Metabolome guided treasure hunt - learning from metabolic diversity

Metabolome guided treasure hunt - learning from metabolic diversity Esra Karakas, Mustafa Bulut, Alisdair Fernie * Max Planck Institute of Molecular Plant Physiology, Am Muhlenberg 1, Golm, 14476, Potsdam, Germany ABSTRACT Metabolomics is a rapidly evolving field focused on the comprehensive identification and quantification of small molecules in biological systems. As the final layer of the biological hierarchy following of the genome, transcriptome and proteome, it presents a dynamic snapshot of phenotype, influenced by genetic, environmental and physiological factors. Whilst the metabolome sits downstream of genes and proteins, there are multiple higher levels—tissues, organs, the entire organism, and interactions with other organisms, which need to be considered in order to fully comprehend organismal biology. Advances in metabolomics continue to expand its applications in plant biology, biotechnology, and natural product discovery unlocking many of nature’s most beneficial colors, tastes, nutrients and medicines. Flavonoids and other specialized metabolites are essential for plant defense against oxidative stress and function as key phytonutrients for human health. Recent advancements in gene-editing and metabolic engineering have significantly improved the nutritional value and flavor of crop plants. Here we highlight how advanced metabolic analysis is driving improvements in crops uncovering genes that influence nutrient and flavor profile and plant derived compounds with medicinal potential. 1. Introduction Metabolomics is a rapidly advancing field focused on the comprehensive identification and quantification of all endogenous and exogenous low-molecular-weight (<1 kDa) small molecules, or metabolites, within a biological system using high throughput techniques. The composition of these endogenous compounds is shaped by upstream factors like proteome and genome, along with environmental influences, lifestyle habits, medications and existing diseases. Metabolomics is described as a reflection of the phenotype, with the metabolome positioned downstream of the transcriptome and proteome, serving as a complementary counterpart to genomics, transcriptomics, and proteomics. Plants are regarded as nature’s most profilic biochemists (Hall, 2006), representing vast, natural compound libraries that remain largely unexplored. Most plant species have not been fully characterized biochemically, and the majority of detected plant compounds have yet to be structurally identified (Fiehn, 2002). This immense natural resource holds tremendous potential for future endeavours, both in scientific research-enhancing our holistic understanding of plant metabolism and its regulationand in an applied context, such as the discovery of novel drugs, flavors, biocides and other valuable compounds (Fernie, 2007; Saito and Matsuda, 2010). Metabolomics has provided plant biologists with the opportunity to explore cell biochemistry in-depth, rapidly establishing itself as a fundamental tool in research (Fiehn et al., 2001). Initial efforts focused on a limited number of “model” species, such as Arabidopsis, tomato and potato. However, these approaches have been extended to a wide variety of species, both cultivated and wild, to address diverse biological questions of significant scientific and industrial relevance. Metabolomics approaches have been extensively utilized to deepen our understanding of the diverse factors influencing plant metabolism. These include the effects of mutations (Bino et al., 2005; Yonekura-Sakakibara et al., 2008), environmental perturbations (Allwood et al., 2010; Choi et al., 2006; Jansen et al., 2008; Kaplan et al., 2004; van Dam and van der Meijden, 2011; Ward et al., 2010) genetic introgression (Fernie and Schauer, 2009; Keurentjes et al., 2006; Schauer et al., 2006) and organ development (Tikunov et al., 2010). Metabolomics has now been carried out for over a quarter-of-acentury, however, its foundations are much older. Metabolic studies have arguably been conducted since the 13th century with Ibn al Nafis’s observation that “the body and its parts are in continuous state of dissolution and nourishment, so that are inevitably undergoing permanent change” (Fernie and Pichersky, 2015; Mahdi, 1974) being remarkably prescient. However, studies in yeast (Buchner, 1897; Kühne, 1877; Sumner, 1968), at the turn of the 20th century that discovered enzymes and thereby mechanistically defined metabolism. In the interim many key findings have contributed to our understanding of the hundreds of thousand metabolites of the Plant Kingdom with the core of chemical reactions comprising primary metabolism being relatively well characterized. This knowledge was founded on the basis of isotope This article is part of a special issue entitled: Challenges&Breakthroughs published in Journal of Plant Physiology. * Corresponding author. E-mail address: [email protected] (A. Fernie). Contents lists available at ScienceDirect Journal of Plant Physiology journal homepage: www.elsevier.com/locate/jplph https://doi.org/10.1016/j.jplph.2025.154494 Received 10 February 2025; Received in revised form 10 April 2025; Accepted 13 April 2025 Journal of Plant Physiology 309 (2025) 154494 Available online 16 April 2025 0176-1617/© 2025 The Authors. Published by Elsevier GmbH. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ). tracer studies and more recently the study of mutant and transgenic plants (Fernie and Pichersky, 2015), but a quantum leap in our ability to study metabolism came on the development of transcriptomics and genome sequencing, swiftly followed by proteomics and metabolomics (Gallardo et al., 2001; Girke et al., 2000; Roessner et al., 2001; Zhu and Wang, 2000). These technologies confirmed a number of long-standing hypothesis as well as providing new insights into plant function. In this review we start by providing a brief technical overview of metabolomics before defining attempts to utilize metabolomics in crop improvement and its role in metabolic engineering and finally describe how it is currently being utilized in natural product discovery. 2. A brief overview of the development and current status of metabolomics Multiple techniques, such as gas chromatography-mass spectrometry (GC-MS) (Godber and Parsons, 1998; Katona et al., 1999), liquid chromatography-mass spectrometry (LC-MS) (Allwood et al., 2010), Fig. 1. The overview of the sample preparation process for the metabolomics and data analysis for the plant biological samples. E. Karakas et al. Journal of Plant Physiology 309 (2025) 154494 2 capillary electrophoresis-mass spectrometry (CE-MS), fourier-trans form-ion-cyclotron-resonance mass spectrometry (FT-ICR-MS) and nuclear magnetic resonance spectroscopy (NMR) (Nicholson et al., 2002; Ward et al., 2003), are frequently employed in plant metabolomics research (Fig. 1). These methods are often used in combination because they are largely complementary, with each technique offering preferential coverage for different types of metabolites. Mass spectrometry is frequently paired with chromatographic separation techniques such as GC and high-performance liquid chromatography (HPLC). GC coupled with MS, particularly electron impact mass spectrometry, offers high sensitivity, excellent resolution, and consistent molecular fragmentation patterns. Commercial databases further support metabolite identification, enhancing the analytical process (Kopka, 2006; Looser et al., 2005; Santos and Galceran, 2003). GC-MS was initially the most commonly employed method in plant metabolomics research (Fiehn et al., 2000; Roessner et al., 2001). Polar metabolites are derivatized to increase their volatility and subsequently separated by GC. Electron impact (EI) enables a stable interface between GC and MS, producing highly reproducible fragmentation patterns. Time-of-flight (TOF)-MS has become the preferred method for detection due to its common advantages, including rapid scan times that enable improved deconvolution or shorter run times for complex mixtures, as well as sometimes offering high mass accuracy (Lu et al., 2017; Taylor et al., 2002). A key advantage of this technology lies in its long-standing use for metabolite profiling, resulting in well-established protocols for machine setup, maintenance, chromatogram evaluation, and interpretation (Lisec et al., 2006). There are pros and cons of using high-resolution mass spectrometry the advantages are that chromatographic resolution is better and thus chances of structure elucidation are also higher, however, this comes at the cost of increased price of machinery as well as much higher complexity of data analysis resulting in greater computational times. That said a major advantage of the use of high-resolution machines is that it renders the use of databases for annotation far more precise (Lei et al., 2011; Putri et al., 2022; Rey-- Stolle et al., 2022). Volatile organic compounds (VOCs), released by plant organs like leaves, flowers and fruits, play diverse biological roles, including defense against pathogens, parasites and herbivores, as well as facilitating interactions with pollinators (Holopainen and Gershenzon, 2010). Solid-phase microextraction coupled (SPME) with GC-MS is a highly sensitive, reproducible and robust analytical method, making it well-suited for metabolomics studies of volatile compounds (Augusto et al., 2000). LC is the most widely used separation technique for coupling with MS in metabolomics research, surpassing GC and CE in popularity. While GC is limited by the need for compound volatilization, LC offers an advantage by eliminating the need for prior sample treatment and separating components in a liquid phase. The selection of columns, such as reversed-phase, ion-exchange and hydrophobic interaction columns, allows for the separation of metabolites based on their distinct chemical properties. The development of ultra-performance liquid chromatography (UPLC) has significantly improved the technique, providing superior resolution, sensitivity and throughput compared to conventional high-performance liquid chromatography (HPLC) (Rogachev and Aharoni, 2012). In LC-MS, various ionization sources such as electrospray ionization (ESI), atmospheric pressure chemical ionization (APCI) and atmospheric pressure photo-ionization (APPI) can be employed. Various types of MS, such as quadrupole (Q), TOF, qTOF, triple quadrupole (QpQ), ion trap (IT) and linear trap quadrupole (LTQ) orbitrap are utilized based on the required sensitivity, mass resolution and dynamic range. On the other hand, Fourier-transform-ion-cyclotronresonance mass spectrometry (FT-ICR-MS) relies exclusively on extremely high-resolution mass analysis, offering the potential to determine the empirical formulas of thousands of metabolites. However, its capability is somewhat constrained by the absence of chromatographic separation (Roessner et al., 2001). Moreover, ion chromatography (with suppression technology) is a powerful approach for analyzing highly polar and charged metabolites (Allwood and Goodacre, 2010; Glauser et al., 2013; Vuckovic, 2012). Capillary electrophoresis (CE) separates polar and charged compounds based on their charge-to-mass ratio. CE is primarily coupled with mass spectrometry using ESI through a sheath-liquid interface. However, achieving adequate sensitivity and robustness in CE-ESI-MS interfacing remains a significant challenge. ESI is widely used for ionization in LC-MS, while TOF-MS is the most commonly employed detector in CE-MS-based metabolomics studies. This combination offers excellent mass accuracy and high resolution (Soga et al., 2003). That said, it is very time-intensive requiring considerably longer runs than LCor GC-coupled MS and also requires highly trained individuals to run it. As such, as opposed to GC-MS and LC-MS, CE-MS is not commonly utilized for high throughput metabolomics (Obata and Fernie, 2012). NMR provides a fundamentally different analytical approach compared to MS-based techniques, as it relies on the magnetic properties of atomic nuclei and their interactions with an external magnetic field, rather than ionization and mass-to-charge analysis. NMR is complementary to MS in that it handles abundant metabolites well, is highly reproducible, and excels at structural elucidation (Markley et al., 2017; Yin et al., 2025). Indeed, Table 1 summarizes the capabilities of the four major techniques alongside new approaches such as MALDI imaging and ambient ionization methods. NMR stands out for its exceptional reproducibility, suitability for high-throughput analysis, cost-effectiveness, and unparalleled capability in quantifying and identifying the structures of unknown compounds. In comparison to GC-MS and LC-MS, NMR offers complementary advantages, especially for studying abundant metabolites that are hard to ionize, require derivatization, or are present in very high concentrations (Kruger et al., 2008). Despite that, MS offers broader metabolite coverage and higher throughput potential in terms of metabolites per minute, thanks to its superior sensitivity. This advantage explains the greater focus on developing novel MS technologies compared to NMR (Fernie et al., 2004). Having defined the major technical approaches used in metabolomics we will dedicate the rest of this review to the insights derived from their application. 3. Using metabolomics as an aid in crop improvement In recent years, substantial advancements have been made across various “omics” disciplines, including transcriptomics, genomics, metabolomics, proteomics, phenomics and epigenomics (Alseekh et al., 2023; Shen et al., 2023; Yang et al., 2012). These cutting-edge “omics” approaches have significantly improved data precision and expedited selection processes in breeding programs, playing a pivotal role in ensuring global food security (Kumar et al., 2017). Research has particularly emphasized the vital role of genomics and phenomics in producing genetically enhanced seeds and boosting crop performance through advanced breeding strategies (Langridge and Fleury, 2011; Wang et al., 2022). Despite this, the potential of “omics” in selecting key traits and improving plant performance through innovative strategies remains immense. Among these approaches, metabolomics stands out as a versatile and promising field but has received comparatively less attention in plant selection and trait mapping, particularly within crop species. Metabolomics, when combined with the functional characterization of genes, enables the investigation of how specific candidate genes influence metabolic pathways and uncovers regulatory relationships among interconnected pathways-insights that are often challenging to achieve through traditional methods such as transcriptome analysis (Fernie and Stitt, 2012). Integrated omics approaches empower researchers to identify potential traits for enhancing crop yield, quality and nutritional value by allowing the identification of previously unknown and unexpected associations such as those identifying metabolic factors underlying plant growth and yield described above. Identification of such associations is, however, not as facile as it sounds. That said E. Karakas et al. Journal of Plant Physiology 309 (2025) 154494 3 the development of machine learning based approaches is beginning to render them considerably easier to delineate (for reviews of the impact of machine learning on metabolomics see (Aksenov et al., 2021; Greener et al., 2022; van Dijk et al., 2021; Xu and Jackson, 2019). Additionally, omics studies can be extended to explore regulatory mechanisms at the levels of post-transcriptional and post-translational modifications, as well as epigenetic regulation. In this context, interactome studies, which examine molecular interactions among biomolecules, may further deepen our understanding of genotype-phenotype relationships (Anguraj Vadivel, 2015; Vidal et al., 2011). Metabolomics approaches have been employed across various crop species, regardless of the availability of successful transgenic systems, demonstrating significant potential for identifying key traits (Sim´ o et al., 2014). The emergence of integrated metabolomics, coupled with advancements in whole-genome accessibility, genotyping assays and identification of genetic variants, represents a major breakthrough in effectively incorporating metabolomics into crop breeding systems (Zivy et al., 2015). Over the past decades, metabolomics has been widely suggested for use in breeding programs for various crop species, owing to its potential for selecting superior traits. The integration of modern metabolomics with next-generation sequencing, whole-genome sequencing, mGWAS and mQTL offers an efficient and powerful approach to advancing sustainable agriculture (Luo, 2015). Moreover, a vast number of plant genomes have been sequenced and integrating metabolomics with next-generation sequencing has enabled comparative genomic analyses in multiple species, uncovering hundreds of genes and pathways. One factor limited the uptake of metabolomics is its relatively high cost, however, recent studies using machine learning alongside metabolomics would likely provide a route by which this could be massively reduced thus greatly enhancing its potential (Colantonio et al., 2022; Fernie and Alseekh, 2022; Tiozon et al., 2023). Crop breeding primarily involves the selection and development of crop varieties with desirable traits, achieved through either traditional breeding methods or advanced genetic technologies. By combining genetic approaches with metabolomics, breeders and researchers can address crucial agronomic challenges, evaluating crop performance across diverse environmental conditions (Langridge and Fleury, 2011). Metabolomics could lead to the development of more systematic models that connect genotype-phenotype or metabolic pathways to yield, specific metabolites, or traits associated with quality (Carreno-Quintero et al., 2013). The application of targeted and untargeted metabolomics facilitates spatial-temporal metabolic profiling in developing plants. These profiles help identify key biomarker metabolites that are crucial for breeding crops with preferred genetic and developmental traits. For example, metabolic approaches have been used to investigate rice tillering, uncovering 83 % metabolic variation (Tarpley et al., 2005). In soybeans, phytochemical metabolomics profiling identified flavonoid kaempferol glycosides associated with the transition from vegetative to reproductive phases (Park et al., 2023). Targeted metabolite profiling demonstrated the interaction between genetic and environmental factors in a group of corn hybrids, as well as the impact of water stress on metabolite levels (Harrigan et al., 2007). Integrated transcript and metabolite profiling identified QTL associated with spider-mite-induced volatile biosynthesis in cucumber (Mercke et al., 2004). Furthermore, extensive metabolite profiling of a tomato introgression line library facilitated the identification of over 880 mQTL and their inheritance patterns (Schauer et al., 2006, 2008). Another key area in which metabolomics has received much recent attention is the use of metabolite levels as predictors for yield (Hajheidari et al., 2022; Meyer et al., 2007; Shi et al., 2020; Sonnewald et al., 2020; Vergara-Diaz et al., 2020). An early study in this regard was the identification of a metabolic signature for biomass that was identified in a population of Arabidodpsis (Meyer et al., 2007). This study identified that one analysis of the small molecular primary metabolite complement of the cell was more effective a predictor of final biomass than any single metabolite. Although it must be said that later studies revealed that assessment of the macromolecules starch and protein provided reasonable assessments of growth (Ishihara et al., 2017; Sulpice et al., 2009). Subsequently, similar approaches have been used in crop species with metabolomics on its own being used to evaluate early onset yield heterosis in young maize plants (Lisec et al., 2009; Meyer et al., 2010; Riedelsheimer et al., 2012) and yield on the basis of metabolomics of spike bracts of wheat (Vergara-Diaz et al., 2020) as well as prediction of yield of maize using a combination of metabolomics, ionomics and transcriptomics data (Hajheidari et al., 2022). Metabolomic data is also used as input into genome scale models where it serves as a constraint which improves the accuracy of its outcomes (Benes et al., 2020; Nikoloski et al., 2015). Alternatively, simpler hints in this direction have been elucidated by comparing metabolite and yield associated traits in mapping populations of many crops including maize, rice and wheat as well as potato, tomato and cassava by means of various types of network analysis (Carreno-Quintero et al., 2013; Chen et al., 2016; Ding et al., 2023; Schauer et al., 2006; Shi et al., 2020). Whilst these studies look promising their adoption in agriculture have Table 1 Comparisons of analytical techniques. Analytical techniques Principle Advantages Limitations GC-MS Analyze both volatile and nonvolatile compounds following derivatization. High sensitivity, selectivity, fast scan speeds as well as high mass resolution. Enables to determine simultaneously a great number of polar metabolites. Limited to volatile, thermally stable and energetically stable compounds. LC-MS Separating compounds in the liquid phase does not necessitate the prior derivatization. High sensitivity and molecular specificity. Enables to detect a huge range of semi-polar compounds, including many key groups of secondary metabolites. Matrix effects, ion suppression, complex sample preparation. CE-MS Separates analytes based on charge and size, making it ideal for analyzing highly polar and ionic metabolites. Requires minimal sample amounts. Enable to detection for ionic and small molecules. Low sensitivity, poor reproducibility and electrochemical reactions of metabolites. NMR Uses magnetic fields to determine molecular structure Provides detailed structural and quantitative information. High for structure elucidation but low for complex mixtures. Highly quantitative and reproducible. Requires large sample quantity, expensive instrumentation Low sensitivity and signal overlap compared to MS. Ambient Ionization methods Ionizes analytes directly from a sample surface under ambient conditions. It provides rapid analysis. Suitable for in situ or real-time analysis. Limited sensitivity compared to LC-MS. Requires optimization for complex mixtures. MALDI Imaging Uses a laser to desorb and ionize molecules from a sample surface for spatially resolved mass spectrometry analysis. Allows direct analysis of tissue and surfaces. Provides spatial distribution of molecules. Limited quantification capabilities. Requires optimization of matrix application. E. Karakas et al. Journal of Plant Physiology 309 (2025) 154494 4 been somewhat restructured by the high costs of chemotyping broad populations. That said, a recent approach that married metabolomics to machine learning revealed that this allowed highly accurate predictions of flavour in tomato and blueberry when only a subset of samples was subject to metabolite profiling (Colantonio et al., 2022). To clarify machine learning reduces costs by machine streamlining data interpretation as well as allowing partial metabolite profiling to reliably predict broader metabolic traits thereby lowering the number of samples requiring full, expensive analyses. As we have previously argued, this bodes well for the uptake of metabolomics not just for such traits, but generally in plant breeding (Fernie and Alseekh, 2022) (Fig. 2). Having detailed the use of metabolomics to define non-metabolic traits in the above paragraphs the next section of the review defines its use for metabolic engineering per se. 4. Identification of genes controlling metabolite content Plants have provided humans with essential metabolites for food and nutrition, biomaterials for daily living, and remedies for pain and disease. Plants generate a wide range of metabolites, showcasing significant diversity across both population and individual levels. Growing research attention has been drawn to uncovering the genetic foundations of metabolic diversity. Advancements in mass spectrometry-based analytical systems and genome sequencing technologies have enabled the extension of metabolic quantitative trait loci (mQTL) and genome-wide association studies (GWAS) to explore the diversity of plant metabolomes using ultra-high-density maps. Notably, these approaches allow researchers to link genetic variation to metabolic profiles across diverse genotypes and environments. In these cases, metabolomics primarily guides the identification of natural allelic variants useful for breeding. One of the most powerful tools for globally identifying the genetic determinants of plant metabolic diversity is the metabolic genome-wide association study (mGWAS). Recently, mGWAS has been conducted across various species with continuous advances offering deeper insights into the genetic foundations of metabolic diversity (see below). To gain a better understanding of the genetic and biochemical foundations of metabolomic diversity through synthetic biology, researchers have utilized linkage mapping based on next-generation sequencing in metabolomics studies (Luo, 2015). A notable aspect of the genetic architecture of metabolism is the presence of hotspots containing major genes or genome regions that influence the natural variation of many metabolites (Chan et al., 2010; Knoch et al., 2024). These regions often harbor key biosynthetic or regulatory genes with broad influence across multiple metabolic pathways. The impact of interactions among genotypes, environment and development on metabolite accumulation has been well documented in structured mapping populations, such as recombinant inbred lines (RILs) and diversity panels, under both controlled and field conditions. For example, a study using mature kernels from 145 wheat RILs mapped 1005 mQTLs and identified 24 candidate genes associated with metabolite variation and agronomic traits such as grain number and plant height (Shi et al., 2020). Similarly, a GWAS in 163 tomato accessions identified 44 loci linked to 19 key fruit metabolites, showcasing GWAS as an effective tool for mapping complex traits and guiding crop improvement in tomato, highlighting GWAS as an effective approach for dissecting complex metabolic traits and informing crop improvement strategies (Sauvage et al., 2014). Plant metabolic diversity reflects genomic evolution, genetic variation and responses to environmental stimuli. Human behavior in crop improvement has also influenced metabolic evolution by acting as a source of selective pressure (Zhao et al., 2023; Zhu et al., 2018). For example, during tomato breeding, humans selected less bitter fruit which led to a reduction in the levels of bitter compounds (Zhu et al., 2018). This is a case where breeding acted on naturally occurring metabolite variation. GWAS has been widely used to investigate a variety of traits, spanning from individual morphological characteristics (like plant height, yield and heading date) to responses to different environmental factors, including biotic and abiotic stresses. In general, agricultural traits exhibit limited variation (Huang et al., 2011; Si et al., 2016), whereas a significant proportion of detectable metabolites show substantial variability (Matsuda et al., 2015; Peng et al., 2017; Wen et al., 2015). Several mGWAS studies have employed a targeted profiling strategy, detecting only a limited number of metabolites. Chan et al., Fig. 2. Schematic representation of how the metabolite prediction model applied in plant breeding, maize is shown as an example. M represents metabolite. E. Karakas et al. Journal of Plant Physiology 309 (2025) 154494 5 conducted a mGWAS using 96 Arabidopsis accessions, analyzing 43 glucosinolate (GSL) phenotypes and ~230 000 single nucleotide polymorphisms (SNPs). A more in-depth analysis is Arabidopsis showed that some of these hotspots are located in genomic regions previously recognized as undergoing recent strong positive selection (selection sweeps). Additionally, these regions exhibit trans-linkage to the putative sweeps, indicating that selective forces have influenced the genome-wide regulation of Arabidopsis metabolism (Chan et al., 2010). However, a comparative mGWAS between rice and maize investigated the shared genetic determinants for the metabolites identified in both species (Chen et al., 2016) and as mentioned above such studies have been now carried out to determine the genetic determinants of key metabolites in an extremely broad range of species including but not limited to wheat, barley, tomato, pepper, cassava, watermelon, citrus, tea and Tartary buckwheat (Burgos et al., 2021; He et al., 2024; Qiu et al., 2024; Ren et al., 2021; von Steimker et al., 2024; Zemach et al., 2023; Zeng et al., 2020; Zhang et al., 2020; Ding et al., 2023; Tieman et al., 2017; Zhao et al., 2024). Collectively, these studies have uncovered candidate genes and metabolic pathways underlying key traits, offering valuable targets for marker-assisted selection. They underscore the power of metabolomics as a data-driven approach for uncovering natural genetic variation and guiding selection of desirable metabolic profiles in breeding programs. Klein et al. were the first to perform a GWAS, uncovering a variant in the Complement Factor H gene that has a strong association with agerelated macular degeneration (Klein et al., 2005). Over the past decade, GWAS has demonstrated its effectiveness in unraveling the genetic foundations of variation in complex phenotypes, including diseases in both humans and animals (Duncan et al., 2017; Jin et al., 2017; Pickrell et al., 2016; Soldner et al., 2016), and physiological and agricultural traits of plants (Bulut et al., 2023; Chen et al., 2016; Huang et al., 2010; Tian et al., 2011; Yano et al., 2016). Nonetheless, variations in population structure and unequal relatedness among individuals can cause misleading associations and false discoveries in GWAS. Significant efforts have been made to address population stratification in GWAS using mixed linear models (MLM), which treat population structures as a fixed effect and account for kinship through random effects (Abecasis et al., 2001; Devlin et al., 2001; Liu et al., 2016). MLM has become the preferred method due to its effectiveness in correcting biases and has led to the development of various models for accurately identifying associations between genetic variants and traits (Korte et al., 2012; Loh et al., 2015; Segura et al., 2012). The use of multi-locus mixed models in GWAS can lead to the discovery of new causative loci (Segura et al., 2012; Wen et al., 2016). While SNP-based GWAS has advanced the understanding of complex traits, it faces challenges like multi-testing penalties and overlooked SNP interactions. To reduce false discovery rates, haplotype-based GWAS, which has higher statistical power, has been developed to identify causal haplotypes with specific combinations. Statistical methods have been used to develop fine-mapping strategies that improve the genomic localization of causal variants identified in GWAS (Schaid et al., 2018). Metabolic engineering and gene-editing approaches-described in other parts of the review-focus on the direct modification of biosynthetic pathways to tailor metabolite profiles. In these cases, metabolomics serves as a key tool to validate the success of the engineered pathways. For example, using synthetic biology and metabolic engineering tools, researchers can introduce specific genetic modifications to alter biosynthesis in plant species, such as modifying the levels of beneficial metabolites like flavonoids or alkaloids. These genetic modifications are guided by the data derived from metabolomic analyses, ensuring that the intended metabolic pathways are successfully altered. In recent years, gene-editing techniques like CRISPR-Cas9 have also been used to modify metabolic pathways. This allows researchers to enhance yields or alter the composition of secondary metabolites, offering new possibilities for improving crop performance and quality. These technological advances are particularly promising given that the current limitations lie not in the ability to manipulate metabolism but rather in our incomplete understanding of the relative importance of individual metabolites-whether in confering tolerance and resistance to environmental stresses or as essential constituents of human diets (Alseekh and Fernie, 2018). Research on the former is currently being carried out in a wide range of studies that perform GWAS under stress and ambient conditions (Zhu et al., 2022), with key metabolites being defined that are particularly important under for example cold, drought or salinity stress (Guo et al., 2015; Janda et al., 2021; Zhang et al., 2024). Similarly, as we detail below research at the interface of crop biology and human nutrition is undergoing a renaissance with strong evidence of the health benefits of a broader range of compounds accumulating in recent years. While we are not quite there yet, there are already several examples of how linking metabolomics to next-generation sequencing, mGWAS, and mQTL can be projected to have a tangible impact on sustainable agriculture. For example, in wheat considerable progress has been made in delivering improved varieties, suggesting that the inclusion of information concerning these metabolites and genes and metabolic pathways enables a more explicit understanding of phenotypic traits and, as such, this procedure could serve as an -omics-informed roadmap for executing similar improvement strategies in wheat and other species with complex genomes (Chen et al., 2021). Similarly, as recently reviewed by Raza et al. metabolomics has been massively employed across multiple species in order to evaluate metabolites that aid in resilience to drought stress (Raza et al., 2025), whilst carrying our metabolomics as a part of a wider multiomics approach has proven highly informative as a way to identify metabolite to phenotype associations with yield associated traits (Luo, 2015). All of these examples offer a clear route by which metabolomics information can be directly taken up into breeding strategies. That said in spite of the fact that the function of a large number of metabolites is well characterized for the majority of metabolites especially secondary metabolites this is not the case and innovative methods to fill this gap are desperately required. This fact notwithstanding, as we define in the following sections, considerable biological knowledge has been attained both in defining pathways important for nutrition or flavor as well as for important plant natural products. 5. Definition of metabolic pathways important for nutrition or flavor Specialized metabolites, particularly flavonoids, exhibit potent antioxidant properties that aid plants in mitigating oxidative stress. Additionally, these metabolites serve as essential phytonutrients for human consumption (Martin, 2018). Significant progress has been made in enhancing the nutritional composition of plants through gene-editing and metabolic engineering. In tomatoes, two major biosynthetic pathways have been modified to enrich phytonutrient content: the anthocyanin biosynthesis pathway (Zhang et al., 2015) and the vitamin D biosynthetic pathway (J. Li et al., 2022). A notable example is the development of the purple tomato, engineered for elevated anthocyanin production. This modification involves SlMYB12, a transcription factor that not only induces flavonoid biosynthesis but also influences carbon allocation from primary metabolism (Zhang et al., 2023). More recently, gene-editing has enabled the production of vitamin D in tomato fruits by disrupting a competing cholesterol biosynthesis pathway via knockout of Sl7-DR2 (J. Li et al., 2022). Beyond tomato, rice has been a major target for phytonutritional improvements aimed at addressing malnutrition and ‘hidden hunger.’ A landmark achievement in this area is the development of Golden Rice, genetically engineered to accumulate high levels of carotenoids (Ye et al., 2000). Additional efforts have focused on fortifying rice with essential vitamins, such as vitamin B9 (folate) through metabolic engineering (Blancquaert et al., 2015), and vitamin B1 (thiamine) by increasing its accumulation in the rice endosperm (Fitzpatrick et al., E. Karakas et al. Journal of Plant Physiology 309 (2025) 154494 6 2024). Various strategies have been proposed to generate vitamin-enriched food crops, as reviewed by Strobbe et al. (2018). In parallel, recent advances in rice improvements have identified low glycemic index and high-protein rice varieties using multi-omics approaches. This discovery links these nutritional traits to allelic variations in STARCH BRANCHING ENZYME 2B (Badoni et al., 2024), opening new avenues for developing functional rice cultivars tailored for improved human health. Furthermore, allelic variation, namely for Wx and GluA2 loci, demonstrated variation in thiamine levels has been suggested that the population’s deficiency in thiamine may be directly related to the thiamine content of local rice varieties in regions where rice is the predominant stable food (Li et al., 2024). In a similar vein, the NET locus was identified in explaining altered levels of among others the quality of rice (Y. Li et al., 2022). This quality improvement includes several nutrients, such as thiamine, lysine, and cinnamic acid. While most of these studies have been performed in polished rice, a recent shift in research on pigmented rice, as a potential source of novel genetic variants to improving nutritional content, has emerged (Mbanjo et al., 2020; Tiozon et al., 2023). Regarding flavor composition, volatile compounds are responsible for flavor in tomatoes and originate from essential nutrients, including carotenoids, amino acids, and fatty acids (Goff and Klee, 2006). While some biosynthetic pathways are identified, many biochemical steps and regulatory mechanisms remain under investigation (Goulet et al., 2012, 2015; Mageroy et al., 2012; Shen et al., 2023). Branched-chain amino acid (BCAA)-derived volatiles are key volatile organic compounds (VOCs) in fruits and vegetables, contributing to both desirable and off-flavors (Bizzio et al., 2022; Schwab et al., 2008). In tomato, major branched-chain volatiles (BCVs) include alcohols, aldehydes, esters, nitriles, and thiazoles (Kazeniac and Hall, 1970; Wang et al., 2016). BCAA degradation pathways, well-characterized in microorganisms, share similarities with those in fruits (Maloney et al., 2010). However, in tomato, BCAA breakdown is linked to respiration rather than volatile synthesis (Kochevenko et al., 2012; Kochevenko and Fernie, 2011). Other fruits, such as melon and apple, employ distinct pathways, including BCAA transamination and the citramalate pathway, to generate BCVs (Gonda et al., 2010; Sugimoto et al., 2021). In tomato, BCAA metabolism begins with conversion to α -ketoacids by branched-chain amino acid aminotransferases (BCATs), followed by three pathways leading to aldehydes, α -hydroxyacids, and alcohols. α -Ketoacid decarboxylase catalyzes aldehyde formation, which can be further reduced by alcohol dehydrogenases (Maloney et al., 2010; Qi et al., 2011). Kochevenko et al. (2012) suggest α -ketoacids, rather than amino acids, are direct precursors of BCAA volatiles. Alcohols may be esterified by alcohol acyltransferases (AATs), with SlAAT1 showing high specificity for 2-methyl-1-butanol in tomato. A comparison with S. pennellii (SpAAT1) revealed greater catalytic efficiency, especially for 3-methyl-1-butanol and butanol (Goulet et al., 2015). Esters can be hydrolyzed back into alcohols via carboxyesterases (CXEs) (Goulet et al., 2012; Kochevenko and Fernie, 2011; Rambla et al., 2017). While esters contribute to fruit aroma, in tomatoes, their lower abundance is linked to high expression of the esterase gene SlCXE1, which reduces volatile ester levels (Goulet et al., 2012). Combining these studies, as highlighted, several different approaches have been demonstrated by gene-editing and metabolic engineering of different metabolic biosynthetic pathways to enrich phytonutritional composition for proposing these approaches highly valuable in combatting hidden hunger. 6. Elucidation of pathways of natural product medicines Natural products from medicinal plants are a diverse group of compounds with significant biological activities, serving essential roles in the pharmaceutical, food and cosmetics industries. Medicinal plants, as a significant source of bioactive small molecules, serve as valuable resources for treating various human diseases. Owing to their remarkable physiological functions, there is growing interest in the biosynthetic pathways of plant natural products (Halder and Jha, 2023; Kawatra et al., 2022). However, the rising market demand for these compounds places significant environmental strain due to reliance on traditional harvesting and plant extraction methods (Singh et al., 2023). Medicinal plants are a rich repository of highly diverse specialized metabolites with significant pharmacological properties. In the past, plant biologists faced challenges in studying the biosynthetic pathways of these metabolites due to limited plant genomic resources. However, recent advancements in high-throughput, large-scale analytical techniques have empowered researchers to uncover the biosynthetic pathways of crucial plant-derived medicinal metabolites. As the global population grows and the demand for medicinal compounds rises given the parallel increase of chronic diseases (Martin and Li, 2017; Sreenivasulu et al., 2023), understanding the full biosynthesis of specialized metabolites is crucial for identifying or developing reliable future sources. Synthetic biology offers a great opportunity in this respect (Facchini et al., 2012; Ye et al., 2024). Plants serve as a rich repository of diverse specialized metabolites, long used for fragrances, flavorings, pigments, insecticides and medicinal purposes (Facchini et al., 2012) These metabolites are considered as to be fundamental mechanism by which plants adapt to diverse ecological niches and environmental conditions, while also protecting themselves from biotic invasions in their habitats (Fürstenberg-H¨ agg et al., 2013; Weng, 2014). Throughout history, plants that produce specialized bioactive compounds have been utilized to treat various diseases, serving as the foundation for the discovery and development of contemporary therapeutics (Butler, 2004). The specialized metabolites in plants possess complex chemical structures, derived from simpler basic units, indicating the involvement of advanced biosynthetic mechanisms and regulatory processes that have been refined and enhanced through natural selection. The investigation of secondary metabolism in plants is hindered by several challenges, including the complexity of the plant genome, large genome size, genetic redundancy, cellular compartmentalization and multifaceted regulatory mechanisms, among other factors (Heinig et al., 2013; Nascimento and Fett-Neto, 2010; Sweetlove and Fernie, 2013). However, in recent years considerable advances have been made in establishing natural product biosynthesis with metabolomics playing a key role. Examples include morphine (Winzer et al., 2015), colchine alkaloid (Nett and Sattely, 2021), mescaline (Watkins et al., 2023), cocaine (Tian et al., 2022), paclitaxel (Zhang et al., 2023), hupezine A (Nett et al., 2023), amaryllidacea alkaloids (Nett et al., 2021), strychinine (Hong et al., 2022) and artemesian (Czechowski et al., 2016). Due to space limitations we will only cover the biosynthesis of artemesian and paclitaxel here. Artemisinin is the most effective treatment against both drugresistant and cerebral malaria caused by Plasmodium falciparum (Nosten and White, 2007). It is utilized throughout Southeast Asia and Africa, where resistance to nearly all other anti-malaria drugs has emerged. Artemisinin (Qinghaosu), is a sesquiterpene lactone, that was isolated in pure form from the shoots of Artemisia annua L. (annual wormwood) plants, with its structure being identified in 1979 (Klayman, 1985). Artemisinin, like most other secondary metabolites, is synthesized through a complex biosynthetic pathway. The process of terpenoid biosynthesis begins with the plastidial methylerythritol phosphate (MEP) and the cytosolic mevalonate (MVA) pathways (Vranov´ a et al., 2013). The biosynthesis of artemisinin and other terpenoids is widely recognized as being highly dependent on the availability of isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP), which are key precursors derived from the mevalonate pathway. The enzyme isopentenyl diphosphate isomerase (IDI) plays a critical role in the interconversion of IPP and DMAPP, ensuring that the balance between these molecules is maintained. This equilibrium is crucial for regulating the flow of isoprenoid precursors, which in turn influences the biosynthesis of terpenoids such as artemisinin. Both plastidial and cytosolic E. Karakas et al. Journal of Plant Physiology 309 (2025) 154494 7 sources of IPP and DMAPP are converted into farnesyl pyrophosphate (FPP) through the action of farnesyl pyrophosphate synthase (FPS). The biosynthesis of artemisinin begins with the production of amorpha-4, 11-diene which is subsequently hydroxylated and oxidized to form artemisinic alcohol, aldehyde and acid. This process is catalyzed by cytochrome P450 monooxygenase (CYP71AV1) and cytochrome P450 oxidoreductase (AACPR) (Ro et al., 2006). Artemisinic aldehyde is then reduced by the enzyme artemisinic aldehyde D11 (13) reductase (DBR22) to form dihydroartemisinic aldehyde, which is further converted into dihydroartemisinic acid by aldehyde dehydrogenase-1 (ALDH1). Finally, a light-induced non-enzymatic photochemical oxidative reaction is believed to represent the last step in the synthesis or artemisinin (Czechowski et al., 2016) (Fig. 3). By expressing the enzymes synthetically in tobacco chloroplasts, Fuentes et al. (2016) were able to produce 120 mg artemisinic acid per kg biomass. This was achieved by not only expressing the core artemisinic acid pathway, rather by the inclusion of additional enzymes affecting flux through the artemisinin pathway (Fuentes et al., 2016). Similarly, paclitaxel is a plant-derived isoprenoid natural compound with powerful anticancer properties. It was first isolated and characterized from the bark of the Pacific yew (Taxus brevifolia) more than 50 years ago (Wani et al., 1971). It was later found to bind to β-tubulin, destabilizing the assembly of microtubules which blocks mitosis at the metaphase–anaphase transition, leading to the induction of apoptotic cell death (Jordan and Wilson, 2004). For many years, Taxaceae plants, especially Taxus trees, have been the main source of the natural product taxol and its derivatives. Although Taxol is well established as a cancer chemotherapeutic agent, its native biological role remains largely uncertain (De Brabander et al., 1981; Manfredi et al., 1982). While traditional breeding and genetic modification approaches have attempted to enhance artemisinin production, metabolomics has played a pivotal role in unraveling its biosynthetic pathway (Liu et al., 2023). identified over 1000 metabolites across 10 categories, compared to 535 metabolites found using untargeted metabolomics. Key compounds included flavonoids, phenolic acids, lipids, amino acids, organic acids, alkaloids and terpenes, with pharmacologically important ones being flavonoids, phenolics acids, artemisinins and coumarins. Furthermore (Graham et al., 2010), mapped QTLs in A. annua, identifying key genetic regions that contribute significantly to variations in traits influencing artemisinin yield. Notably, transcriptome profiling showed complex gene regulation behind artemisinin production. Using these insights (Shen et al., 2018) engineered transgenic A. annua lines with enhanced artemisinin yield, paving the way for large-scale production to meet rising global demand. With the growing demand for artemisinin due to its role in malaria treatment, genetic studies on medicinal plants are increasingly important. Building on this foundation, the availability of high-resolution, haplotype-resolved genomes enables more precise identification of candidate genes and regulatory elements involved in artemisinin biosynthesis (Liao et al., 2022). These genomic resources can accelerate marker-assisted selection, gene editing and metabolic engineering strategies aimed at improving artemisinin content in A. annua, thereby supporting more sustainable and scalable production methods. Paclitaxel is one of the examples of more than 400 identified toxoids, all being characterized by their unique taxane skeleton (Croteau et al., 2006). In addition to paclitaxel, several other toxoids have demonstrated potent anti-tumor activity (Appendino, 1995; Yang et al., 2019). To date, an increasing number of compounds with the taxane core have been isolated and identified from various Taxaceae trees (Siegle and Pietsch, 2018). In recent years advances in genomic and bioinformatic resources have led to the identification of the key enzymes governing its biosynthesis (Fernie et al., 2024). High-quality reference genomes of three Taxus species have provided great insight into the genetic architecture of paclitaxel biosynthesis (Cheng et al., 2021; Song et al., 2021; Xiong et al., 2021) revealing considerable expansions of cryptochrome P450 and BAHD acyltransferase families (Kuang et al., 2019). Initially natural sources of paclitaxel were extracted directly from the twigs, bark or needles of Taxus species. However, even at full tree maturity, the levels of this metabolite are vanishingly small, ranging from 0.0012 % of the dry weight in twigs to 0.015 % in bark (Zhang et al., 2021). The poor yield, the substantial quantity needed for a complete therapeutic treatment (requiring approximately 2–3 g of paclitaxel per patient) (Zhang et al., 2021), the slow growth of Taxus trees and the increasing propensity of cancer in the world population (Bray et al., 2024) have prompted the exploration of alternative production methods (Li et al., 2019). Integration of co-expression and gene cluster analyses exposed the missing steps in the pathway facilitating identification of new enzymes. Recent genomic advances led to the reconstruction of the pathway in model organisms such as Nicotiana bentamiana. For example, Liu et al., reconstituted an early paclitaxel biosynthetic network by expressing six Taxus genes in this species (Liu et al., 2024), while Li et al. (2019) used chloroplast engineering to achieve the biosynthesis of taxadiene and taxadiene-5α -OH by the expression of three enzymes. More recently, the biosynthesis of baccatin III, a key paclitaxel intermediate was achieved by two different routes albeit at far greater yields via the route proposed by the Fernie group (Zhang et al., 2023) than by the Yan group (Li et al., 2019), discussed in detail (Fig. 4). Metabolomics has played a crucial role in advancing our understanding and production of paclitaxel, a highly valuable anti-cancer compound. One common strategy involves the use of jasmonic acid and its derivative, methyl jasmonate (MeJA), to stimulate secondary metabolite production in plant cell cultures. In Taxus species, MeJA treatment combined with optimized culture conditions has significantly increased paclitaxel yields at an industrial scale (Pauwels et al., 2008). In addition, an integrated proteomic and metabolomic study showed that short-term exposure to high-dose ultraviolet-A (UV-A) radiation can further enhance paclitaxel production in Taxus mairei, offering an alternative method to boost metabolite accumulation (Zheng et al., 2016). Metabolomics has also helped map the complex biosynthetic pathway of paclitaxel, which involves more than 20 enzymatic steps. Notably, untargeted metabolomics revealed previously unknown intermediates and branch points within this pathway (Zhou et al., 2019). Furthermore (Yu et al., 2018), used metabolomics and UPLC-MS/MS to compare the metabolic profile of two endangered Himalayan Taxus species, T. fuana and T. yunnanensis, revealing how environmental differences influence their metabolic variation-highlighting the ecological dimension of paclitaxel biosynthesis. Despite the massive recent advances in both of these pathways future studies that characterize the bioactivities of compounds exhibiting similar chemistries to artemesian and paclitaxel will likely prove highly beneficial that said these prove highly compelling examples of the type of treasures that can be unlocked by metabolomics. Indeed, currently 25 % of our medicines are sourced from plants and the use of synthetic biology will likely reveal more economically and ecologically routes to access these in the volumes required. As detailed in the proceeding paragraphs, however, the utility of plant metabolomics is not just in prospecting for medicines and unraveling the complex biosynthetic pathways that nature uses to produce them but also includes enhancing crop nutrition, improving stress resistance, optimizing flavor profiles, and increasing yields to meet the demands of a growing global population. 7. Conclusion and future prospects In this review we have provided a basic overview of the techniques used in contemporary metabolomics as well as providing examples of the considerable “treasures” that metabolomics has aided in uncovering. These are either indirect treasures in which the influence of metabolites on other important traits such as growth and crop yield, and direct treasures wherein alteration of metabolite content directly impacts the flavor or nutritional content of our foodstuffs here examples of E. Karakas et al. Journal of Plant Physiology 309 (2025) 154494 8 Fig. 3. Artemisinin biosynthesis pathway. The MEP pathway products are shown within the pink box. HMGR, 3-hydroxy-3-methylglutaryl coenzyme A reductase; GA-3P, glyceraldehyde 3phosphate; DXS, 1-deoxy-D-xyluse 5-phospate; MEP, 2-C-methyl-D-erythritol-4-phospate; HMBPP, 4-hydroxy-3-methyl-but-2-enyl diphosphate; DMAPP, dimethylallyl diphosphate; IPPI, isopentenyl diphosphate; FPPS, farnesyl diphosphate synthase; ADS, amorpha-4,11-diene synthase; CYP71AV1, cytochrome P450 monooxygenase; CPR, cytochrome P450 reductase; DBR2, artemisinic aldehyde Δ11(13) reductase; ALDH1, aldehyde dehydrogenase 1. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) E. Karakas et al. Journal of Plant Physiology 309 (2025) 154494 9