Authors: Susan T. Ovbude (1Department of Chemistry, University of Nebraska-Lincoln, Lincoln, NE 68588-0304, USA), Sadia Sharmeen (1Department of Chemistry, University of Nebraska-Lincoln, Lincoln, NE 68588-0304, USA), Isaac Kyei (1Department of Chemistry, University of Nebraska-Lincoln, Lincoln, NE 68588-0304, USA), Harshana Olupathage (1Department of Chemistry, University of Nebraska-Lincoln, Lincoln, NE 68588-0304, USA), Jacob Jones (1Department of Chemistry, University of Nebraska-Lincoln, Lincoln, NE 68588-0304, USA), Richard J. Bell (1Department of Chemistry, University of Nebraska-Lincoln, Lincoln, NE 68588-0304, USA), Robert Powers (1Department of Chemistry, University of Nebraska-Lincoln, Lincoln, NE 68588-0304, USA; 2Nebraska Center for Integrated Biomolecular Communication, University of Nebraska-Lincoln, Lincoln, NE 68588-0304, USA), David S. Hage (1Department of Chemistry, University of Nebraska-Lincoln, Lincoln, NE 68588-0304, USA)
Categories: Article, Metabolomics, Liquid chromatography, Gas chromatography, Supercritical fluid chromatography, Mass spectrometry, Nuclear magnetic resonance spectroscopy
Source: Journal of chromatography. B, Analytical technologies in the biomedical and life sciences
Authors: Susan T. Ovbude, Sadia Sharmeen, Isaac Kyei, Harshana Olupathage, Jacob Jones, Richard J. Bell, Robert Powers, David S. Hage
Chromatography is a robust and reliable separation method that can use various stationary phases to separate complex mixtures commonly seen in metabolomics. This review examines the types of chromatography and stationary phases that have been used in targeted or untargeted metabolomics with methods such as mass spectrometry (MS) and nuclear magnetic resonance (NMR) spectroscopy. General considerations for sample pretreatment and separations in metabolomics are considered, along with the various supports and separation formats for chromatography that have been used in such work. The types of liquid chromatography (LC) that have been most extensively used in metabolomics will be examined, such as reversed-phase liquid chromatography and hydrophilic liquid interaction chromatography. In addition, other forms of LC that have been used in more limited applications for metabolomics (e.g., ion-exchange, size-exclusion, and affinity methods) will be discussed to illustrate how these techniques may be utilized for new and future research in this field. Multidimensional LC methods are also discussed, as well as the use of gas chromatography and supercritical fluid chromatography in metabolomics. In addition, the roles of chromatography in NMR- vs. MS-based metabolomics are considered. Applications are given within the field of metabolomics for each type of chromatography, along with potential advantages or limitations of these separation methods.
Metabolomics is a field that involves the analysis of a metabolome, or a collection of low-mass compounds (typical mass, < 1500 g/mol) that are produced through metabolic processes in a biological system [1–6]. Metabolites have diverse physical and chemical structures. Examples of structural classes of metabolites include small peptides, steroids, vitamins, carbohydrates, lipids, fatty acids, amino acids, and organic acids [1–9]. Many of these metabolites are the intermediates or end products of enzymatic reactions in biological systems and can be found in cells or tissues and complex biological fluids such as blood, urine, saliva, or cerebrospinal fluid [1–9].
Use of the term “metabolomics” began in the late 1990s [10–12]. Work in this field has grown significantly over the last few decades [2–8]. This growth is demonstrated in Figure 1, which shows the number of publications that have appeared and included the terms “chromatography” plus “metabolomics” or “metabolites” between the years 2000 and 2022. As this figure indicates, chromatography has long been an important tool for the separation and analysis of metabolites (e.g., in drug metabolism studies) and, as a result, has also been a key component in much of the work conducted in metabolomics [3,8]. In addition, the rapid growth in metabolomics over the last 20 years reflects the many areas that have been impacted by this field [13]. For instance, metabolomics has been used for drug discovery and development [14–16] and the detection or analysis of disease biomarkers [17,18]. Metabolomics has further been utilized as a tool in nutritional studies [19,20], animal breeding [21], plant biology [22], and characterization of the effects of various environmental factors on human health [15,23–27].
There are two major experimental approaches in targeted and untargeted methods [7,8,28,29]. Untargeted metabolomics, or global metabolomics, is based on a top-down strategy that ideally provides a comprehensive, unbiased analysis of all the metabolites in a biological sample [13,30]. This method is typically used in discovery-based research. The information that is obtained in untargeted metabolomics is used to compare patterns or fingerprints of metabolites, such as found in normal systems vs those that have been modified in response to an abnormal biological process, exposure to a drug, or a change due to genetic alterations in a biological system [31]. Untargeted metabolomics has also been used to help identify metabolites that may act as biomarkers for disease states [16,32–35]. However, this approach can be challenging because of the time required to process the complex data that is generated, the need for adequate coverage of the metabolites in a system, and the requirement for accurate and reproducible measurements for many types of metabolites [36–38]. Targeted metabolomics is an alternative approach that is used to investigate a specific set of metabolites with related pathways of interest [7,28,39]. The emphasis on a given set of metabolites allows for better detection and absolute quantification of these compounds, as well as for more reproducible and stable conditions for data analysis. However, targeted metabolomics cannot detect unknown compounds and is generally used for only a limited number of metabolites [7,8,28,39].
One challenge in metabolomics is the diverse range of structures, properties, and concentrations that may occur for metabolites in biological samples [5,40,41]. At present, there is no single analytical method that can identify and detect the many types and levels of metabolites that may occur in living systems [42–52]. Instead, such work typically involves a combination of chromatography or a related separation method (e.g., extractions) with a second analytical technique, which is usually nuclear magnetic resonance (NMR) spectroscopy or mass spectrometry (MS) [4–6,53–62]. NMR has been used in just under a third of recent publications in metabolomics, with MS being employed in around 70% of these reports [5,57]. In addition, some of these studies have used both NMR and MS [6,55,63,64].
Both NMR and MS have advantages and disadvantages when they are applied to metabolomics [6–9,28,29,58,62]. Advantages of utilizing NMR are the variety of information it can provide on compound structure, its good reproducibility, its relatively straightforward use for compound quantitation (e.g., without the need for derivatization or calibration), its ability to work with metabolites that are not readily ionizable, and its need for only minimal sample manipulation [55]. These features have allowed NMR to often be used with only extractions or simple pretreatment steps besides chromatography; however, chromatography has also been used in a growing subset of this work, as is indicated in Figure 2. In addition, hybrid methods such as liquid chromatography-NMR (LC-NMR) and LC-NMR-MS have been used for some research in metabolomics [55,58,63,64]. A key disadvantage of NMR is its sensitivity, which tends to limit this method in metabolomics to the study of relatively abundant metabolites (i.e., those with concentrations > 1 μM) [5]. The highly complex spectra and large amount of information provided by both NMR and MS (e.g., in LC-MS) can also make it time consuming and challenging to process such data for the accurate identification and annotation of metabolites.
An advantage of MS is its higher sensitivity than NMR for detecting low-concentration metabolites [6–9,28,29,58]. The mass information that is provided by MS is also valuable in identifying compounds and in monitoring specific metabolites at low levels in biological matrices [28,29,42,43]. A possible limitation of MS when used in metabolomics is a general requirement for LC, gas chromatography (GC), or a related separation method [28,42,55,57,58,65–70]. This need is demonstrated in Figure 2 by the large number of reports that have combined mass spectrometry with chromatography for metabolomics. The use of chromatography makes it possible for MS to distinguish and examine the many compounds that occur over a relatively narrow mass range in the metabolome [71] and to avoid effects by matrix components on ionization suppression [46].
Matrix effects can be an issue with both NMR and MS. The complex biological samples examined in metabolomics often require some level of pretreatment to minimize interferences from the sample matrix [5,28,41,42,44,46,53,54,58,59,60,72–74]. This often requires at least an extraction or separation step followed by filtration or centrifugation to remove cell/tissue debris and precipitated biomolecules. Again, more advanced separation or chromatographic techniques are typically necessary for MS in metabolomics studies to enable reliable metabolite detection and annotation [46,65–70]. Although the application of direct-infusion MS methods to metabolomics has been reported [75,76], the use of direct-infusion vs. chromatography with MS for metabolomics is still a topic of debate in this field [77,78].
This review will examine the types of chromatography that have been used for sample separations or pretreatment in the field of metabolomics over the last two decades. A discussion will first be presented on how the different modes of LC have been employed and combined with techniques such as MS or NMR for metabolomics. This will include LC methods that are commonly used in metabolomics, such as reversed-phase liquid chromatography (RPLC) and hydrophilic liquid interaction chromatography (HILIC). In addition, other LC techniques will also be discussed that have been used on a more limited basis but that may provide new opportunities for future research in this field. The use of methods such as GC, supercritical fluid chromatography (SFC), and planar chromatography will also be considered. Examples of applications will be provided for each of these techniques, along with a discussion of their potential advantages or limitations when employed in metabolomics.
Sample pretreatment is often a key component in the workflow of an analytical method for metabolomics [16,47,48–50,61]. This step may involve sample clean-up, target enrichment, or analyte pre-concentration to enhance the analysis of metabolites. A good sample pretreatment plan can minimize matrix effects, remove interfering compounds, and/or convert the metabolites of interest into a medium that is compatible with the methods of analysis that will be used to examine the sample’s contents [46,48–50]. Although sample pretreatment and preliminary separations are often required in metabolomic studies, these steps can produce errors in such work (e.g., through the creation of variations in MS ionization) [47–52]. In addition, some forms of sample pretreatment can be tedious and time-consuming to perform [16].
One factor to consider in sample pretreatment is whether the final analysis is to be used for untargeted vs. targeted metabolomics [35,47,79–81]. In untargeted metabolomics, the goal is to obtain a broad coverage of metabolites while reducing or eliminating loss of any compound; thus, only minimal sample pretreatment is desirable [31]. The sample preparation steps in this case should allow the recovery of a broad range of metabolites while also discarding biomolecules and sample debris [82,83]. Solvent extraction techniques often applied in metabolomics include the use of only water, 1 v/v methanol, or a Folch extraction consisting of a 4:8 v/v methanol:chloroform mixture [84,85]. It is also common to combine multiple separation steps and solvent conditions because two or three such steps may be necessary to maximize the recovery of all detectable metabolites [82–85]. In targeted metabolomics a more elaborate sample separation or pretreatment scheme may be needed, as the focus is instead on the analysis of only a given set of metabolites [35,47,79–81].
It is important that the pretreatment steps used in metabolomics provide a stable sample for analysis (e.g., by quenching or removing enzymes that may alter the metabolites) and give a homogeneous sample for analysis that is an accurate representation of the relative levels and forms of the metabolites in their original matrix [61,86–89]. In addition, it is desirable to have a high rate of recovery of the desired metabolites with few interferences from other sample components. Proteins are examples of agents that may lead to significant interferences in metabolomics if they are not removed by sample pretreatment [59,90–93]. For instance, proteins and other matrix components can alter the level of ionization and/or the overall apparent concentration of metabolites (i.e., total vs. bound forms) during the use of LC with MS [35,47,55,57,79–81]. Proteins and sample components such as lipids can also affect spectral quality in NMR by producing broad background signals or additional peaks that may interfere in the detection and identification of metabolites [6,59,91–94].
Various supports and separation formats for chromatography have been employed for metabolomics. This is especially true for LC, the most common chromatographic method used in metabolomics. Early reports in metabolomics employed LC in a planar form by employing the method of thin-layer chromatography (TLC) [1,4,10–12,95,96]. The advantages of TLC are its ability to process several samples simultaneously, its low cost, and the relative ease with which it can be combined with other separation methods for multidimensional separations [97–100]. Early work in metabolomics utilized two-dimensional (2D) TLC to characterize metabolites in microorganisms and to study how variations in the concentrations of these metabolites affected cellular activity [1,4,10–12]. Several more recent studies have used TLC or high-performance TLC (or HPTLC) with imaging based on matrix-assisted laser desorption/ionization (MALDI) and MS for lipidomics (i.e., a field closely related to metabolomics) [101–106]. Examples include the use of HPTLC with imaging based on MS to examine phospholipids in ovine bone marrow mesenchymal stem cells [103] and the use of TLC with imaging MS to profile extracted lipids from various portions of the human brain (see Figure 3) [104]. HPTLC has also been employed with MS to monitor individual classes of lipids in human plasma [105] and lipids such as fatty acids, cholesterol, sphingolipids, and glycerophospholipids that were extracted from porcine brain [106].
The employment of LC as an on-line component in analytical methods for metabolomics is typically done by using packed columns in the methods of high-performance liquid chromatography (HPLC) or ultra-high-performance liquid chromatography (UHPLC, which is also referred to as ultra-performance liquid chromatography or UPLC) [67,70,107–129]. A few metabolomic studies using LC have also employed monolithic supports [130–132]. Modern HPLC columns are operated at pressures up to 5000-6000 psi and are typically packed with 3.5-5 μm diameter particles [42,98]. UHPLC is a subset of HPLC that uses smaller diameter particles (i.e., less than 2 μm) and even higher operating pressures (i.e., 17,400 psi) [98,125]. UHPLC has seen increasing use in metabolomics because it can provide a higher throughput with better resolution and sensitivity than conventional HPLC [126–128,133–137]. For instance, UHPLC gave a three-fold improvement in speed and an increase in signal-to-noise ratio for most of the examined compounds when both HPLC and UHPLC were used with MS, under comparable flow rate and mobile phase conditions, to analyze soy isoflavones or tamoxifen plus their corresponding metabolites [127]. This increase in sample throughput can be valuable for large-scale metabolomics studies, which can be comprised of hundreds to thousands of samples [138–140].
Current estimates suggest the metabolome comprises more than 270,000 compounds with a diverse set of chemical and physical properties [141]. At the simplest classification level, metabolites and lipids can be grouped into either polar or non-polar classes, which dictates the application of different types of solvents and LC support depending on the desired targets. This chemical complexity often leads to the use of several types of chromatographic separations [142–147], which is often referred to as a multidimensional separation method. Both 2D LC and GC have been utilized by the metabolomics community [100,148–151]. Conversely, a single chromatographic method is insufficient for obtaining complete coverage of the metabolome. Instead, multidimensional separations (e.g., based on LC) are commonly employed to increase the coverage of the metabolome and to account for the range of metabolite components that varies widely in terms of polarity and classes of chemicals [142,143].
A multidimensional chromatographic method can be created by combining two or more separate chromatographic steps either off-line or on-line with one another. A common example in the use of LC for metabolomics is the combination of RPLC with HILIC [121–123,152–159]. An off-line approach can be relatively simple to develop; however, this approach will probably also involve many manual sample manipulations and transfer steps, leading to a long overall analysis time [160]. One way two chromatographic methods can be combined on-line is to use a precolumn for the first separation step. The components eluting from this first column are then transferred directly to a second, analytical column for further separation [161]. This on-line combination of columns can be helpful for automated sample pretreatment and untargeted metabolomics [142,161–167]. A major obstacle to the routine implementation of multidimensional chromatographic methods or multiple, distinct column types is throughput when the study size reaches thousands or more samples [100,138–140,151]. Practical considerations such as cost and experience also need to be considered when using multidimensional methods.
Methods in LC are often categorized in terms of the general type of stationary phase employed and the mechanism of separation provided by this stationary phase. As demonstrated by Figure 4, the two main types of LC used for metabolomics are RPLC and HILIC [98,168,169]. However, other categories of LC have also been used in metabolomics for sample preparation or analysis. These other LC methods include ion-exchange chromatography (IEC), size-exclusion chromatography (SEC), normal-phase liquid chromatography (NPLC, of which HILIC is a subset), adsorption chromatography, and affinity chromatography (AC) [98,99]. In addition, there are some LC methods that combine several of these separation modes within a single column or as part of a multidimensional separation scheme. The following section will examine each of these forms of LC and the ways they have been employed in metabolomics.
RPLC (also known as reversed-phase chromatography) is the most common form of LC that has been used in metabolomics [98,168–170]. This is demonstrated in Figure 4 by the number of reports that have used RPLC when compared to other types of LC. RPLC is a type of partition chromatography that separates compounds based on polarity and by employing a non-polar stationary phase, such as octyl (C8) or octadecyl (C18) groups [96–98,170,171]. The fact that water is a weak mobile phase in RPLC has made this method particularly appealing for separating and analyzing targets such as metabolites in biological specimens and in aqueous samples [97–99].
RPLC has been used in both targeted and untargeted metabolomics and is often combined with MS [42]. RPLC and RPLC-MS may be done as part of a system that employs either HPLC or UHPLC [42,125–128]. RPLC is a key tool for metabolomics because it works easily with aqueous samples and separates chemicals based on the general property of polarity [98,99]. A few examples of the many classes of metabolites that have been resolved and examined by RPLC include amino acids, carbohydrates, lipids, organic acids, nucleotides, peptides, and vitamins [98]. Specific examples of compounds that have been separated by RPLC for metabolomics include triglycerides, cholesterol esters, sphingomyelins, phosphatidylcholines, diacylglycerols, phosphatidylethanolamines, fatty acids, and eicosanoid [80]. Metabolites and compounds with low polarity will show the highest retention. In contrast, compounds with high polarity (e.g., sugars and amino acids) may be only weakly retained and even require derivatization to increase their retention time on a column for RPLC [98].
RPLC has also been used in metabolomics for sample pretreatment. Examples are the use of C18 columns or pre-columns for the extraction of steroids and neurosteroids from several types of biological samples [172–174] and the isolation of various classes of metabolites from urine [175,176]. For instance, a C18 column has been used to extract and preconcentrate metabolites produced by Lactobacillus sakei in growth media [177]. Multiple RPLC columns (C8 and C18) have also been combined with LC-MS to extract metabolites such as amino acids, phospholipids, carnitines, bile acids, free fatty acids, and sphingomyelins from plasma and liver tissue [178]. In addition, RPLC has been often used directly in combination with many other types of LC in multidimensional separations for metabolomics. Examples that will be described in more detail in the following sections are the use of RPLC with HILIC and other forms of LC [121–123,143,152–158,179].
Another type of partition chromatography is normal-phase liquid chromatography (NPLC, also known as normal-phase chromatography) [97–99,170]. NPLC differs from RPLC by using a polar stationary phase, such as diol, cyano, or aminopropyl groups [97–99]; thus, NPLC will have its strongest retention for polar compounds. NPLC is not as common as RPLC in metabolomics, clinical analysis, and related fields because NPLC requires sample components to be injected in a nonpolar solvent [99]. However, normal-phase columns are useful when dealing with compounds that may be present in an organic solvent (e.g., through liquid-liquid extraction) and/or that contain one or more polar functional groups. Some examples include lipids, steroids, and sugars [99,174]. For example, NPLC has been used to fractionate various classes of lipids in rat brain tissue and to separate lipoidal steroids [174].
HILIC is another type of partition chromatography that uses a polar stationary phase [98,168,180]. After RPLC, HILIC is the second most common type of LC that is utilized in metabolomics (see Figure 4). This method separates chemicals based on their ability to disperse between an organic-rich region of the mobile phase and a more polar water-enriched layer that is located at or near the surface of a polar support. HILIC often begins with a mobile phase that contains a small amount of water (e.g., 5%) in an organic solvent such as acetonitrile to form the water-enriched layer. The water content of the mobile phase is then gradually increased to promote elution [98,181]. The surface of the support in HILIC may contain a polar, noncharged surface (e.g., silanol groups on unmodified silica) or surfaces that have been modified to contain diol, amide, or cyano groups [98,182]. Charged groups may also be present, as occurs in the use of supports that have surfaces with protonated amines, carboxylates, or zwitterionic groups (e.g., with the latter being referred to as zwitterionic HILIC or ZIC-HILIC) [98].
HILIC has been used in metabolomics and other areas as a tool to retain and separate polar and charged compounds found in biological samples [168,183,184]. Compounds in this group range from small organic acids and amino acids to nucleosides, sugars, and phosphate-containing compounds, as illustrated in Figure 5. The ability of HILIC to retain and separate these polar compounds makes this method complementary to RPLC, which instead has its highest retention for nonpolar analytes [185–188]. One way HILIC has been used in metabolomics is as a pretreatment method for polar metabolites [7,189]. This has included the use of HILIC cartridges to extract phenolic metabolites from plasma and urine samples [189] and the use of a HILIC column to enrich hydrophilic metabolites and amino acids in blood for analysis by targeted metabolomics [7].
Many studies have also used HILIC in analytical separations for metabolomics. Like RPLC, HILIC is often used with MS for this purpose [70,153,190]. The fact that HILIC has multiple modes of interaction can make this method more versatile than RPLC and potentially allow for improved metabolite coverage [190,191]. However, the same feature makes it more difficult to use HILIC with regard to predicting analyte retention [192]; this factor has led to HILIC being used in metabolomics mainly for untargeted studies [190,191]. HILIC has been used with MS to simultaneously analyze short-, medium-, and long-chain aromatic amines and acylcarnitines that are responsible for insulin resistance in type 2 diabetes in several human population studies [193,194]. A method to examine acylcarnitine and amino acid metabolism was developed based on HILIC combined with MS and used to profile samples such as plasma, urine, cerebrospinal fluid, and tissue extracts [195].
A growing number of researchers are using HILIC in combination with RPLC and MS [121–123,152–159]. For instance, this combination of techniques has been employed to separate and resolve over 150 metabolites in rice [152]. This set of tools has also been used together in the non-targeted metabolomics analysis of other food products [157]. The use of HILIC plus RPLC with MS was found to increase the coverage of the metabolome in urine and plasma samples by approximately 44% and 108%, respectively, when compared to the use of RPLC along with MS [153]. UPLC and MS have been used with both RPLC and HILIC for untargeted metabolomics studies to identify potential biomarkers for liver cancer or breast cancer [122,154].
Despite the complementary nature of RPLC and HILIC, many metabolomics studies employing LC with MS still use a single type of column. This is believed to be linked to the increase in time and cost when using multiple columns. RPLC is also more commonly used with MS than HILIC in metabolomics [191,196,197]. This is partly due to RPLC being a more established method with a long history and high degree of familiarity in clinical analysis and biomedical research [97–99]. However, part of the preference for RPLC may be linked to the perception that RPLC-MS provides better performance than HILIC-MS and is a more robust analytical approach in areas such as metabolomics [153,198–200]. In addition, HILIC typically produces broader peaks than RPLC, resulting in a reduced peak capacity and greater reliance on the resolution capabilities of the mass spectrometer in HILIC [181,183].
RPLC works well for many nonpolar metabolites or metabolites with medium polarity. However, this separation method may be difficult to use with solutes that contain a charge. In this situation, an alternative is to use IEC [97–99]. IEC is a form of LC in which a support containing fixed charges on its surface is used to bind and separate solutes with complementary charges. IEC can be divided into two categories based on the type of charge present on the support and stationary phase. In cation-exchange chromatography, the stationary phase consists of negatively-charged groups that are used to retain positive ions. In anion-exchange chromatography, the stationary phase has positively-charged groups that are used to bind and separate negative ions [97–99,170].
Both cation- and anion-exchange chromatography have been used for sample pretreatment and fractionation prior to the analysis of metabolites by other LC methods combined with MS [168]. For example, strong cation-exchange supports have been used to extract more than 140 charged and water-soluble metabolites from cultures of Escherichia coli and Saccharomyces cerevisiae prior to a further separation and analysis by HILIC and MS [143]. In another report, both anion- and cation-exchange chromatography were used to fractionate components of the steroid metabolome (e.g., bile acids, oxysterols, and hormonal steroids) before these compounds were separated and analyzed further by LC-MS (see Figure 6) [201].
IEC has been directly used in other studies in the specialized form of ion chromatography (IC) for analytical separations and in combination with detectors based on conductivity, fluorescence, amperometry, or MS [98,99,168]. The need to address the salt content and relative volatility of the mobile phase contents in IEC is an issue that must always be considered when combining this method with MS [202–204]. For instance, the coupling of IC with MS requires the use of an ion-exchange column with an on-line electrolytic eluent suppressor system to convert a mobile phase that may have a high concentration of competing ions to an aqueous solution that contains little or no competing ions [97,98]. An organic solvent can also be added to the mobile phase through a mixing tee to provide an eluent that is more compatible with electrospray ionization for MS [168].
IC in the form of capillary IC-MS has been used as the basis for several metabolomics studies. Capillary IC-MS with a strong anion-exchange medium has been employed to analyze organic acids and sugar di/triphosphates from the protozoan Trypanosoma brucei [205] and to examine metabolites related to carbon metabolism in glioblastoma cells [206]. This combination of methods has further been used to profile agents such as sugar phosphate and nucleotides in extracts of human kidney HEK293 cells [207] and to examine nucleotides, organic acids, sugars, and sugar phosphates in SW480 cancer cells [208]. IC-MS with anion-exchange has been used to evaluate phosphorylated metabolites after their enrichment using metal oxide affinity chromatography [209]. A comparison has also been made of LC-MS methods based on RPLC, HILIC, or anion-exchange LC in terms of their detection and coverage of 354 metabolites from 19 compound classes in mammalian samples [179]. HILIC and RPLC gave retention factors greater than one for 54% and 41% of these metabolites, respectively. Combining HILIC with RPLC increased the overall metabolite coverage to 92%, and including anion-exchange chromatography (e.g., for short chain fatty acids) further increased the coverage to 97% [179].
Several reports have considered the direct coupling of IEC with other forms of LC for multidimensional separations. RPLC and anion-exchange chromatography have been used with MS for metabolomic analysis as related to environmental exposure (i.e., the exposome) [210]. A scheme based on RPLC followed by IEC and MS was used to detect over 80 metabolites in plasma and urine samples [211]. HILIC has been used in a mixed-mode format with weak anion-exchange chromatography and MS to determine glyphosate and its metabolite aminomethylphosphonic acid in fruits and vegetables [212]. Strong cation-exchange chromatography was used with RPLC and a chip-based format for the analysis of metabolites and peptides in spleen samples [213].
Another way in which a separation based on ion-exchange can be combined with a second separation mode is through the method of ion-pair chromatography (IPC), which is also referred to as ion-pair liquid chromatography (IPLC) [28,67,98]. IPC typically uses a column for RPLC and a mobile phase that contains an ion-paring agent with a non-polar tail (e.g., for interacting with a RPLC column) and a charged section for interacting with an analyte that contains the opposite charge [98,214,215]. Examples of ion-pairing agents in IPLC are perchlorate or alkane-sulphonic acids (e.g., as occurs in sodium dodecyl sulfate) for the separation of cations and tetrabutylammonium or related surfactants for the separation of anions [96,214].
Several reports have used IPC for metabolomics, either alone or in combination with other LC separation methods. One study compared the use of HILIC-MS, RPLC-MS, and IPC-MS for the analysis of polar and/or ionic metabolites, with a particular emphasis on compounds related to central carbon metabolism [215]. Of these three methods, only IPC-MS was found to provide a sufficiently robust method (when suitable precautions and conditions were followed), making it possible to profile over a hundred organic metabolites in serum, urine, and tissue extracts [215]. IPC-MS was used to detect small polar molecules despite contamination, including the use of diamyl ammonium acetate as an ion-pairing reagent for the retention of acidic polar compounds [216]. A C18 based column was used either directly or with IPC to give separations based on reversed-phase, cation-exchange, anion-exchange, and normal-phase modes for the separation of polar metabolites [81].
Another type of LC based on stationary phase and separation mode is adsorption chromatography. In this method, solutes are separated based on their adsorption to the surface of a support [97–99,172,174]. Adsorption chromatography is not usually used for the analytical separation of biological samples for metabolomics but is instead employed for sample pretreatment. Supports that are frequently used for sample pretreatment by adsorption chromatography and SPE include polymeric supports (e.g., polystyrene) and non-derivatized silica [97]. Similar supports for adsorption chromatography are also often used in planar chromatography [98,99]. Examples of adsorption chromatography in metabolomics include the use of polymeric resins to extract steroids and neurosteroids [172,174]. In addition, silica on TLC plates has been used for the separation of lipids in brain tissues and their analysis by MS [104].
Separation of compounds based on their size/shape and ability to partition into a porous support forms the basis for SEC [97–99]. Although most of the compounds examined in metabolomics have relatively small masses and cannot be separated from one another by SEC, this method is still important in removing proteins and other higher-mass substances from smaller metabolites [98]. For instance, the removal of proteins in metabolomics is often required to avoid matrix effects and ionization suppression in the use of LC with MS [35,47,79–81]. The retention of biomolecules in a metabolomics sample has been shown to non-uniformly perturb metabolite concentrations primarily through protein binding interactions [59,217,218]. Protein precipitation with an extraction solvent (e.g., methanol) is a preferred approach to remove biomolecules and release the bound metabolites [218]. While ultrafiltration is also used to remove biomolecules, the technique may also selectively deplete protein-bound metabolites from the MS or NMR metabolomics analysis. Sample preparation by SEC is often carried out using low-performance size-exclusion media in solid-phase extraction cartridges or spin columns. Alternatively, SEC can be employed to isolate a large biological agent that may contain bound or encapsulated metabolites. Examples of this latter application are the use of SEC for the isolation of metabolites that are associated with proteins or extracellular vesicles [219–223].
Alternatively, SEC has been used in metabolomics to specifically analyze and characterize the interactions of metabolites with proteins and other biomacromolecules [4,220,224–226]. In this type of work, SEC can be used to separate free metabolites and metabolite–protein complexes based on their large differences in size. One recent study used SEC to examine metabolite-protein interactions in cell-free lysates from cultures of the plant Arabidopsis thaliana [220]. SEC may also be used with known mixtures of a metabolite and a protein to separate and measure the free and bound fractions of a metabolite; this information can then be used in binding studies to provide information on the equilibrium constant for the metabolite–protein interaction [4,224–226].
Another way size-based separations can be utilized in metabolomics is through the use of restricted access media (RAM) [163,227–231]. This type of material consists of a porous support in which only the interior contains the desired stationary phase, such as non-polar groups or ion-exchange sites [79,232–238]. The pore size is set so that only small compounds such as metabolites are allowed to enter the interior region while prohibiting the entry of larger agents such as proteins. This ability makes RAM columns useful in on-line sample pretreatment for work in areas such as untargeted metabolomics [163,227–231].
Turbulent flow chromatography (TFC) is an additional way in which size-based separations can be combined with other forms of LC for use in metabolomics. TFC is also known as ultra-high flow or high-flow chromatography [90,239–241]. TFC often makes use of a relatively large diameter RPLC support (e.g., 25-50 μm) that is operated at a high flow rate (e.g., 1.5-5.0 mL/min). These conditions can generate turbulent flow within the column that enables macromolecules such as proteins to be readily washed away, while allowing smaller molecules to enter the pores of the particles and to be retained for further separation and analysis [241,242]. A common application for this method is to remove proteins and to isolate drug metabolites for the analysis of complex matrices such as urine, plasma, tissue homogenates, cerebrospinal fluid, and environmental samples [243–248].
Another type of LC is affinity chromatography (AC) [73,97,249]. This method is based on the selective and reversible interactions between biomolecules [73,249]. Immunoaffinity chromatography (IAC) is a subset of AC in which an antibody or related binding agent is employed [73,98,250,251]. The high selectivity and strong binding of antibodies for their targets have made IAC an important tool for sample pretreatment and the isolation of specific sample components for applications related to targeted metabolomics [73]. This form of IAC is often referred to as immunoextraction and can be used either off-line or directly with other analytical techniques [251]. This has included the on-line use of IAC with further LC separation and MS for the trace analysis of specific compounds and their metabolites in matrices such as food, plant or animal tissues, clinical specimens, and environmental samples [73,250–264].
IAC has been used in several studies to isolate mycotoxins and related metabolites for the measurement or characterization of these components [252–255]. For instance, IAC has been used to enrich and detect deoxynivalenol and fungal-derived metabolites from grains [255]. An IAC column has been employed to isolate and aid in the detection of residues and metabolites related to the steroids methandrostenolone and brassinosteroid [256,259]. Aflatoxins, ochratoxins, and related metabolites have been isolated from samples such as cocoa, oil, spices, cereals, nuts, and infant formula by using immunoaffinity columns [257,258]. Broad-specificity monoclonal antibodies were used to purify and enrich brassinosteroid metabolites from minute plant samples before further analysis [259]. IAC has been coupled on-line with RPLC for the trace analysis of triazine herbicides and their metabolites or degradation products [265–268]. In addition, IAC has been used on-line with LC-MS for trace analysis of LSD and its metabolites in samples such as urine or plasma [260,261].
Other binding agents have also been used in AC for metabolomics. For example, ergosterol biosynthetic enzymes were tagged with an immunoglobulin G-binding domain [269]. Binding by metabolites to these enzymes was then examined after the affinity-based isolation of the tagged proteins [269]. Metal oxide AC based on zirconia or titania was used to enrich phosphorylated metabolites prior to their separation and analysis by IC-MS [209]. The use of traditional immobilized metal-ion AC for sample pretreatment was considered in the same report [209].
AC and high-performance affinity chromatography (HPAC) have been used to study interactions by proteins and other biomacromolecules with metabolites of drugs and hormones [4,225,226,270–272]. Examples include the application of HPAC to compare and characterize the binding of thyroxine, triiodothyronine, and related metabolites with HSA [273,274]; the interactions of phenytoin and its major metabolites with HSA [275,276]; and the binding of warfarin and L-tryptophan and related coumarin or indole compounds at Sudlow sites I or II of HSA [277,278]. HPAC has also been used to examine the rate constants for the interactions of the individual chiral forms of phenytoin metabolites with HSA [279].
GC is a chromatographic method in which the mobile phase is a gas and sample components are separated in the gas phase [97–99]. Although LC is more commonly used in metabolomics, GC is also an important separation tool in this field [280–317]. This is especially true when the goal is to work with volatile compounds or compounds that can be readily converted into a volatile form [98,281,282,305,306]. The relative extent of use for GC vs LC in this area is illustrated in Figure 7 by the number of papers that have appeared between 2000 and 2022 that have contained the terms “metabolomics” plus “gas chromatography” vs “liquid chromatography”. In this comparison, GC was used in about one third of the reports over 2000-2022 that used either LC or GC for metabolomics. In addition, the use of GC to analyze metabolites goes make many decades and predates the modern era of metabolomics. For instance, one of the first applications of GC occurred in the early 1950’s when it was used to separate methyl esters of fatty acids [98,281]. GC is most often used in metabolomics in the form of GC-MS [280,282,283,305], although multidimensional GC-MS (e.g., 2D GC-MS) has also been used for untargeted and targeted metabolomics [148–150,306]. Examples of metabolite classes that have been analyzed by GC-MS include amino acids, organic acids, catecholamines, fatty acids, hormones, hydroxyl acids, sterols, and sugars, among others [282,306].
Advantages of GC-MS are its high efficiency and good resolution for volatile compounds and the availability of well-established libraries of mass spectra that can be used for compound identification [98,282,283,305]. One limitation of GC-MS is the requirement that its analytes be both sufficiently volatile and thermally stable at the temperatures that are employed for sample injection and elution [98,99,306]. Meeting this requirement often means that analytes must first be derivatized to convert them into a volatile and thermally-stable form (e.g., through the use of silylation or the formation of methyl esters) [58,98,99,284–289,305,306]. This form of pretreatment is important in metabolomics because it significantly expands the range of compounds that can be examined by GC-MS [305,306]. Figure 8 gives a typical workflow for GC-MS in metabolomics and that includes a derivatization step [305].
Two fields in which GC-MS has been used for metabolomics are clinical analysis and biomedical research [280,292,299–314]. For instance, GC-MS has been used to analyze metabolites in human plasma [299] and for profiling of endogenous agents such as organic acids, amino acids, and sugars in urine, serum, or plasma [292]. The use of GC-MS in the study of cancer is another important application in the field of metabolomics [307–311]. For instance, GC-MS has been used in metabolomics to compare hepatocellular carcinoma with healthy surrounding tissue [307] and to examine metabolites in plasma for patients with Medullary thyroid cancer vs the plasma of healthy individuals [308]. GC-MS has been used to identify metabolites in urine that may be used for diagnosis non-Hodgkin’s lymphoma [310] and to compare healthy vs cancerous prostate epithelial cells [311]. A 2D GC method has been employed with untargeted metabolomics to analysis the contents of serum from patients with colorectal cancer [309]. Both GC-MS and 2D GC-MS have also been used in metabolomics studies involving yeast and microbial systems, as well as in other biomedical applications [288,312,313].
GC-MS has also been frequently employed to examine both volatile metabolites and low mass non-volatile metabolites (e.g., through derivatization) in plants and food-related matrices [290–298,305,312–317]. One report used GC-MS for plant metabolomics to examine changes in sugars, amino acids, organic acids, and some inorganic acids that were present in radish roots stressed by exposure to lead or cadmium [291]. Application in food science have included the metabolic profiling, quality analysis, and authentication of foods, along with studies of the growth, processing, stressing, and storage of foods [305]. For instance, GC-MS has been employed to investigate flavor development in shrimp paste [314]. Metabolomics based on both GC and LC have been used to study the volatile plus nonvolatile compounds involved in the withering process of black tea [315]. GC-MS and metabolic profiles have been used to characterize the regions in which certain foods or their components have originated, such as has been used for some varieties of rice or whole grain liquor [316,317].
SFC is a form of chromatography in which a supercritical fluid is the mobile phase [97–99,318–324]. SFC has not been used as extensivity as LC or GC in metabolomics, with the number of papers in this area being about 1% of those shown in Figure 7 for the use of LC and GC in metabolomics. However, there has been increasing and ongoing interest in the unique capabilities SFC may offer for metabolomics [318,319,322,323]. For instance, a supercritical fluid has a lower viscosity than a liquid, along with a higher density than a gas [98,99,318,319]. These properties mean SFC will have retention behavior and efficiencies that are between those of LC and GC. Furthermore, SFC can often be used with modified forms of equipment, stationary phases, and supports that were originally created for either GC or LC [98,99], including the use of both HPLC and UHPLC columns [125,319–324]. Stationary phases for SFC have included those developed for RPLC, NPLC, and HILIC [98,319–324].
SFC has been combined with MS for various applications in metabolomics. As an example, SFC-MS has been used for comprehensive lipid profiling and the separation of polar metabolites [318,322,323]. The use of SFC-MS in pharmaceutical analysis has been considered [319], and the use of SFC-MS with UHPLC-type columns and various stationary phases has been evaluated for the analysis of medium and high polarity metabolites in urine [324]. SFC-MS using UHPLC columns has been demonstrated to separate both polar and non-polar lipids [323]. Along with the possible addition of methanol to the mobile phase, it has been found that ionizable compounds such as poly acids, aliphatic amines, and other polar metabolites can be separated by SFC when utilizing small amounts of acids, bases, or salts as mobile phase modifiers [318,319]. A small amount of water can also be used with CO2 and methanol to elute highly polar compounds and metabolites in SFC [318,319].
Many of the prior examples given in this review have involved the combination of LC, GC, or SFC with MS for metabolomics. This reflects the fact that LC-MS is currently the most common approach used to analyze samples for metabolomics, with GC-MS also playing important or growing roles in this field [290–306,325]. In each of these methods, a separation based on some form of chromatography is generally required prior to MS to resolve the many compounds that occur over the relatively narrow molar mass range of the metabolome [71]. This separation step can further be employed to minimize effects by the matrix or co-eluting compounds that may lead to ion suppression in MS [35,46,47,79–81,326]. However, neither the presence of metabolites over a narrow mass range nor ion suppression are issues in NMR, although sample matrix effects due to proteins or lipids may still occur [6,59,91–93]. Thus, chromatography has not been as frequently used in NMR as it has been in MS for metabolomics, as shown earlier in Figure 2.
There are several other factors that have led to chromatography not being part of many prior NMR-based studies. One primary reason is to minimize sample handling and the overall time it takes to acquire an NMR spectrum for a sample. This is required because metabolites are readily interconverted due to residual enzymatic activity or oxidation and may be subject to such issues as thermodynamic instability or degradation and solubility [86–89,327,328]. Another concern is the change in relative concentration that may occur for a metabolite during pretreatment [329]. Compound decomposition during the separation process, inefficient or non-uniform compound derivatization, and less than 100% compound recovery are other ways chromatography may impact the outcome of metabolomics studies, as seen in LC-MS [329–331]. In addition, the impact of these factors may not be uniform on metabolites and give an incorrect impression of the relevance of a given change on the system under investigation. For example, such effects may lead to a metabolite being incorrectly identified as a potential biomarker for cancer [332].
NMR does face challenges in metabolomics with peak overlap, low sensitivity, and a limited dynamic range; however, these issues can be partly overcome by employing multidimensional NMR experiments instead of chromatography. For instance, NMR methods such 2D ^1^H-^13^C heteronuclear single quantum coherence spectroscopy (HSQC), heteronuclear multiple bond correlation (HMBC), ^1^H-^1^H total correlation spectroscopy (TOCSY), and HSQC-TOCSY are frequently utilized in metabolomics to assist with metabolite annotations and can be used in combination with LC-MS [6,55–57,333–337]. Analogous schemes based on MS/MS are used in metabolomics [338,339], although these methods are not replacements for chromatography in the same way as when NMR peaks are spread across two dimensions using data based on ^1^H and other nuclei (e.g., ^13^C, ^15^N, or ^31^P) [340,341]. In addition, non-uniform sampling, rapid acquisition methods and super sequences based on NMR by ordered acquisition using ^1^H detection can be used for rapid (< ms) 2D NMR collection for metabolomics [57,342–346].
Despite the differences in the need for chromatography in NMR and MS, there is a growing interest in combining LC with NMR for metabolomics. This is demonstrated in Figure 2 by the increasing number of publications in NMR-based metabolomics that have employed chromatography. The main reason for this interest is the same as it is for MS, in that using LC with NMR can simplify complex heterogeneous mixtures before their analysis [347–350]. In principle, all the chromatographic techniques discussed in Section 4 can be employed with NMR in either an on-line or off-line mode. The easiest of these two modes is to use off-line coupling of LC with NMR, as illustrated in Figure 9 for a scheme that couples RPLC with ^1^H NMR in metabolomics [348]. The use of on-line LC with NMR requires interfacing LC with an NMR flow probe [351,352] to rapidly transfer sample aliquots into a stationary chamber positioned inside the magnet of the NMR spectrometer. In either off-line or on-line modes, the metabolomics samples will need to be placed into an NMR-friendly solvent that includes a deuterated buffer and has a low salt concentration (i.e., < 50-100 mM for use with cryoprobes) [353].
This review examined the use of chromatography and its various forms in the field of metabolomics. This included the use of chromatography with either MS or NMR and the use of LC, GC, or SFC for the pretreatment or separation of compounds in samples for metabolomics. General considerations for sample pretreatment and separations in metabolomics were discussed along with the various supports and separation formats for chromatography that have been used in metabolomics. Many types of LC methods were considered regarding their applications and possible advantages or disadvantages for metabolomics. These methods included RPLC and HILIC, which are the two forms of LC used the most in metabolomics. Other LC methods (e.g., IPC, NPLC, IEC, SEC, AC, and adsorption chromatography) were also discussed that have been used in fewer applications for metabolomics but offer unique abilities that may be valuable for future work in this field. The combination of these techniques in multidimensional LC for metabolomics was further considered, as well as applications of GC and SFC.
Several trends and areas for possible future development were noted in this review. For example, there remains room for significant growth in the use of LC methods other than RPLC or HILIC and in the continued use of GC or SFC with MS for metabolomics. The same is true for NMR-based metabolomics, which has usually been conducted without the use of chromatography in most previous studies and for which the inclusion of chromatography may allow for the analysis of even more complex samples and sets of metabolites. The need for methods that can examine a greater range of metabolites will also probably involve the use of more forms of multidimensional separations with either MS or NMR and combining RPLC and HILIC to a great degree with other forms of chromatography. Finally, little work has been done up to the present in using chromatography with microfluidics for metabolomics. This is another area expected to see future development, as microfluidics has recently been employed for single cell metabolomics [354–357] and in other fields for LC separations or for sample analysis by MS [358–360]. Developments in any of these areas should continue to increase the capabilities and impact that chromatography offers for new research and applications in metabolomics.