Off-target identification by chemical proteomics for the understanding of drug side effects

KEYWORDS:

• Chemical proteomics
• ABPP
• CCCP
• drug targets
• side effects

1. Introduction

Drug side effects (also known as adverse events) are noxious and unintended responses of drugs that occur at doses normally used in humans [1]. It has been shown that drug side effects should be responsible for as much as 11% of hospital admissions, 20% of clinical trial failures, several high-profile drug withdrawals (such as Vioxx, Lipobay, and Thalomid), and a large portion of therapeutic noncompliance incidents [2]. Unraveling the side effect mechanisms will undoubtedly promotes the development of intervention strategies to circumvent drug side effects in future clinical application. During the past decades, the rapid development of chemical proteomics approaches and mass spectrometry (MS) technologies has enabled the unbiased deconvolution of drug off-targets, contributing greatly to our understanding of the side effect mechanisms.

2. Chemical proteomics in the identification of drug off-targets

Chemical proteomics approach is capable of identifying the binding proteins of small molecules in cell lysates or living cells via small chemically synthesized probes. A typical workflow of chemical proteomics starts with probing the proteome of interest, followed by enrichment of probed drug-target complexes and subsequent MS analysis. The core of chemical proteomics is to design proper probes that can immobilize small molecules and specifically link to target proteins [3]. Here, we will summarize two commonly used chemical proteomics strategies in drug off-targets identification for unraveling side effects.

2.1. Activity-based protein profiling (ABPP)

ABPP uses activity-based probes (ABPs) to covalently react with the catalytic residues of specific enzyme families from complex proteomes. A functional ABP usually consists of three components, a reactive group (also called warhead) that can covalently bind to the active sites of target proteins, a reporter tag (such as fluorophore, biotin, or alkyne) for the downstream purification and detection of target proteins, a linker (such as PEG, alkyl, or peptide) which connects the reactive group and tag as well as avoids steric hindrance [4]. According to the different characteristics of reactive groups, ABPs can be used to target different types of enzymes. For example, the fluorophosphonate (FP) probe has been well developed to capture serine hydrolases, while the epoxide electrophile probe has been designed for the binding of cysteine proteases [5]. However, the commonly used biotin or fluorophore tags are bulky and membrane impermeable, these ABPP approaches are unable to detect the small molecule–target interactions in situ. To this end, a two-step tag-then-capture approach called click chemistry-ABPP (CC-ABPP) has been developed, which is based on the copper-catalyzed azide-alkyne cycloaddition reaction (CuAAC), a typical bioorthogonal ligation. Using this two-step post-labeling CC-ABPP, target proteins can be probed in living cells, the CuAAC reaction can then be carried out in cell lysates [6]. In addition to covalent binding, ABPP can also investigate noncovalent interactions between small molecule and their targets through photo-affinity labeling (PAL), in which benzophenone, aryl azide, or diazirine is applied as a photocrosslinker [7].

During the past decades, ABPP has been successfully employed to identify the off-targets of many drugs. A recent example is BIA 10–2474, an inhibitor of fatty acid amide hydrolase (FAAH). BIA 10–2474 was a potential drug for the treatment of anxiety and pain but failed in phase I clinical trial due to unknown severe neurotoxicity. To uncover the mechanism of this side effect, competitive ABPP was performed to profile the serine hydrolase interactome of BIA 10–2474. As a result, BIA 10–2474 was found to interact with several other lipases and cause lipid network alterations in human cortical neurons, which may contribute to its neurotoxicity [8]. In another study, quantitative acid-cleavable ABPP (QA-ABPP) was developed to map the acetylated target proteins of aspirin. In total, 523 acetylated proteins were identified as potential targets of aspirin, which may explain its multiple drug actions or side effects [9]. However, due to the specific characteristics of different probes, ABPP is only available to investigate defined enzymes families, which represent a small portion of the whole proteome.

2.2. Compound-centric chemical proteomics (CCCP)

Different from ABPP, CCCP is a more unbiased approach to interrogate drug targets, regardless of enzymatic functions. CCCP relies on drug affinity chromatography, in which drug molecules will be immobilized on commercially available matrix, such as agarose or magnetic beads. This immobilization requires specific functional groups on drug molecules, such as amino, sulfhydryl, carboxyl, or hydroxyl groups. If not, the structure of drug molecules of interest can be chemically modified to introduce these functional groups. Alternatively, the drug molecules can be chemically linked with affinity tags (such as biotin or fluorescent tags, or in combination with click chemistry) for better target enrichment and identification in affinity chromatography [7,10]. A general principle is that drug immobilization cannot affect its pharmacological activity. The immobilized drug probes are then incubated with protein lysates. Target proteins will be captured and enriched on the probes for further proteomics analysis.

Recently, CCCP has been applied as an effective strategy to interrogate not only the on-targets-related mechanism of action (MOA) but also the off-targets-related side effects of many drugs. For example, Ito et al. investigated the teratogenic mechanism of thalidomide by CCCP, in which carboxylic thalidomide was conjugated to ferrite-glycidyl methacrylate beads. The results showed that thalidomide directly bound to cereblon (CRBN) and inhibited the associated E3 ligase activity, which accounted for its teratogenicity [11]. Recently, our group employed CCCP to investigate the mechanism of bleomycin-induced lung fibrosis, a severe side effect that occurrs in the clinical use of bleomycin. Bleomycin was covalently coupled to NHS-activated Sepharose beads through its inert amino group at the end of its bithiazole tail. MS analysis combined with surface plasmon resonance (SPR) and isothermal titration calorimetry (ITC) analysis identified annexin A2 as a direct binding target of bleomycin. Bleomycin bound to the Glu139 residue of annexin A2 and impeded TFEB-induced autophagic flux, leading to the induction of lung fibrosis. We further found that the sugar moiety of bleomycin was required for annexin A2 binding [12]. Indeed, deglycosylated bleomycin did not induce pulmonary fibrosis, but retained its anticancer activity [13], suggesting deglycosylated bleomycin as a potential surrogate for bleomycin in clinical use.

3. Expert opinion

As mentioned above, chemical proteomics strategies, especially ABPP and CCCP, have contributed greatly to the identification of the off-targets and understanding of the side effect mechanisms of many drugs. However, it must be recognized that there are several limitations to chemical proteomics. For example, the synthesis of probes in ABPP or drug immobilization in CCCP may alter the pharmacological activities of drugs. For some drugs, especially those nature products, no active sites exist in their structure for chemical modification. In this case, drug affinity responsive target stability (DARTS) and thermal proteome profiling (TPP) approaches can be methods of choice, which depend on reduced protease susceptibility and increased thermodynamic stability of the target proteins upon drug binding, respectively, without chemically modifying the drug molecule [14,15]. Using TPP combined with affinity enrichment-based chemoproteomics, the marketed histone deacetylase (HDAC) inhibitor panobinostat (used for the treatment of multiple myeloma), was found to directly engage phenylalanine hydroxylase (PAH) to increase cellular phenylalanine level, decrease tyrosine level and finally cause hypothyroidism [16]. Proteome integral solubility alteration (PISA) approach is then developed to further increase the throughput and simplify the data analysis of TPP [17]. Functional identification of target by expression proteomics (FITExP) is another modern chemical proteomics approach for drug target identification with no need for chemical modification. FITExP is based on the observation that the abundance alteration of drug target proteins in late apoptosis is exceptionally large compared to those co-regulated proteins. Therefore, FITExP analysis requires long-term treatment (often 48 h) of cultured cells at a concentration of LC50 value, which is in contrast to TPP [18]. Based on FITExP, the ProTargetMiner database containing the proteome signature library of over 50 drugs is then established, which serves as a chemical proteomics resource for highly specific target identification [19].

Another limitation of ABPP is that ABPP can only identify the protein targets with enzyme activities. In addition, different preparation approaches for protein lysates can yield different protein compositions; thus, some membrane or hydrophobic proteins may not be captured. The high abundance proteins may also mask the identification of low abundance proteins or real targets. Moreover, many nonspecific drug–protein interactions usually occur. With the recent and future improvement in probe design, cell lysis methods, as well as MS techniques, chemical proteomics will be more feasible for both on-targets and off-targets identification to facilitate our understanding of the MOA and side effect mechanisms, respectively.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This work was supported by grants from the Chinese NSFC (81872277, 81790251, 81821002, 82073081, 82002963), Guangdong Major Project of Basic and Applied Basic Research (2019B030302012), Sichuan Applied Basic Research Project (2020YJ0107), and National Training Program of Innovation and Entrepreneurship for Undergraduates (C2020109620).