Chemical Biology 2025: Highlights From the Ch/Bi145 Course at Caltech

Abstract

This review highlights key advances in chemical biology reported in 2025, curated by students in the Advanced Chemical Biology course (Ch/Bi145) at the California Institute of Technology (Caltech). The selected studies span advances in bioconjugation and bioorthogonal chemistry, visualization and sensing, interaction profiling, precision control of biological systems, emerging therapeutic modalities, and synthetic chemical biology. Together, they illustrate how innovative molecular design continues to expand our ability to interrogate and manipulate biological systems.

This student‐authored review highlights notable advances in synthetically enabled chemical biology from 2025, spanning bioconjugation and bioorthogonal chemistry, visualization and sensing, precision control of biological systems, synthetic chemical biology, emerging modalities in drug discovery, and interaction profiling.

1. Introduction

This review highlights selected advances in chemical biology reported in 2025. The selection of papers, manuscript preparation, and figure design were carried out through the collective effort of students and instructors in the Advanced Chemical Biology 2026 course (Ch/Bi145) at Caltech. The review is organized into six major areas of chemical biology research that together reflect the breadth of the field:

• 1.Bioconjugation and bioorthogonal chemistry
• 2.Visualization and sensing
• 3.Interaction profiling
• 4.Precision control of biological systems
• 5.Emerging modalities in drug discovery
• 6.Synthetic chemical biology

Peter Dervan's pioneering work at the chemistry–biology interface was motivated by the unique ability of synthetic organic chemistry to design new “molecules with function” [1]. In this spirit, Ch/Bi145 explored the design and application of synthetic molecules to probe, control, and reprogram biological systems. Topics included covalent modulators, fluorophores and molecular probes, crosslinkers, “bumped” small molecules for analog‐sensitive chemical genetics, protein editors, photocages and photoswitches, and bifunctional molecules for targeted degradation, posttranslational modification, or cellular relocalization, as well as bitopic inhibitors, molecular glues, and synthetic metabolites capable of reprogramming cellular networks. We note that modern chemical biology extends well beyond small‐molecule approaches and includes major advances based on engineered proteins, nucleic acids, and synthetic biological systems. These areas lie outside the scope of Ch/Bi145 and this review. Although the course did not rely on a single resource, we frequently referred to the recently published textbook Advanced Chemical Biology (Wiley VCH) [2] alongside extensive discussion of the primary literature.

Each student in the class reviewed the chemical biology literature, selected a chemical biology paper published in 2025, and presented it to the class. The selection criteria were broadly defined to identify contributions that reflect diverse journals, topics, and authors in synthetically enabled chemical biology. The papers were discussed collectively, and input from the entire class shaped the final selection of studies included in this review. Each student prepared a written highlight of their selected paper in the style of a News & Views article, which was reviewed by the instructor and teaching assistants. Following individual meetings to provide guidance on manuscript preparation, we conducted an internal peer‐review process in which each draft was evaluated by two other students in the course. The revised highlights were then compiled into the present review.

The concept of this review was developed in discussion with the editors of ChemBioChem, and we thank them for supporting the idea. The selection reflects the interests and discussions of the course and is not a comprehensive survey of the literature, and many outstanding contributions could not be included. We hope these highlights provide readers with a timely perspective on recent developments and inspire further exploration of the rapidly evolving field of chemical biology.

2. Advances in Bioconjugation and Bioorthogonal Chemistry

2.1. Covalent Inhibitor for AKT1 (E17K)

Covalent inhibitors represent a common strategy in drug design. By employing a reactive functional group on the inhibitor molecule, covalent interactions are formed with a target protein to modulate its activity for a therapeutic effect. Although this modality has been known since the late 18th century, the past three decades have seen considerable advancements in its development. Currently, many highly successful pharmaceuticals including ibrutinib, osimertinib, and sotorasib are covalent inhibitors [3, 4]. Nonetheless, certain targets still present challenges in achieving proper reactivity and/or selectivity for inhibitors. Writing in Nature, Craven et al. [5] describe selective covalent lysine targeting of AKT1 (E17K), an oncogenic mutant in various solid tumor cancers. During experimentation, they uncovered a chelation effect by endogenous zinc that enhanced the inhibitor's selectivity for the mutant form.

Prior studies in the laboratory of Taunton have focused on selective targeting with sulfonyl fluorides [6] and more recently with salicylaldehydes [7] as electrophilic substrates for coupling with lysine residues. In the present work, they employed a salicylaldehyde moiety for imine formation at the aldehyde function (Figure 1). The scaffold for their salicylaldehyde warhead was modeled after the clinical allosteric inhibitor ARQ092 [8]. This compound is an effective AKT (E17K) inhibitor. However, due to additional lysine residues in the binding pocket that are conserved across the wild‐type (WT) and other AKT paralogs, pan‐AKT inhibition is observed, ultimately leading to hyperglycemia as an on‐target side effect and dose‐limiting toxicity.

FIGURE 1.

Improved inhibitor of mutant AKT1 (E17K). (A) Chemical structure of clinical AKT inhibitor and covalent salicylaldehyde inhibitor with relevant amino acid residue interaction shown. The mutant lysine forms a covalent imine bond with the aldehyde function of the covalent inhibitor. Tetrahedral Zn²⁺ coordination is observed with the imine nitrogen, the alcohol of the salicylaldehyde, and two proximal cysteine residues. (B) Various experiments reveal selectivity of the improved inhibitor for AKT1 (E17K) over WT AKT1 and other paralogs via “residence‐time‐based selectivity.” The structure of AKT1 (E17K) is shown conjugated with the salicylaldehyde inhibitor and zinc chelation. Nanobody‐41 was included for crystallization.

The salicylaldehyde's critical “neo‐zinc” chelation was fortuitously discovered while obtaining crystal structures due to suspected trace zinc contamination. It uniquely allowed endogenous zinc to coordinate with the imine, phenol, and two proximal cysteines (Figure 1A). Furthermore, the chelation effect rotated the compound for more favorable H‐bonding and π‐stacking interactions with tyrosine and arginine residues respectively located within the binding pocket.

Craven et al. compared their chelate‐assisted inhibitor to the clinical drug in solutions of WT or mutant AKT1, and found greater thermal stabilization for the complex between their inhibitor and AKT1 (E17K) over WT AKT1. The clinical compound, by contrast, showed greater thermal stabilization for the WT complex. Next, in cellular studies, an alkyne‐derivatized inhibitor was click conjugated to TAMRA‐azide to enable in‐gel fluorescence. When a mixture of salicylaldehyde‐AKT complexes was analyzed, the inhibitor modified AKT1 (E17K) with 2.4‐fold selectivity over WT AKT1, and 140 fold selectivity over AKT2. When subjected to washout by ARQ092, the compound was removed from the non‐mutant AKT at significantly higher rates. This “residence‐time‐based selectivity,” described earlier by the Taunton lab for a similar system [9], further enhanced the effectiveness of mutant inhibition. In quantitative chemoproteomic experiments using TMT6 labelling, AKT1 was clearly the most enriched protein.

The authors continued validation studies by in vivo experimentation with tumor‐bearing mice. A fluorinated analog of the inhibitor was synthesized to achieve desired pharmacokinetics. In patient‐derived, triple‐negative breast cancer xenograft models, it successfully suppressed tumor volume growth by up to 96% at 20 days posttreatment without appreciable increase in blood glucose levels. Additionally, treatment of mice bearing SkBr3 tumor cells, a line that is WT for AKT1, did not result in the same effectiveness of suppression, thus supporting the proposed mechanism of treatment.

The “neo‐zinc” chelation ability greatly contributed to the covalent inhibitor's potency in binding AKT1 (E17K). Overall, this discovery points to potential future use in rational drug design where lysine targeting is coupled with proximal metal chelation for improved selectivity and efficacy of treatment.

2.2. Clickable Serotonin Enables Identification of Transglutaminase 2 as a Writer, Eraser, and Exchanger

Posttranslational modifications (PTMs) of histones are a form of epigenetic markers that affect the architecture of chromatin and include the addition of monoamines like serotonin or dopamine to Gln5 on the histone H3 tail (H3Q5) [10, 11]. Transglutaminase 2 (TG2) is the only enzyme currently known to catalyze these monoaminylations [12, 13]. Recent studies by Zheng et al. have revealed that TG2's functionality extends beyond monoamine addition; this enzyme also erases and exchanges serotonin, dopamine, and histamine adducts on H3Q5 [14]. This discovery advances the mechanistic understanding of these PTMs that influence neural transcriptional programming and offers insight into their contributions to gene expression.

TG2, encoded by TGM2, helps deposit monoamines onto H3Q5; however, the mechanism of their removal was previously unknown [12, 13]. To probe H3 monoaminylation, 5‐propargyltryptamine (5‐PT), a synthetic analog of serotonin, was metabolically fed to cells. Depending on the experiment, histones were then either isolated, or the cells were incubated in 5‐PT‐free media followed by histone isolation. The fluorescent probe cyanine 5 (Cy5) azide was subsequently clicked onto 5‐PT via a Cu(I)‐catalyzed azide–alkyne cycloaddition (CuAAC) reaction [10, 15] (Figure 2A). Fluorescent imaging demonstrated TGM2 ^−/− cells were not serotonylated at H3Q5, but 5‐PT addition was restored following overexpression of WT TG2 [10]. The stability of H3 serotonylation was explored in WT HeLa cells, and H3Q5ser was detected in the presence of 5‐PT but diminished following 5‐PT removal. This erasure was not observed in cells treated with TG2 inhibitors, supporting TG2's role as a monoamine eraser.

FIGURE 2.

Clickable serotonin enables visualization of H3Q5ser, catalyzed by the writer, eraser, and exchanger enzyme TG2. (A) To track serotonin incorporation on H3Q5, a clickable analog of serotonin was metabolically fed to cells before the fluorescent marker Cy5 azide was added via a CuAAC reaction. (B) Schematic depicting addition of the modified serotonin to H3Q5, catalyzed by TG2, followed by addition of Cy5 that validated TG2's role as a writer enzyme. TG2 was further identified as an eraser and exchanger enzyme of monoamines in vitro.

One mechanistic hypothesis for TG2's removal of monoamines involves the formation of a thioester intermediate between Cys277 of TG2 and the gamma‐amide of H3Q5. Cleavage of the thioester bond by water or an amine generates H3Q5E or a new monoaminylation adduct, respectively. Synthetic H3 peptides underwent stoichiometric removal or deamidation of the monoamine adduct when exposed to HEK293T cell lysates containing WT TG2. Observation of the TG2‐H3 thioester complex with WT TG2, but not TG2(C277A), indicated Cys277 is necessary for monoamine addition and exchange (Figure 2B).

In addition to serotonylation, Zheng et al. demonstrated histaminylation of H3Q5 was also catalyzed by TG2. H3Q5 and H3Q5E synthetic peptides were prepared, as well as those with 5‐PT, dopamine, or histamine added to H3Q5. Conversion into H3Q5E, H3Q5ser, H3Q5dop, or H3Q5his was observed with the requisite monoamines and either cellular lysates expressing WT TG2 or recombinant WT TG2 (Figure 2B). Peptides incubated with monoamines and corresponding TG2(C277A) mutants did not undergo monoamine conversion or removal, and monoamine exchange was validated in physiologically relevant nucleosome core particles. Thus, this work expands TG2's role from a writer to a remover and exchanger of monoamines.

Zheng et al. subsequently studied monoaminylation dynamics in vivo. Among other notable discoveries, H3Q5his, in contrast to H3Q5ser, weakened binding of the WDR5 complex to the H3 tail with trimethylated H3K4. WDR5 is a core component of the H3K4 methyltransferase complex, and this reduced interaction appeared to alter circadian gene expression. Therefore, this study supports dynamic H3 monoaminylation as an important component in the epigenetic regulation of circadian rhythmicity.

2.3. A New Mutant‐Selective Covalent Inhibitor of K‐Ras(G12D)

Ras proteins are GTPases that regulate cell proliferation. In response to mitogen signaling, Ras switches from an inactive GDP‐bound state to an active GTP‐bound state, activating downstream components of the mitogen‐activated protein kinase (MAPK) pathway. Oncogenic missense mutations in K‐Ras typically trap the GTP‐bound “on” state of the protein, driving constitutive MAPK signaling and tumor growth. The high prevalence of K‐Ras mutations in nonsmall cell lung cancers (21%), colorectal carcinomas (38%), and pancreatic adenocarcinomas (82%) [16] has fueled the development of small‐molecule K‐Ras inhibitors in recent years. Zheng and Shokat report a new covalent inhibitor that exclusively targets K‐Ras(G12D), the most common mutant, by leveraging the chemoselectivity of α‐diazoacetamide warheads [17]. Beyond its therapeutic potential, this discovery lays the groundwork for future efforts in covalently targeting carboxylate residues in biological systems.

Ras was long considered “undruggable” due to the lack of deep binding pockets on its surface and its high nucleotide affinity, precluding both allosteric and competitive inhibition [18]. In 2013, Shokat discovered a new cryptic allosteric pocket, termed Switch‐II Pocket, and covalent inhibitors of oncogenic K‐Ras(G12C) are now clinically approved [19, 20, 21]. However, the G12D mutation has remained challenging to covalently engage due to the comparatively weaker nucleophilicity of the aspartate side chain, necessitating the development of a new targeting strategy.

Previously reported covalent inhibitors of K‐Ras(G12D) employ ring‐strained electrophiles to target the mutant aspartate, but these compounds also react with G12C mutants, suggesting that they are susceptible to off‐target binding with cellular thiols [22, 23]. In an earlier work, Shokat and coworkers demonstrated that diazo warheads react with carboxylates but at very slow rates [24]. With this knowledge in hand, Zheng and Shokat designed Diazo‐G12Di‐1, which features the reversible K‐Ras(G12D) SII‐P binding ligand MRTX1133 [25] with an α‐diazoacetamide warhead to covalently engage the D12 residue.

Using isotopic labeling and mass spectrometry experiments, the authors demonstrated that Diazo‐G12Di‐1 modifies both GDP‐ and GTP‐bound K‐Ras D12 through a denitrogenative alkylation reaction (Figure 3A). The ability of Diazo‐G12Di‐1 to covalently bind both nucleotide states of K‐Ras(G12D) has clinical significance, as it may enable more rapid and complete MAPK pathway inhibition than compounds that only target one nucleotide state. The authors also found that Diazo‐G12Di‐1 displayed no reactivity toward other K‐Ras mutants, including G12C, G12E, G13D, or wild‐type G12. Consistent with this selectivity, Diazo‐G12Di‐1 treatment exclusively inhibited proliferation and ERK phosphorylation in K‐Ras(G12D) cell lines, indicating effective suppression of MAPK signaling (Figure 3B). The exceptional chemoselectivity of Diazo‐G12Di‐1 positions it as a promising candidate for K‐Ras(G12D)‐positive cancers while underscoring its potential utility as a tool for chemical biologists to use in expanding the covalently targetable proteome to Asp/Glu.

FIGURE 3.

Diazo‐G12Di‐1 is a mutant‐selective covalent inhibitor of K‐Ras(G12D). (A) Diazo‐G12Di‐1 features an α‐diazoacetamide warhead appended to a Switch‐II Pocket binder. The compound selectively and irreversibly cross‐links the K‐Ras D12 residue via a denitrogenative alkylation reaction. (B) K‐Ras(G12D)‐harboring cell lines constitutively activate the MAPK signaling pathway, resulting in tumor growth. When these cell lines are treated with Diazo‐G12Di‐1, MAPK signaling is repressed, resulting in tumor cell growth inhibition.

2.4. A Brighter Future for Bioimaging With SNAP‐tag2

Self‐labeling protein tags (SLPs) are powerful chemogenetic tools that enable the covalent attachment of synthetic fluorescent probes to proteins of interest [26]. Among the most popular SLPs is SNAP‐tag, an engineered human O⁶‐alkylguanine‐DNA alkyltransferase (hAGT) that covalently reacts with benzylguanine or chloropyrimidine‐linked fluorescent probes. SNAP‐tag has been widely used in bioimaging since its introduction in 2003 [27]. Despite its popularity, SNAP‐tag exhibits slower labeling kinetics [28] and lower fluorogenicity [29] with rhodamine‐based substrates compared to another widely used SLP, HaloTag. These limitations have constrained SNAP‐tag performance, particularly in high‐resolution live‐cell imaging applications.

Johnsson and colleagues report the development of SNAP‐tag2, a reengineered SNAP‐tag variant that markedly improves both reaction kinetics and fluorescence output when paired with optimized substrates [30]. To optimize the substrate, the team synthesized 17 different heterocyclic scaffolds and 12 linker variants conjugated to tetramethylrhodamine (TMR) to screen for the highest reactivity with SNAP‐tag. The top‐performing combination, trifluoromethyl fluorobenzyl pyrimidine (TF‐TMR), exhibits 3.9‐fold faster labeling kinetics in vitro and 1.7‐fold higher fluorescence in live cells compared to the conventional substrate (Figure 4A).

FIGURE 4.

Cooptimization of enzyme and substrate enables faster and brighter labeling by SNAP‐tag2. (A) Reaction scheme illustrating the labeling of protein of interest (POI) fused to SNAP‐tag2 via nucleophilic substitution reaction by the reactive cysteine of SNAP‐tag2 on the optimized substrate, trifluoromethyl fluorobenzyl pyrimidine (TF‐TMR). (B) Engineering strategy schematic in SNAP‐tag2 development. Computational design, directed evolution, and deep mutational scanning were used to improve protein stability and reactivity, while substrate screening identified an optimized scaffold with improved kinetics performance. The combined enzyme‐substrate pair yields substantially faster labeling rates and higher fluorescence intensity.

In parallel, the protein was extensively reengineered. The authors first used PROSS [31] and RosettaRemodel [32] to design variants that improve the protein's thermal stability. They then performed saturation mutagenesis near the substrate‐binding site, followed by deep mutational scanning across the entire protein to screen for beneficial substitutions. The resulting SNAP‐tag2 contains 11 residue substitutions and an 8‐residue replacement of a previously unstructured 18‐residue region. When paired with TF‐TMR, SNAP‐tag2 achieves an apparent rate constant approaching 10⁷ M⁻¹s⁻¹, approximately 100‐fold faster than the previous enzyme‐substrate pair and similar to the kinetics of HaloTag (Figure 4B).

To demonstrate the utility of SNAP‐tag2 in live‐cell imaging, the authors evaluated its performance in multiple bioimaging applications. The enhanced brightness of SNAP‐tag2 in super‐resolution stimulated emission‐depletion (STED) microscopy led to improved signal‐to‐noise ratios and imaging resolution. Interestingly, SNAP‐tag2 shows efficient fluorescent labeling in yeast cells—a traditionally challenging system due to the presence of a cell wall and multidrug efflux pumps. These results suggest that SNAP‐tag2 expands the applicability of SLP‐based labeling across diverse biological systems.

2.5. Labeling aGPCRs Reveals Conformational Changes From Allosteric Modulation

Adhesion G protein‐coupled receptors (aGPCRs) are a large class of membrane proteins involved in many physiological pathways. Characterized by its 7 transmembrane domains with a large extracellular N‐terminal fragment (NTF), aGPCRs have been important drug targets, but the NTF often obscures the binding pocket of the protein. Therefore, understanding the mechanism of how aGPCRs modulate and engineering peptides to enable allosteric modulation would be crucial for therapeutic applications. Zheng and coworkers report that a bivalent nanobody, Nb23‐bi, can be engineered such that an aGPCR crucial for male fertility, Adhesion G Protein‐Coupled Receptor G2 (ADGRG2), can be allosterically modulated by binding to the NTF. This modification increases the potency of the endogenous ligand, dehydroepiandrosterone (DHEA). Here, by utilizing chemical and structural biology techniques, Zheng et al. elucidated the mechanism by which the nanobody increases the potency of DHEA, which allowed them to explore in vivo application in mice [33].

Utilizing yeast display [34], the authors screened through a large library of nanobodies that would bind selectively to the actively‐DHEA‐bound form of ADGRG2 and selected against nanobodies that would bind to the inactive or truncated forms of ADGRG2 through magnetic‐activated sorting and fluorescence‐activated cell sorting. Two nanobodies were isolated, Nb23 and Nb32, which were further subjected to more validation experiments. The authors also included the corresponding bivalent nanobodies (Nb23‐bi and Nb32‐bi) since it's been shown that multivalent nanobodies may increase binding affinity [35].

A peptide tagging approach is utilized to understand the conformational dynamics of the GPCR. More specifically, a fluorescein arsenical hairpin binder specific for the tetracysteine tag (CCXXCC) is used in tandem with bioluminescence resonance energy transfer (FlAsH‐BRET) to optically observe conformational changes upon the binding of the nanobody [36, 37, 38]. Here, the authors inserted the fluorescent motif NanoLuc (NLuc) at extracellular loop 2 (ECL2) and FlAsH peptide motifs at ECL3 site 1 and site 2 through chemical labeling (Figure 5). Upon conformational changes that modulate the distance between NLuc and the FlAsH motif, changes in BRET occur that can be optically recorded. By leveraging FlAsH‐BRET to label ADGRG2, Zheng and coworkers were able to conclude that the binding of the nanobody led to the ECL3 and ECL2 to shift away from each other, leading to an increased potency of DHEA.

FIGURE 5.

(A) Tagging of Peptide with FlAsH‐EDT₂. (B) Conformational change caused by the binding of Nb23‐bi where FlAsH motifs are denoted in red on ECL3. Binding of Nb23‐bi led to increased potency of DHEA 3.6‐fold.

2.6. Profiling Electrophile Selectivity Across the Proteome

Over the past decade, covalent inhibitors have become a popular choice for drug discovery due to their ability to form highly specific, irreversible bonds with single amino acid residues. Although long avoided for their potential off‐target effects [39], covalent inhibitors have gained traction as drug candidates and a number of covalent inhibitors are now FDA‐approved. This shift was marked by the development of tuned electrophilic warheads. Historically, cysteine's highly nucleophilic thiol side chain has been the primary focus of covalent inhibitor development; however, cysteines are fairly uncommon in binding pockets [40], thus limiting the number of proteins that can be targeted. This has motivated growing interest in targeting other amino acid side chains, but direct comparisons of electrophile selectivity have remained difficult as published studies often rely on different methods and analysis pipelines.

Hacker and colleagues address this challenge by developing a robust and unbiased workflow to profile amino acid selectivity of covalent probes on a proteomic scale. The group exposed two identical samples of Staphylococcus aureus lysate to a broadly reactive alkyne probe and then labeled each with either a heavy or a light isotopically encoded desthiobiotin azide (IsoDTB) tag [40]. The samples were combined in a 1:1 ratio, enriched, digested, and analyzed by liquid chromatography tandem mass spectrometry. The ratio of heavy to light isotopes was used to determine alkyne competition across the proteome. To identify modification sites without assuming the targeted residue in advance, the authors extended the MSFragger FragPipe computational pipeline and performed an open search, followed by a focused search to assign and localize modification masses in an unbiased manner. The resulting residue selectivity data were summarized in a letter plot to provide a systematic view of probe reactivity.

This novel workflow removes previous methodological variability and provides a side‐by‐side comparison of the reactivity profiles of many electrophilic motives targeting different amino acids (Figure 6). Well‐characterized probes such as IA‐alkyne were shown to be reactive with cysteines [41], while new probes such as OxMet2‐alkyne and PhGO‐alkyne were identified as potential candidates for methionine and arginine targeting, respectively. One notable difference between the bacterial and human proteomes was the reactivity of PCA‐alkyne with the N‐termini of proteins. This discrepancy may be due to variations in N‐terminal acetylation levels [42], thus highlighting the need to validate organism‐specific reactivity. By providing a map to standardize electrophile reactivity studies, Hacker and colleagues move the field forward and provide the scientific community with both a common reference point and a roadmap to continue targeting the previously undruggable proteome.

FIGURE 6.

Unbiased profiling of electrophilic warhead selectivity across the proteome. (A) Electrophilic warheads with the highest reactivity for cysteine, lysine, tyrosine and methionine with sites of electrophilicity highlighted in red. (B) The reactivity of 56 alkyne probes was profiled in bacterial and human proteomes. Following probe labeling of lysates, samples were conjugated to heavy or light IsoDTB tags, combined, enriched, and analyzed by LC‐MS/MS. Data were processed using a modified MSFragger FragPipe workflow to identify modification sites and quantify residue selectivity, which was summarized in the letter plot.

2.7. Engineering Covalent RNA Ligands

RNA has long been viewed as a challenging target for chemical probe development [43]. Most existing approaches for probing RNA are non‐covalent [44, 45, 46] and compared to proteins, there are fewer tools and approaches to target RNA [47]. As a result, covalent RNA labeling techniques remain insufficiently explored [48, 49, 50, 51, 52, 53, 54]. In a recent study, Bereiter et al. present a structure‐guided strategy that begins to close this gap by engineering covalent small molecules targeting specific RNA aptamers [55]. Using an existing RNA‐ligand binding system, the authors leverage the nucleophilicity of the G5 residue and incorporate electrophilic handles into ligands to achieve covalent attachment, as shown in Figure 7.

FIGURE 7.

Covalent Binding of RNA–ligand Complex. Small molecule Brc3DPQ₁ (highlighted in red) with a short electrophilic handle tethers RNA aptamer through a specific reaction to a guanine (A) General workflow of an engineered small molecule covalently binding to Tt C15U preQ1 RNA aptamer (B) Schematic of covalent RNA aptamer binding of small molecule.

Specifically, the authors first use the bacterial preQ1 class I riboswitch as a testbed and demonstrate that an electrophilic handle attached to a specific amine group enables efficient and site‐specific alkylation of a single guanine within the RNA binding pocket. They evaluate multiple electrophilic warheads and rank them based on HPLC and mass spectrometry analyses (conversion rate comparison). Brc3DPQ1, as shown in Figure 7, emerges as the most effective small molecule, as it has the fastest conversion rate for the RNA–ligand complex. Importantly, the modified ligand retains its cellular function, downregulating translation through the riboswitch and validating that covalent engagement is compatible with RNA function in vivo.

The authors further extend their method to fluorogen‐activating aptamers. By converting the Pepper aptamer ligand HBC, a small molecule mimic of the GFP fluorophore, into MsOc3‐HBC and MsOc3‐HBC‐vinyl, they overcome the limitation of noncovalent RNA imaging probes being washed out during fluorescence imaging. Based on in vivo staining experiments, the covalent fluorophore remains bound to its RNA target after harsh washing steps, enabling robust live‐cell imaging. In addition, a vinyl‐functionalized derivative of MsOc3HBC remains functionally active while providing compatibility with pull‐down experiments. Notably, the choice of electrophilic handle must be optimized based on specific targets, as precise structure‐based interactions are a prerequisite for this strategy.

Finally, the authors convert a weak‐affinity binder of the Tt preQ1 aptamer, DPQ1, into a covalent binder, thereby significantly enhancing its interaction with the aptamer. This result indicates that the method can transform weak to moderate RNA binders from large libraries, which may lack therapeutic efficacy, into long‐lasting binders suitable for therapeutic or research applications. However, given the structural constraints of both ligands and aptamers and the reliance on specific interactions with guanine residues, further studies are needed to evaluate the broader applicability of this approach.

3. Advances in Visualization and Sensing

3.1. Visualizing Lipid Trafficking

Eukaryotic cells synthesize thousands of distinct lipid species, and each organelle maintains a characteristic lipid composition essential for its structure and function. How cells preserve membrane identities despite continuous lipid synthesis and flux remains a fundamental question in biology. A key obstacle in studying this question has been technical: Unlike proteins, lipids cannot be genetically tagged, and fluorescent analogs often perturb native behavior [56]. Nadler and colleagues overcome this limitation by combining bifunctional lipid probes with time‐resolved imaging, mass spectrometry, and kinetic modeling [57].

The authors developed a library of bifunctional lipid probes that are both photoactivatable and clickable [58, 59], enabling the UV‐induced crosslinking and subsequent fluorescent labeling via click chemistry. The probe library spans major lipid classes, including sphingomyelin (SM), phosphatidic acid (PA), phosphatidylethanolamine (PE), and phosphatidylcholine (PC), with varying degrees of acyl chain saturation. After insertion into the outer leaflet of the plasma membrane (PM), the authors tracked intracellular redistribution using fluorescence microscopy and quantified metabolic conversion using mass spectrometry (Figure 8). Kinetic modeling was then used to extract transport rate constants. These experiments revealed that retrograde transport from the PM to the endoplasmic reticulum (ER) is dominated by nonvesicular pathways, accounting for approximately 85%–95% of lipid flux.

FIGURE 8.

Bifunctional lipid probes for measuring lipid transport. (A) Representative bifunctional lipid probe classes, including sphingomyelin (SM), phosphatidic acid (PA), phosphatidylethanolamine (PE), saturated phosphatidylcholine (PC), and unsaturated PC. (B) Experimental workflow for measuring lipid transport and metabolic rates. Lipid probes were inserted into the outer leaflet of the plasma membrane using α‐methyl‐cyclodextrin‐mediated exchange, followed by crosslinking using UV light and fluorescently labeled for confocal imaging or extracted and analyzed using ultra‐high‐resolution mass spectrometry.

Furthermore, transport was also found to be species‐specific. Polyunsaturated phosphatidylcholine was transported more rapidly than saturated variants. This suggests lipid transport proteins can distinguish between different lipid types, imparting selectivity to intracellular transport.

However, selectivity alone cannot explain how cells maintain lipid gradients. While non‐vesicular transport is largely passive, membrane asymmetry is preserved by ATP‐dependent flippases. Knockdown of TMEM30A, a subunit of P4‐ATPase flippases, slowed retrograde transport of phosphatidylethanolamine, indicating that transbilayer lipid flipping is functionally coupled to inter‐organelle transfer to generate directional flux [60, 61].

The authors also compared the kinetics of lipid transport and metabolism. By measuring probe turnover with mass spectrometry, they found that metabolic conversion occurs much more slowly than inter‐organelle transport. This separation of timescales suggests that steady‐state lipid distributions in the secretory pathway arise primarily from selective non‐vesicular transport rather than from local metabolic remodeling. Collectively, this work provides the first quantitative map of lipid flux at molecular resolution.

3.2. Illuminating Glycans With Boronic Acid Based Fluorescent Probes

The composition and structure of cell surface glycans convey information about cell type and cell state, hence can serve as a critical biomarker for distinguishing normal and cancerous cells. The overexpression of two sialic acid‐containing glycans, Sialyl–Lewis A (sLe^A) and Sialyl–Lewis X (sLe^X), is known to be associated with various cancers. While boronic acid based fluorescent probes have been reported for the bioconjugation of cell surface sialylated glycans [62, 63], their low selectivity has limited their use in high‐quality imaging and in vivo applications.

Writing in JACS, Ko et al. [64] reported Sialyl Lewis Yellow (SLY), a Rhodamine B and oxaborole‐based fluorescent probe that specifically labels cells overexpressing sLe^A and sLe^X. Instead of the boronic acid used in previous studies, the authors use oxaborole as the glycan recognition moiety for its superior affinity for glycans at physiological pH (Figure 9) [65]. They explored the avidity effect by varying both the number and spatial arrangement of the oxaborole moieties and also leveraged a surface phenomenon wherein electrostatic interactions between the positively charged dye and the negatively charged cell membrane enhance binding. The potent and specific targeting effect of SLY is a result of careful design that integrates these factors. After binding to its glycan target on the cell surface, SLY is internalized via caveolae‐mediated endocytosis and trafficked to the mitochondria, leading to a sustained fluorescent signal. The authors demonstrated that SLY clearly labels tumor tissue in a diethylnitrosamine‐induced liver cancer mouse model, both ex vivo in cryosections and in vivo following intravenous injection.

FIGURE 9.

Mechanisms of Sialyl Lewis Yellow (SLY) and Rhobo6 glycan probes. (A) Sialyl Lewis Yellow (SLY) is a Rhodamine B (highlighted in red) based dye that specifically recognizes sialic acids in Sialyl Lewis A and Sialyl Lewis X through covalent engagement. Rhobo6 is a Rhodamine 110 (highlighted in red) derived dye that reversibly binds diols from various glycans through its 2 “Wulff‐Type” boronic acid groups (highlighted in gray). The carboxylic acid modification on 6‐position makes the dye membrane impermeable. (B) After binding to cell surface glycan, SLY is internalized to mitochondria. Upon glycan binding of Rhobo6, its emission maximum experiences a 14‐nm red shift and the molar absorptivity increases.

Glycan labeling is closely related to the study of extracellular matrix (ECM). ECM is intrinsically difficult to be labeled comprehensively due to the high heterogeneity in its building blocks, where antibodies and genetic tagging can only visualize one or a few of the ECM components simultaneously. However, glycosylation is a feature shared by nearly all ECM components [66].

In Nature Methods, Fiore et al. [67] utilized a boronic acid‐based fluorescent probe, Rhobo6, to visualize ECM components en masse. Rhobo6 is derived from a previously reported molecule, Rhobo [68], but with carboxylic acid installation at the 6‐position to make the molecule cell‐impermeable, reducing background due to intracellular staining. In contrast to the SLY design, Rhobo6 contains “Wulff‐type” boronic acids that have lower affinity for glycans [65], to generate a more promiscuous probe. Indeed, Rhobo6 labels 98 out of 100 glycans on a commercially printed glycan array, demonstrating its ability to detect the diverse glycans present within the ECM. The lower affinity of Rhobo6 also promotes reversible binding between the probe and its glycan targets. Furthermore, Rhobo6 exhibits a 14‐nm redshift in its emission maximum and an increase in molar absorptivity upon glycan binding, enabling wash‐free imaging. The combination of reversible binding and wash‐free properties allows samples to be stained with excess Rhobo6, creating a large reservoir of unbound probes that can replace photobleached bound probes. This dynamic exchange sustains a stable fluorescent signal over time—a particularly advantageous feature for ECM studies, which often require repeated imaging over extended timescales. Both SLY and Rhobo6 utilize boronic acid‐based glycan binding to enhance our ability to probe these molecules in different contexts.

3.3. Nanoreactors for Neurotransmitter Sensing

Efficient methods for detecting neurotransmitters (NTs) are of interest due to their essential role in the function of the central and peripheral nervous system, as well as their ability to serve as biomarkers for neural diseases [69, 70]. While a number of NT probes have been reported in the literature, current methods often exhibit poor affinity and specificity [69, 71]. Presented with these challenges, Kozibroda et al. developed a method for detecting dopamine reliably and selectively [72]. They describe a nanoreactor design combining concepts of molecular recognition and dynamic covalent chemistry, achieving an optimized limit of detection for dopamine at micromolar concentrations.

While traditional probing methods rely on a standard “recognition yields signal” approach, the key of the described design is the incorporation of a two‐part sensing system into a nanoemulsion of oils dubbed the nanoreactor [73]. The researchers employed a common structural motif, a boronic acid derivative known to be selective for the catechol structure found in dopamine and norepinephrine, as a recognition ligand [71, 74]. In the first step of the detection process, the boronic acid housed within the nanoreactor covalently binds to the dopamine molecule, capturing it in the apolar interior and preventing outward diffusion. Subsequently, a fluorescent solvatochromic aldehyde forms a Schiff base with the amino group on the dopamine‐recognition ligand complex. Upon imine formation, the emission of the complex blueshifts, indicating the presence of the bound dopamine (Figure 10). The emission change results from a push–pull system within the probe, in which the imine is a weaker electron acceptor compared to the aldehyde. Addition of a phenol catalyst improved the speed of detection to the minute timescale.

FIGURE 10.

Nanoreactor‐based sensing of dopamine. (A) Chemical sensing mechanism illustrating boronate‐mediated recognition of dopamine followed by imine formation with a solvatochromic aldehyde. (B) Schematic depiction of the stepwise sensing process within the nanoreactor, where imine formation induces a blueshift in fluorescence.

Use of a nanoreactor offers several advantages for dopamine detection. First, only dopamine was found to be capable of crossing the apolar boundary into the nanoemulsion, making the probe highly selective over related neurotransmitters. Second, the core of the nanoreactor protects the imine bond from hydrolysis, which can confound sensing in an aqueous environment. Finally, the increased localized concentration of dopamine within the reactor allows for a higher effective concentration as well as proximity to the phenol catalyst, speeding up the rate of the reaction. With this system, the researchers were able to selectively detect dopamine in biological media (serum, plasma, and urine) and cell lysates. Overall, this work introduces a nanoreactor‐based sensing paradigm that combines spatial confinement with dynamic covalent chemistry to achieve improved selectivity, stability, and reaction efficiency in neurotransmitter detection.

3.4. New Tools in Visualizing Ferroptosis

Ferroptosis is a form of cell death that is distinct from apoptosis, necrosis, and autophagy [75, 76]. Ferroptosis is triggered when lipid hydroperoxides accumulate [77]. Polyunsaturated fatty acids (PUFAs) are the substrates of pro‐ferroptotic lipid peroxidation products [78]. Accumulation of oxidized PUFA‐containing phospholipids ultimately drives membrane permeabilization and ferroptotic cell death. Experimentally, ferroptosis can be artificially induced through disruption of the GSH‐GPX4 pathway by chemical modulators. The small molecule erastin inhibits the cystine/glutamate antiporter system X_c ⁻ depleting intracellular cysteine and GSH, whereas RSL3 directly inhibits GPX4 enzymatic activity (Figure 11B).

FIGURE 11.

Organelle‐resolved RTA probes for ferroptosis imaging. (A) Mechanism of fluorogenic activation of radical‐trapping antioxidant probes via suppression of PeT quenching upon lipid radical trapping, and structures enabling organelle‐specific localization. (B) Activation of ferroptosis through the inhibition of the Glutathione (GSH)‐Glutathione Peroxidase 4 (GPX4) pathway.

In a recently published study, Xu et al. report a series of radical‐trapping antioxidant (RTA) probes that localize to distinct organelles to enable real‐time monitoring of lipid peroxidation during ferroptosis [79]. These probes are fluorogenic and are activated through the trapping of lipid radicals leading to PeT quenching and fluorescent turn on. The probe was conjugated to organelle targeting moieties to achieve selective localization to the mitochondria, lysosome, ER/lipid droplet, and PM (Figure 11A). As RTAs, these probes serve dual roles by partially inhibiting lipid radical propagation while simultaneously reporting local lipid peroxidation levels.

Using these probes, the authors quantified organelle‐specific sensitivity to ferroptotic stress. In rescue experiments, ER and lysosome‐targeted probes provided the strongest protection against induced ferroptosis, as reflected by the highest EC50 values after either erastin or RSL3 treatments. Time‐resolved imaging further revealed the spatial progression of lipid peroxidation. RTA fluorescence localized to the ER and Golgi‐associated vesicles increased before detectable cell rounding, whereas PM RTA signal rose nearly simultaneously with cell rounding and propidium iodide uptake. These results support a ferroptosis model in which lipid peroxidation accumulates in the ER during early onset and propagates outward to the PM.

3.5. A Fluorogenic Sensor for Hydrogen Peroxide

Peroxides comprise a portion of the reactive oxygen species (ROS) found within biological systems and are transient, with half‐lives of roughly 1 ms, making them difficult to detect and quantify [80]. In 2004, Chang et al. reported a highly selective, cell‐compatible fluorescein‐based boronate probe for visualizing peroxides in mammalian cells [81]. In 2006, another breakthrough was made when Belousov et al. disclosed the first genetically encoded peroxide sensor, HyPer [82]. Nevertheless, the probe remained sensitive to pH changes and possessed low sensitivity.

Potekhina et al. recently described a novel chemogenetic sensor for hydrogen peroxide [83]. Termed HyPerFLEX, the probe is composed of a redox‐sensitive OxyR‐RD unit from N. meningitidis and an engineered circularly permuted Y‐FAST fluorogenic protein. When OxyR‐RD is oxidized, an exogenous fluorogen is conformationally restricted in the Y‐FAST binding pocket, providing a fluorescence readout of peroxide levels (Figure 12). This design bypasses key limitations of earlier probes, including pH sensitivity and oxygen‐dependent chromophore maturation.

FIGURE 12.

HyPerFLEX activation and application in quantification of peroxide levels. (A) Schematic illustrating fluorogen compatibility and the mechanism of HyPerFLEX. Fluorogens bind to the cpFAST domain under basal conditions. Upon oxidation of the OxyR domain by hydrogen peroxide, conformational changes restrict the bound fluorogen into a more planar configuration, resulting in enhanced fluorescence. (B) Schematic of in vitro and in vivo workflows for HyPerFLEX‐based peroxide quantification. HyPerFLEX‐UnaG is expressed genetically via transfection of mammalian cells; vector delivery for in vivo neural studies is done via stereotactic injection. Hydrogen peroxide is then introduced into the system along with an external fluorogen, providing a ratiometric readout of peroxide levels.

Targeted variants fused to organelle localization motifs enable compartment‐specific peroxide sensing. Compared to HyPer7 [84], HyPerFLEX demonstrates enhanced sensitivity, including nanomolar detection under hypoxic conditions. Its compatibility with diverse synthetic fluorogens allows modular tuning of spectral properties.

To enable ratiometric imaging, the authors developed HyPerFLEX‐UnaG, which provides a dual readout of sensor expression (green UnaG signal) and peroxide levels (far‐red fluorogen signal). Overlapping two‐photon excitation profiles allow simultaneous excitation with a single laser, facilitating deep‐tissue imaging. Coupled with a genetically encoded peroxide generator, this system enables quantitative, compartmentalized imaging of H₂O₂ dynamics in live murine neurons.

Overall, HyPerFLEX establishes a versatile chemogenetic platform for sensitive, tunable, and spatially resolved detection of hydrogen peroxide in living systems.

3.6. A 3D View of Covalent Drug Distribution

Covalent drugs contain reactive warheads that form covalent bonds with their targets [85]. Although this strategy can produce highly potent and durable target engagement, the same irreversible chemistry can also lead to off‐target effects. A detailed understanding of where covalent drugs bind in vivo is therefore essential for improving efficacy while minimizing toxicity. Pang et al. report volumetric clearing‐assisted tissue click chemistry (vCATCH), a method that links covalent drug engagement and whole‐body imaging with cellular resolution [86].

To enable this approach, the authors built on earlier advances in tissue clearing and click‐based drug labeling. Ye and colleagues had previously developed HYBRiD, a hydrogel‐based reinforcement of 3D imaging solvent‐cleared organs, to support whole‐body clearing [87]. Pang et al. then improved their earlier CATCH method [88] by addressing the shallow and uneven labeling that limited its use in large 3D tissues. Specifically, they introduced two key steps: prereaction copper saturation (PRCS) and repeated iterations of reaction (RIR). PRCS uses excess Cu(II) to occupy endogenous copper‐binding sites and preserve catalyst availability, while RIR refreshes click reagents through multiple shorter reaction cycles, improving labeling homogeneity and helping maintain reaction efficiency throughout thick tissues. Together with HYBRiD and light‐sheet microscopy, these advances enabled vCATCH to generate unbiased, cellular‐resolution maps of covalent drug engagement across large tissue volumes and the intact mouse body (Figure 13).

FIGURE 13.

Visualizing covalent drug engagement across the intact mouse body. (A) An alkyne‐tagged drug reacted with an azide fluorophore via copper‐catalyzed azide–alkyne cycloaddition (CuAAC), forming a stable triazole linkage. (B) Alkyne‐tagged drugs are administered in vivo. Following tissue fixation and HYBRiD clearing, target‐bound drug molecules are fluorescently labeled by CuAAC. Presaturation of endogenous copper‐binding sites using PRCS and refreshment of the click chemistry reagent using RIRs improve homogeneous labeling. Whole‐body drug engagement is then mapped at cellular resolution by light‐sheet microscopy.

As a proof of principle, the authors first applied vCATCH to pargyline, a covalent monoamine oxidase inhibitor. They used a clickable alkyne derivative, pargyline‐yne, which could subsequently be reacted with an azide fluorophore. Applying the Allen Brain Atlas and the group's AI‐based Cartography of Ensembles pipeline [89], quantitative and region‐by‐region analysis of drug abundance and drug‐bound cell density recapitulated known drug targets, demonstrating that uniform, deep labeling was achieved through vCATCH.

Scaling the approach to whole‐body mapping, Pang et al. applied vCATCH to juvenile and adult mice to analyze covalent cancer drugs: afatinib, an epidermal growth factor receptor (EGFR) inhibitor used in nonsmall cell lung cancer, and ibrutinib, a BTK inhibitor for B cell malignancies. Organ binding analysis showed that afatinib binding was enriched in the lung and ibrutinib preferentially bound in the heart, aligning with FDA autoradiography data. In addition, vCATCH also revealed sub‐organ distribution: afatinib was enriched in lung alveolar and kidney glomeruli, and ibrutinib was located in lung bronchial regions and kidney tubules.

Together, these studies illustrate that vCATCH provides a powerful method for visualizing covalent drug engagement, potentially revealing both on‐ and off‐target interactions of covalent drugs and their metabolites.

3.7. Precision Sensing of Acetaldehyde

Unlike traditional sensing methods, which rely on reversible binding, activity‐based sensing (ABS) [90] exploits a chemical reactivity to trap a target analyte, enabling high selectivity and sensitivity. While it is often straightforward to design reactions that recognize a specific functional group, discriminating between closely related metabolites that share the same functionality remains a major challenge. A notable example is formaldehyde ABS, which leverages rapid imine formation with a secondary amine followed by a rearrangement that is uniquely favored by formaldehyde due to its minimal steric demand [91, 92].

In this work, Li et al. [93] extend this concept to develop a selective sensor for acetaldehyde. Acetaldehyde (AA) is a reactive metabolite of ethanol oxidation, but accumulation of AA can lead to an increase in oxidative stress and DNA damage [94, 95]. Tools to monitor acetaldehyde are therefore critical, but existing methods suffer from low selectivity. Here, the authors address this challenge by designing a probe with a secondary amine handle and a tetrazine motif. This probe, termed AAP‐1, enables real‐time imaging of acetaldehyde through fluorescence microscopy [93].

Mechanistically, acetaldehyde first condenses with the secondary amine of the probe to form an enamine intermediate (Figure 14A). Under physiological conditions, the electron‐rich olefin moiety reacts with the tetrazine via an inverse electron‐demand [4 + 2] cycloaddition reaction. Subsequently, the bridged intermediate undergoes a retro [4 + 2] cycloaddition reaction, releasing nitrogen gas. After an elimination reaction, the fluorescent pyridazine product is formed. AAP‐1 was shown to be selective for AA over other common aldehyde metabolites, including methylglyoxal and formaldehyde. This selectivity was proposed to stem from two major reasons: the efficient enamine formation and steric hindrance. While formaldehyde is more reactive than AA, it cannot form the enamine intermediate and fails to trigger the subsequent cycloaddition reaction.

FIGURE 14.

Fluorescent sensing of acetaldehyde. (A) Fluorescence turn‐on mechanism of AAP‐1 probe. (B) Demonstrations of acetaldehyde sensing in A375 and Huh7 cell lines.

The authors validated AAP‐1 in A375 and Huh7 cell lines (Figure 14B). Within the range of physiological acetaldehyde concentration, the fluorescence intensity of the probe was found to have a positive correlation with concentration. Additionally, the authors showed that the probe can track ethanol metabolism in real‐time. Inhibition of alcohol dehydrogenase decreased fluorescence, consistent with reduced acetaldehyde production, whereas inhibition of aldehyde dehydrogenase increased fluorescence due to acetaldehyde accumulation. In summary, the new probe has excellent selectivity and allows real‐time sensing of acetaldehyde, providing a promising way to study diseases linked to acetaldehyde, such as cancer and cirrhosis, particularly in populations carrying the ALDH2*2 variant where acetaldehyde detoxification is impaired.

4. Advances in Interaction Profiling

4.1. MultiMap Resolves Signaling Neighborhoods

Cell‐surface receptors are dynamic and known to be modulated by the interaction with various neighboring proteins [96]. This activation goes through a series of time‐staged signaling processes ranging from the first minute to hours, making it hard to accurately assess them. Lin et al. introduce a temporally resolved photoproximity‐labelling strategy that captures how these signaling environments reorganize during activation of the EGFR [97]. Their work builds on the recently developed MultiMap platform, first described in Science by the same authors in 2024, which enables both time‐ and distance‐resolved photoproximity‐labeling [96, 98].

In the MultiMap platform, a single photocatalyst, eosin Y, is combined with multiple biotinylated photoprobes to map protein environments at adjustable radii [97, 98]. Upon light activation, eosin Y activates these probes to generate reactive intermediates with distinct lifetimes. For example, aryl‐diazirine probes produce short‐lived carbene intermediates that label nearby proteins within a narrow radius, whereas phenol‐based probes generate longer‐lived phenoxyl radicals that interact with broader neighborhoods [97, 99] (Figure 15A). By utilizing these lifetime differences, MultiMap enables spatial mapping of cell‐surface protein environments with tunable resolution. However, because receptor signaling is inherently dynamic, capturing these temporal changes has remained a major challenge.

FIGURE 15.

Schematic of the MultiMap workflow. (A) Three photoprobes with distinct labeling radii, aryl‐diazirine, aryl‐azide, and phenol are activated by photoactivated eosin Y to generate reactive intermediates that covalently tag nearby proteins with biotin. (B) In the MultiMap workflow, photocatalyst, eosin Y is attached to either the extracellular domain (ECD) or intracellular domain (ICD) of EGFR via engineered tags (FLAG tag and Halo tag). Upon light activation, photoproximity labeling captures protein neighborhoods at different spatial scales from both sides of the membrane, enabling spatially and temporally resolved mapping during receptor activation.

To overcome this limitation, Lin et al. modified the MultiMap strategy so that the photocatalyst could be positioned on either the extracellular or intracellular side of EGFR, enabling selective anchoring of the Eosin Y photocatalyst to either location (Figure 15B). After ligand stimulation, short pulses of blue light were applied at defined time points to trigger proximity labeling. This design enables snapshots to be captured of EGFR‐associated protein neighborhoods from both the extracellular and intracellular perspective, allowing the dynamic reorganization of signaling complexes to be tracked with minute‐scale temporal resolution.

Strikingly, the protein neighborhoods observed from the extracellular and intracellular sides of EGFR upon EGF activation differed substantially, with less than 10% overlap between the two datasets [97]. This highlights how spatial context shapes receptor interactions and underscores the effectiveness of the MultiMap approach. For example, the receptor phosphatase, PTPRF, was primarily observed in the extracellular‐side labeling experiments, whereas Sec61, an essential protein complex for translocation and membrane insertion into the ER, was captured predominantly from the intracellular side [97]. The method also mapped time‐dependent changes in the EGFR neighborhood. Shortly after EGF activation, proteins involved in phosphorylation and membrane trafficking are rapidly recruited. At later time points, factors linked to downstream signaling processes such as endocytosis, degradation, and transcriptional regulation are recruited and are captured by the probes. Mutational studies further indicate that many of the proteins identified by MultiMap actively regulate EGFR signaling, rather than simply residing in its vicinity.

Together, these findings illustrate the power of multiscale mapping to further our understanding of extracellular and intracellular signaling processes.

4.2. Lipid Photoaffinity Labeling Reveals Biological Role of Rare Triacylated Phospholipid

N‐acylphosphatidylethanolamine (NAPE) is a rare, triply acylated glycerophospholipid [100]. In cells, NAPE is formed through the acylation of phosphatidylethanolamine (PE), with its third acyl tail donated from phosphatidylcholine (PC) by several N‐acyltransferases. Multiple catabolic pathways convert NAPE into the fatty acid amide N‐acylethanolamine (NAE) [101]. Most insights into NAPE biology have come indirectly from studies on NAE, whose roles as an endocannabinoid signaling molecule modulating CB1 and CB2 receptors are well established [102]. By contrast, the biological functions of NAPE itself remain poorly understood. Chiu et al. now address this gap through the development of an elegant photoaffinity labeling (PAL) strategy to interrogate the NAPE interactome [103].

PAL probes typically combine two key features: a photoreactive group that covalently captures protein–ligand interactions and a bioorthogonal handle that enables enrichment and identification of the labeled proteins [104]. Diazirines are commonly used as the photoreactive group and terminal alkynes as the clickable handle, and for lipids, both functionalities are often installed within the same acyl chain. Because NAPE can be metabolized through multiple routes into diverse lipid derivatives that may retain substantial structural similarity, placing both functionalities on a single acyl chain risks generating false‐positive signals from metabolites that still contain both the diazirine and alkyne. To address this problem, the authors strategically separated the photoreactive group and clickable handle onto different acyl chains, minimizing the likelihood that downstream metabolites would retain both features simultaneously (Figure 16A). This design was central to designing a NAPE PAL probe with minimal false‐positive results.

FIGURE 16.

NAPE photoaffinity labeling. (A) Synthetic route toward NAPE photoaffinity probe. (B) Fluorescent labeling of proteins using NAPE photoaffinity probe.

Upon crosslinking, the authors identified 189 proteins across triplicate proteomics experiments, of which 30 were detected in at least two of the three replicates. Through gene ontology analysis, they found NAPE‐associated proteins to be enriched in biological processes related to protein synthesis and transport, as well as metabolic regulation in response to stimuli or stress. They proceeded to validate labeling of select proteomic hits by demonstrating that probe labeling is outcompeted by excess native NAPE.

Further studies focused on two‐hit proteins, CD44 and CD147, that are each involved in regulating transport of small monocarboxylate metabolites such as lactate and pyruvate [105]. The observation of these hits led the authors to test whether NAPE might influence lactate transport. Indeed, they found that NAPE promotes monocarboxylate‐mediated lactate efflux in a manner dependent on CD44 and CD147, revealing a new role for this lipid in the regulation of cellular lactate homeostasis.

4.3. Profiling the Developmental Proteome In Vivo

Studying protein–protein interactions (PPIs) allows us to better understand protein function and regulation, which is especially important for understanding complex systems, such as the brain. However, current methods for studying PPIs in vivo face limitations, including the detection of transient interactions that cannot withstand lysis [106] or involve coexpressing a large enzyme, which may affect the function and proximal proteome of the target protein [107, 108]. Addressing these limitations requires the development of proximity‐labeling strategies capable of mapping protein interactions in vivo with increased spatiotemporal resolution.

In a recent study, Takato et al. [109] developed PhoxID, a light‐triggered photoproximity labeling method that combines ligand‐directed acyl imidazole chemistry [110] with proximity‐based labeling. Using the AMPAR‐targeting ligand PFQX (Figure 17A), the authors demonstrated that PhoxID successfully characterized the AMPAR‐proximal proteome. PhoxID was able to detect 58 hippocampal proximal proteins, where close to 60% of hit proteins were previously associated with synapses. Three subunits of AMPAR (GRIA1, GRIA2, and GRIA3) but not GRIA4 were identified in the hippocampus, which is consistent with previous studies showing that GRIA4 is only present in small amounts in the hippocampus but is largely present in the cerebellum [112].

FIGURE 17.

PhoxID can profile transient changes to the developmental proteome in vivo using ligand‐directed acyl imidazole chemistry [110, 111] and photoproximity labeling. (A) Mechanism of ligand‐directed acyl imidazole chemistry and probe design. (B) PhoxID demonstrates how the AMPAR‐proximal proteome changes from postnatal stages to adulthood, where AMPAR shifts from a LRRC4B‐rich proteome at P8 to a NECTIN3‐rich proteome at P13.

After initial validation, PhoxID was used to probe interactions in the brain during postnatal development. The authors studied how the AMPAR‐proximal proteome in the cerebellum shifted from 8 days old (P8), 13 days old (P13) to 5 weeks at adulthood (Figure 17B) [109]. Most hit proteins increased in abundance over time, but 14 proteins were more enriched in neonatal stages than in adulthood. These included LRRC4B and NECTIN3, where LRRC4B was enriched at P8 and P13, whereas NECTIN3 was only enriched at P13. Since PhoxID labels proximal proteins with a nanometer radius, the authors hypothesized that AMPAR is in a LRRC4B‐rich proteome at P8 but migrates to a NECTIN3‐rich proteome at P13. Using confocal microscopy, the authors confirmed that NECTIN3 was localized in the molecular layer (ML) at P8, whereas AMPAR was localized in the Purkinje cell layer (PCL), and at P13, both AMPAR and NECTIN3 were both localized in the ML.

Combined, PhoxID is a powerful technique that can profile proximal proteomes in vivo with high spatiotemporal resolution and minimal perturbation to study molecular interactions in complex biological environments.

4.4. Natural Product‐Based Probes Expand the Ligandable Proteome

Chemoproteomics is a powerful method for identifying ligands for proteins implicated in human disease [113]. Conventional chemoproteomic probes utilize electrophilic groups to form covalent bonds with nucleophilic residues, and in recent years, photoaffinity‐based methods have enabled scientists to study noncovalent interactions even for binding sites that lack nucleophilic residues [114, 115, 116].

Chaheine et al. describe fully functionalized natural product probes (FFNPs) as a strategy to expand chemoproteomic discovery into structurally distinctive regions of chemical space [117]. Seven commercially available terpenoids and alkaloids with well‐documented transformations were elaborated into 30 FFNPs, consisting of a natural product (NP)‐derived binding group and a fully functionalized tag containing a photoactivatable diazirine and alkyne for click chemistry (Figure 18A). Compared to previous libraries, these FFNPs have higher three‐dimensionality and are more drug‐like based on Lipinski's rules [118].

FIGURE 18.

FFNP photoaffinity labeling. (A) Chemical components of FFNPs. (B) Gel‐based and MS‐based workflows for evaluation of FFNPs and chemoproteomic hit identification.

To assess protein‐FFNP interactions, the authors employed gel‐based and mass spectrometry‐based workflows (Figure 18B). Cells were incubated with selected FFNPs, irradiated with UV light, lysed, and then subjected to CuAAC with either rhodamine azide for in‐gel fluorescence analysis or biotin azide for affinity enrichment, digestion, tandem mass tag labeling, and LC–MS analysis.

The authors identified 182 proteins as FFNP targets, with 147 being preferentially enriched by a single FFNP. Of the targets identified, 63% did not have known ligands, and 77% were specific to FFNPs when compared to targets of photo‐stereoprobe and fragment enantioprobe libraries. Additionally, many targets belong to protein classes considered difficult to drug (e.g., chaperones, transcription factors, and scaffolding proteins), and half are reported in the Cancer Dependency Map as essential for cancer growth and development.

Overall, the study shows that natural‐product‐derived photoaffinity probes can broaden the chemical tractability of the human proteome by introducing greater topological and stereochemical diversity into chemoproteomic screening libraries.

4.5. ER Chaperone Target Triggers Tau Cleanup

Neurodegenerative diseases such as Alzheimer's and Pick's disease are characterized by the buildup of toxic tau aggregates that are linked to impaired neuronal function [119]. The cause of this accumulation is unknown, and therapeutic strategies targeting tau proteins via phosphorylation and autophagy have so far shown limited clinical success [120].

The tau protein helps stabilize cellular microtubules, which are important for the movement of cargo across the neuron. In neurodegenerative diseases, tau is abnormally modified (often through hyperphosphorylation), detaches from microtubules, and aggregates inside neurons and glia, forming neurofibrillary tangles [121] (Figure 19). These protein aggregates disrupt cellular architecture and impair axonal transport, contributing to neuronal dysfunction and degeneration.

FIGURE 19.

(A) Discovery of compounds 1a and 2a, and their proposed mechanism of targeting P4HB, which is implicated in the proper folding of tau proteins through disulfide oxidation (PDB: 8EOJ) [122]. (B) In a healthy neuron, tau binds to and stabilizes microtubules, maintaining cytoskeletal integrity. In tauopathy, tau is hyperphosphorylated, detaches, and aggregates into neurofibrillary tangles.

Rather than targeting proteins that are known to be implicated in this process, Conway et al. used a target‐agnostic phenotypic screen, allowing cellular biology to reveal mechanisms capable of clearing tau aggregates [123]. They screened a library of 383 chemoproteomic‐enabled photoaffinity probes, discovering hit compound 1a and optimizing it to aminoindole 2a by removing the enrichment handle and photoreactive groups. This optimized compound cleared over 90% of tau aggregates in iPSC‐derived neurons.

The authors then used a chemoproteomic workflow to elucidate the molecular targets of 2a and found that it engages the ER chaperone enzyme P4HB, targeting Cys397 and Cys400 in the enzyme's active site. They subsequently found that 55% reduction in P4HB protein levels led to a 25% decrease in tau aggregate formation, the most significant effect among all tested targets.

Inhibition of P4HB to reduce aggregated tau appeared to be counterintuitive, as P4HB is important for binding pathogenic hyperphosphorylated tau and reducing their ability to form aggregates. Other studies show that inhibition of P4HB enzymatic activity contributes to neurodegeneration rather than providing therapeutic benefit [124]. So how does blocking a “good” chaperone from doing its job actually help clear “bad” tau proteins? The authors propose two theories that could explain this phenomenon: (i) P4HB may occasionally act as an anti‐chaperone [125] and (ii) ER‐associated degradation is upregulated upon P4HB modulation.

Together, this study identified P4HB as a potential therapeutic target for promoting the clearance of toxic tau aggregates and illustrates the powerful interplay of phenotypic screening and chemoproteomic target elucidation to uncover a new therapeutic avenue in neurodegenerative disease.

5. Advances in Precision Control of Biological Systems

5.1. The BEAR Necessities of Protein Backbone Extensions

Polypeptides, polymers composed of α‐amino acids, demonstrate remarkable functional and structural diversity. Despite their natural diversity, canonical α‐backbones limit the chemical reactivity and design possibilities of proteins. Extended backbone linkages can be engineered to contain one (β‐linkage) or more (γ‐, δ‐linkages) methylene groups between peptide bonds [126]. These semi‐synthetic protein backbones could be used to alter protein–protein interactions, improve enzymatic catalysis, and develop thermostable protein‐like biomaterials [126].

The addition of extended backbone linkages into polypeptides has traditionally been framed as a biosynthetic in vitro translation problem that relies on stoichiometric transfer RNA (tRNA) mis‐acylation agents, flexizymes, or the engineering of the translational machinery (elongation factors (EF‐Tu), tRNA, or the ribosome) [126]. In the past decade, this approach has been extended to in vivo systems, with β‐linkage monomers added via mutant ribosomes [127]. However, these approaches are not readily scalable, difficult to reengineer, and rigidly selective [126].

Now, Roe et al. approach peptide backbone extension as an in vivo posttranslational problem using the backbone extension acyl rearrangement (BEAR) reaction [126]. This biosynthesis method enables scalable, high‐fidelity formation of site‐specific β‐, γ‐ and δ‐ linkages within folded proteins without requiring changes to the ribosomal active site or EF‐Tu [126].

The authors use an engineered pyrrolysyl tRNA synthetase (PylRS)/tRNA pair from archaea, which, in its natural form, incorporates the amino acid pyrrolysine into proteins using the amber (UAG) stop codon [126, 128]. In Escherichia coli, the mutated PylRS/tRNA pair incorporates α‐hydroxy acid building blocks carrying a masked nucleophile into proteins [129]. After translation, a chemical or photochemical unmasking agent reveals a nucleophilic amine that is well‐positioned for an intramolecular attack on the backbone ester linkage to produce an amide group in the protein backbone (Figure 20A) [117]. BEAR adds one or more methylene units to form a β‐, γ‐ or δ‐ linkage, depending on the specific α‐hydroxy amino acid derivative incorporated (Figure 20B) [126]. Although BEAR chemistry is analogous to native chemical ligation reactions previously seen in short synthetic peptides, the authors are the first to perform this reaction within folded proteins [130].

FIGURE 20.

Extended linkage formation with BEAR. (A) The BEAR reaction begins with the unmasking of a side chain nucleophile in an α‐hydroxy acid unit within a protein via light (hv) or 1,3,5‐triaza‐7‐phosphaadamantane (PTA). A nucleophilic attack occurs between the unmasked amine and the indicated ester to create a backbone extension. (B) Scheme illustrating the unmasking conditions and intermediate monomers for the biosynthesis of β²‐, γ‐, and δ‐linkages. A masked α‐hydroxy acid is translated into E. coli proteins using a PylRS/tRNA pair, proteins are purified, and the nucleophile is unmasked to allow the BEAR reaction to produce an extended peptide linkage. Peptide digests followed by MS/MS spectra comparison with a synthetic peptide confirm the integration of the extended linkage.

The group first computationally modeled their proposed β‐, γ‐ or δ‐BEAR reaction schemes to verify that they were thermodynamically favorable compared to native chemical ligation of cysteine/serine analogs. They confirmed that the expected BEAR reaction occurred by digesting the purified target protein to produce a high‐recovery peptide containing the extended linkage and compared it to synthetic standards using tandem mass spectrometry (MS/MS). Notably, the in vivo biosynthesis of γ‐ and δ‐linkages, which was reported here for the first time, proceeded with robust ~91% and moderate ~27% yield, respectively.

This work could be of interest for protein engineering, biotherapeutics, and next‐generation biomaterials. Photomasked‐monomer BEAR reactions can be used to produce spatiotemporal mechanistic probes that trigger precise protein–protein interaction inhibition or provide access to occluded small molecule–protein interaction sites [126, 131]. The therapeutic potential of noncanonical polymers is profound—judiciously placed extended linkages have already been shown to increase cell permeability, enhance protease resistance, and alter immunogenic recognition of peptides [126, 131, 132]. This work provides a novel biosynthetic tool for posttranslational β‐, γ‐, or δ‐linkage extensions that has the flexibility to be generalized to other nucleophiles or caging groups [126, 133].

5.2. Protein Editing in Living Cells

Tracking protein dynamics in living cells is essential for understanding cell biology. A widely used strategy is to fuse one or more epitopes to target protein genes and label the protein with either peptide tags or small molecules to enable visualization or functional modulation. However, this strategy has several limitations, including inefficient temporal control and potential disturbance to protein function.

Now Beyer et al. utilize intein chemistry to overcome these challenges [134]. Inteins are protein domains that were first discovered in Saccharomyces cerevisiae [135]. Under physiological conditions, they can catalyze their own excision from precursor proteins and ligate the flanking proteins together. Some inteins are naturally split into two fragments that associate and undergo protein trans‐splicing [136]. Because of these properties, inteins are widely used as removable handles in protein purification and semi‐synthetic protein chemistry [137]. Moreover, in 2003, the Muir lab achieved protein semi‐synthesis in live cells by installing a short peptide epitope at the C‐terminus of an overexpressed protein [138], demonstrating the compatibility of intein chemistry in living cells.

In this work, Beyer et al. applied two orthogonal pairs of split inteins to edit endogenous proteins (Figure 21) [134]. An intein acceptor sequence is genetically inserted into the genomic locus of the target protein, maintaining native expression levels and regulation. In parallel, a recombinant intein donor protein containing intein modules and the desired cargo, which could be a peptide tag or noncanonical amino acid, is produced in vitro and delivered into cells by electroporation or lipid nanoparticles. The split intein pairs then undergo trans‐splicing, inserting the cargo into the target protein. After splicing, the intein sequences are excised, leaving minimal sequence “scar” in the protein.

FIGURE 21.

Intracellular protein editing. (A) Structure of unnatural amino acid and functionalization by click chemistry. (B) Protein editing workflow: intein donor expression and modification, intein tagging on the target protein, and intein splicing and cargo installation. POI: protein of interest, FG: functional group.

To validate the method, the authors showed that β‐actin can be edited within 10 min of donor delivery, enabling temporal control. The transcription factor c‐Myc and kinase Chk1 were also successfully edited, highlighting the applicability of this method to dynamic proteins. Additionally, this approach enables incorporation of noncanonical amino acids into endogenous proteins. A donor protein containing p‐azido‐phenylalanine (pAzF) is introduced in E. coli by genetic code expansion, functionalized with fluorophores or biotin using click chemistry, and successfully inserted into endogenous calnexin.

Overall, this technique provides a platform for site‐specific insertion of protein tags or non‐natural amino acids into endogenous proteins. These modifications are modular, nearly traceless, and preserve native protein expression, thereby enabling precise chemical manipulation of proteins in living systems. Additionally, a complementary study by Hua et al. describes a protein transposition system that also leverages intein chemistry to manipulate protein primary sequences posttranslationally [139], underscoring the broader potential of intein‐based protein editing.

5.3. Controlling Stem Cell Fate With a Glycoconjugate

Stem cells have been identified as an attractive therapeutic target for organ regeneration, treating neurodegeneration, and the development of immune cell‐based therapies [140]. However, it has been challenging to spatiotemporally direct specific factors to distinct cell populations to induce their differentiation. Heparan sulfate (HS) is found on the surface of stem cells and modulates growth factor signaling. During stem cell development, HS recruits fibroblast growth factor 2 (FGF2) to its receptor and induces subsequent differentiation into Sox1⁺ neuroectodermal cells [141]. Previous methods to modulate HS for stem cell differentiation have relied on genetic engineering or recombinant protein expression and lack broad therapeutic potential [142, 143, 144].

Purcell et al. developed a modular chemical strategy to manipulate the cell surface and control the developmental fate of stem cells [145]. They utilized a previously developed DNA aptamer that binds alkaline phosphatase (Alpl), a stem cell pluripotency marker, to specifically target undifferentiated embryonal stem cells (ESCs) [146]. By conjugating polydisperse heparin (Hep) to this aptamer, they were able to selectively target a defined cell population at a specific developmental stage and induce differentiation.

Purcell et al.'s aptamer‐glycan chimera approach is a highly modular platform to probe the many roles of HS. The authors first synthesized a Hep‐biotin conjugate using a bifunctional 3‐(N‐methylaminooxy)propylamine linker that reacts with the reducing end of Hep to generate a primary amine that was subsequently reacted with NHS‐PEG4‐Biotin (Figure 22A). Next, Hep–biotin was complexed with streptavidin, and finally, a 5′‐biotinylated DNA aptamer was introduced to yield the final aptamer–Hep chimera.

FIGURE 22.

(A) Synthetic strategy for Hep‐Aptamer conjugate. Biotin‐streptavidin conjugation was used to overcome the electrostatic repulsion between the negatively charged polydisperse heparin and DNA aptamer. (B) Differentiation of Alpl+/Sox‐ ESCs into neural progenitor cells was efficiently induced by the Hep‐Aptamer chimera.

The authors next evaluated the specificity of the aptamer‐Hep chimera for pluripotent ESCs. Through a series of fluorescence microscopy imaging experiments, they saw that the Alpl targeting aptamer‐Hep was specific for Alpl by a decrease in fluorescence intensity as Ext‐/‐ ESCs differentiated with exogenous Hep stimulation. They concluded that their aptamer‐Hep chimera specifically targeted Alpl‐expressing pluripotent stem cells because Alpl expression is reduced upon differentiation.

They further evaluated if the chimera stimulated differentiation through FGF2 signaling. Western blot analysis revealed an increase in phospho‐Erk1/2 (downstream of FGF2) levels in cells treated with Hep‐containing chimeras, signifying the ability of their aptamer‐Hep chimera in restoring FGF2 dependent signaling in HS‐depleted ESCs.

Having established both specificity and signaling activity, the authors assessed whether the chimera could promote neural differentiation more effectively than exogenous Hep alone. The authors assessed the expression of Sox1 and Alpl after HS‐deficient ESCs were stimulated with FGF2 and treated with the chimera or HS alone. The results showed that all Hep‐containing constructs could induce Sox1 expression; however, the aptamer‐Hep chimera was the most potent. These results strongly support the conclusion that the DNA aptamer provides enhanced affinity for the ESC surface and augments Hep‐induced FGF2 recruitment and subsequent differentiation.

This creative and powerful new approach to glycocalyx engineering lays the foundation for future glycan‐aptamer chimeras that can elicit phenotypic responses from specific cell populations. These modular, bifunctional chimeras have potential to not only probe complex biological processes but also as potential therapeutic modalities.

5.4. Cell‐Surface Editing With Cellular Resolution

Precise control of protein activity at the cell surface remains a central challenge in chemical biology. Current strategies deliver preformed effectors via antibodies or ligands, but because these are active upon arrival, their function cannot be confined to the intended cellular context. Kofoed et al. now address this limitation with SMART (splicing‐modulated actuation upon recognition of protein targets), a programmable protein actuator that uses conditional protein trans‐splicing to synthesize functional proteins directly on designated cell surfaces from otherwise inactive precursors [147].

Split inteins ligate flanking polypeptides through protein trans‐splicing [137] and naturally associate with high affinity, meaning splicing occurs constitutively and cannot be controlled. To enable control, Muir and colleagues engineered “caged” split inteins that block premature association; colocalization restores activity via domain swapping [148, 149]. SMART exploits this principle on the surface of living cells. Each precursor contains a targeting module directed against a cell‐surface antigen, an output protein fragment, and a caged split‐intein half. SMART uses NrdJ‐1 split inteins [150], which leaves a serine at the splice site and is compatible with extracellular conditions. Splicing occurs only when both antigens are codisplayed and produces the ligated output protein exclusively on the target cell surface (Figure 23).

FIGURE 23.

SMART enables logic‐gated protein ligation on cell surfaces. (A) Two inactive precursors bearing anti‐Ag1 and anti‐Ag2 targeting modules colocalize on cells that display Ag1 and Ag2, enabling uncaging and trans‐splicing to generate a tethered ligated extein output while excising the intein. In one implementation, SMART assembles SpyCatcher003, which subsequently covalently captures SpyTag003‐linked effectors through isopeptide bond formation. (B) Antigen‐dependent proximity drives conditional splicing, which can be wired to different functions, including AND‐gated cell depletion, AND‐gated proximity labeling, and AND‐gated cytokine release.

SMART implements Boolean logic (“AND,” “OR,” and “NOT”) and is modular across inputs and outputs. In SMART‐SpyCatcher, splicing produces SpyCatcher003, which covalently captures SpyTag003‐fused cargos, creating a programmable recruitment platform [151]. This enables selective delivery of toxins or labeling enzymes, achieving up to ~150‐fold discrimination and ~92% depletion of target cells in mixed populations.

SMART‐gated proximity labeling was demonstrated using both SpyTag003–APEX2, which produces long‐lived phenoxyl radicals, and SpyTag003 conjugated to a μMap iridium photocatalyst [152], which generates short‐lived carbene species for higher spatial resolution. Importantly, the authors engineered a range of caged variants with tunable response dynamics, enabling NOT logic via competitive decoys. The platform accommodates a diverse range of targeting modalities, including DARPins, nanobodies, scFvs, peptides, and small molecules.

Beyond surface‐retained outputs, SMART supports ligation‐and‐release. By rearranging the domain architecture of the precursors, the authors generated SMART‐IL‐1β, which ligates and releases the pro‐inflammatory cytokine IL‐1β in response to defined antigen inputs, which could be used for highly localized immunotherapies. Together, SMART establishes a general framework for encoding spatially restricted protein activity through cell‐surface logic, with applications in targeted cell ablation, proximity proteomics, and localized immunotherapy.

6. Emerging Modalities in Drug Discovery

6.1. A Mini Protein Escapes the Endosome

Biological barriers prevent therapeutics from reaching their intended target, leading to diminished efficacy and off‐target toxicity [153, 154]. While many small‐molecule therapeutics can permeate through the PM, protein therapeutics typically enter cells via endocytosis and subsequent endosomal escape [155, 156]. Previous work identified ZF5.3, a short‐helical zinc‐finger mini‐protein of 27‐amino acid containing a penta‐arginine motif, that can efficiently escape from late endosomes in an HOPS‐dependent manner. ZF5.3's ability to gain access to the cytosol has made it a viable delivery tag for enzymatic and protein payloads [157, 158]. However, how ZF5.3 efficiently exits the late endosomal membrane is unclear. In a new report, Giudice et al. show that ZF5.3 escapes from the late endolysosome by cooperatively unfolding at the low pH of the late endolysosomal lumen [159].

The late endolysosome exhibits an acidic environment that denatures proteins for enzymatic degradation [160]. This led Giudice et al. to postulate that, like other Cys2His2 (C2H2) zinc‐finger proteins, ZF5.3 may unfold under acidic conditions. They tested this idea by synthesizing ZF5.3 using solid‐phase peptide synthesis (Figure 24A) and observed a strong increase in the circular dichroism ellipticity of ZF5.3 at 208 nm at pH 4.6. To identify a mechanism for the observed pH‐dependent unfolding, the authors solved an NMR structure to show that ZF5.3 adopts a canonical ββα fold that is stabilized by Zn(II) coordination by Cys3, Cys6, His19, and His23 residues at neutral pH. NMR titration experiments further indicated that protonation of a Zn(II)‐binding histidine side chain occurs at lower pH, destabilizing Zn(II) coordination and driving unfolding.

FIGURE 24.

pH‐Dependent Unfolding Enables ZF5.3 Endosomal Escape. (A) Solid‐phase peptide synthesis of mini‐protein ZF5.3. A rink‐amide or precoupled Wang resin (yellow circle) was first coupled to Fmoc‐protected L‐threonine (red highlighted). The coupled resin was then deprotected and subsequently coupled to the next 26 amino acids of ZF5.3. When the N‐terminal (tyrosine) amino acid was coupled, the peptide chain was cleaved off into solution and purified. (B) Cartoon of pH‐dependent ZF5.3 endosomal escape. ZF5.3 is first endocytosed. Then, as the pH of the maturing late endosome decreases, ZF5.3 unfolds, which both prevents maintenance of the native zinc‐finger fold and promotes interaction with BMP‐rich membranes, enabling escape into the cytosol.

To test whether ZF5.3 unfolding promotes escape from the late endosomal lumen in cells, the authors incubated rhodamine‐labeled ZF5.3 and an analog, BBA5, that is designed to remain folded at low pH, in cells. The relative fluorescence of rhodamine‐ZF5.3 was significantly higher in the cytosol compared to the BBA5 confirming that ZF5.3 endolysosome entry is unfolding‐dependent. The authors also utilized unilamellar vesicles coated with bis‐monoacylglycero‐phosphate (BMP), an anionic lipid enriched in late endolysosome membranes, to see if pH dependent ZF5.3 unfolding facilitates an interaction with BMP. Using fluorescence correlation spectroscopy, a pH‐dependent ZF5.3 specific interaction with BMP was observed. Together, these findings show that pH‐triggered unfolding of ZF5.3 is required for efficient endosomal escape and may help guide the design of cargo‐delivery systems that are tuned to specific stages of endosome maturation.

6.2. Bifunctional Molecules to Target p53 Abundance

p53 is a tumor suppressor protein, acting in response to various cellular stress signals to induce apoptosis, cell‐cycle arrest, DNA repair, and other anti‐oncogenic activities [161]. Mutations in TP53, the gene encoding p53, are the most common in human cancers, with ~80% of TP53 mutations being dominant‐negative missense mutations [162]. These are often encoded within the DNA‐binding domain of p53 and lead to compromised activity or conformational instability [162]. Despite its essential antitumor activities and implications in cancer and disease, p53 has long been considered undruggable due to the lack of suitable binding pockets for small‐molecule drugs and the inherent challenge of selectively refolding or rescuing function in aberrant p53 [163]. Sadagopan et al. propose an alternative therapeutic framework to target mutant p53 using heterobifunctional molecules [164]. By designing a proximity‐induced cytotoxic drug, they show the pathological accumulation of p53 in cells harboring TP53 missense mutations can be exploited.

Comprehensive genomic and proteomic analyses indicate that a key distinguishing feature between cells expressing wild‐type and mutant p53 is the elevated abundance of mutant p53. This enrichment arises from disruption of the MDM2‐mediated negative feedback loop, in which p53 induces MDM2, leading to ubiquitin‐mediated proteasomal degradation of p53 [165]. Taking advantage of this elevated p53 abundance, Sadagopan et al. designed a bivalent molecule, in which one moiety targets mutant p53 and the other delivers a cytotoxic payload, thereby selectively concentrating the cytotoxin in cells with high p53 levels. To target mutant p53, the authors used an analog of rezatapopt, PMV6, a small molecule reported to selectively bind the p53 Y220C mutant [166]. On the contrary, they used BI2536, a potent inhibitor of polo‐like kinase 1 (PLK1), an essential regulator of mitosis. The resulting compound, PMV6‐PEG4‐BI2536 (p53‐01; Figure 25A), was shown to induce complex formation between p53^Y220C and PLK1 and to drive antiproliferative effects selectively in TP53 Y220C mutant cells, with minimal effects in wild‐type or other p53 variants (Figure 25B). PLK1 was also shown to mislocalize in response to p53‐01 administration, colocalizing with p53^Y220C in the nucleus or chromatin. Lastly, their work reveals p53‐01 is functionally distinct from PMV6, which when administered alone restores the expression of p53 target genes in TP53 Y220C mutation‐bearing cancer cell lines. p53‐01 did not show such effects, suggesting its mechanism of action is independent of p53 transcriptional reactivation and rather induces toxicity through its bivalent, proximity‐inducing design.

FIGURE 25.

Design of heterobifunctional molecule targeting mutant p53 overabundance. (A) Chemical structure of bifunctional compound p53‐01 consisting of PMV6, a p53 Y220C mutant binder, and the PLK1 inhibitor BI2536, connected by a PEG linker. (B) Schematic representation of the mechanism of action of p53‐01, where cells bearing the p53 Y220C missense mutation display prolonged p53 half‐life and accumulation. Elevated mutant p53 levels increase intracellular binding sites for the bifunctional compound, effectively amplifying cytotoxic PLK1 inhibitor concentration and selectively reducing cell viability in mutant cells.

Combined, Sadagopan et al. demonstrate that protein abundance can serve as a new druggable parameter, expanding the scope of heterobifunctional molecules to induce cytotoxic effects that scale with intracellular target levels.

6.3. Making It Stick: Covalent Targeting of RNA

RNA is a versatile biomolecule that can adopt defined secondary and tertiary structures essential for biological function [167]. RNA has historically been more challenging to target with small molecules than proteins, in part because RNA is composed of only four chemically similar nucleobases compared to the 20 natural amino acids of proteins. That being said, only a modest fraction of the proteome is considered “druggable,” and many disease‐associated proteins lack deep and well‐defined binding pockets favored by small molecules [168]. Intriguingly, these same proteins are encoded by mRNAs that fold into stable structural elements [169], fueling growing interest in RNA itself as a drug target [54]. Previous work to develop covalent RNA‐targeting compounds relies largely on conjugating promiscuous electrophiles, such as nitrogen mustards or N‐acylimidazoles, to RNA‐specific noncovalent binding elements, such as heteroaromatic rings [170]. While effective in some cases, this approach risks off‐target reactivity. Small molecules capable of binding covalently and selectively to RNA structures remain largely unexplored.

Now, a team led by Matthew D. Disney describes a new high‐throughput method to identify small molecules that form covalent adducts with RNA structures [171]. Optimization of a SPRI magnetic bead cleanup enables efficient RNA purification from 5 µL reactions in 384‐well plates. Rather than starting with known nucleic acid‐reactive warheads, the authors use a MALDI‐TOF mass spectrometry‐based screen to detect covalent RNA modifications. By screening a library of 2,000 electrophiles against structured RNA derived from the r(CUG)exp repeat associated with myotonic dystrophy type 1 (DM1) [172], the authors uncover 34 RNA‐reactive compounds.

The screen identifies 3‐chloropivalamides, a surprising finding given that neopentyl halides are classically considered poor S _N 2 electrophiles [173]. Strikingly, compounds containing this group show a hit rate of 49% compared with ~0.5% for the remainder of the library. Follow‐up experiments demonstrate that these compounds alkylate guanine, likely at the reactive N7 position in the RNA major groove. Reactivity drops markedly in the absence of noncovalent RNA binding motifs, indicating that both the electrophile and noncovalent elements are required for efficient binding.

Leveraging these insights, the authors rationally design an RNA‐specific covalent ligand by conjugating a well‐known DNA‐binding Hoechst derivative to the 3‐chloropivalamide electrophile (Figure 26A). This compound shows enhanced selectivity for a DM1‐associated RNA structure mimic compared with control RNA and DNA, demonstrating heightened specificity absent in earlier, more promiscuous RNA‐targeting strategies. Molecular modeling suggests that in the DM1‐associated RNA structure, Hoechst binds in the major groove, positioning the electrophile near the nucleophilic N7 position of guanine. In DNA, Hoechst binds in the minor groove away from productive nucleophiles, enabling selectivity for RNA over DNA. Together, these results highlight the importance of electrophile identity, noncovalent RNA‐binding elements, and positioning of the reactive group in achieving RNA‐selective covalent modification.

FIGURE 26.

Identification of Covalent RNA Probes. (A) A MALDI‐TOF‐based mass spectrometry screen identifies 3‐chloropivalamides as RNA‐reactive electrophiles that form selective covalent adducts with structured RNAs. Guided by these findings, the authors rationally design a small molecule that combines the validated electrophile, a noncovalent RNA‐binding element, and positioning to enable nucleophilic attack within the RNA structure. (B) Although proteins have historically dominated drug discovery, structured RNAs may offer a largely untapped set of targets that can be engaged through covalent screening strategies.

While fundamentally a proof‐of‐concept study, the team's unbiased screening approach has the potential to expand how RNA is framed as a drug target (Figure 26B). The results suggest that RNA‐selective electrophiles may be more general than previously appreciated, with high‐throughput screens serving as powerful tools for uncovering this underexplored reactivity. More broadly, the work implies that RNA selectivity may be engineered through the same principles that transformed covalent protein drug discovery: proximity, orientation, and controlled reactivity. Rapid advances in RNA structural biology continue to reveal structured RNAs, from riboswitches to viral RNAs, that present opportunities for targeting via small molecules [174]. While translation to broadly applicable drugs will require further work, scalable screening approaches such as this one could help reposition RNA as an expandable target class for small molecules.

6.4. Synthetic Mucins as Glycan‐Defined Prebiotics

The human gut is home to trillions of microorganisms whose activities profoundly influence host health [175]. Among these, probiotic bacteria such as Lactobacillus species have been found to improve gut barrier function, suppress pathobiont growth, and modulate immune responses [176]. Yet, despite their widespread use, many probiotics fail to persist in the gut and are typically cleared from the body within three to ten days due to poor mucosal adhesion and colonization efficiency [176, 177]. A major barrier to their persistence lies in the mucosal layer, where interactions between bacteria and host‐secreted mucins determine whether these microbes adhere. Kiessling and colleagues now report a chemical strategy to investigate these interactions using synthetic mucin mimics with defined glycan structures [176].

Mucins are heavily glycosylated proteins that form the structural scaffold of mucus [178]. Their dense array of O‐linked glycan residues provides both binding sites and nutrient sources for commensal microbes, and mucins are therefore considered endogenous prebiotics that regulate microbial organization and metabolism [179]. However, the structural heterogeneity of native mucins has made it difficult to determine which glycan features govern specific probiotic behaviors [176]. To address this challenge, the authors designed synthetic polymers that mimic the extended, linear structures and the multivalent glycan display of native mucins [175], while still allowing for precise control over the glycan epitopes used.

The authors synthesized a collection of amine‐terminated O‐glycan epitopes representing the most prevalent neutral glycan building blocks found in native mucins, including α‐fucose (α‐Fuc), β‐galactose (β‐Gal), β‐N‐acetyl galactosamine (β‐GalNAc), β‐N‐acetyl glucosamine (β‐GlcNAc), β‐lactose (β‐Lac), and α‐mannose (α‐Man) (Figure 27A). These glycans were installed on a polymer scaffold previously developed by the authors using a cis‐selective ring‐opening metathesis polymerization of a norborene derivative. This polymer backbone forms an extended, bottlebrush‐like structure that closely mimics the morphology of native mucins while improving steric accessibility and water solubility (Figure 27B) [176, 180].

FIGURE 27.

Design of glycan‐functionalized synthetic mucins. (A) Mucin surrogate synthesis: cis‐poly(norborene) scaffolds bearing defined O‐glycan epitopes, a linker (ethanolamine, triethylene glycol, or aryl), and one Alexa Fluor 405 fluorophore per chain. (B) Synthetic mucins adopt linear structures that mimic the linear bottlebrush structure of native mucins.

Using these mucin mimetics, the authors observed that different Lactobacillus species displayed distinct glycan preferences. For example, Lactiplantibacillus plantarum preferentially bound mannose‐ and fucose‐containing glycans, whereas Limosilactobacillus fermentum exhibited strong affinity for lactose‐derived motifs. Interestingly, these preferences were not fixed but depended on environmental conditions such as nutrient availability and growth phase, suggesting that mucin recognition is regulated by the metabolic state of the bacteria. Together, these results indicate that probiotic–mucin interactions are dynamic and responsive to both microbial and environmental contexts.

Multivalent glycan presentation further influenced probiotic behavior by promoting bacterial clustering and adhesion to mucin‐coated surfaces. Synthetic mucins displaying preferred glycans enhanced retention of certain probiotic strains, indicating that glycan identity can regulate microbial organization within the mucosal layer. Beyond adhesion, the authors demonstrate that these interactions also carry metabolic consequences. Exposure to specific glycans stimulated activity of bacterial glycosidases, enabling microbes to cleave these glycan epitopes and use them as nutrient sources, thus supporting bacterial growth.

These findings position synthetic mucins as powerful chemical tools for probing host–microbe interactions at mucosal interfaces. The authors’ approach reveals how subtle variations in glycan presentation can shape probiotic metabolism and persistence. More broadly, the work opens new opportunities for designing glycan‐defined prebiotics that enhance probiotic retention and function, offering promising strategies for promoting human health.

6.5. Antibody‐Bottlebrush Conjugates Expand Targeted Drug Delivery

Antibody‐drug conjugates (ADCs) have emerged as a compelling modality for the targeted delivery of anticancer therapeutics [181, 182]. By directly conjugating drug molecules to antibodies that recognize tumor epitopes, ADCs ensure drug molecules specifically hit diseased cells, reducing off‐target cytotoxicity. Yet, existing ADCs are constrained by the number of surface‐exposed lysines and cysteines on the antibody available for functionalization, limiting the number of drug molecules delivered per antibody [172, 181]. Liu et al. now propose a solution to this problem by introducing antibody‐bottlebrush prodrug conjugates (ABCs), a reimagining of the ADC that shows great promise in advancing the field of cancer therapy [183].

Unlike traditional ADCs, ABCs first conjugate drug molecules onto hydrophilic PEG chains grafted along a rigid central backbone [184]. The result is a densely loaded scaffold resembling a bottlebrush with drug molecules buried among the bristles, known as a bottlebrush prodrug (BPD) [185]. This BPD is then attached to an antibody through a tetrazine‐TCO inverse‐electron‐demand Diels‐Alder reaction (Figure 28A).

FIGURE 28.

Synthesis and Mechanism of ABCs. (A) Chemical synthesis of ABCs through ROMP polymerization of BPD monomers and IEDDA click to Ab‐TCO (B) ABC‐mediated targeted intracellular drug delivery.

This design enables the tuning of both the drug‐antibody ratio (DAR) and the biophysical properties of the conjugate. Because the antibody no longer bears individual payloads directly, the DAR is no longer limited by the number of accessible residues, expanding the possible DAR range from <8 in traditional ADCs to over 150 [181, 182, 183]. In addition, because the length and identity of both the side chains and central polymer backbone can be varied synthetically, properties such as hydrophobicity can be tuned, reducing the propensity for aggregation, decomposition, and accumulation in off‐target organs.

Functionally, the constructs exhibit excellent selectivity and potency in vitro. Liu et al. demonstrate this by generating a panel of ABCs with clinically approved antibodies targeting HER2 and MUC1 and equipping them with payloads including microtubule inhibitors, topoisomerase inhibitors, and protein degraders (PROTACs) [186, 187]. Across antigen‐positive cell lines, these constructs show strong target‐dependent cytotoxicity while sparing antigen‐negative controls.

Encouraged by these results, the authors then evaluated ABCs in vivo. Human antigen‐positive tumor cells were implanted subcutaneously into immunodeficient mice, with therapeutics administered intravenously. Across studies, ABC‐treated mice showed marked tumor growth suppression and, in some cases regression, relative to controls and benchmark ADCs. Pharmacokinetic analyses indicated that the ABC remains stable in circulation, allowing sustained systemic exposure over time without rapid clearance of the payload. Biodistribution studies showed preferential tumor accumulation compared to untargeted bottlebrush prodrug controls consistent with antibody‐mediated targeting. Importantly, these results were observed without overt systemic toxicity.

In sum, Liu et al. introduce ABCs as a structurally distinct alternative to conventional ADCs that overcomes key limitations. By shifting drug conjugation onto a tunable polymer scaffold, ABCs enable substantially higher drug–antibody ratios while maintaining serum stability, favorable pharmacokinetics, and tumor‐selective accumulation in vivo. The demonstrated efficacy across multiple payload classes, including agents not typically amenable to ADC formats, highlights the platform's versatility. Together, these results position ABCs as a promising strategy to expand the scope and performance of targeted cancer therapeutics.

6.6. First in Class Phospholipid Degrader

Despite significant advances in our scientific understanding of cancer biology, the efficacy of cancer treatments remains limited by adaptive cellular processes that promote drug resistance. Sub‐populations of cancer cells evade conventional cancer therapies using nongenetic mechanisms by entering a drug‐tolerant persister (DTP) cell‐state [188]. One hallmark of the DTP cell state is the high expression of the iron‐uptake protein CD44, which enables cells to dynamically alter between cell states through epithelial‐mesenchymal transitions, enabling metastasis [189].

The high levels of iron in DTP cancer cells cause them to be susceptible to ferroptosis, a non‐apoptotic form of cell death in which iron catalyzes the peroxidation and degradation of membrane lipids [190]. Cañeque et al. use chemical probes to demonstrate that lysosomal iron initiates lipid peroxidation, triggering ferroptosis. This work establishes catalytic lysosomal iron as an attractive target for triggering ferroptosis in cancer cells that are dependent on high levels of iron to maintain a DTP cell‐state [191].

Building on this insight, Cañeque et al. develop a first‐in‐class bifunctional phospholipid degrader, which induces proximity between catalytic iron (II) and phospholipids to promote oxidative degradation (Figure 29). The phospholipid degrader, Fentomycin‐1, is composed of a fluorescent lipophilic natural product, Marmycin, a short linker, and the catalytic Chen‐White ligand [192, 193]. Under mildly acidic aqueous conditions and in the presence of hydrogen peroxide, the Chen‐White ligand is suspected to activate iron (II) to form a reactive iron‐oxo species capable of oxidizing nearby lipids [193].

FIGURE 29.

Development of phospholipid degraders. (A) Overview of Fentomycin‐1 synthesis and the formation of the redox active iron catalyst. (B) Fentomycin‐1 is endocytosed into cells and localized to the lysosome. It then activates lysosomal iron with the help of hydrogen peroxide and oxidizes proximal lipids, leading to their degradation. Lipid peroxidation propagates and eventually spreads to the outer membrane, which ultimately leads to membrane rupture and ferroptosis.

Marmycin was selected to localize the catalytic ligand to lysosomal phospholipids because it is known to accumulate in the lysosomal membrane after being endocytosed. Together, Fentomycin‐1 takes advantage of the abundant levels of iron (II) in DTP cells to induce proximity between catalytic iron and phospholipids to promote the oxidative degradation of phospholipids and ultimately trigger ferroptosis [192].

The authors confirmed with fluorescent imaging and lipidomic studies that Fentomycin‐1 localizes to the lysosome and oxidizes phospholipids. In mice, Fentomycin‐1 selectively killed cells with high CD44 levels and reduced tumor growth. Proteomic analyses further showed that sublethal doses shift cancer cells from a CD44‐high DTP state to a CD44‐low state that is more susceptible to apoptosis.

Together, Cañeque et al. identified lysosomal iron as a tractable target in drug‐resistant cancer cells and introduced a strategy that extends proximity‐induced degradation to lipids.

7. Advances in Synthetic Chemical Biology

7.1. Opening the Doors to Noncanonical Amino Acids

Genetic code expansion (GCE) enables the site‐specific incorporation of non‐canonical amino acids (ncAA) into proteins through reassignment of the amber stop codon and the use of orthogonal translation components such as aminoacyl‐tRNA synthetase/tRNA pairs (aaRS/tRNA) [194]. This platform facilitates the installation of diverse chemical functionalities, including mimics of posttranslational modifications [195], bioorthogonal handles [196], and chemical or photo‐crosslinkers [197]. Despite efforts to optimize existing approaches, practical implementation remains constrained by inefficient uptake of many ncAAs. Because many synthetic ncAAs are difficult and expensive to synthesize, and because bulky or charged derivatives often show poor cellular uptake, high extracellular concentrations are frequently required to achieve useful intracellular incorporation, limiting broader application.

Building on earlier work [197, 198] on a site‐specific azide‐caged diglycine acceptor motif, Iype, Fottner et al. developed isopeptide‐linked tripeptides designed as a general Z–XisoK scaffold [199], where Z and X could be either a natural or non‐canonical residue. The authors showed for glycine‐alanine‐isoleucine (G‐AisoK) that after import into the cell, the N‐terminal glycine is proteolytically removed, generating the active AisoK species, which is subsequently incorporated via amber stop codon into proteins using the aaRS/tRNA from Methanosarcina barkeri. In contrast, free AisoK derivatives alone show minimal cellular uptake and negligible incorporation. Highly efficient intracellular accumulation and incorporation into the protein of interest was observed for G‐AisoK (Figure 30), which outperformed the commonly used BocK as a “golden standard” of ncAA.

FIGURE 30.

Strategic hijacking of the Opp ABC transporter for expanded genetic code functionality. (A) Z–XisoK tripeptide strategy from delivery to protein incorporation, G‐AisoK structure, scope of ncAA incorporation, and OppA–G‐SisoK structure (PDB: 9RD1) highlighting substrate recognition. (B) Proposed mechanism for the active transport of non‐canonical amino acid (ncAA) precursors into the bacterial cell via the engineered and WT ABC transporter system.

Genetic analysis further showed that uptake of G‐AisoK depends on the oligopeptide permease (Opp), a bacterial ATP‐binding cassette (ABC) transporter. Opp comprises the periplasmic substrate‐binding protein OppA, the transmembrane domains OppB/OppC, and the ABCs OppD/OppF. Structural characterization of the OppA‐G–SisoK complex revealed that substrate recognition primarily depends on interactions with the tripeptide backbone and terminal groups rather than the side chain identity. This binding mode explains the transporter's tolerance for diverse synthetic ncAA‐containing tripeptides. The uptake could be further enhanced through genetic engineering.

Overall, the study establishes transporter engineering as a powerful and general route to expand the usable chemical space of GCE.

7.2. Programmable Pseudouridine (Ψ) Codons for ncAA Encoding

The incorporation of ncAAs allows for the functional manipulation of proteins of interest [200]. GCE achieves this by replacing stop codons with blank codons; however, this process often results in off‐target translational readthrough in endogenous stop codons [201, 202]. In this paper [203], Liu et al. introduce an RNA codon expansion (RCE) strategy to incorporate pseudouridine (Ψ) into mRNA codons to enable highly specific ncAA incorporation in living cells.

RCE is a two‐step process consisting of encoding and decoding. For encoding site‐specific Ψ codons, the authors utilized three components: the programmable pseudouridylation tool RESTART [204], a CRISPR‐free base‐editing technology that incorporates Ψ into mRNA to override premature stop codons; guide small nucleolar RNAs (gsnoRNAs), which are short, non‐coding RNAs found in nuclei; and DKC1, a gene encoding a protein involved in the assembly of snoRNPs. In tandem, these components produce Ψ codons (ΨGA/ΨAA/ΨAG) from endogenous stop codons (UAA/UAG/UGA) (Figure 31A). Transcriptome‐wide, single‐base resolution quantitative analysis of Ψ (PRAISE) showed high Ψ installation ratios, with an average of 71%, for the UGA target. Furthermore, PRAISE detected only 9 off‐target locations, each with low Ψ ratios and no incorporation in endogenous stop codons. As such, their system is highly specific for Ψ codon generation.

FIGURE 31.

Overview of the RNA codon expansion system (RCE). (A) Method for encoding in RCE and representative ncAAs utilized to demonstrate proof of concept. (B) RCE components and method for decoding in RCE.

For decoding, they screened wild‐type and engineered tRNA^Pyl structures to identify decoders specific for the stop codons ΨGA, ΨAA, and ΨAG. These tRNAs were generated from tRNA^Pyls with single point mutations, especially at position 37 in the anticodon stem‐loop, as these variants showed strong preference for the Ψ codon with highly efficient ncAA incorporation. Of those tested, the tRNA^Pyl(UCA)‐37G from Methanosarcina mazei (Mm) was selected.

Ribosome profiling experiments were conducted to determine off‐target modifications for the RCE(ΨGA) system. Although RCE provided a similar on‐target readthrough as GCE, their data suggest the RCE demonstrated a fourfold increase in the specificity of ncAA incorporation for the target transcript compared to GCE. Additionally, RNA‐seq experiments demonstrated no interference of the RCE system with gene expression.

To demonstrate proof of concept, they utilized the RCE system to incorporate representative ncAAs, including TCOK‐α, a chemically caged lysine analog, CbzK, an aromatic lysine derivative, and the bioorthogonally clickable AzK (Figure 31A) in order to modulate enzyme activity, label proteins of interest, and control protein localization. The incorporation of TCOK‐α and AzK at these positions allowed for a decaging‐mediated reactivation of SRC kinase activity in situ and the installation of azide handles for strain‐promoted click chemistry used in the selective labeling of modified proteins.

Overall, RCE establishes a programmable framework for transcript‐specific codon reassignment, substantially improving the precision of ncAA incorporation and enabling more selective manipulation of proteins in living systems.

Caltech Advanced Chemical Biology Class 2026 (Ch/Bi 145) Students enrolled in chemistry and biology programs across Caltech. PhD candidates Chloe Cerione and Jonathan Farhi were Teaching Assistants, and Assistant Professor Johannes Morstein was the instructor for the course.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.