Metabolomics in drug research and development: The recent advances in technologies and applications

Abstract

Emerging evidence has demonstrated the vital role of metabolism in various diseases or disorders. Metabolomics provides a comprehensive understanding of metabolism in biological systems. With advanced analytical techniques, metabolomics exhibits unprecedented significant value in basic drug research, including understanding disease mechanisms, identifying drug targets, and elucidating the mode of action of drugs. More importantly, metabolomics greatly accelerates the drug development process by predicting pharmacokinetics, pharmacodynamics, and drug response. In addition, metabolomics facilitates the exploration of drug repurposing and drug-drug interactions, as well as the development of personalized treatment strategies. Here, we briefly review the recent advances in technologies in metabolomics and update our knowledge of the applications of metabolomics in drug research and development.

Graphical abstract

Metabolomics greatly facilitates drug research and development from understanding disease mechanisms and identifying drug targets to predicting drug response and enabling personalized treatment.

1. Introduction

Metabolomics, a thriving field that depicts the comprehensive characterization of metabolites and metabolism in biological systems, allows scientists to decipher the mystery of biology1, 2, 3, 4. Accumulated evidence has demonstrated that complicated diseases and disorders, including cancer, heart disease, diabetes, obesity, infectious diseases, and others, are characterized by metabolic alterations. Metabolomics improves our understanding of the physiological and pathophysiological processes⁵. In this regard, metabolomics enables the identification of disease biomarkers and provides new insight into disease mechanisms6, 7, 8. These essential metabolic insights, in combination with advanced analytical techniques and computational methods, make metabolomics a vital tool for drug research and development9, 10, 11. Basic drug research is the initial and fundamental step before drug development. Metabolomics helps advance the field of drug research by facilitating the understanding of the mechanisms of diseases, identification of drug targets, and elucidation of the mode of action of drugs. Furthermore, the drug development process can be costly and time-consuming, which typically consist of preclinical studies, clinical development, and post-market evaluation¹². Fortunately, metabolomics has the potential to accelerate the drug development process at several stages, such as evaluating the safety and efficacy of the drug and uncovering the toxicity and adverse events of the drug¹³.

Emerging evidence demonstrates the vital role of metabolomics in drug research and development. A case in point is the pharmaceutical research and development of Ivosidenib and Enasidenib, which target the mutated isocitrate dehydrogenase (IDH) and consequently inhibit the generation of d-2-hydroxyglutarate (D-2HG). D-2HG is an oncometabolite identified by metabolomics, which contributes to the disease processes of acute myeloid leukemia (AML) and gliomas14, 15, 16, 17, 18, 19, 20, 21 (Fig. 1). Glutamine metabolism has been recognized as a hallmark in cancer cell metabolism and a promising drug target in cancer therapy²². CB-839 (Telaglenastat) first displayed antitumor activity in both cell line and xenograft models in triple-negative breast cancer (TNBC) by inhibiting glutaminase, evidenced by reducing the levels of glutamate and its downstream metabolites that were identified by metabolomics²³. Thereafter, it entered several clinical trials and exhibited safety and efficacy in multiple tumors²⁴. Excellent reviews have summarized the applications of metabolomics in unveiling disease mechanisms and facilitating early drug discovery^11,13. More recently, lipidomics has been discussed as a separate discipline for its specific developments in techniques and applications for the drug research and development²⁵. Here, we present a brief overview of the recent advancement in metabolomics technologies and update the recent applications of metabolomics in drug research and development.

Figure 1.

Metabolomics drove the discovery and development of Ivosidenib and Enasidenib. IDH1 and IDH2 mutations were identified as potential drug targets for the treatment of gliomas and AML in 2009 and 2010, evidenced by elevated levels of D-2HG in tumors. Consequently, D-2HG was indicated as a diagnostic and prognostic biomarker in IDH-mutant gliomas and AML in 2012 and 2013. Eventually, Ivosidenib and Enasidenib were developed and approved by FDA to treat IDH-mutant AML in 2017 and 2018. Reprinted with permission from Ref. 26. Copyright @ 2023 John Wiley & Sons, Inc.

2. Metabolomics technologies

2.1. Mass spectrometry (MS)-based metabolomics

Metabolomics involves the measurement of the whole set of small-molecule metabolites, which are produced and consumed by metabolic enzymes and reactions in the biological system, such as cells, tissue, organ, and organism⁴. Typically, the metabolomics workflow contains the following steps: experimental design, sample collection, sample preparation, data acquisition, data analysis, and biological interpretation²⁶. Over the past decades, advanced technologies, mainly including mass spectrometry (MS) and nuclear magnetic resonance (NMR), greatly propelled the metabolite measurement^27,28. Although NMR spectroscopy is often valued for its ability to shed light on molecular structure, the relatively lower sensitivity limits its application in metabolomics. Owing to the high sensitivity, broad-coverage, and flexibility to be combined with diverse chromatographic methods, such as liquid chromatography (LC) and gas chromatography (GC), to improve the separation and quantification of metabolites, MS-based metabolomics has become the workhorse in metabolomics researches28, 29, 30. Chromatographic separation is generally used before MS and achieved by columns packed with diverse materials. The most commonly used LC columns in metabolomics are reversed-phase columns (RP) and hydrophilic interaction chromatography (HILIC) columns, which are suitable to separate nonpolar metabolites and water-soluble metabolites, respectively.

In MS-based metabolomics, high-resolution mass analyzers such as orbitrap and time of flight (TOF) are frequently adopted in non-targeted metabolomics for their excellent mass accuracy, while low-resolution mass analyzers such as single quadruple or triple quadruple are usually preferred in targeted metabolomics for their high sensitivity³¹. In addition, a comprehensive pseudotargeted metabolomics method was developed by using the combination of UHPLC/Q-TOF MS and UHPLC/QQQ MS^32,33. Metabolite annotation is a major bottleneck in nontargeted metabolomics. Novel strategies were developed to improve the accuracy and expand the coverage of metabolite annotation, such as metabolic reaction network (MRN)-based recursive algorithm (MetDNA)³⁴, ion mobility CCS atlas (AllCCS)³⁵, knowledge-guided multi-layer network (KGMN)³⁶. Meanwhile, unremitting efforts were made to broaden the coverage and improve the sensitivity in targeted metabolomics. For example, Hu group developed an LC‒MS-based targeted metabolomics platform with broad coverage and high sensitivity, which enabled simultaneous monitoring of more than 250 functional metabolites in a single run. By utilizing this effective platform, they unveiled several metabolic reprogramming events in cancer37, 38, 39, 40, virus-infected diseases41, 42, 43, 44, cardiovascular diseases⁴⁵, and other metabolic diseases^46,47.

In addition, metabolomics coupled with other MS-based multi-omics techniques such as proteomics and lipidomics enables a comprehensive understanding of disease processes at different molecular levels, thus offering more convincible biomarkers and therapeutic targets in drug discovery^8,48,49. For instance, Pang et al.⁴¹ revealed aberrant NAD⁺ metabolism in Zika virus-induced microcephaly, evidenced by altered levels of metabolites and proteins (metabolic enzymes) involved in the NAD⁺ salvage pathway that were identified by metabolomics and proteomics, respectively. Shen et al.⁵⁰ established a molecular classifier to classify severe COVID-19 patients based on their proteomics and metabolomics measurements with an overall accuracy of 93.5%.

2.2. Spatial metabolomics

Spatial metabolomics has been increasingly used in basic drug research for its unique ability in providing regional information on metabolites in cells and tissues, thus offering researchers spatially resolved metabolic profile information in situ⁵¹. Spatial metabolomics is greatly achieved by mass spectrometry imaging (MSI) technologies. A timely review by Hang et al.⁵² comprehensively introduced the high spatial resolution mass spectrometry technologies in MSI. Here, we mainly focused on three MS-based technologies that have been frequently adopted in MSI-based metabolomics, including secondary ion mass spectrometry (SIMS)⁵³, matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS)⁵⁴, and desorption electrospray ionization (DESI-MS)⁵⁵. SIMS is a mass analyzer that measures the secondary ions generated on the sample surface through the bombardment by the primary ion beams with high energy. SIMS offers the highest spatial resolution (nanometer scale), whereas, the high energy of primary ion beams may cause molecular fragmentation which complicated the data analysis⁵⁶. MALDI is an ionization technique that requires a matrix to help with the ionization and desorption of analytes by a laser. MALDI can achieve high spacial resolution with the range of 5–20 μm⁵². DESI is an ambient ionization technique where the electronically charged solvent is sprayed onto the sample surface and analysts are charged and desorbed from the sample surface and enter the mass analyzer. The resolution of DESI is typically 50–200 μm⁵⁷(Fig. 2A).

Figure 2.

Principles of main MSI technologies and typical workflow of MALDI. (A) Illustration of the ionization methods of three major ion sources used in MSI technologies: SIMS, MALDI, and DESI. (B) The typical workflow of MALDI-based MSI metabolomics. Take brain tissue as an example (Created by Biorender, https://biorender.com).

Among them, MALDI-MS was the most widely used MSI approach in accessing the different metabolic states between tissues and the metabolic heterogeneity between locations in a single tissue, or even a single cell. The typical workflow of MSI-based metabolomics usually contains sample preparation, MS detection, and imaging (Fig. 2B). Sample preparation is a crucial step and largely determined the reliability and reproducibility of the data in MALDI⁵⁸. The sample preparation in MALDI typically includes sample collection, tissue sectioning, and matrix coating. A tissue section with a thickness of 5–10 μm is recommended and can be achieved by a freezing microtome. Matrix implementation should be optimized before the experiment since it can interfere with the detection of metabolites with low mass to charge (m/z).

In addition, other new ionization methods, such as laser ablation electrospray ionization mass spectrometry (LAESI-MS)⁵⁹, matrix-assisted laser desorption electrospray ionization (MALDESI)⁶⁰, and MALDI-2⁶¹, have been proved to be powerful MSI tools in recent years. Moreover, the combined platform, such as air flow-assisted desorption electrospray ionization (AFADESI) and MALDI, can achieve both broad coverage and high spacial resolution of detected metabolites⁶².

2.3. Metabolic flux analysis

Metabolomics, giving valuable information on metabolite concentrations, however, is limited in understanding the metabolic pathway activity. Metabolic flux analysis (MFA) in contrast explores metabolic activities in a dynamic manner, which can illustrate whether the accumulation of a certain metabolite is caused by faster production or slower consumption⁶³. Thus, metabolomics and MFA jointly provide more comprehensive and complementary biological insights into the metabolic regulatory network in biological systems. Stable isotope tracing is the typical method used in MFA experiments, which explores the flux rates by measuring the isotopic enrichment ratio of downstream metabolites. Many tracers with variously labeled elements or atoms have been used in isotope tracing experiments for specific metabolic activity measurements, i.e., [1-¹³C]-glucose, [1-²H]-glucose, [3-²H]-glucose. Liang et al.⁶⁴ previously summarized the most commonly used isotope tracers in probing metabolic activities of diverse pathways. The isotopic enrichment ratio is usually interpreted by mass isotopomer distribution (MID), which is measured by MS^64,65. Considerable efforts have been made to simplify the MFA experiments, especially for data acquisition and analysis. For example, SWATH MS/MS, a data independent acquisition (DIA) technique, was introduced in MFA experiments, which allowed quantification of a large number of precursor and product MIDs in a single run and improves the precision of many measured fluxes when compared to those only precursor MIDs⁶⁶. PRM-MRM (parallel reaction monitoring‒multiple reaction monitoring) combination mode was proposed to detect various isotopologues and isotopomers of ¹³C-labeled nonessential amino acids (NEAAs) in cells incubated with ¹³C5-glutamine (Gln)⁶⁷. A novel comprehensive isotopic targeted mass spectrometry (CIT-MS) method was developed to improve the reliability and coverage of MFA⁶⁸. In addition, a global stable-isotope tracing metabolomics workflow–MetTracer–was developed to expand the coverage of the isotopically labeled metabolites⁶⁹. More recently, Meng et al.⁷⁰ developed a 3-nitrophenylhydrazine (3-NPH) derivatization strategy based on LC‒MS/MS, simultaneously targeting metabolites with carbonyl, carboxyl, and phosphoryl groups, which greatly facilitates the MFA for glycolysis, the tricarboxylic acid (TCA) cycle, and pentose phosphate pathway (PPP) in biological samples.

2.4. Single-cell metabolomics

Given the cell heterogeneity in bulk cells, the collective results obtained from conventional metabolomics cannot reflect the real metabolic state of individual cells⁷¹. Single-cell metabolomics analysis enables an understanding of the metabolic cell–cell heterogeneity in many biological processes, such as drug resistance, tumor metastasis, and cell fate. Limited sample size, low metabolite abundance, and fast turnover of metabolites in single cells raised high technical demanding on the sampling and detection methods in single-cell metabolomics. Versatile approaches for preparing single cells have been investigated, including manual aspiration, microdissection, liquid microjunction, and chip-based methods⁷². MS-based technologies are currently considered to be the most powerful approaches to obtaining the metabolic profiles of single cells. Recently, a collection of electrospray ionization (ESI)-MS-based strategies have been developed to improve the detection sensitivity and increase the number of metabolites detected in single cells, such as probe electrospray ionization (PESI)⁷³, pico-ESI⁷⁴, induced nano-electrospray ionization (InESI)⁷⁵. Notably, Zhu et al.⁷⁵ developed a single-lysosome mass spectrometry (SLMS) platform which allows to simultaneous explore the metabolic and electrophysiological states in a single enlarged lysosome by integrating the induced nano-electrospray ionization (nanoESI) based-MS and patch-clamp technologies. As a result, this technique may provide much higher resolution at the subcellular level, thus facilitating the depiction of metabolic fingerprints of the intracellular organelles, as well as their complicated interaction networks⁷⁶. Alternatively, imaging MS enables to capture of the metabolic profiles of single cells spatially and temporally and thus provides new dimensional insights into the hierarchical processes. MALDI-TOF-MS⁷⁷, TOF-secondary ion mass spectrometry (SIMS)⁷⁸, and vacuum ultraviolet photoionization (VUVDI)-MS⁷⁹ have been commonly used for the imaging analysis of single cells. In addition, a novel technique named SpaceM was recently proposed by Alexandrov and Heikenwalder's teams to assign metabolite intensities to individual cells by integrating MALDI-imaging with light microscopy⁸⁰.

3. Applications of metabolomics in drug research and development

3.1. Understanding disease mechanisms

Metabolic dysregulation plays a vital role in the pathogenesis and progression of several important chronic and complex diseases, such as cancer, obesity, diabetes, cardiovascular diseases, neurological diseases, infectious diseases, and others. Metabolic mechanisms under several pathological and physiological processes were unveiled by LC‒MS/MS-based targeted metabolomics. For instance, Pang et al.⁴¹ found reprogrammed NAD⁺ metabolism in Zika virus (ZIVK)-induced microcephaly, evidenced by decreased levels of NAD⁺ and increased levels of NAD⁺ precursors. Further integration with transcriptomics, proteomics, and phosphoproteomics revealed the potential regulatory mechanism of MAPK-NMNAT2-NAD⁺ in Zika virus (ZIVK)-induced microcephaly. A global metabolic landscape of healthy controls, mild and severe COVID-19 patients revealed perturbed levels of metabolites involved in arginine, tryptophan, and purine metabolism, which were tightly associated with the cytokine release induced by SARS-CoV-2, suggesting the metabolic mechanisms under cytokine release syndrome (CRS)⁴⁴. A comprehensive metabolomics profiling of two-cell and blastocyst-stage embryos demonstrated decreased L-2HG levels and increased αKG levels in blastocyst embryos, which contributed to the discovery and understanding of the histone methylation erasure mechanism during embryo development since αKG is the substrate of dioxygenases such as histone demethylases⁴⁶.

Particularly, recent advancements in metabolomics have profoundly improved our understanding of the tumor initiation, progression, metastasis, and drug resistance of lung cancer (Fig. 3). Tumor-initiating cells (TICs) are responsible for tumor initiation. LC‒MS-based non-targeted metabolomics, along with LC‒MS/MS based targeted metabolomics, revealed the enrichment of methionine, SAM, and S-adenosyl homocysteine (SAH) in patient-derived lung tumor-initiating cells. Following metabolite tracing analysis confirmed the highly elevated methionine cycle activity in TICs, which contributed to the tumor initiation via the upstream regulation mechanism driven by methionine adenosyltransferase II alpha (MAT2A)⁸¹. Cancer cell metabolism dysregulation underlies the development and growth of tumors. By using an LC‒MS/MS targeted metabolomics method, Huang et al.³⁸ identified two metabolically heterogeneous subtypes of small cell lung cancer (SCLC) cells with differential levels of purine nucleotides (i.e., IMP, GMP, XMP, and AMP) identified by metabolomics and nucleotide synthesis rates identified by metabolic flux analysis. These two metabolic phenotypes were regulated by inosine-5′-monophosphate dehydrogenase (IMPDH) and proved to be responsible for their different proliferation rates. Metastasis is the end result of the tumor evolutionary process. Succinate was found to be a differential metabolite between the control medium and conditioned medium of lung cancer cells (LLC-SCM) by LC‒MS-based metabolomics. Succinate then promoted tumor-associated macrophages (TAM) polarization and cancer metastasis in the tumor microenvironment via succinate receptor (SUCNR1)-triggered PI3K-hypoxia-inducible factor 1α (HIF-1α) axis⁸². In addition, cancer cell metastasis to the lung requires the metabolic adaptation of tumor cells to new environments. Breast cancer cells preferentially metastasized to the lung through upregulation of PCG-1α by enhancing global bioenergetic capacity, which has been proved by GC–MS based metabolomics and MFA with altered levels and enrichment ratios of metabolites in glycolysis and TCA cycle, e.g., pyruvate, lactate, citrate, and malate⁸³. Drug resistance in lung cancer therapy contributes to the fatal outcome of cancer, including chemoresistance, targeted drug resistance, and immunotherapy resistance. For instance, a comprehensive metabolomics platform containing LC‒MS, LC‒MS/MS, and GC–MS, uncovered that cisplatin-resistant cancer cells relied on glutamine metabolism for nucleotide biosynthesis, evidenced by rescued levels of succinyl adenosine, AMP, ADP and ATP in cisplatin-resistant cancer cells post glutamine treated⁸⁴. AXL overexpression enhanced drug-tolerant persister (DTP) survival and accelerated erlotinib and osimertinib resistance in EGFR-mutant lung cancer. LC‒MS-based metabolomics and MFA confirmed the upregulation of purine metabolism in AXL-KO cells with decreased levels of GMP and IMP and increased levels of their precursors, such as adenine, guanine, and adenosine, suggesting the downstream mechanism of purine metabolism in AXL mediated resistance⁸⁵. More recently, by using a large-scale LC‒MS/MS-based targeted metabolomics platform, Nie et al.³⁹ found the accumulated level of acetylcholine (ACh) in DTP cells, which triggered the discovery of a new mechanism under the osimertinib resistance in EGFR-mutant non-small cell lung cancer (NSCLC)–ACh/M3R/WNT axis. In addition, metabolomics deepened our understanding of the drug resistance to anti-PD-1 therapy. For instance, A CE‒MS-based metabolomics and MFA study revealed the decreased intracellular and extracellular levels of GABA and the abolished flux from ¹³C-labelled glutamine to ¹³C-GABA in shGAD1 cells, which confirmed the role of GAD1 in rewiring glutamine metabolism for the synthesis of γ-aminobutyric acid (GABA). GABA then promoted resistance to anti-PD-1 therapy via β-catenin activation⁸⁶.

Figure 3.

Metabolomics provides novel insights into the mechanism underlying lung cancer initiation, progression, metastasis, and drug resistance. In the top panel, metabolic processes and key regulators (shown in the brackets) involved in the mechanisms were indicated. In the bottom panel, metabolic processes and resistant drugs (shown in the brackets) involved in the mechanisms were indicated (Created by Biorender, https://biorender.com).

It is worth noting that metabolism is a complicated kinetic process during disease progression. The metabolomic alterations are variable during different stages of the disease, reflecting different functional or pathological changes in disease mechanisms or metabolomic pathways. For instance, by using a targeted metabolomics method, Nie et al.³⁷ depicted the metabolic evolutionary processes from atypical adenomatous hyperplasia (AAH) to adenocarcinoma in situ (AIS), minimally invasive adenocarcinoma (MIA), and invasive adenocarcinoma (IAC), and revealed several progressive and unique metabolic vulnerabilities from preneoplasia to invasive lung adenocarcinoma, e.g., nicotinate and nicotinamide metabolism, β-alanine metabolism, glutathione metabolism, arginine, and proline metabolism. Xiao et al.⁴⁴ characterized the metabolic profiles of healthy controls, mild, and severe COVID-19 patients and found several common metabolic changes in both mild and severe patients, e.g., primary bile acid biosynthesis, as well as some unique metabolic changes in severe patients, e.g., tryptophan metabolism. These consistent or variable metabolomic alterations during different stages of diseases offer researchers great opportunities in deciphering the complexity of disease mechanisms and development of precision treatment strategies.

3.2. Identifying drug targets

Metabolic vulnerability of diseases offers novel treatment strategies and promotes the discovery of drugs targeting specific metabolic enzymes.

Emerging evidence highlighted the crucial role of metabolomics in identifying drug targets against cancer therapy. The discovery and development of IDH-1/2 mutation inhibitors targeting the accumulation of D-2HG in AML and gliomas mark the significance of exploiting a metabolic vulnerability in cancer therapy. LC‒MS and GC‒MS-based metabolomics revealed the accumulation of D-2HG in gliomas and AML harboring IDH-1/2 mutation^14,15. D-2HG, an oncometabolite generated by mutant IDH-1/2 enzyme, contributes to the formation and malignant progression of gliomas and AML. Targeting the mutant IDH-1/2 enzyme, Ivosidenib and Enasidenib were developed and approved by FDA^19,21 (Fig. 1). Several previous investigations identified promising drug targets against cancer and infectious diseases. For instance, LC‒MS/MS-based targeted metabolomics revealed that a subgroup of SCLC cells (ASCL1^Low) possessed elevated levels of purine nucleotides, i.e., IMP, GMP, XMP, and AMP. Further MFA confirmed that the accumulated purine nucleotides were caused by high rates of de novo purine biosynthesis. Pharmacological inhibition of IMPDH, the rate-limiting enzyme in guanine nucleotide biosynthesis, selectively retarded the ASCL1^Low tumor growth, suggesting the therapeutic potential of IMPDH as a promising drug target³⁸. LC‒MS/MS based targeted metabolomics found that ACh was specifically accumulated in DTP cells in NSCLC. High levels of ACh promoted DTP formation via the M3R receptor. Pharmacologically targeting Ach metabolism with darifenacin, a potent M3R inhibitor retarded tumor relapse in vivo, thus offering a therapeutic target for overcoming EGFR TKI drug tolerance in the treatment of NSCLC³⁹. Cancer cells require upregulated glucose metabolism to meet their high proliferation demands. LC‒MS/MS based targeted metabolomics found that knockdown of 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase 2 (PFKFB2) increased the levels of fructose-2,6-bisphosphate, fructose-1,6-bisphosphate, and pyruvate in glycolysis in ovarian and breast cancer cells with wtTP53, and increased the levels of 6-phosphoglyconate and ribulose 5-phosphate in pentose phosphate pathway in cancer cells with mutant TP53. Silencing PFKFB2 inhibited tumor cell growth and enhanced paclitaxel sensitivity in ovarian and breast cancer cells with wtTP53. Thus, PFKFB2 can be exploited as a new drug target in the treatment of ovarian and breast cancers harboring wtTP53⁸⁷. Distinct metabolic signatures were found by LC‒MS based metabolomics, wherein tyrosine kinase inhibitor (TKI) resistant chronic myeloid leukemia (CML) cells were characterized by enriched glycolysis pathway and sensitive CML cells were characterized by enriched TCA cycle, suggesting a metabolic shift from mitochondrial metabolism-dependency towards glycolysis-dependency was underlies in TKI persistent leukemic stem cells (LSCs). Targeting the rate-limiting enzyme in glycolysis–pyruvate kinase M2 (PKM2)–specifically eliminated the TKI-persistent LSCs, suggesting PKM2 as a targetable metabolic vulnerability in TKI-persistent LSCs⁸⁸. Targeting glutamine metabolism is another promising anticancer therapy. Stable isotope-resolved metabolomics revealed reprogrammed glutamine metabolism in cancer cells and glutaminase (GLS) appeared to be a promising target for cancer therapy⁸⁹. Based on the findings, CB-839 (telaglenastat), a GLS inhibitor, has entered clinical trials⁹⁰. GC/TOF-MS-based metabolomics analysis revealed reprogrammed glutathione metabolism KRAS-mutant lung cancer cells, with elevated levels of cystine, glutamate, and glutathione. SLC7A11, a cystine/glutamate antiporter, is responsible for cystine uptake. HG106, a potent SLC7A11 inhibitor, decreased cystine uptake and intracellular glutathione biosynthesis in KRAS-mutant lung cancer cells and marked tumor suppression and prolonged survival mouse models. These findings suggested SLC7A11 as a therapeutic target for the treatment of KRAS-mutant lung cancer⁹¹.

Similarly, metabolic profiling uncovers promising metabolic targets for the treatment of infectious diseases, obesity, and diabetes. Pang et al.⁴¹ revealed NAD⁺ metabolic reprogramming in ZIKV-infected brains by a large-scale LC‒MS-based metabolomics method with a decreased level of NAD⁺ and increased levels of NAD⁺ precursors, such as tryptophan, kynurenine, NMN and NAM. Targeting NAD⁺ metabolism by supplementing mice with NR, an effective NAD⁺ precursor, alleviated the microcephaly induced by ZIKV infection in mice, suggesting that the NAD⁺ metabolism could be exploited as a therapeutic strategy against ZIKV-induced microcephaly. ¹H NMR-based metabolomics revealed an elevated level of 5′-AMP in mice with type 2 diabetes. Metformin lowered 5′-AMP-induced hyperglycaemia, offering a targetable potentiality of adenine nucleotides in type 2 diabetes⁹². Aberrant glucose and nucleotide metabolism were identified by GC‒MS-based metabolomics in host cells infected with chlamydia trachomatis, characterized by elevated levels of G6P, R5P, and GMP. Pharmacological inhibition of IMPDH decreased biomass synthesis that is used for cell proliferation and prevented chlamydia infection⁹³. Despite the great potential of exploiting metabolic enzymes as drug targets, it still needs to be cautious when identifying the metabolic vulnerability as therapeutic targets since the loss of metabolism homeostasis may either be a pathogenic causal or pathological consequence in specific diseases.

3.3. Elucidating the mode of action

Insufficient understanding of the drug mode of action (MoA) hinders the progress of drug research and development. Metabolomics studies at the early stage of drug development allow a systematic understanding of the drug mode of action by direct regulation of metabolic pathways or indirect regulation mediated by gut microbiota (Fig. 4).

Figure 4.

Drug mode of action uncovered by metabolomics. (A) Direct regulation of metabolic pathways. Drugs, perturbed metabolic pathways, and (or) key modulators were indicated. (B) Indirect regulation mediated by gut microbiota. Drugs and metabolites produced by gut microbiota that mediated the therapeutic effects were indicated.

Metabolites and metabolic pathways are reported to be involved in the mode of action of drugs. Metformin and phenformin depleted the TCA cycle and glycolytic-related metabolites during cell transformation, and nucleotide triphosphates in breast cancer stem cells (CSCs), leading to the impediment of nucleotide synthesis and cancer cell proliferation⁹⁴. NMR- and MS-based stable isotope-resolved metabolomics analyses revealed that surviving tumor cells were reliant on glycolysis and glycogen synthesis following treatment with glutaminase inhibition–bis-2-(5-phenylacetamido-1,2,4-thiadiazol-2-yl)ethyl sulfide (BPTES) nanoparticles, evidenced by elevated levels of labeled-lactate from ¹³C6-glucose and glycogen in residual tumors. Thus, Metformin was selected for combination therapy with BPTES and a better pancreatic tumor reduction effect was obtained, for its regulatory effects on glucose metabolism⁹⁵. Several anti-infectious drugs also exert metabolic-regulatory effects. LC‒MS/MS based targeted metabolomics uncovered that antibiotic treatment, i.e., ciprofloxacin, systemically altered purine metabolism in the host, and supplementation of metabolites retarded drug efficacy⁹⁶. K13 mutations altered mitochondrial metabolism in Plasmodium falciparum, which was identified by untargeted LC‒MS metabolomics with changed levels of metabolites involved in the TCA cycle, purine, glutamate, and pyruvate metabolism, and further drove artemisinin resistance. Atovaquone reversed the artemisinin resistance by modulating mitochondrial metabolism⁹⁷. LC‒MS-based metabolomics identified a strong metabolic linkage between glutamine and ATP in Mycobacterium tuberculosis (Mtb) that was treated with bedaquiline (BDQ). Inhibition of glutamine synthetase (GS) improved the efficacy of BDQ in killing Mtb, revealing GS as a secondary or downstream component of BDQ's bactericidal MoA⁹⁸.

Gut microbiota also plays a vital role in the mode of action of drugs that could elucidate by metabolomics. Ginseng polysaccharides (GPs) sensitized the antitumor response to αPD-1 monoclonal antibody (mAb) in non-small cell lungs by altering the microbial kynurenine/tryptophan ratio, which was identified by LC‒MS/MS-based metabolomics and contributed to the T-cell mediated immune regulatory effects⁹⁹. Untargeted metabolomics revealed that Ginseng extract (GE) has beneficial effects on body weight management by inducing Enterococci faecalis to produce myristoleic acid (MA), a metabolite that contributes to brown adipose tissue (BAT) activation and beige fat formation¹⁰⁰. GC‒MS-based non-targeted metabolomics uncovered that Rhein alleviated chronic colitis by decreasing uric acid levels, which was mechanically mediated by probiotic Lactobacillus¹⁰¹. Methotrexate (MTX) is the first-line rheumatoid arthritis (RA) treatment targeting mammalian dihydrofolate reductase (DHFR). LC‒MS non-targeted metabolomics identified purine metabolism as a conserved pathway in gut microbiotas post MTX treatment, suggesting that MTX acts on bacterial DHFR¹⁰².

3.4. Predicting drug response

Pharmacometabolomics was first proposed by Nicholson in 2006, which was defined as the prediction for drug responses based on pre-dose metabolic characteristics¹⁰³. The integration of metabolomics pharmacokinetics and pharmacodynamics provides an endogenous response to the drug, contributing to precision medicine¹⁰⁴ Integration of pharmacokinetic and pharmacometabolomic in early-phase clinical trials uncovered the possible association between the pharmacokinetics of Metformin and a series of metabolic pathways, including arginine and proline metabolism, BCAA metabolism, glutathione metabolism and others¹⁰⁵. Remoxipride is an antipsychotic drug targeting dopamine D2. An integrated pharmacometabolomics with pharmacokinetic/pharmacodynamic (PKPD) modeling highlighted that glycine, serine, and threonine pathway was associated with Remoxipride pharmacology¹⁰⁶. More recently, the microregional pharmacokinetics and pharmacodynamics of Olanzapine (OLZ), a first-line drug for the treatment of schizophrenia, were investigated by a temporospatial pharmacometabolomics method based on ambient mass spectrometry imaging. Ninety endogenous metabolites were identified to be associated with the microregional effects of the OLZ¹⁰⁷. Besides, metabolomics improves our understanding of drug bioavailability and drug clearance. For instance, Lorlatinib is a drug used for the treatment of brain metastasis from non-small cell lung cancer. LC‒MS-based metabolomics, along with multilayer perceptron analysis, identified that nine biomarkers were related to Lorlatinib concentration in the brain, suggesting the valuable role of metabolites for predicting the concentration of drugs in the brain¹⁰⁸. Metabolites may serve as an intermediate bridge between gut bacteria and drug metabolism, which contributes to drug availability by altering its bioaccumulation. Duloxetine is an antidepressant drug whose bioaccumulation is mechanistically unknown. Two complementary metabolomics platforms–flow-injection analysis mass spectrometry (FIA-MS) and HILIC‒MS/MS were performed to explore the metabolic mechanisms under the bioaccumulation of Duloxetine. Results showed that Duloxetine bound to abundant metabolic enzymes in gut bacteria and thus led to its intracellular storage⁹. Urine metabolomics based on GC–MS demonstrated that a panel of 28 endogenous metabolites, including a set of substrates of organic anion transporters (OATs), in patient urine at baseline could predict individual MTX clearance¹⁰⁹.

In addition, metabolites are compounds with small molecular weight which may serve as competitors that interact with the enzymes, in this regard, metabolomics may predict the drug response phenotypes. Gemcitabine is the first-line chemotherapy in pancreatic ductal adenocarcinoma (PDA). LC‒MS/MS based targeted metabolomics found that the resistance to gemcitabine was driven by deoxycytidine, a metabolite released by tumor-associated macrophages (TAMs), which inhibited gemcitabine activity through molecular competition¹¹⁰. Warburg effect (WE) is the hallmark of cancer cell metabolism. Comparative metabolomics based on LC‒MS depicted the metabolic profiles of HCT116 cells treated with several putative glyceraldehyde-3-phosphate dehydrogenase (GAPDH) inhibitors, an enzyme that controls metabolism during the WE, and nominated Koningic acid (KA) as a selective GAPDH inhibitor. The quantitative extent of the WE mediated by GAPDH could predict KA response¹¹¹. Metformin as a cancer therapy is undergoing late-phase clinical trials. Distinct mitochondrial metabolism was identified in primary breast tumors post Metformin treatment through LC‒MS-based metabolomics. Patients characterized by an increase in OXPHOS gene transcription with increases in proliferation were prone to resist to Metformin¹¹². In a phase 2 randomized placebo-controlled trial of patients with nonalcoholic steatohepatitis (NASH), the safety and efficacy of GS-0976 were evaluated. GS-0976 is an inhibitor of acetyl-coenzyme A, which catalyzes the rate-limiting step in de novo lipogenesis. LC‒MS/MS-based non-targeted metabolomics revealed decreased plasma levels of acylcarnitines in patients given GS-0976 20 mg, indicating the improved fatty acid catabolism and efficacy of GS-0976¹¹³.

Moreover, metabolomics helps us understand the toxicity of potential drug candidates, which is crucial to drug research and development. Perhaps, COMET (the Consortium for Metabonomic Toxicity) is the best-known example of the application of metabolomics to drug toxicity. Nearly 150 studies in 5 years clearly demonstrated that metabolomics is a powerful tool in probing the liver and kidney toxicity of drug candidates¹¹⁴. Recent metabolic findings updated our knowledge of drug toxicity. For instance, an overdose of acetaminophen (AP) may induce liver injury, which could be attenuated by vancomycin (Vac) pretreatment. GC‒MS-based non-targeted metabolomics identified that Vac pretreatment elevated cecum and serum 2-hydroxybutyric acid (2-HB). Mechanically, 2-HB protected AP-induced liver injury by reducing the bioavailability of AP¹¹⁵. Sunitinib, a tyrosine kinase (TK) inhibitor for the treatment of renal, gastrointestinal, and pancreatic cancers, carries a black box warning for hepatotoxicity. Liver metabolome enabled by LC‒MS identified 55 altered metabolites, including 35 acylcarnitines, suggesting impaired fatty acid β-oxidation (β-FAO) in hepatic mitochondria¹¹⁶. MSI-based in situ metabolomics unveiled spatially resolved metabolic profiles in rat kidneys after the administration of aristolochic acid, which is known as a nephrotoxic drug. Thirty-eight metabolites, involved in amino acid, nucleoside, and lipid metabolism, were significantly changed post aristolochic acid treatment and spatially matched with histopathological renal lesions, demonstrating the utility of metabolomics in the elucidation of drug toxicity⁵⁵. Based on a combined targeted and non-targeted metabolomics platform, two groups of metabolites were screened out from pre-dose serum, which could be used to predict the toxicity of Irinotecan (CPT-11)¹¹⁷.

3.5. Exploring drug repurposing

Drug repurposing expands our knowledge of treatment of diseases and bypasses common challenges in drug development. Metabolomics helps with drug repurposing in the field of oncology (Table 1). Metformin is the first-line drug for the treatment of type II diabetes. Since the first report on the anticancer effect in 2001, repurposing Metformin for other diseases is of significant interest for its desirable safety profile and promising clinical trials¹¹⁸. An LC‒MS-based metabolomics analysis highlighted the therapeutic potential of Metformin against ovarian cancer through modulating mitochondrial metabolisms, such as nucleotide metabolism, redox, and energy status. Stable isotope tracing analysis further unveiled that the response or resistance to Metformin depends on the availability of specific mitochondrial nutrients and the flexibility of substrate utilization in mitochondria¹¹⁹. Phosphoseryl-tRNA kinase (PSTK) mediated the resistance to Sorafenib in hepatocellular carcinoma (HCC) treatment by suppressing ferroptosis. Inhibition of PSTK with punicalin, an agent used to treat hepatitis B virus (HBV), synergized with Sorafenib in vitro and in vivo. Mechanistically, LC‒MS-based non-targeted metabolomics revealed that PSTK suppressed ferroptosis by maintaining GSH metabolism and folate biosynthesis, evidenced by altered levels of γ-glutamylcysteine, cysteinylglycine, and cysteine in the GSH synthesis and folic acid and 5-methyltetrahydrofolic acid in folate biosynthesis. These results suggested that repurposing punicalin may serve as a promising candidate for overcoming drug resistance in HCC¹²⁰.

Table 1.

Repurposing drugs for the treatment of cancer identified by metabolomics.

Repurposing drugs	Metabolomics platforms	Metabolic findings	Conventional use	Repurposing use	Ref.
Metformin	LC‒HRMS	Mitochondrial metabolism was affected post metformin administration	Type II diabetes	Ovarian cancer	119
Punicalin	LC‒MS/MS	PSTK regulated GSH metabolism and folate biosynthesis	Hepatitis B virus	Hepatocellular carcinoma (HCC)	120
Darifenacin	LC‒MS/MS	Acetylcholine (ACh) was accumulated in DTP cells	Overactive bladder	EGFR-mutant non-small cell lung cancer (NSCLC)	39
Mizoribine	LC‒MS/MS	High rate of de novo purine biosynthesis in ASCL1Low cell	Organ transplantation and autoimmune diseases	Small cell lung cancer (SCLC)	38
Sulfasalazine	GC/TOF‒MS	Increased intracellular cystine levels and glutathione biosynthesis.	Chronic inflammatory disease	KRAS-mutant lung adenocarcinoma (LUAD)	91

As noted above, metabolic enzymes are attractive drug targets. In parallel, many approved drugs targeting metabolic enzymes are repurposed as anti-tumor candidates. Darifenacin, an M3-selective receptor antagonist, is commonly used for the treatment of overactive bladder. A recent metabolomics study achieved by LC‒MS/MS found the accumulation of ACh levels in DTP cells and minimal residual disease (MRD) tumors in NSCLC. Muscarinic receptor 3 (M3R) is a key regulator of Ach metabolism. Pharmacological inhibition of M3R with darifenacin impeded tumor relapse in vivo through the ACh/M3R/WNT axis, suggesting the repurposing of Darifenacin for adjuvant therapy in the treatment of NSCLC³⁹. Mizoribine is a drug used as an immunosuppressant in organ transplantation and autoimmune diseases. By using metabolomics and metabolic flux analysis, Huang et al.³⁸ identified a subset of SCLC characterized by accumulated levels of metabolites involved in purine metabolism and a high rate of de novo purine biosynthesis. Inhibition of IMPDH, the rate-limiting enzyme in purine biosynthesis, by Mizoribine impeded ASCL1Low tumor growth, emphasizing the significance of repurposing Mizoribine in SCLC. GC‒MS-based metabolomics found strikingly increased intracellular cystine levels and glutathione biosynthesis KRAS-mutant lung adenocarcinoma (LUAD). SLC7A11, a cystine/glutamate antiporter, is responsible for cystine uptake. Pharmacological inhibition SLC7A11 with Sulfasalazine, an FDA-approved drug for the treatment of chronic inflammatory disease, killed KRAS-mutant cancer cells in vitro and retarded tumor growth in vivo, implying the therapeutic strategy of repurposing Sulfasalazine in KRAS-mutant LUAD⁹¹.

3.6. Investigating drug‒drug interactions

Extremely attractive advantages are recognized for the combined use of drugs, including improved efficacy, decreased toxicity, and reduced risk of drug resistance, while the blocked or reduced effectiveness caused by drug antagonism is nonnegligible. Thus, it is necessary to fully understand drug‒drug interactions. A high-throughput metabolomics platform based on FIA-TOFMS was used to investigate the metabolome profiles post antimicrobial drug exposure and predict drug–drug interactions, unveiling unprecedented insights into drug repurposing¹²¹. A stable isotope-tracer direct-infusion mass spectrometry (SIT-DIMS) based metabolomics drug screening platform was developed to evaluate drug combinations. A novel drug combination with synergic effects between CB-839 and docetaxel was discovered to treat prostate cancer¹²². The metabolomics-based phenotypic screening revealed novel synergy between the oxidative phosphorylation inhibitor IACS-010759 and the FMS-like tyrosine kinase 3 (FLT3) inhibitor AC220 (Quizartinib) in the treatment of AML¹²³. A phase 1 clinical trial of sirolimus and hydroxychloroquine for the treatment of lymphangioleiomyomatosis (LAM) was conducted using LC‒MS-based metabolomics. Plasma metabolomics revealed altered polyamine metabolic pathways upon treatment with this drug combination in patients with LAM¹²⁴. LC‒MS/MS-based metabolomics revealed the synergic effects of β-sitosterol and Fluoxetine on anxiety reduction, evidenced by elevated levels of 5-HT and amino acids in brain regions¹²⁵.

3.7. Enabling personalized treatment

Metabolomics-based patient stratification provides a novel and complementary perspective of patient information and propels personalized treatment or precision medicine based on their metabolic profiles.

Compelling findings arising from metabolomics highlighted the role of metabolomics in patient stratification. For instance, by employing LC‒MS/MS-based targeted metabolomics, Nie et al.³⁷ defined three metabolic subtypes of IAC patients with distinct clinical features, including S-I with the lowest mutation frequency of EGFR, S-II with the highest mutation frequency of EGFR, and S-III with unique BRAF mutation. Importantly, different clinical outcomes were observed among patients in different subtypes, wherein patients in S-II and S-III had a relatively poor prognosis than those in S-I, suggesting the prognostic value of metabolomics clustering. In another study, Wang et al.⁴⁵ stratified patients with hypertrophic cardiomyopathy (HCM) into three metabolic subtypings based on their metabolic profiles monitored by LC‒MS/MS based targeted metabolomics. Patients with distinct glutathione metabolism, TCA cycle, and purine and pyrimidine metabolism were observed to be associated with distinct clinical outcomes. In addition, a recent study classified TNBC patients into three metabolomics subgroups based on their metabolic profiles provided by LC‒MS-based non-targeted metabolomics, including C1 with upregulated ceramides and fatty acids, C2 with upregulated oxidation reaction and carbohydrate metabolism, and C3 with mild metabolic dysregulation compared with normal tissues¹²⁶. These metabolomics-based patient stratification allows the development of personalized treatment strategies and contributes to the achievement of precision medicine.

In summary, metabolomics greatly facilitates drug research and development spanning basic biology to the pharmaceutical industry (Fig. 5).

Figure 5.

Metabolomics techniques and applications in drug research and development.

4. Conclusions and outlook

Over the past few years, great advancement has been made in the development of techniques and technologies used in metabolomics, as well as their extensive applications in drug research and development. However, there are still many multifaceted challenges to be addressed.

Regarding the analytical techniques and technologies, unremitting efforts are still needed to improve the sensitivity of metabolite detection, reliability of metabolite annotation, and accuracy of metabolite quantitation²⁸. The point of achieving high sensitivity in MS is to improve ionization efficiency, ion transmission efficiency, and ion detection efficiency. In fact, the ionization efficacies of many methods are relatively low (10⁻⁵−10⁻³)¹²⁷. Thus, seeking ways to improve ionization efficacy should be the top priority, such as modification of ion source. Metabolite annotation is one of the major obstacles in the development of non-targeted metabolomics. Cross-validation between different platforms, accompanied by database and literature data, can improve the reliability of metabolite annotation. Accuracy of metabolite quantitation is of great challenge in MSI-based metabolomics, which would benefit from the standardized and unified sample preparation procedures, reasonable data normalization strategies, and involvement of internal standards. Improving the standardization from sample collection to data analysis is another technical challenge in metabolomics and is crucial to obtain comparable and reliable metabolic results from different laboratories and different experiments¹²⁸. In addition, single-cell metabolomics techniques are still far from being established due to their limited metabolite coverage and sophisticated pretreatment methods¹²⁹.

In parallel with intracellular metabolomics which has been extensively discussed in this paper, extracellular metabolomics, such as culture medium and extracellular vesicles (EVs), is another attractive field, which is particularly vital for the understanding of intracellular metabolism and cell–cell communications. Particularly, EVs metabolomics has emerged as a promising tool for disease mechanism elucidation and disease biomarker identification in recent years^130,131. However, whilst the value of extracellular metabolomics is clear, there are challenges inherent to sample preparation where targeted metabolites often appear as low concentrations in cell medium or EVs. With the aid of advanced metabolomics technologies and efficient isolation and enrichment preparation methods, extracellular metabolomics will be more increasingly used in drug research and development.

Integration with other omics, such as genomics, transcriptomics, proteomics, lipidomics, and microbiome, can offer a comprehensive understanding of disease mechanisms and MoA of drugs, as well as other orthogonal evidence for metabolic findings during drug research and development. Over the past years, multi-omics studies have achieved great advancement in basic drug research, such as drug target discovery and drug MoA elucidation. With the aid of advanced mass spectrometry technologies and optimized integration methods, multi-omics will be increasingly important during drug research and development in the future. Further, genetically encoded fluorescent biosensors are capable of monitoring biological processes in live cells with high temporal and spatial resolution¹³². Several metabolite-based biosensors were developed and shown to be valuable tools for real-time monitoring of metabolic dynamics in live cells, e.g., NADH sensor for monitoring NADH levels¹³³, SoNar for tracking cytosolic NAD (+) and NADH redox states¹³⁴, iNap for the visualization of NADPH¹³⁵, FiLa for the monitoring lactate fluctuations¹³⁶. Multi-omics Integrated with such biosensors will provide unprecedented insights into disease mechanisms and offer valuable opportunities for drug research and development in the future.

Application-wise, promising biomarkers and drug targets identified by metabolomics will need to be further validated and used in clinical practice. It is worth noting that due to the complexity of disease pathogenesis and metabolism, different diseases may share similar metabolic phenotypes, e.g., metabolites and metabolic enzymes. Care must be taken when exploiting treatment strategies that targeted similar metabolic vulnerabilities shared in different diseases to avoid potential side effects. Moreover, to accelerate the process of drug research and development, metabolomics researchers will need to deliberate on how to convert a discovery into a drug and work more closely with clinicians and drug developers.

Acknowledgments

We appreciate many other excellent publications that have contributed to the metabolomics field and its applications in drug research and development. Due to the space limitation, we regret that we were unable to include all the relevant works in this review. Zeping Hu is supported by Tsinghua University Spring Breeze Fund (2021Z99CFY031) and National Natural Science Foundation of China (32150024 and 92057209). Huanhuan Pang is supported by the Fundamental Research Funds for the Central Public Welfare Research Institutes (ZZ16-YQ-046 and ZZ16-ND-10-13, China).

Author contributions

Huanhuan Pang and Zeping Hu contributed to the writing and editing of this review article.

Conflicts of interest

The authors declare no competing interest.