Metabolites, long recognized as essential reactants and products of cellular metabolism—serving as energy carriers, structural components, and biosynthetic precursors —are increasingly appreciated for their roles as vital signaling molecules that orchestrate intricate cellular responses [1]. These signaling metabolites, exert precise control over protein function via mechanisms including allosteric regulation, direct enzyme substrates, drivers for post-translational modifications (PTMs), and gene expression control through epigenetic or transcription factor modulation [2]. Furthermore, the complexity of metabolite-regulated signaling arises from concentration-dependent dynamics and combinatorial effects, such as simultaneous or competitive binding to proteins, or convergent regulation of shared processes like transcription [3]. Thus, the interests in studying PMIs surges and the knowledge of PMIs is anticipated to yield profound insights into intracellular metabolic networks and their dynamic plasticity under physiological perturbations (e.g., nutrient availability) and pathological insults (e.g., oxidative stress), unravel metabolic dysregulation-related disease mechanisms at a molecular level and hence open avenues for innovative therapeutics, such as targeting metabolite–enzyme interfaces for drug design [4,5].

Despite their significance, PMIs remain far less characterized than protein–protein interactions, primarily due to the lack of high-throughput methodologies for comprehensive profiling in complex biological contexts. Classical approaches such as affinity purification-mass spectrometry (AP-MS) have been informative—for example, the Emili group used AP of 6xHis-tagged proteins based on ASKA E. coli K12 strain library to identify 296 high-confidence PMIs across 114 proteins [6]. However, these methods are limited by the need for cell lysate and in vitro purification, which hinders large-scale PMI profiling in physiologically relevant samples [7]. Recently, advanced MS-based chemoproteomic approaches have begun to overcome these barriers, enabling systematic identification of PMIs, even including low-affinity and transient ones. Thus, this review summarizes recent progress in derivatization-based and derivatization-free chemoproteomic strategies (Figure 1) and highlights their effectiveness in uncovering previously unrecognized PMIs. These developments deepen our understanding of cellular metabolism and will ultimately enhance our ability to address metabolism-relevant disease mechanisms and developing targeted therapies.

Classical derivatization-based approaches employ specialized probes (Table 1) to capture covalent or, with appropriate modifications, non-covalent PMIs. Targets are enriched via bioorthogonal handles or affinity tags and subsequently identified by MS [8]. These well-established strategies provide robust platforms for systematic PMI characterization.

Certain metabolites containing electrophilic groups can covalently modify nucleophilic amino acid residues on proteins. This property has been harnessed to design covalent PMI probes based on the active structures of metabolites, facilitating identification of the metabolites-modified target proteins. For example, recognizing that itaconate can covalently modify proteins, the Wang group designed a bioorthogonal probe, ITalk (Table 1), incorporating an alkyne handle derived from itaconate [9]. This probe was later employed by the O'Neill group to show that itaconate and its derivatives target JAK1, inhibiting macrophage alternative activation, providing a potential therapeutic strategy for treating type 2 immunity-driven diseases such as asthma [10]. In addition, the Wang group designed the C3A probe (Table 1), which replaces the ester linkage with an amide bond to connect the α, β-unsaturated carboxylic acid moiety to a propynyl group, thereby preventing the potential cleavage of the ester linkage by bacterial-specific hydrolases. This modified probe revealed that itaconate binds to isocitrate lyase in S. enterica, inhibiting its enzymatic activity and stability, thus exerting an antimicrobial effect [11]. Further applications of this probe led to global mapping of itaconate-modified targets in three pathogens, uncovering the PMIs primarily occurred within the de novo purine biosynthesis pathway, making this pathway a promising target for new antibacterial drugs [12]. PMI mapping for other electrophilic metabolites has been extensively reviewed elsewhere [13, 14, 15].

Covalent PMIs can also be studied through metabolic labeling probes designed with the structures of the donor precursor as the template and applicable to living cell labeling. For instance, a lactate analogue, YnLac, was designed by linking an alkyne group to the methyl terminus of lactate (Table 1). This analogue was incorporated into cellular proteins, leading to the identification of PARP1 as a lactate-modified protein and led to further observation that the lactate-derived modification on PARP1 further modulates its ADP-ribosylation activity [16]. Metabolic glycan labeling has also been used in similar studies and has been comprehensively reviewed elsewhere [17].

Given that many covalent PMIs typically modify reactive residues such as cysteine and lysine, competitive profiling has been adopted for PMI mapping as a way to bypass metabolite derivatization. In this approach, proteins from metabolite-pretreated and untreated samples are covalently modified with residue-reactive probes, and differences in target enrichment reveal metabolite-reactive sites. For instance, the Meier group employed iodoacetamide-alkyne (IA-alkyne, Table 1) probes to competitively profile the targets of fumarate, an oncometabolite that covalently modifies cysteine thiols. This work identified functionally relevant cysteine sites affected by fumarate engagement in hereditary leiomyomatosis and renal cell carcinoma, revealing a mechanism of hereditary cancer driven by bioactive metabolites [18]. Notably, a limitation of this strategy is that, while it effectively pinpoints residues sensitive to metabolite engagement, it does not directly determine whether these sites represent primary binding locations or regions influenced through allosteric effects.

In contrast to covalent PMIs, some metabolites modulate biological functions through affinity binding to proteins. To capture such PMIs that may be transient and unstable, photoaffinity probes have been developed. These probes contain photoreactive groups, such as aryl azides, aryl ketone, or diazirines, which are adopted to form covalent bonds with nearby proteins upon UV irradiation [19]. Photoaffinity probes have been widely used to capture the binding targets of lipid (e.g., polyunsaturated fatty acids [20], cholesterol [21], phospholipids [19]), glycolytic metabolite (e.g., fructose-1,6-bisphosphate FBP [22]), enzyme cofactor (e.g., vitamin B12 [23]) and bile acid metabolites [24]. Recent advances in bile acid-derivatized photoaffinity probes (Table 1) have enabled the identification of direct targets for microbial metabolites in host cells [25] and microbiota [26,27]. Building on this methodology, the Hang group designed a photoaffinity probe for a secondary bile acid lithocholic acid (LCA) and utilized it to elucidate its PMIs in E. faecium [28]. Through subsequent collaboration with the Shen group, they applied this probe to identify BapR, a novel LCA-sensing transcription factor in C. difficile that mediates metabolic adaptation via gene regulation, revealing mechanistic insights into gut microbiome–pathogen interactions [29]. In a recent study, the Lin group investigated the anti-aging mechanisms of LCA by designing a multifunctional probe (Table 1) containing a diazirine group for photoaffinity capture of PMIs and an alkyne handle for enrichment. Using this probe, chemoproteomic analysis revealed that LCA does not directly bind to SIRT1. Instead, pull-down and knockout experiments identified TULP3 as the primary target, with LCA binding to TULP3 allosterically activating SIRT1. The TULP3–SIRT1 complex deacetylates the V1E1 subunit of v-ATPase, inhibiting its activity. This inhibition activates the AMPK pathway, reducing calorie intake and extending lifespan. These findings uncover a novel anti-aging mechanism and broaden our understanding of the diverse roles of bile acid metabolites in humans [30].

While derivatization-based methods have provided valuable insights into PMIs, they require introducing chemical modifications to the metabolites of interest. Such modifications can potentially alter the native physicochemical properties of the metabolites, thereby biasing the interaction profile and leading to the capture of skewed PMIs. These limitations have driven the development of derivatization-free chemoproteomic methods that recognize metabolites-binding target proteins at the proteome-level without altering the metabolites themselves. This section introduces such derivatization-free strategies, which detect PMIs by monitoring changes in protease resistance, thermal stability, and protein accessibility upon metabolites binding (Table 2).

Early studies on gene promoters revealed that DNA binding sites, when bound by transcription factors, exhibit resistance to nuclease degradation. This observation inspired the hypothesis that drug binding could similarly confer protease resistance to target proteins. The first method developed to explore this concept was Drug Affinity Responsive Target Stability (DARTS). In DARTS, a ligand is incubated with cell lysate, followed by treatment with non-specific proteases. Protein degradation is then assessed using gel electrophoresis or liquid chromatography-mass spectrometry (LC-MS), identifying proteins that exhibit significant stability changes as potential targets [31]. Recently, the Tang group applied DARTS to demonstrate that malate directly binds to the binding immunoglobulin protein (BiP), a key mediator of malate's anti-inflammatory effects in macrophages [32].

Another widely used protease resistance-based approach is limited proteolysis-mass spectrometry (LiP-MS), which provides detailed insights into ligand-induced conformational changes and binding pockets. This method involves partial digestion of protein samples with non-specific proteases, followed by secondary digestion with trypsin. Using LiP-MS, the Picotti group identified 1447 previously unknown PMIs and their associated binding sites in E. coli [33]. LiP-MS has since been applied to investigate PMIs in the tumor microenvironment. The He group used this technique to discover that the purine metabolite inosine binds to UBA6 in tumor cells, inhibiting its activity. This PMI sensitizes tumor cells to T cell-mediated killing, thereby enhancing tumor immunogenicity and generating an immunologically active microenvironment [34]. LiP-MS has also used to elucidate the binding targets of arginine, and uncovered three essential transcriptional regulators for arginine-dependent T cell survival [35].

Alternatively, the Ye group recently developed PEptide-centric Local Stability Assay (PELSA), which amplifies ligand-induced protein stability changes by using a high enzyme-to-substrate ratio with destructive trypsin digestion [36]. PELSA is effective for both traditional drug target discovery and low-affinity PMIs, as demonstrated by its ability to detect weak PMIs such as folate-DHFR (Ka = 98 ± 20 mM [37]) and leucine-LARS1 (Kd = 96.1 ± 37.7 μM [38]). Application of PELSA to α-ketoglutarate (α-KG) and its structurally related oncometabolite R-2-hydroxyglutarate (R2HG) revealed distinct binding preferences, with pyruvate carboxylase showing higher affinity for R2HG. Given that pyruvate carboxylase is a key enzyme in tricarboxylic acid (TCA) cycle anaplerosis, these findings provide mechanistic insight into TCA-cycle dysregulation in IDH2-mutated cancers, where R2HG accumulates to high levels.

Cellular Thermal Shift Assay (CETSA) and Thermal Proteome Profiling (TPP) are widely used for monitoring changes in protein thermal stability upon small-molecule binding [39,40]. TPP integrates temperature gradients and MS to generate protein melting curves, enabling high-throughput proteome-wide analysis of thermal stability shifts induced by ligand engagement. A recent application of TPP to PMI studies by the Geng group identified fructose-1,6-bisphosphate (FBP)-binding proteins, revealing that FBP functions as a phosphate donor to activate phosphoglycerate mutase 1 (PGAM1) via histidine phosphorylation, thereby promoting glycolysis and tumor proliferation. Building on this discovery, they designed the FBP analogue 1-DMeFBP, which disrupts phosphate transfer and inhibits PGAM1 activity, providing a potential therapeutic strategy to target cancer metabolism [41]. More broadly, TPP has been increasingly applied to study endogenous ligands, such as metal ions [42] and dinucleotide [43], demonstrating its utility as a universal approach for PMI identification and for gaining mechanistic insight into metabolite actions.

To simultaneously assess both thermal stability and binding affinity, the two-dimensional TPP (2D-TPP) method was developed. This approach incorporates multiple ligand concentrations alongside temperature gradients, improving the resolution of PMI detection [44]. For example, the Savitski group applied 2D-TPP using 5 ligand concentrations and 10 temperatures to characterize ATP-mediated PMIs in Jurkat cell lysate. Their study revealed ATP's dynamic roles—serving as a substrate at low concentrations and as a protein solubilizer at high concentrations, highlighting the versatility of 2D-TPP in elucidating multifunctional metabolites [45].

Traditional stability-based methods are limited in studying proteins with unusual structural or functional features, such as extreme heat resistance, thermally instability, or hydrolases insensitivity. To address these challenges, our group developed Target Responsive Accessibility Profiling (TRAP), a complementary target discovery approach that maps proteome-wide accessibility changes upon ligand binding through lysine accessibility profiling via isotope-coded dimethylation [∗∗46, 47, 48]. Unlike stability-based methods such as DARTS or TPP, TRAP pinpoints ligand-binding regions at the residue level, enabling effective analysis of proteolytically or thermally stable proteins [46].

Applying TRAP to cancer cells, we systematically profiled PMIs of ten key glycolytic metabolites, identifying 2487 interactions spanning metabolic enzymes, transcriptional regulators, and post-translationally modified proteins. Mechanistic studies revealed that: (i) glycolytic metabolites cooperatively modulate PKM2 activity in a composition-dependent manner; (ii) lactate engages TRIM28, potentially promoting NLRC3 suppression and PI3K-AKT-mTOR pathway activation; and (iii) pyruvate attenuates trichostatin A—induced histone acetylation, protecting cancer cells from apoptosis. These results establish TRAP as a powerful platform for uncovering PMI—driven regulatory mechanisms and identifying metabolic vulnerabilities in cancer.

Expanding the toolbox for protein accessibility-based target identification, several chemical labels have been adopted for lysine labeling and show potential for PMI discovery [49]. For example, Reactive Amino acid Profiling by Inverse Detection (RAPID) uses membrane-permeable ortho-phthalaldehyde (OPA) probes to label lysine residues in living cells, enabling drug target discovery by monitoring ligand-induced accessibility changes [50]. Other reagents, such as O-cyanobenzaldehyde [51] and cationic pyridinium activated esters [52], can profile lysine accessibility but have not yet been applied to drug-target discovery or PMI mapping. Together with labeling reagents targeting other amino acids [53,54], these chemical labeling strategies hold strong potential for PMI detection via accessibility profiling and could substantially broaden the scope of metabolite target identification.