their targets, relatively low stability, and potential immunogenicity. To address these limitations and improve the therapeutic potential of protein therapeutics, extensive research has been devoted to developing innovative chemical reactions that enable the modification of therapeutic proteins with diverse payloads.

The first-generation chemical modification of therapeutic proteins aims to enhance their potency by performing chemical reactions to create protein-small molecule conjugates. The most commonly used methods for chemical modification include N-hydroxysuccinimide (NHS) and maleimide-mediated chemistry, which attach functional molecules non-specifically to lysine or cysteine residues, respectively (Fig. 1A)¹¹. Other classic modification methods utilizing reactive lysine and cysteine residues include isothiocyanate reaction, sulfonyl chloride reaction, vinyl sulfone reaction, amination reaction with aldehyde, and squaric acid reaction¹¹. These methods are compatible with aqueous solutions, neutral pH, and room temperature. One of the major applications of these reactions is in the preparation of antibody-drug conjugates (ADCs) by conjugating toxins to antibodies. Several ADCs have been developed and approved by the FDA for various cancer treatments using the first-generation chemical modification approach^12,13. However, these methods can only result in heterogeneous products with variable drug-to-antibody ratios (DARs) due to the random labeling nature of the modification, which cannot be further purified. As a result, these heterogeneous ADCs exhibit suboptimal therapeutic efficacy compared to their homogeneous counterparts^14,15. This issue also applies to other therapeutic proteins, as the random labeling can lead to appendage blocking bio-reactive residues and causing a loss of protein activity, significantly limiting the utility of this non-specific strategy.

Figure 1.

Protein modification approaches and representative reactions. (A) The first-generation protein modification approach employs N-hydroxysuccinimide (NHS) and maleimide-mediated reactions. (B) The second-generation protein modification approach incorporates copper-catalyzed azide-alkyne cycloaddition (CuAAC) and strain-promoted alkyne-azide cycloaddition (SPAAC) reactions. (C) Proximity-induced chemistry involves reactions between proximity-enabled reactive groups and specific natural amino acids in close proximity.

Due to the limitations associated with the first-generation protein modification, the second-generation approach has gained significant attention. This approach involves site-specific modification of target proteins using bioorthogonal click chemistry (Fig. 1B). Bioorthogonal chemistry comprises a set of chemical reactions that can occur within living systems without disrupting normal biochemical processes^16,17. These reactions are bioorthogonal, meaning they are selective and do not react with other biomolecules in the system. Some commonly employed reactions in biological systems include oxime ligation^18,19, Staudinger ligation^20,21, copper-catalyzed azide-alkyne cycloaddition (CuAAC) reaction^22,23, strain-promoted alkyne-azide cycloaddition (SPAAC) reaction^24,25, and tetrazine-based inverse electron-demand Diels-Alder (IEDDA) reaction^26,27. These methods exhibit selective reactivity towards target moieties, offering excellent regio-selectivity and compatibility with living organisms^16,28. However, the second-generation protein modification approach requires the integration of two specific moieties onto the conjugation partners, often necessitating extensive engineering of the labeling proteins. Consequently, this poses additional limitations for the biomedical application of these techniques in vivo, particularly due to the difficulties associated with selectively labeling protein targets in live organisms.

The next-generation protein modification approach overcomes the requirement of engineering labeling proteins by enabling the site-specific covalent bond formation of a native amino acid residue through proximity-induced chemistry (Fig. 1C). Proximity, or the close physical closeness of molecules, plays a crucial role in driving two reactants to approach each other at the Ångström level. Bringing molecules into close proximity significantly enhances reaction rates, as the likelihood of an effective collision between molecules is a function that increases exponentially with proximity^29,30. Proximity-induced chemistry facilitates chemical transformations between a proximity-enabled reactive group and a specific natural residue of proteins through the complex-induced proximity effect. To achieve high selectivity of the proximity-induced reaction, the proximity-enabled reactive group must remain stable and unreactive with residues on the protein surface, free amino acids, and other biomolecules under physiological conditions. This ensures that the proximity-enabled reactive moiety remains inactive until it reaches close proximity to the target site, where it selectively reacts with the desired natural residue in the binding pocket. Consequently, the use of proximity-induced chemistry for site-specific protein labeling eliminates the need for protein engineering to introduce unique chemical moieties. This advancement opens up exciting opportunities for preparing and developing novel protein therapeutics²⁹. Furthermore, proximity-induced chemistry has been used to prepare covalent protein drugs. The concept of incorporating a proximity-enabled reactive group, typically a Michael acceptor, to selectively react with a cysteine residue in the binding site of the target protein has been previously employed to enhance the affinity of small molecule drugs. Initially, this approach was investigated to develop selective inhibitors for kinase mutants and has more recently been applied to optimize various inhibitors³¹. Several reviews have already covered the recent advancements in the development of these covalent inhibitors^29,32–34, so we will not delve into those in this particular review. Instead, our focus will be on providing a comprehensive summary of the recent progress in utilizing proximity-induced chemistry to advance protein therapeutics.

1. THE CHEMISTRY BEHIND PROXIMITY-INDUCED PROTEIN THERAPEUTICS

The diversity of chemistry within natural proteins is limited due to the presence of only 20 canonical amino acids. Nucleophilic residues such as lysine, cysteine, and tyrosine are the most commonly used residues for carrying out chemical reactions (Fig. 1C and Table SI). The reactions with other weaker nucleophiles like histidine, aspartate, and glutamate have also been reported but to a lesser extent. These good nucleophilic residues are capable of participating in substitution reactions and nucleophilic addition reactions. When designing proximity-induced chemistry, the proximity-enabled reactive group is chosen to covalently react with one of these specific amino acid residues only in proximity. Among various proximity-induced chemistries, reactions with lysine residues are preferred because lysine is relatively abundant on protein surfaces and often found at the protein-protein interfaces of interest³⁵. On the other hand, strong nucleophiles like cysteine are less common on protein surfaces and are frequently involved in disulfide bond cross-linking³⁶. Apart from their intrinsic chemical properties, the nucleophilicities of amino acid residues can be profoundly influenced by their microenvironments.^37,38 This means that amino acids nearby or adjacent to a nucleophilic residue can modulate its reactivity through various interactions, such as electrostatic forces, hydrogen bonding, metal ion coordination, and hydrophobic effects. For instance, while free cysteine has a pKa value of 8.3, its pKa can vary from as low as 3 to over 10 depending on its specific microenvironment in a protein. Consequently, the protein’s microenvironment can drastically alter the reactivity of nucleophilic amino acids, underscoring the importance of carefully selecting positions to introduce proximity-enabled reactive groups.

For an ideal proximity-enabled reactive moiety, it should exhibit high reactivity toward specific natural amino acid residues when they are in close proximity. However, it is crucial for this chemical moiety to remain inert under physiological conditions. Therefore, it is important to carefully modulate the reactivity of the chemical moieties used for proximity-induced chemistry. This involves finding a balance between reactivity and stability since stronger reactivity may make the moiety susceptible to broad chemical species in the natural system, leading to non-specific protein labeling.

1.1. Substitution reaction with fluoride as a leaving group

One major class of proximity-induced chemistry involves nucleophilic substitution reactions, where nucleophilic amino acid residues such as lysine, cysteine, and others attack weak electrophiles, leading to the displacement of a leaving group (Fig. 1C). In proximity-induced chemistry, the leaving groups predominantly consist of halogen atoms (F, Cl, and Br), as well as p-fluorophenyl. These leaving groups remain inert under aqueous and physiological conditions but become reactive when in proximity to nucleophilic residues, such as lysine or cysteine.

Fluoride, due to its small size and high charge density, is considered the least favorable leaving group among the reactive residues. In traditional S_N2 reactions, F⁻ rarely acts as a leaving group due to its low polarizability. However, when two residues come into close proximity, the reactivity of cysteine residues towards weak electrophiles containing fluoride as a leaving group is notably enhanced. To further improve the reactivity, α-carbonyl groups are introduced to facilitate the attack of cysteine residues with very high specificity (Fig. 1C and Table S1, 1)³⁹. Despite the high specificity of these proximity-induced groups for cysteine, the yield is relatively low, with less than 50% efficiency due to the weak reactivity of fluoride⁴⁰. Previous studies have demonstrated that perfluoroaromatic molecules react with cysteine thiolate via an S_NAr mechanism⁴¹. Consequently, in a proximity-induced setting, the pentafluorophenyl-diazenyl (Fig. 1C and Table S1, 2) group has proven effective in achieving site-specific modification of cysteine residues, yielding complete conversion after 6 h at 25°C^42,43.

The sulfur (VI) fluoride exchange (SuFEx) represents the latest set of click chemistry transformations^44,45. S^VI-F motifs, including arylfluorosulfates (ArOSO₂F) and iminosulfuroxyfluoride (RN═S(O)F₂), are two common forms that can be readily synthesized using sulfuryl fluoride (SO₂F₂) and thionyl tetrafluoride (O═SF₄), respectively⁴⁶. SuFEx reactions have recently been utilized for proximity-induced protein crosslinking, where S^VI-F motifs react with lysine, histidine, or tyrosine residues on proteins via SuFEx chemistry when the two reactants are in close proximity. Building upon this principle, fluorosulfonyl (Fig. 1C and Table S1, 3) and arylfluorosulfonyl (Fig. 1C and Table S1, 4) were designed and successfully incorporated into proteins, demonstrating exceptional biocompatibility and proximity-induced activities against multiple residues, such as tyrosine, lysine, and histidine^47,48. The crosslinking reactions with these residues exhibited outstanding specificities, resulting in high yields with an efficiency of over 75%^47,48.

1.2. Substitution reaction with phenyl carbamate as a leaving group

Aryl carbamates, characterized as moderate electrophiles, are stable under physiological conditions but become susceptible to being attacked when in proximity to good nucleophiles. As a result, aryl carbamates have emerged as promising candidates for proximity-induced chemistry. Exploiting this property, phenyl carbamate and fluorophenyl carbamate (Fig. 1C and Table S1, 5) have been developed as proximity-induced crosslinkers for native protein residues⁴⁹. The reactivities of these molecules were assessed in proximity-induced conditions, revealing that the phenyl carbamate group exhibits suboptimal reactivity, and the addition of a fluoro substituent to phenyl carbamate demonstrates remarkable reactivity towards cysteine, lysine, and tyrosine residues. The reaction is very impressive, resulting in more than 95% yields after 4–24 hours⁴⁹. Another phenyl carbamate-based leaving group (Fig. 1C and Table S1, 6) is connected to an affinity peptide through an amide bond at the para position. The reactivity of proximity-enabled reactive moiety 6 is slightly lower than that of moiety 5, resulting in a labeling efficiency of 80% after 1 hour⁵⁰. This reduced reactivity is likely attributed to the absence of an electron-withdrawing group on the phenyl carbamate moiety⁵⁰.

1.3. Substitution reaction with chloride, bromide, and iodide as leaving groups

In addition to fluoride, other halogen atoms can be utilized as more reactive leaving groups in proximity-induced chemistry. When comparing reactive residues with different halogen atoms linked by the same alkyl chain, the covalent bond formation efficiency follows the order of I > Br > Cl > F, which aligns with the leaving ability of halides in S_N2 reactions. Chlorine, bromine, and iodine substituents exhibit good reactivity for cysteine residue without the need for adjunct α-carbonyl groups. The alkyl halides (Fig. 1C and Table S1, 7, 8, 9, 10, 11) exhibit moderate cross-linking efficiency, ranging from 24% to 48% for the cysteine residue^51,52. Among these chemical moieties, bromoethyl (Fig. 1C and Table S1, 9) has been extensively studied and recognized for its versatility, primarily attributed to its long aliphatic side chain that enhances reactivity⁵¹. Alkyl chloride (Fig. 1C and Table S1, 8) also served as a proximity-enabled reactive moiety in the Halo-tag system for site-specific protein labeling⁵³. The Halo-tag protein, derived from bacterial haloalkane dehalogenase, was utilized in this process. Upon binding to the Halo-tag protein, the Halo-tag ligand with an alkyl chloride chain exhibited remarkable selectivity, reacting exclusively with the Asp residue in the binding pocket of the Halo-tag protein, resulting in over 90% labeling efficiency.⁵⁴.

1.4. Addition reaction

Extensive research has been conducted on the irreversible, covalent inhibition of kinases and proteases through alkylation, acylation, phosphorylation, and other modifications of reactive amino acid residues⁵⁵. Among the various strategies employed, the Michael addition of reactive residues with Michael acceptors is one of the widely used strategies^56,57.

Michael acceptors, such as α, β-unsaturated ketones and vinyl sulfones, are frequently utilized for conjugate addition with nucleophiles like lysine and cysteine residues. Vinyl carbonyl, vinyl amidyl, and vinyl sulfonamidyl groups (Fig 1C and Table S1, 12–14) have been introduced into proteins⁵⁸. These groups exhibit stability under physiological conditions and display relatively low non-specific reactivity towards other serum and cellular proteins. Notably, the protein with a vinyl sulfonamidyl group (Fig. 1C and Table S1, 14) demonstrates the highest reactivity, resulting in >95% covalent cross-linking of human epidermal growth factor receptor 2 (HER2) protein and its antibody⁵⁸. Conversely, the protein with a vinyl amidyl group (Fig. 1C and Table S1, 13) exhibits moderate reactivity, while the protein with an N-alkylacrylamide group (Fig. 1C and Table S1, 12) shows relatively lower reactivity. These findings align with S^VI-F probes, suggesting that the presence of an electron-withdrawing sulfone moiety with a sulfur (VI) center significantly enhances proximity-induced chemistry reactivity.

Phenyl isothiocyanates have been employed in Edman degradation peptide sequencing. Under mildly alkaline conditions, phenyl isothiocyanate reacts with an uncharged N-terminal amino group, forming a cyclic phenylthiocarbamide derivative^59,60. The center carbon of the isothiocyanate group is electron-deficient, enabling attack by the amine group and the formation of a thiourea bridge. An isothiocyanate derivative (Fig. 1C and Table S1, 15) based on tyrosine demonstrates high efficiency (>80%) when proximal to a lysine residue⁶¹.

2. APPROACHES TO INTRODUCE PROXIMITY-ENABLED REACTIVE GROUPS INTO PROTEINS

The reactivity of proximity-induced chemistry relies on the distance and spatial geometry between the proximity-enabled reactive moieties within the binder and the specific natural residues of targeted proteins. Therefore, it is crucial to introduce these proximity-enabled reactive moieties into the binder in a site-specific manner. When the binder is a protein, the most commonly employed strategy involves using Genetic Code Expansion technology^62–65. This technique allows for the incorporation of proximity-enabled reactive noncanonical amino acids (ncAAs) that can react with neighboring natural amino acid residues in close proximity^{39,49,51,58,66,67}. Other strategies, such as split intein-mediated protein ligation, solid-phase peptide synthesis, and similar approaches, are utilized to bring the proximity-enabled reactive moieties closer to natural amino acids^68–72. When the binder consists of small molecules, a structure-activity relationship analysis can be conducted to optimize the spatial geometry of the proximity-enabled reactive moieties in relation to the specific natural residue of the targeted protein. These efforts have resulted in the development of covalent-based protein tags like SNAP-tag, Halo-tag, and others^53,73–75. Additionally, proximity-induced reactive peptides have been prepared using solid-phase peptide synthesis for antibody labeling purposes^69,76–78. In this section, we provide a summary of these innovative approaches for the precise incorporation of proximity-induced reactive moieties into proteins.

2.1. Genetically encoded ncAAs with proximity-enabled reactive moieties into proteins

To achieve the site-specific introduction of unnatural reactive moieties into proteins in vivo, Genetic code expansion technology offers a highly effective approach^62–65. Typically, a noncanonical amino acid (ncAA) is designed with a distinct chemical handle based on a tyrosine or pyrrolysine scaffold. This engineered ncAA can then be incorporated into proteins at specific sites in both prokaryotic and eukaryotic cells using a bioorthogonal aminoacyl-tRNA synthetase (aaRS)/tRNA pair that responds to the amber stop codon. Through the genetic incorporation of a ncAA equipped with a proximity-enabled reactive moiety, it becomes possible to establish precise and stable covalent linkages between proteins, enabling tailored crosslinking between ncAA and targeted amino acid residues in a site-specific manner (Fig. 2A). Furthermore, a diverse array of photo-activatable crosslinking ncAAs have found utility in protein engineering, propelling the development of innovative protein therapeutics. Several reviews have summarized the principles and utilities of these photo-activatable crosslinking ncAAs^79,80. In this review, we will focus on proximity-enabled ncAAs that don’t require any post-treatment to elicit proximity-induced reactions.

Figure 2.

Genetic incorporation of ncAAs with proximity-enabled reactive moieties into proteins. (A) Genetic Code Expansion technology enables the incorporation of bioreactive ncAAs with proximity-enabled reactive moieties into proteins. These moieties can form covalent bonds with native residues on interacting proteins through proximity-induced reactions. (B) Structural of bioreactive ncAAs utilized in proximity-induced chemistry, based on either a tyrosine or pyrrolysine scaffold. (C) F_fact selectively reacts with the cysteine residue when two residues are in close proximity. (D) VSF selectively reacts with the lysine residue when two residues are in close proximity.

Based on the tyrosine scaffold, the first proximity-enabled reactive ncAA with a fluoroacetylphenyl moiety, fluoroacetylphenylalanine (F_fact) was designed and incorporated into proteins in Escherichia coli (Fig. 2B and 2C, Table 1)³⁹. F_fact was designed to selectively react with nearby cysteine residues, allowing for precise crosslinking with the protein targets. In order to showcase its proximity-induced chemistry, F_fact was incorporated at position Asp36 within the Z_SPA affibody (Afb) using the bioorthogonal Methanococcus jannaschii tyrosyl-tRNA synthetase (MjTyrRS)/tRNA_CUA system³⁹. The wild-type Afb and the Z protein exhibited a binding affinity of 6 μM. Upon introducing F_fact, the resulting Afb mutant showed a crosslinking yield of over 60% with the Z protein containing an Asn6Cys mutation^39,81. This study is particularly noteworthy as it marks the first successful example of Genetic Code Expansion technology to introduce a proximity-induced chemical moiety for protein crosslinking. Expanding on these discoveries, a series of haloalkyl ncAAs with varying side chain lengths and reactivities, including O-(3-chloropropyl)-tyrosine (CprY), O-(3-bromoethyl)-tyrosine (BetY), O-(3-bromopropyl)-tyrosine (BprY), and so on, were synthesized and integrated into proteins using a specific mutated Methanosarcina mazei pyrrolysyl-tRNA synthetase (MmPylRS)/tRNA_CUA variants (Fig. 2B and Table 1)⁵¹. Notably, the efficiency of crosslinking varied depending on the halogen atom present in the ncAAs, following the order of I>Br>Cl, which aligns with the halide leaving ability in S_N2 reactions. Moreover, when comparing ncAAs with different side chain lengths, BetY, which possesses a two-carbon linker, demonstrated the highest crosslinking efficiency between the Afb mutant and Z protein⁵¹. Additionally, intramolecular proximity-induced crosslinking was employed to enhance protein stability. Intriguingly, the formation of a novel intramolecular covalent bond between BetY and a cysteine residue of Afb protein resulted in a remarkable increase in its T_m value from 46.7 ± 0.2 °C to 60.4 ± 0.7 °C⁵¹. This observation highlights the significant impact of proximity-induced crosslinking on bolstering protein stability.

Table 1.

NcAAs with proximity-enabled reactive moieties

No	ncAA	Modified protein/Interacting protein	Target residues	Expression systems	Reference
1	Ffact	Afb/Z protein	Cys	Bacteria	39
2	VSF	Anti-HER2 Fab of Herceptin/Extracellular domain of HER2	Lys	Bacteria	58
3	AcrF	Anti-HER2 Fab of Herceptin/Extracellular domain of HER2	Lys	Bacteria	58
4	pNCSF	Z protein/Afb	Lys	Bacteria	61
5	Haloalkyl ncAAs	Afb/Z protein	Cys	Bacteria & Mammalian cells	51
6	Cl-PSCaa	The central helix of calmodulin (CaM)/CaM	Cys	Bacteria & Mammalian cells	52
7	F-PSCaa	CaM/CaM	Cys	Bacteria & Mammalian cells	43
8	EB3	Z protein/Afb	Cys	Bacteria & Mammalian cells	85
9	FAcK	Afb/Z protein	Cys	Bacteria & Mammalian cells	40
10	BrC6K	Afb/Z protein or Afb/Afb	Cys, Lys, His, Glu, Asp	Bacteria & Mammalian cells	66
11	FSY	Z protein/Afb	Lys, His, Tyr	Bacteria & Mammalian cells	47
12	FFY	nanobody mNb6/Receptor binding domain (RBD) of the SARS-CoV-2 viral spike protein	Lys, His, Tyr	Bacteria & Mammalian cells	86
13	mPyTK	glutathione-S-transferase (GST) homodimer/GST homodimer	Glu, Asp	Bacteria & Mammalian cells	87
14	o-NBAK	GST homodimer/ GST homodimer	Lys	Bacteria & Mammalian cells	88
15	AcrK	Anti-HER2 Fab of Herceptin/Extracellular domain of HER2	Lys	Bacteria & Mammalian cells	58
16	FPheK	Z protein/Afb	Cys, Lys, Tyr	Bacteria & Mammalian cells	49
17	FnbY	GST homodimer/ GST homodimer	Cys, Lys, His, Tyr, Trp, Met, Arg, Asn, Gln	Bacteria & Mammalian cells	89
18	FSK	Nanobody 7D12/Epidermal growth factor receptors (EGFR)	Lys, His, Tyr	Bacteria & Mammalian cells	48
19	FmnbY	Ubiquitin (Ub)/ Ubiquitin (Ub)	Cys, Lys, His, Tyr, Trp, Met, Arg, Asn, Gln	Bacteria & Mammalian cells	90

To harness the benefits of proximity-induced Michael addition chemistry, various tyrosine and lysine analogs with Michael acceptors have been designed and integrated into proteins⁵⁸. These ncAAs, such as p-acrylamido-phenylalanine (AcrF), p-vinylsulfonamido-phenylalanine (VSF), and N-acryloyl-lysine (AcrK), display selective reactivity towards nearby lysine residues (Fig. 2B, 2D, and Table 1). Among them, VSF has shown the most favorable proximity-induced reactivity, achieving up to 95% crosslinking efficiency. In order to demonstrate the utility of VSF, it was employed to crosslink an antibody with its antigen. Based on the co-crystalized complex structure, Tyr92 residue within the anti-human epidermal growth factor receptor 2 (HER2) fragment antigen-binding (Fab) region closely interacts with two lysine residues (Lys569 and Lys593) of HER2 at the binding interface. Accordingly, VSF was incorporated site-specifically at Tyr92 within the anti-HER2 Fab using a mutated MjTyrRS/tRNACUA system in E. coli (Table 1). The resulting mutant anti-HER2 Fab exhibited rapid covalent crosslinking with HER2 on the cell surface under physiological pH conditions⁵⁸.

Sulfur (VI) -fluoride exchange (SuFEx) reactions have emerged as a novel class of click chemistry transformations involving the rapid replacement of an S^VI-F bond with an S-O or S-N bond^44,45. To achieve site-specific incorporation of S^VI-F motifs into proteins, two designed compounds, namely fluorosulfate-tyrosine (FSY) and fluorosulfonyloxybenzoyl-lysine (FSK), enable proximity-induced sulfur-fluoride exchange (SuFEx) reactions with native Lys, His, and Tyr residues (Fig. 2B and Table 1)^47,48. In a Z_spa affibody (Afb) and Z protein binding model, the modification of Afb with FSY demonstrates cross-linking efficiencies of 59%, 53%, or 35% for Lys, His, or Tyr residues at position 7 of the Z protein, respectively. However, FSK exhibits a longer and more flexible aryl fluorosulfate-containing side chain compared to FSY. This unique property of FSK allows for the crosslinking of protein sites that are inaccessible with FSY⁴⁸. Based on the Afb/Z protein pair, authors also studied the kinetic character of proximity-induced covalent bond formation, which consists of two steps, initial non-covalent protein binding and covalent bond formation (Fig. 2A).⁸² The Afb/Z protein complex demonstrated non-covalent binding with a dissociation constant (Kd) of approximately 6 uM. Upon the initiation of covalent bond formation, the equilibrium is disrupted, leading to a decrease in the Kd over time. This phenomenon is attributed to the relationship Kd = K_off/K_on, where K_off represents the dissociation rate constant and K_on represents the association rate constant. As the covalent bond forms, K_off decreases, contributing to the observed decline in Kd. The second-order rate constant for Afb/Z protein covalent protein complex formation was determined to be (1.32 ± 0.04) × 10⁻² uM⁻¹ h⁻¹, indicating that the binding affinity is dramatically increased.

Building upon the versatility of Genetic Code Expansion and the utilization of proximity-induced chemistry towards natural residues, a wide range of proximity-enabled reactive ncAAs have been designed and genetically integrated into proteins (Fig. 2 and Table 1). In addition to the aforementioned proximity-enabled reactive ncAAs, other ncAAs containing phenyl isothiocyanate, fluorophenyl carbamate, and various other moieties have also been successfully introduced into proteins using Genetic Code Expansion for proximity-induced crosslinking of proteins (Fig 2B and Table 1)^79,80,83,84. By expanding the repertoire of protein modification, this innovative approach opens up new avenues for precise and specific protein modification, facilitating the creation of advanced protein-based therapeutics with enhanced functionalities and therapeutic potential.

2.2. Innovative protein tags based on proximity-induced chemistry

Inspired by the reactions of natural self-labeling proteins or auto-processing proteins, several proximity-induced chemistry-based tags have been designed^53,73,75,91. These proteins utilize their three-dimensional structures to bring proximity-induced chemical moieties to native residues in close proximity, thereby enabling covalent bond formation. These mechanisms have been utilized to create various protein tags (Fig. 3A).

Figure 3.

Strategies for Incorporating Proximity-Enabled Reactive Moieties into Proteins. (A) The proximity-induced chemistries between SNAP Tag or Halo Tag and their ligands. (B) Employing trans-splicing of split inteins to generate protein conjugates. (C) Post-modification of the affibody with acrylamide to enable proximity-induced chemistry with cysteine residues in close proximity. (D) Introduction of a proximity-enabled reactive group, fluorophenyl carbamate, into an antibody-binding peptide using solid-phase peptide synthesis. The resulting peptide enables proximity-induced chemistry with antibodies. (E) Two cysteine residues allow for the incorporation of proximity-responsive functional groups and this arrangement facilitates the detection of specific covalent binding agents through the utilization of display technologies.

One popular proximity-induced chemistry-based protein tag, called the SNAP-tag, was derived from human O⁶-alkylguanine-DNA alkyltransferase (hAGT)⁹¹. It facilitates the site-directed attachment of small molecules to proteins. When bound to O⁶-alkylguanine derivatives substituted at the 4 position of the benzyl ring, the SNAP-tag irreversibly transfers the benzyl moiety (Fig. 1C and Table S1, 17) of the substrate to its reactive cysteine residue, resulting in a covalent linkage between the SNAP-tag and the desired compounds (Fig. 3A). Expanding on the foundation of the SNAP-tag, the hAGT enzyme has also been evolved to enable covalent reactions with O²-benzylcytosine derivatives. This bioorthogonal tag, known as the CLIP-tag⁷⁵, specifically reacts with O²-benzylcytosine substrates in close proximity, while the SNAP-tag selectively reacts with nearby O⁶-benzylguanines. Thus, these two innovative technologies enable simultaneous bioorthogonal labeling, thereby expanding the prospects for the development of innovative double-labeled protein drugs⁷⁵. In addition to SNAP- and CLIP-tags, the Halo-tag is another extensively used self-labeling protein tag derived from bacterial haloalkane dehalogenase (Fig. 3A)⁵³. When ligands with alkyl chloride (Fig. 1C and Table S1, 8) moiety bind to it, the Asp106 residue of the Xanthobacter dehalogenase mutant nucleophilically attacks the terminal chloride, resulting in the formation of a covalent alkyl-dehalogenase conjugate (Fig. 3A). Self-labeling enzyme tags have significantly transformed protein labeling techniques and the development of novel protein drugs. However, these tags do present certain challenges, such as limited labeling positions and the comparatively bulky size of the self-labeling protein.

Proximity-induced chemistries offer another category of protein reactions, exemplified by intein-mediated protein splicing^70,92–94. Protein splicing is a naturally occurring process wherein an intein excises itself from a host protein through the cleavage and reformation of amide bonds in close proximity. This process is initiated by the attachment of an amide bond (Fig.1C and Table S1, 16) to a nearby nucleophilic residue (such as cysteine, serine, or threonine). The resulting thioester intermediate then undergoes trans-thioesterification and an S to N acyl shift, leading to the regeneration of the amide bond. Building upon this chemistry, various methods have been developed for therapeutic protein modifications, such as expressed protein ligation (EPL) and protein trans-splicing (PTS) (Fig. 3B)^92–94. EPL is commonly employed to generate proteins with site-specific modifications, including phosphorylation, glycosylation, lipidation, ubiquitination, acetylation, and other desired alterations^70,95. On the other hand, PTS has found applications in the synthesis of therapeutic cyclic proteins and protein conjugations. For instance, PTS has been utilized for the biosynthesis of cyclic peptide libraries, enabling the selection of α-synuclein inhibitors, as well as the production of site-specific antibody-drug conjugates and bispecific antibodies⁹⁶.

2.3. Other strategies to introduce proximity-enabled reactive moieties into proteins

To create a protein that can form the covalent bond with specific residues of a target protein, a key step is to introduce proximity-enabled reactive groups in a site-specific manner. Besides Genetic Code Expansion, other strategies have been employed, such as post-chemical protein modification and solid-phase peptide synthesis, to synthesize polypeptides with well-defined proximity-enabled reactive moieties^{69,71,72,81,81,97}.

Because of their low abundance in proteins, cysteine residues have been widely utilized for site-specific protein modification^98–100. By employing the cysteine-based Michael addition reaction, acrylamide, a weak electrophile, was introduced specifically at the Asp36 residue of an affibody⁸¹. The post-chemical modification of acrylamide on the affibody facilitated the formation of covalent bonds with nearby nucleophilic residues (Cys, His, and Lys) on Z_SPA (Fig. 3C). Importantly, this modified affibody displayed remarkable specificity towards the targeted protein, Z_SPA, both in vitro and when expressed on mammalian cell surfaces. As a result of the enhanced specificity, the enzyme-linked immunosorbent assay (ELISA) demonstrated significantly improved detection sensitivity. Additionally, the modified affibody exhibited increased stability for cell surface imaging applications. These findings provide valuable insights into the development of post-chemical modification strategies to incorporate proximity-induced chemistry for protein engineering purposes.

Solid-phase peptide synthesis (SPPS) offers a versatile approach for incorporating proximity-enabled reactive groups into small polypeptides^69,101. This can be achieved by either directly introducing a noncanonical amino acid as the proximity-enabled reactive group or performing post-chemical modification on a native lysine residue. In 2013, the small peptide BI-107D1 was discovered as a modest inhibitor for Siah, and a post-chemical modification using acrylamide was employed as the proximity-enabled reactive moiety⁷¹. The introduction of Lys-acrylamide onto the peptide facilitated the formation of a crosslink with a nearby cysteine residue situated in the binding pocket of the Siah mutant, resulting in a significant enhancement of its inhibitory efficiency. A similar strategy has been applied to prepare peptides with proximity-enabled reactive groups. Various electrophiles, such as acrylamide, chloroacetamide, and propiolamide, have been introduced into the Bim peptide to enhance its binding affinity against the oncogenic protein Bcl2A⁹⁷. The introduced electrophiles in the modified peptide exhibited the ability to form stable covalent bonds with Cys55 of Bcl2A1, which is located in close proximity to the binding site. Consequently, the modified peptide effectively inhibited the activity of Bcl2A1 in U937 lymphoma and HeLa cells, surpassing the inhibitory efficacy of the native Bim peptide ligand. Furthermore, SPPS was utilized to introduce a proximity-enabled reactive group, sulfonyl fluoride, into SAHp53-8 peptides for inhibiting the p53-Mdm2/Mdm4 interaction⁷². The sulfonyl fluoride modification of SAHp53-8 enables the formation of covalent bonds with lysine and histidine residues of Mdm2 or Mdm4 upon binding. In addition to enhancing the affinity between therapeutic proteins and their targets, proximity-enabled reactive moieties have also been incorporated into antibody-binding peptides to enable the site-specific conjugation of antibodies. By employing SPPS, 4-fluorophenyl and N-hydroxysuccinimide ester were introduced into the FB and Fc-III peptides^69,76,77. These modified antibody-affinity peptides can form covalent bonds with the lysine residues of antibodies upon binding, facilitating the site-specific modification of antibodies with drugs, fluorophores, and other proteins (Fig. 3D).

The cysteine bridging strategy has been employed to introduce novel chemical functionalities through the formation of covalent connections between two cysteine residues.¹⁰² This approach has also been harnessed for the integration of proximity-enabled reactive groups, thereby enabling the identification of covalent inhibitors using display technologies. To elaborate, a cyclic peptide library targeting the tobacco etch virus (TEV) protease was constructed using this strategy. The library was assembled by linking rebridging reagents, featuring a cysteine-reactive vinyl sulfone moiety, with two cysteine residues within the peptide library. Subsequent to this construction, the resulting cyclic peptide library underwent selection through a phage display process. The obtained peptide candidates adopted properly folded conformations that closely approached the TEV protease. This close proximity facilitated the reaction of the vinyl sulfone group with cysteine residues in the vicinity of the TEV protease, leading to the formation of covalent bonds. A similar methodology was employed, using a serine-reactive diphenylphosphonate moiety, to identify covalent inhibitors for fluorophosphonate-binding hydrolases F (Fig. 3E). This highlights the way in which incorporating proximity-enabled reactive groups into potential binders markedly streamlines the process of screening for specific covalent inhibitors using phage display technology.

3. HARNESSING PROXIMITY-INDUCED CHEMISTRY FOR PROTEIN THERAPEUTICS

Proximity-induced chemistry offers reaction selectivity by facilitating covalent bond formation between proximity-enabled reactive groups and specific natural residues within targeted proteins that are in close proximity. This approach eliminates the requirement for targeted protein engineering, UV irradiation, or additional chemical treatments, making it an optimal strategy for site-specific crosslinking of targeted proteins with desired molecules. It has found applications in preparing antibody modifications with payloads in a site-specific manner, as well as the development of covalent protein therapeutics with “infinitely high” affinity against specific protein targets. Taking advantage of binding between antibodies and peptides, a proximity-enabled reactive group has been introduced into antibody-binding peptides for the preparation of antibody-drug conjugates, bispecific antibodies with well-defined structures^{67,69,76–78}. The therapeutic efficacy of these constructs has been successfully validated in vivo. Furthermore, proximity-induced chemistry allows for the formation of irreversible covalent bonds between therapeutic proteins and their targets, thereby overcoming the inherent low affinity between them. This strategy has been instrumental in the development of covalent protein inhibitors and nanobodies, leading to improved therapeutic efficacies in vitro and in vivo^86,103,104. Most recently, this chemistry has been utilized to evolve epitope-specific antibodies capable of targeting specific antigenic epitopes, further expanding its application in antibody engineering¹⁰⁵.

3.1. Preparation of well-defined antibody drugs using proximity-induced chemistry

Wild-type antibodies typically lack inherent cytotoxic activity. To enhance their therapeutic efficacy, antibodies have been conjugated to various potent payloads, including small molecular drugs, radio agents, immune cell-targeting antibodies, and others^106–108. This strategy has led to the development of several FDA-approved ADCs. However, current approaches for preparing ADCs rely on N-hydroxy succinimidyl ester and maleimide conjugation strategies, resulting in heterogenous products with varying DARs^106,107.

Recently, proximity-induced chemistry has emerged as a promising method for generating well-defined antibody conjugates. In an initial application of proximity-induced antibody conjugation (pClick) technology, a lysine-based proximity probe called 4-fluorophenyl carbamate lysine (FPheK) was employed to covalently link to a nearby lysine residue within an antibody (Fig. 2B and 4A)^67,69. Specifically, the FPheK probe was introduced at the Glu25 position of an affinity peptide known as FB peptide, which possesses a defined binding site within the Fc region of the antibody. Upon binding to the antibody, the FPheK-modified FB peptide undergoes a spontaneous covalent bond formation reaction with a lysine residue on the antibody (Fig. 4A)^76,77. This approach successfully enabled site-specific conjugation of fluorescein or toxins to Trastuzumab, an antibody used in the treatment of breast and gastric cancers^67,69,109. Moreover, using this method, a prostate-specific membrane antigen (PSMA) inhibitor called DUPA was site-specifically conjugated to an anti-CD3 antibody, resulting in a bispecific antibody that targets both PMSA-positive prostate cancer cells and T cells⁶⁹. The therapeutic efficacy of this bispecific antibody was demonstrated in a prostate xenograft mouse model. In addition, another proximity-enabled reactive group called phenyl azidoacetate was introduced into a different antibody-binding peptide known as Fc-III⁵⁰. Upon binding to the antibody, the phenyl azidoacetate substitution at Glu8 of the Fc-III peptide can react with proximal Lys248 of the antibody. Although the reaction is rapid and proceeds to over 50% completion within 1 hour, it does not reach 100% yield, possibly due to the relatively low reactivity between phenyl azidoacetate and lysine residues.

Figure 4.

Preparation of site-specific antibody conjugates using proximity-induced chemistry. (A) The introduction of site-specific labeling and precise modifications in the antibody structure is achieved through proximity-induced chemistry by employing FPheK-modified FB protein and the affinity moiety Fc-III peptide. This crosslinking process occurs with nearby lysine residues on the antibody, utilizing pClick and CCAP technologies. (B) Utilizing a Fc-binding dirhodium metallopeptide catalyst, a cross-linking reaction is facilitated between asparagine residues on the antibody and an alkyne-bearing diazo reagent in close proximity. This strategy allows for the site-specific conjugation of functional moieties to the antibody.

The concept of proximity-induced chemistry involves the utilization of chemical moieties that remain inert under physiological conditions but exhibit high reactivity towards specific natural amino acid residues when in close proximity. This reactivity can be modulated by altering the pH or introducing catalysts in proximity. For instance, the NHS ester group demonstrates rapid reactivity with lysine residues at neutral pH, but its reactivity diminishes at lower pH levels¹¹⁰. Therefore, NHS ester has been incorporated into the Fc-III peptide. Under optimal pH conditions of 5.5, the Fc-III peptide modified with NHS ester exhibits selective reactivity against lysine residues while minimizing non-selective reactions (Fig. 4A)^76,77. Upon binding to the antibody, it specifically reacts with the Lys248 residue, resulting in 100% labeling efficiency of 95%. This modification of trastuzumab antibody with emtansine demonstrates significant efficacy in treating HER2-positive tumors in vivo studies. Instead of relying on pH modulation, a metallocatalyst capable of facilitating the reaction between carboxamide and diazo groups was employed by bringing it into proximity with the desired asparagine residue of the antibody using an antibody-binding peptide⁷⁸. An antibody-binding dirhodium metallocatalyst was created by introducing three glutamate residues into the Z domain of protein A via SPPS and subsequently metalating it with a heteroleptic dirhodium complex, Rh₂(tfa)(OAc)₃. When bound to the antibody, this metallocatalyst catalyzes the derivatization of the nearby Asn312 residue with an alkyne-bearing diazo reagent (Fig. 4B). This method proved to be compatible with human, pig, rabbit, and dog antibodies, demonstrating its broad applicability and compatibility.

3.2. Covalent Protein Therapeutics based on Proximity-induced chemistry

Previously, the concept of utilizing a proximity-enabled reactive group to selectively react with a natural residue within the target protein’s binding site has been successfully employed in the design of covalent small molecule drugs, offering advantages such as enhanced target selectivity, prolonged target engagement, the ability to overcome resistance mutations, and potentially increased potency¹¹¹. More recently, this approach has been extended to the realm of protein-based therapeutics.

The first well-characterized covalent protein drug in both in vitro and in vivo is a human programmed cell death protein-1 (PD-1) fused with a proximity-enabled reactive ncAA called fluorosulfate-tyrosine (FSY, Fig. 2B and 5A)¹⁰³. PD-1 is a transmembrane receptor that plays a crucial role in regulating T cell function^112,113. Among the complex network of immune responses, programmed death-ligand 1 (PD-L1) emerges as a key ligand for PD-1, often found in an overexpressed state in various tumor types^114–116. The affinity between PD-1 and PD-L1, with a Kd value of approximately 8.2 μM^117,118. Incorporating FSY into PD-1 using the Genetic Code Expansion technology offers a fundamentally different mechanism compared to antibody blockage. PD-1 with FSY irreversibly forms a covalent bond with the histidine residue of PD-L1 on cancer cells upon binding. Once the bond is formed, therapeutic PD-1 is no longer able to dissociate from the PD-L1 target (Fig. 5A). Under normal circumstances, the interaction between T cell PD-1 and cancer cell PD-L1 inhibits T cell proliferation and activity. However, the irreversible binding of soluble PD-1 protein with FSY to PD-L1 effectively disrupts this interaction. As a result, PD-1 modified with FSY has been shown to significantly enhance the functional activities of human T cells and chimeric antigen receptor T cells in laboratory studies, demonstrating comparable efficacy to atezolizumab, an FDA-approved anti-PD-L1 antibody¹⁰³. On the other hand, the noncovalent wild-type PD-1 did not exhibit similar effects. To further evaluate its therapeutic potential, PD-1 modified with FSY was administered in xenograft mouse models that were immune-humanized using either human peripheral blood mononuclear cells, human CAR-T cells, or a systematic bone marrow-liver-thymus humanization approach. In all three models, the PD-1 mutant with FSY showed potent inhibition of tumor growth, surpassing the therapeutic effect of PD-1 alone. These findings indicate the significant promise of covalent protein drugs and their potential to revolutionize cancer treatment.

Figure 5.

Preparation of covalent protein drugs using proximity-induced chemistry. (A) FSY-modified PD-1 covalently captures PD-L1 on tumor surfaces, leading to the restoration of T cell responses and the inhibition of tumor growth. (B) Antibodies with proximity-enabled reactive groups demonstrates stable crosslinking with the targeted antigen in close proximity, leading to superior efficacy in inhibiting the growth of tumor cells. (C) Proximity-enabled reactive moieties are integrated into antibody recruiting molecules, featuring target and antibody binding domains, to attract immune cells to tumor sites. (D) Immunization of mice with antigens with site-specific Kcr modification results in focused and enhanced antibody responses, specifically targeting the desired epitopes.

Creating antibodies with infinite affinity that possess excellent specificity against antigens without dissociation is an intriguing goal. This is particularly interesting because targeted therapeutic antibody drugs typically require high specificity and prolonged targeting to enhance efficacy. Previous approaches involved modifying antigens with proximity-enabled reactive groups to form covalent bonds with corresponding antibodies¹¹⁹. Expanding this concept in vivo, it is highly desirable to modify antibodies with proximity-enabled reactive groups instead of antigens due to the challenges associated with selectively labeling protein antigens in vivo. To address this, proximity-enabled reactive ncAAs have been incorporated into nanobodies using Genetic Code Expansion technology^48,86. The resulting modified nanobodies can establish irreversible bonds with targeted protein antigens on cells, offering a potential strategy to enhance affinity, counteract mutational resistance, and improve pharmacodynamic properties. The global outbreak of COVID-19, caused by the novel coronavirus SARS-CoV-2, has emphasized the importance of developing innovative strategies to combat virus infection¹²⁰. A crucial step in the viral cell entry process involves the binding of the viral spike protein to the human cellular angiotensin-converting enzyme 2 (ACE2) receptor¹²¹. To block the viral binding to the human ACE2 receptor, nanobodies incorporating a proximity-enabled reactive ncAA, fluorine-substituted fluorosulfate-tyrosine (FFY), have been developed⁸⁶. Upon binding to the viral spike protein, these nanobodies with FFY initiate a proximity-induced covalent reaction between the fluorosulfate group of FFY and natural residues of ACE2. These covalent nanobodies targeting ACE2 demonstrate exceptional neutralization capabilities against various strains of SARS-CoV-2, surpassing the efficacy of their noncovalent wild-type counterparts. The covalent nanobodies exhibit potent neutralization activity against not only the wild-type SARS-CoV-2 strain (with a 41-fold increase in potency) but also its variants, including Alpha (23-fold), Delta (39-fold), Epsilon (38-fold), Lambda (24-fold), and Omicron (8- to 10-fold)⁸⁶. Importantly, the superior potency of these covalent nanobodies effectively impedes viral infection by blocking the binding of SARS-CoV-2 to the ACE2 receptor. Utilizing a similar methodology, researchers have developed a covalent nanobody with a radionuclide that effectively targets HER2 for the treatment of HER2-positive breast cancer¹⁰⁴. By incorporating a proximity-enabled reactive ncAA, FSY, into the nanobody, it acquires a high affinity for its target protein (Fig. 5B). Upon binding to HER2, the nanobody establishes a covalent linkage with the Lys150 residue of the HER2 receptor, resulting in irreversible cross-linking both in vitro and in vivo. Remarkably, the resulting covalent nanobody demonstrates significantly enhanced tumor accumulation in vivo, surpassing the accumulation of the wild-type nanobody. Building upon this success, researchers have also generated a radiolabeled covalent anti-HER2 nanobody. The 225Ac-labeled covalent nanobody exhibits superior efficacy in inhibiting the growth of HER2-expressing tumors in mice compared to the wild-type nanobody¹⁰⁴. Importantly, this enhanced therapeutic agent demonstrates minimal toxicity in vital tissues. These findings underscore the immense potential of integrating proximity-enabled reactive ncAAs into nanobodies, opening up new avenues for targeted cancer therapy. Additionally, they provide valuable insights into the development of safer and more effective treatments.

Instead of modifying antibodies themselves, an approach involves introducing proximity-enabled reactive moieties into bifunctional molecules housing both a target binding domain (TBD) and an antibody binding domain (ABD, Fig. 5C)^122–124. These hybrid constructs, termed antibody recruiting molecules (ARMs), possess the ability to form covalent connections with antibodies present in human serum, thereby enlisting them to target cancer cells. Within the tumor microenvironment, this complex interaction is expected to eliminate cancer cells through mechanisms such as antibody-dependent cellular cytotoxicity (ADCC) or antibody-dependent cellular phagocytosis (ADCP). Illustrating this concept with an initial example, an acyl imidazole was strategically positioned within an ARM featuring a dinitrophenyl moiety as the ABD¹²². This configuration enabled the recruitment of anti-antidinitrophenyl (DNP) antibodies, in addition to incorporating glutamate urea ligands known for binding to Prostate Specific Membrane Antigen (PSMA). The ARM, once linked to anti-DNP antibodies, underwent an interaction where the acyl imidazole was targeted by Lys59 of the anti-DNP antibody. This event facilitated the formation of a complex between the ARM and the anti-DNP antibody, offering a potential therapeutic avenue for treating PSMA-positive cancer cells. Despite achieving success in vitro experiments, further studies are needed to ascertain the efficacy of this approach in vivo.

The mouse hybridoma technology has established itself as a prominent and widely utilized method for generating monoclonal antibodies with high affinities^125–127. However, selectively targeting a specific conformational epitope in a whole antigen in vivo remains a challenge. To overcome this challenge, the Genetic Code Expansion technology has been employed to genetically incorporate N-acryloyl-lysine (AcrK) or N-crotonyl-lysine (Kcr), both possessing cross-linking activities with natural protein residues, into specific epitopes of antigens (Fig. 2B)¹⁰⁵. Upon binding to the B cell receptors (BCRs), the proximity-enabled reactive ncAAs, AcrK, and Kcr on the antigen exhibit an impressive capability to form covalent bonds with Lys or Cys residues situated in the binding pocket of BCRs (Fig. 5D). Consequently, the affinity between antigens and BCRs is significantly enhanced, resulting in a tighter and more robust binding. By immunizing mice with antigens containing AcrK or Kcr, researchers have successfully elicited and enriched antibody responses that selectively target the specific epitope with ncAA modification. By incorporating AcrK or Kcr into human interleukin-1β (hIL-1β) and Phaeodactylum tricornutum nucleoside triphosphate transporter 2 (PtNTT2) peptide, two antigen mutants have demonstrated the remarkable ability to induce robust epitope-directed antibody responses in mice¹⁰⁵.

Conclusion

Protein therapeutics have revolutionized the medical field, offering promising treatment options for complex diseases. Over time, they have significantly advanced and contributed to the growing number of FDA-approved drugs^1,5,6. However, despite their transformative impact, protein drugs encounter certain limitations compared to small-molecule drugs. These limitations encompass factors such as low stability, relatively low cytotoxicity, high manufacturing costs, potential immunogenicity, low cellular permeability, and challenges associated with targeting undruggable proteins. To overcome the challenges posed by weak affinity interactions between binding partners and undruggable protein targets, proximity-induced chemistry has emerged as a powerful tool in the development of covalent protein therapeutics with infinitely high affinity against their targets^86,103,104. This innovative approach facilitates chemical transformations between a proximity-enabled reactive group and specific natural residues within targeted proteins that are in close proximity. By introducing proximity-induced chemistry, covalent protein therapeutics with reactive groups demonstrate enhanced binding affinity and therapeutic efficacy compared to their wild-type counterparts. Furthermore, proximity-induced chemistry has been utilized for site-specific protein labeling, enabling the introduction of diverse functional payloads to antibodies^{67,69,76–78,128}. These modified antibody conjugates exhibit improved stability and cytotoxicity both in in vitro assays and in vivo mouse models. While these successes are notable, there is still much to explore regarding the mechanistic perspective of enhanced efficacy of covalent drugs and the additional benefits they offer, such as targeting undruggable surfaces, potentially increased selectivity, and reduced immunogenicity. Comprehensive studies on covalent protein therapeutics and their advantages in diverse organisms will unlock the full potential of proximity-induced chemistry in the development of innovative protein therapeutics.

The development of protein therapeutics using proximity-induced chemistry relies on the creation of novel proximity-enabled reactive groups and the implementation of effective strategies to incorporate these groups into proteins. Currently, more than 20 proximity-enabled reactive groups have been developed, enabling the targeting of a wide range of natural residues on proteins. When designing proximity-enabled reactive groups, striking a balance between reactivity and stability is crucial. Achieving a high reactivity is generally desirable for efficient proximity-induced crosslinking. However, higher reactivity can lead to reduced specificity, potentially resulting in random reactions with undesired residues. Additionally, the protein microenvironment strongly influences the reactivity between proximity-enabled reactive groups and natural protein residues. Consequently, the screening of proximity-enabled reactive groups with varying reactivity becomes a time-consuming and labor-intensive process in order to identify the most suitable group for a specific target. To expedite this process, computational approaches can be employed to enhance efficiency and streamline the selection of optimal proximity-enabled reactive groups for a specific target. To introduce proximity-enabled reactive groups into proteins in a site-specific manner, solid-phase peptide synthesis and Genetic Code Expansion technology have been widely employed for covalent peptide and protein preparation, respectively. Although Genetic Code Expansion technology enables the incorporation of proximity-enabled reactive groups, it may lead to lower protein yields compared to wild-type proteins. Therefore, further synthetic biology strategies are necessary to address these challenges and enhance protein yields in proximity-induced chemistry.

In conclusion, the advancement of protein therapeutics through proximity-induced chemistry relies on the synergistic contributions of chemistry, synthetic biology, and protein engineering. Initial studies on protein therapeutics based on proximity-induced chemistry have demonstrated promising therapeutic efficacy. Further evolution of protein therapy using proximity-induced chemistry holds great potential for the treatment of cancer and other diseases.

Highlights.

• The molecular mechanism behind proximity-induced chemistry is thoroughly explored.

• The proximity-enabled reactive groups are classified and summarized.

• Approaches to introduce proximity-enabled groups into proteins are summarized.

• The development of protein drugs by proximity-induced chemistry is discussed.

Significance.

Protein therapeutics have revolutionized the medical field, offering promising treatment options for complex diseases. Thriving protein therapeutics have largely contributed to a growing number of FDA-approved of over 200 therapeutic proteins. In this context, we made a comprehensive summary of mechanisms and achievements of proximity-induced chemistry on protein therapeutics to advance the field and guide future development. We first introduced the concept of proximity-induced chemistry, discussed the molecular mechanism behind proximity-induced chemistry, function groups that are able to carry out this reaction, and the methodology to incorporate reactive groups into proteins. Then we focused on the transition of this chemistry to boost the development of protein therapeutics and their therapeutic efficacy in pre-clinical and clinical studies. We summarized current achievements of proximity-induced chemistry as a novel strategy to develop covalent protein therapeutics with infinitely high affinity against their targets and site-specifically modified therapeutic protein with potent payload without requiring extensive target protein engineering. This review will be useful as a toolbox for the modification of native antibodies to generate more potent homogenous antibody conjugates and develop new antibody conjugation methods without requiring protein engineering of native antibodies. Moreover, it will open a new era of covalent protein drugs that drastically improve the therapeutic outcome of these drugs that suffer from low target affinity or stability and will boost the development of protein drugs that target undruggable targets.