Quantitative NMR Methods in Metabolomics

Abstract

Nuclear Magnetic Resonance (NMR) spectroscopy is one of the two major analytical platforms in the field of metabolomics, the other being mass spectrometry (MS). NMR is less sensitive than MS and hence it detects a relatively small number of metabolites. However, NMR exhibits numerous unique characteristics including its high reproducibility and non-destructive nature, its ability to identify unknown metabolites definitively, and its capabilities to obtain absolute concentrations of all detected metabolites, sometimes even without an internal standard. These characteristics outweigh the relatively low sensitivity and resolution of NMR in metabolomics applications. Since biological mixtures are highly complex, increased demand for new methods to improve detection, better identify unknown metabolites, and provide more accurate quantitation, continues unabated. Technological and methodological advances to date have helped to improve the resolution and sensitivity, and detection of a larger number of metabolite signals. Efforts focused on measuring unknown metabolite signals have resulted in the identification and quantitation of an expanded pool of metabolites including labile metabolites such as cellular redox coenzymes, energy coenzymes and antioxidants. This chapter describes quantitative NMR methods in metabolomics with an emphasis on recent methodological developments, while highlighting the benefits and challenges of NMR based metabolomics.

1. Introduction

The field of metabolomics represents the parallel analysis of large numbers of metabolites in biological systems. Metabolites provide information on the instantaneous biological state of an organism or system along with the functions of upstream cellular molecular species such as genes, transcripts and proteins, in health and pathological conditions. Using a variety of advanced methodologies, comprehensive analysis of metabolite data enables understanding biological phenotypes, deciphering mechanisms, and identifying disease biomarkers or drug targets (Raftery 2014; Nagana Gowda & Raftery 2019a). Metabolomics applications span a wide range of disciplines including human health and diseases, pharmacology, drug development, toxicology, environment, plants, food and nutrition. However, a majority of the studies to date is focused on improving the mechanistic understanding, prevention, early diagnosis, and management of human diseases (Kodama et al. 2020; Goldman et al 2019; Johnson et al. 2016; Wishart 2016).

Nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry (MS) are the two most widely used methods in the metabolomics field. MS typically provides 2 to 3 orders of magnitude higher sensitivity than NMR and thereby enables analysis of several hundreds to thousands of metabolites from a single measurement. Generally, in MS analysis, metabolites from biological mixtures are subjected to separation using methods such as liquid chromatography, gas chromatography or electrophoresis prior to detection. Separation using liquid chromatography, however, is the most popular and nearly 80% of the metabolomics methods use liquid chromatography resolved MS method (Edison et al. 2021). Absolute quantitation of metabolites in MS involves using internal or external standards, ideally, for each metabolite. However, finding isotopically labeled internal standards for each metabolite is challenging and hence, one standard that represents a class of metabolites is often used (Djukovic et al. 2020). This approach, however, can result in a loss of accuracy. In contrast, and as will be described below, NMR provides several approaches for accurate quantitation.

Although NMR spectroscopy is less sensitive than MS, it exhibits numerous unique and favorable characteristics that are beneficial to the field of metabolomics (Edison et al. 2021; Wishart 2019; Nagana Gowda and Raftery 2014a, 2015a, 2017a; 2019a). Notably: (1) NMR is highly reproducible and has excellent linearity (Mo, 2008); (2) NMR provides absolute quantitation of all metabolites in the spectrum using a single internal standard or even without the need for an internal standard; (3) it provides the gold standard in establishing the identity of unknown metabolites; (4) it enables the analysis of intact biofluid and tissue samples with little to no need for sample preprocessing; (5) it is non-destructive, which means the sample remains intact after the analysis and can be reused for analysis using NMR or using other methods such as MS; (6) it enables tracing of metabolic pathways and measuring metabolic fluxes utilizing stable isotope labeled precursors; (7) using NMR, the same metabolites can be detected through one or more types of atomic nuclei such as ¹H, ¹³C, ³¹P or ¹⁵N, which provides flexibility to measure metabolite levels; (8) NMR’s ability to detect essentially all molecular species with a given nucleus makes it extremely useful for following methods development; and (9) NMR offers new avenues to measure unstable metabolites that are fundamental to cellular functions. Such characteristics far outweigh the poor sensitivity and resolution of NMR and have been exploited extensively in the metabolomics field.

Human blood serum/plasma, urine and tissue continue to be the most widely used biological specimens in the metabolomics field. However, other biological specimens including saliva (Lohavanichbutr et al. 2018), cerebrospinal fluid (Albrecht et al. 2020), gut aspirate (Bala et al. 2006), bile (Nagana Gowda 2011), amniotic fluid ( Orczyk-Pawilowicz et al. 2016), synovial fluid (Anderson et al. 2020), fecal samples (Zierer et al. 2018), exhaled breath condensate (Maniscalco et al. 2020), tear (Yazdani et al. 2019) and sperm-seminal fluid (Engel et al. 2019) have also been analyzed. In addition, specimens from animal models, cell lines, yeast (Airoldi et al. 2015), bacteria (Lussu et al. 2017), tumor cells (Lane et al. 2017), tumor spheroids (Kalfe et al. 2015), exosomes (Zebrowska et al. 2019) and isolated mitochondria (Xu et al. 2018) have been used.

The key steps involved in almost all metabolomics investigations include metabolite detection, unknown peak identification and quantification. Relative or absolute concentrations of metabolites thus obtained are then subjected to statistical and/or metabolic pathway analysis focused on a wide variety of applications in the areas of basic and medical sciences. Typically, metabolite data are analyzed using univariate and multivariate statistical analysis focused on the discovery and validation of putative metabolite biomarkers. Alternatively, metabolite levels or isotope labeled metabolites are used for identifying the perturbed metabolic pathways, which provide mechanistic understanding of cellular functions including information on drug targets for therapeutic development.

2. Quantitation Approaches Using NMR

The quantitative ability of NMR makes it an important platform complementary to mass spectrometry in metabolomics. NMR can be used for quantitative analysis of metabolites in intact samples, extracted samples, live organisms, cells or subcellular organelles such as mitochondria. In NMR, generally, metabolite peaks are identified prior to their relative or absolute quantitation. The identities of metabolites are established using databases of standard compounds, the comprehensive analysis and 1D and 2D NMR spectra, and/or spiking with authentic compounds.

Quantitation generally involves either (a) relative quantitation, in which metabolite levels are measured relative to one another; or (b) absolute quantitation, in which molar concentrations of metabolites are determined using an internal or external standard. Currently, relative quantitation is the mostly widely used approach owing to its ease of use combined with challenges associated with absolute quantitation, especially for some sample types, such as cells, tissue, fecal samples, and etc. However, absolute quantitation promises a number of benefits. Importantly, it provides a basic platform of metabolite levels for a specific type of biological specimen. This is important for assessment of data quality, such as to compare samples across different geographical regions, different batches or analysis times, or perhaps most importantly to compare to known values, such as clinical ranges for blood or urine metabolites. Considering the increased interest for absolute quantitation, there have been numerous efforts in recent years focused on establishing reference standards for absolute quantitation using NMR as described in the following section.

2.1. Internal Reference Standards for Absolute Quantitation:

Many compounds (>25) have been evaluated as potential internal standards for applications in numerous areas including organic chemistry, natural product chemistry, agriculture, drug discovery and pharmaceuticals (Maniara et al. 1998; Holzgrabe 2010; Pauli et al. 2012; Rundlöf et al. 2010; Salem & Mossa 2012). These compounds exhibit favorable physical characteristics, such as unique chemical shift, purity, stability, solubility and suitability for accurate gravimetry. However, most of these are not suitable for metabolomics due to aqueous solubility concerns or chemical overlap. Chemical shift reference compounds, such as TSP (trimethylsilylpropionic acid) and DSS (trimethylsilylpropanesulfonic acid) have been used as internal standards for absolute quantitation of metabolites. It was realized some years ago, however, that these compounds are unsuitable for quantitation owing to their peak suppression arising from the interaction with proteins. One alternative, formic acid, was evaluated as an alternative to TSP for quantitation of metabolites in intact serum many years ago (Kriat et al. 1992). However, formic acid is unsuitable as a reference since it is an endogenous metabolite; the endogenous concentration in serum varies significantly from person to person (~ 40 to 350 μM) and hence, it interferes with externally added formic acid (Kubáň & Boček 2013; d’Alessandro et al. 1994; Kapur et al. 2007). In another study, DSA (4,4-dimethyl-4-silapentane-1-ammonium trifluoroacetate), which is a derivative of DSS, was evaluated as a potential internal standard using intact rat plasma (Alum et al. 2008). However, it is also unsuitable as a reliable internal standard since several factors including the increased line broadening by a factor of > 2 at pH 7.4 relative to pH 3.0, indicate that DSA interact with sample matrix. One remedy for analysis of metabolites in samples such as blood serum/plasma that contain copious macromolecules is to removing macromolecules effectively by ultrafiltration; in such a case, TSP or DSS, can still be used as standards for absolute quantitation (Psychogios et al. 2011; Barding et al. 2012; Simón-Manso et al. 2013). The challenge with ultrafiltration, however, is that it attenuates many metabolite peaks (Nagana Gowda & Raftery 2014), requires larger sample volumes, and is cumbersome for large-scale studies. In addition, ultrafiltration cannot be used for analysis of samples such as tissue and whole blood. Ultrafiltration is also incompatible with MS analysis, the other major analytical platform used in metabolomics, since MS analysis invariably employs protein precipitation to remove macromolecules, prior to analysis (Nagana Gowda et al. 2018a).

Protein precipitation that removes macromolecules from samples provides an alternative approach to quantitate metabolites and is well suited for large-scale studies. However, even in such samples, peaks from the traditional internal standards, TSP and DSS, are attenuated by up to 35% and hence they are unsuitable as internal standards. More recently, two compounds, maleic acid and fumaric acid, were evaluated for their utility as potential internal standards for quantitation of metabolites since both provide a single peak in NMR spectrum and their peaks do not overlap with peaks from bio-specimen spectra (Nagana Gowda et al. 2021) (Figure 1). It was shown that fumaric acid is a robust standard for protein precipitated serum, plasma and whole blood; and maleic acid is suitable for plasma and serum, but it overlaps with coenzyme peaks in whole blood samples. These findings provide new opportunities for improved and accurate quantitation of metabolites in human plasma, serum and whole blood using NMR spectroscopy. The potential utility of maleic acid and fumaric acid as internal standards may be extended to other biological specimens, as long as they do not overlap with bio-specimen peaks.

Figure 1.

Typical 800 MHz ¹H NMR spectra of a protein precipitated sample of human (a) whole blood, (b) plasma and (c) serum, solubilized in D₂O buffer containing a mixture of three internal standards (TSP, 238 μM; maleic acid, 350 μM; fumaric acid, 293 μM). Each spectrum is overlaid with a spectrum from the blank buffer consisting of the same three standards (spectrum shown in red) to enable the visualization of peak heights for the three internal standards; the spectra of the bio-specimens are slightly right shifted relative to the blank spectrum for clarity. Peaks from the blank are marked with asterisks. Heights for the fumaric acid peaks from the bio-specimen and blank are approximately matched; however, a significant attenuation of the TSP peak in all three bio-specimens spectra is noticeable. TSP: Trimethylsilylpropionic acid-d₄. $ indicates residual water signal from the buffer solution (reproduced with permission from Nagana Gowda et al. 2021).

2.2. Alternative Reference Standards for Absolute Quantitation:

An altogether different approach is to determine metabolite concentrations without the need for an internal standard. One such method is ERETIC (electronic reference to access in vivo concentrations) (Akoka & Barantin 1999). In the ERETIC method, a synthetic signal is generated in NMR spectra with the desired peak intensity, line width and chemical shift. This peak is then calibrated and used as a reference for quantitation. A drawback of this method is that the quantitation error can be large when NMR peaks are attenuated due to altered pulse widths arising, for example, from lossy biological solutions. More recently, a method known as PULCON (pulse length based concentration determination) alleviates the limitation of ERETIC and promises a robust approach to quantitation without the need for an internal standard (Wider & Dreier 2006). PULCON, also known as ERETIC 2, works based on the principle of reciprocity (Hoult 1976, 2000, Van der Klink, 2001) and allows the correlation of signal strength from a reference spectrum with the spectrum of interest. This method shows immense promise for metabolomics applications (Jiménez et al. 2018; Goldoni et al. 2016). The method, however, requires that reference and test spectra are obtained using the same probe, probe tuning and matching for the test samples should be identical to the reference sample, and the same RF power needs to be delivered to the coil for each NMR spectrum. Poor probe matching for salty test samples, for example, will lead to inaccurate quantitation of metabolites. Despite the limitation, however, by using standardized operating procedures quantitative data can be obtained that can potentially enable sharing of inter-laboratory results.

Another approach for quantitation is to use the solvent signal as a concentration reference (Mo & Raftery 2008). Most solvents can be observed by NMR and solvent concentrations can be readily determined independently. In particular, a widely used solvent such as water can serve as a primary concentration standard for metabolite quantitation. The potential problems of radiation damping associated with a strong NMR signal can be alleviated by small pulse angle excitation. The fact that solvent signal can be detected by the NMR receiver with the same efficiency as analytes enables their accurate quantitation. It is shown by this approach that analyte concentration can be accurately determined from 4 µM to more than 100 M.

2.3. Quantitation of Metabolites Using Intact Samples

The ability to analyze intact samples with no need for sample preparation or separation using NMR is an important characteristic that continues to drive NMR-based metabolomics. Initially, widely used bio-specimens including blood serum and plasma were used only in their intact forms and this approach continues to be widely used. In the following sections, analyses of intact bio-specimens that are most widely used such as serum/plasma, urine and tissue are described. The methods presented here are also applicable for other specimen types.

2.3.1. Intact Serum and Plasma Analysis:

Analysis of intact serum or plasma enables quantitation of aqueous metabolites as well as lipids and various classes of lipoproteins in serum and plasma (Würtz et al. 2017). Two widely used, one-dimensional (1D) NMR pulse techniques are 1D NOESY (nuclear Overhauser enhancement spectroscopy) and CPMG (Carr-Purcell-Meiboom-Gill) with water signal suppression (often using presaturation) (Nicholson et al. 1995). The 1D NOESY detects both small molecules such as metabolites as well as macromolecules such as lipids and lipoproteins. On the other hand, the CPMG experiment detects only small molecules; the unwanted macromolecule signals from proteins and lipoproteins are suppressed based on T₂ (transverse relaxation) filter (Beckonert et al. 2007); metabolites exhibit longer T₂ relaxation times compared to macromolecules and hence they are selectively retained in the CPMG spectra. Numerous large-scale epidemiological studies have demonstrated quantitation of 50–70 metabolite peaks and over 200 metabolic measures (which include ratios of metabolite peaks) on a routine basis (Soininen et al. 2015; Würtz et al. 2017). As described above, recent advances in NMR enable absolute quantitation using external reference, with no need for an internal standard (Wider & Dreier 2006). This is remarkable considering that internal standards largely cannot be used for absolute quantitation since they interact with copious proteins present in the samples. However, a challenge for reliable analysis of metabolites in intact samples is that metabolite binding to proteins causes signal attenuation (Nicholson and Gartland 1989, Chatham and Forder, 1999; Bell et al. 1988, Nagana Gowda and Raftery 2014). Moreover, exchange of metabolites between free and protein bound forms results in broader NMR peaks. Further, residual macromolecules signals cause distorted spectral baseline in CPMG spectra, which, together adversely affect metabolite quantitation.

2.3.2. Intact Urine Analysis:

Urine provides a rich source of information as it contains a significantly higher number of detectable metabolites, compared to serum/plasma, with a vast concentration range (~10⁶). In addition, urine has a relatively low concentration of proteins and hence macromolecular interference is minimal for metabolite analysis. A step-by-step procedure for NMR analysis of urine is provided as a guide for routine applications (Beckonert et al. 2007; Emwas et al. 2016). The pH of normal human urine varies widely, from approximately 5 to 8 (Hernandez et al. 2001; Rylander et al. 2006; Welch et al. 2008) and the salt concentration also varies significantly from sample to sample. Such pH and salt concentration variations alter chemical shifts of many peaks in the urine NMR spectra. Such peak shifts are significant for metabolites with functional groups with pKa’s near the physiological pH. This causes a challenge for peak identification, comparison of different spectra and quantitation of metabolites. Therefore, urine samples are generally mixed with buffer solution typically in a 1:1 (v/v) ratio (at pH = 7.4). Using Chenomx software and authentic compound spiking, > 200 metabolites in urine have been identified (Bouatra et al. 2013). However, considering the high complexity of the urine NMR spectrum and the sensitivity of chemical shifts to factors such as pH and salt concentration, the number of metabolites that can be analysed on a routine basis is restricted to ~ 60 to 70. Factors such as diet, medications, physical activity, smoking, gender, age, gut microbe diversity greatly affect the metabolome and they should be carefully accounted for disease biomarker identification (Emwas et al. 2015, 2016). Large-scale (>1000 samples) high-throughput studies now enable quantitative analysis of urinary metabolites using automated or semi-automated regression-based spectral analysis (Tynkkynen et al. 2019). In a large and impressive study, it was shown that prediction of metabolite concentrations, including many invisible inorganic ions, could be made based on the interrelationships between chemical shifts and concentrations, for automated urine analysis (Takis et al. 2017). Such advances promise new applications to areas including clinical, epidemiological, and pharmaceutical research.

2.3.3. Intact Tissue Analysis:

NMR spectra of intact tissue are obtained using high resolution magic angle spinning (HR-MAS) techniques (Tilgner et al. 2019). HR-MAS provides highly resolved spectra, which are comparable to those of bio-fluids. Tissue specimens typically collected from a surgical procedure or biopsy are often snap-frozen and stored for later analyses. The use of fresh samples for direct analysis, however, is advantageous for sensitive and structurally delicate biopsy samples. Resected or biopsied tissue is washed by quickly rinsing, typically with D₂O, to remove any blood contamination prior to freezing or direct analysis. The use of fresh samples avoids any deleterious effects caused by the freeze/thaw process, and protects tissue integrity. Care should, however, be exercised to ensure fresh samples specifically from biopsy are kept under cold and humid conditions until the analyses are performed to retain the integrity of the metabolite profiles and reduce the possibility of metabolic changes. The ability to recover tissue after NMR analysis provides an opportunity to use the same specimens for other studies such as proteomic and genomic analysis or even histology. Advances in probe technologies with a ²H field-frequency lock channel and a magnetic field gradient coil offer spectral stability and resolution sufficient for routine metabolomics studies of tissue samples as small as a few ng (Wong et al. 2012). Such capabilities, combined with minimal sample preparation and fast data acquisition, promise to extend the application of metabolic profiling of biopsied tissue to clinical applications. As examples, studies have shown that HR-MAS NMR of core needle biopsy tissue can predict breast tumor aggressiveness prior to surgery (Choi et al. 2012). Tissue metabolite profiles offer numerous benefits owing to the close association of tissue with disease pathologies. For example, alteration in tissue metabolite profiles has been shown to differentiate breast cancer tumors from normal tissue (Paul et al. 2018; Sitter et al. 2010). Importantly, HR-MAS NMR potentially enables diagnosis, prognosis and staging of cancers (Dinges et al. 2019; Chen et al. 2017).

2.4. Metabolite Quantitation Using Processed Samples.

Sample processing involves separation of metabolites from the macromolecular matrix. Such an approach enables detection of a significantly expanded pool of metabolites. A number of sample processing methods exist, to date, focused on analysis of aqueous metabolites or lipids or both. Further, new methods are being continuously developed that focus on improving the extraction efficiency, preserving metabolite integrity and simplifying the extraction process.

2.4.1. Analysis of Aqueous Metabolites:

Methods such as ultra-filtration, solid phase extraction and protein precipitation using organic solvents such as methanol, acetonitrile, acetone, perchloric acid or trichloroacetic acid have been explored for many years to extract metabolites (Wevers et al 1994; Daykin et al. 2002; Tiziani et al. 2008; Fan 2012). Among them, ultrafiltration using low molecular weight (~3kDa) cut-off filter removes proteins most effectively. Using this method, nearly 50 aqueous metabolites could be identified and quantified (Psychogios et al. 2011). However, ultrafiltration attenuates many metabolite peaks (Nagana Gowda & Raftery, 2014), requires larger sample volumes, and is particularly cumbersome for large-scale studies. Nearly half of the detected metabolites in ultra-filtered serum exhibited lower concentrations ranging from nearly 10 to 75% (Nagana Gowda & Raftery, 2014). Further, ultra-filtration is incompatible with analysis of samples such as whole blood and tissue as well as with analysis using mass spectrometry, which generally employs protein precipitation using organic solvents (Nagana Gowda et al. 2018a).

Detailed studies have focused on increasing the number of detected metabolites, identifying unknown metabolites and optimizing their quantitation in blood serum and plasma (Nagana Gowda & Raftery, 2014; Nagana Gowda et al. 2015). These studies have shown that protein precipitation using methanol in a 2:1 ratio (v/v) with the sample offers an optimal approach for analysis of aqueous metabolites in blood serum/plasma. The use of acetonitrile for protein precipitation, on the other hand, revealed a surprisingly poor performance; one-third of the detected metabolites were attenuated by up to 70% compared to methanol precipitation at the same solvent to serum ratio of 2:1 (v/v) (Figure 2). A further attenuation of nearly two-third of the metabolites was observed for acetonitrile to serum ratio of 4:1 (v/v). As the analysis of metabolites using mass spectrometry invariably employs protein precipitation prior to analysis, methods developed for NMR analysis also help analysis using mass spectrometry. The performance of sample processing for MS analysis is typically evaluated using the total number of ions detected, which is problematic (Ivanisevic et al. 2013), and is an inaccurate approach as far as quantitation is concerned.

Figure 2:

Comparison of absolute concentrations (in μM) of metabolites detected in pooled human blood serum and quantitated using 800 MHz NMR spectroscopy after protein precipitation using methanol (MeOH) (a, b, c and d) or acetonitrile (ACN) (e, f, g and h) at a solvent to serum ratios of 2:1, 3:1 and 4:1. Methanol performs most optimally over a wide range and a methanol to serum ratio of 2:1 provides the best performance (reproduced with permission from Nagana Gowda et al. 2015).

Protein precipitation, however, does not remove macromolecules completely and the residual macromolecules (~2%) are water-soluble, which cause broad baselines in NMR spectra when obtained using the one-pulse or 1D NOESY pulse sequence (Nagana Gowda et al. 2015, 2021). The use of the CPMG sequence helps to suppress signals from these residual proteins (~2%) and provides a flat baseline. The CPMG sequence, however, causes a small attenuation for many signals due to differential T₂ relaxation rates; for example, an evaluation of 20 metabolite peaks revealed an average of ~ 6% attenuation for plasma and serum when a 32 ms CPMG 180° echo pulse train was used. The peak attenuation increased with increasing duration of the echo pulse train and it exceeded 10% for 256 ms echo pulse train. Hence, for accurate quantitation, signal attenuation due to T₂ relaxation in the CPMG spectra should be carefully accounted. Potential alternatives to the CPMG sequence, including the use of stimulated echo (STE) pulse sequence (Lucas et al. 2005) have proved unsuitable for metabolomics applications.

2.4.2. Analysis of Coenzymes and Antioxidants:

Coenzymes, including coenzyme A (CoA), acetyl coenzyme A (acetyl-CoA), coenzymes of redox reactions and energy, and antioxidants mediate biochemical reactions fundamental to the functioning of all living cells. The most common redox coenzymes include NAD⁺ (oxidized nicotinamide adenine dinucleotide), NADH (reduced nicotinamide adenine dinucleotide), NADP⁺ (oxidized nicotinamide adenine dinucleotide phosphate) and NADPH (reduced nicotinamide adenine dinucleotide phosphate). The coenzymes of energy include ATP (adenosine triphosphate), ADP (adenosine diphosphate) and AMP (adenosine monophosphate). Major antioxidants include GSSG (oxidized glutathione) and GSH (reduced glutathione). Conventional enzymatic assays are suboptimal, as separate protocols are needed for analysis of each coenzyme or their ratios. The interference from sample matrix and the finite linear range of these assays further add to the challenges. Although mass spectrometry is extensively used, ion suppression, interference due to the unit mass difference in targeted analysis, and in-source fragmentation pose challenges for reliable coenzyme analysis (Evans et al. 2010; Trammell & Brennera 2013). Hence, the ability to analyze these coenzymes in one-step using NMR represents an important advancement in the metabolomics field. A major challenge unconnected with any analytical method, however, is the notoriously unstable nature of these compounds. Enzyme activity and oxidation affects their levels, deleteriously. Somewhat recently, sample harvesting, processing and analysis conditions were optimized for heart tissue from mouse models and first showed that a simple NMR experiment can simultaneously measure NAD⁺, NADH, NADP⁺, NADPH, ATP, ADP and AMP in one step apart from other metabolites (Nagana Gowda et al. 2016, 2018). Later, the scope of NMR was extended to the analysis of CoA, acetyl CoA and antioxidants (GSH, GSSG) along with a large pool of other metabolites and coenzymes, in one step (Nagana Gowda et al. 2019) (Figure 3). Further, as an important alternative to serum/plasma metabolomics, it was shown that using whole blood, the coenzymes and antioxidants can be measured simultaneously in addition to the nearly 70 metabolites that can be quantitated in serum/plasma with essentially no additional effort (Nagana Gowda & Raftery, 2017). The analysis protocols and the annotated characteristic fingerprints for these newly identified coenzymes and other metabolites are provided for easy identification and absolute quantification using a single internal reference. The ability to measure the unstable but ubiquitous coenzymes fundamental to cellular functions, simultaneously and reliably, offers a new avenue to investigate the mechanistic details of cellular function in health and diseases.

Figure 3.

(a) Typical 800 MHz ¹H NMR spectrum of a mouse heart tissue extract with labeling of some of the metabolites: BCCA: branched chain amino acids; TSP: reference peak; (b-e) expanded spectral regions highlighting characteristic peaks for (b) coenzyme A (CoA), acetyl coenzyme A (acetyl-CoA) and coenzyme A glutathione disulfide (CoA-S-S-G); (c) CoA, acetyl-CoA , oxidized nicotinamide adenine dinucleotide (NAD⁺), oxidized nicotinamide adenine dinucleotide phosphate (NADP⁺), reduced nicotinamide adenine dinucleotide (NADH), reduced nicotinamide adenine dinucleotide phosphate (NADPH), adenosine triphosphate (ATP), adenosine diphosphate (ADP) and adenosine monophosphate (AMP); (d) reduced glutathione (GSH) and oxidized glutathione (GSSG); and (e) creatine (Cr) and phosphocreatine (PCr) (reproduced with permission from Nagana Gowda et al. 2019).

2.4.3. Analysis of Lipids:

NMR spectroscopy is widely used for analysis of lipids and lipoprotein particles in serum and plasma (Mallol et al. 2013). Identification and quantitation of lipoprotein particles by NMR exploits the characteristic chemical shifts of the methyl resonances of fatty acid chains of lipids from different particle sizes, with peaks from smaller particles appearing at lower frequencies. Methodologies used to characterize lipoprotein particles based on methyl resonances utilize either deconvolution (Jeyarajah et al. 2006, Kaess et al. 2008) or statistical (Soininen et al. 2009) methods. These methods have enabled determination of particle size and number for lipoprotein classes such as VLDL (very low-density lipoprotein), LDL (low-density lipoprotein) and HDL (high-density lipoprotein) and up to fourteen (or more) lipoprotein subclasses. The ability to quantitate a variety of lipoprotein particles using NMR has opened avenues for clinical assessment and management of cardiovascular disease risk. In view of the fact that such lipoprotein classification and sub-classification using NMR is superior to the conventional methods, the method has been commercialized to manage the risk of heart diseases. Somewhat recently, a diffusion based method was proposed to characterizing lipoprotein particles (Mallol et al. 2015). Here, two-dimensional diffusion-ordered ¹H NMR spectroscopy (DOSY) was used to measure diffusion coefficients, which provide information on the particle sizes of lipoproteins (Johnson, 1999). The lipoprotein particle numbers are then calculated by dividing the peak volume by the size of lipoprotein particles. The ability to directly calculate lipoprotein sizes using the DOSY method was purported (Mallol et al. 2015) to provide a more accurate results for the particle numbers than the commercialized methods, which are based on 1D NMR.

After extraction, typically using a mixture of organic solvents, the analysis of tissue or blood samples provides quantitative information on individual lipids or lipid classes. The Folch extraction, consisting of chloroform/methanol/water in a volumetric ratio of 8:4:3 (v/v/v) is one of the earliest and most popular methods (Folch et al. 1957). Since then, numerous different lipid extraction protocols with modification to Folch ((Folch et al. 1957) or Bligh and Dyer method (Bligh & Dyer, 1959) have been proposed for biological specimens such as blood, tissue and cells. A more recent method, involving butanol-methanol (BUME), eliminates the need for chloroform, which is considered hazardous (Löfgren et al. 2012; Cruz et al. 2016). More recently, the BUME method was modified to suit the analysis of lipids using NMR spectroscopy (Barrilero et al. 2018). This method replaces heptane with diisopropylether as the organic solvent, since peaks from the residual heptane overlap with lipid signals. Notably, this method has enabled identification and quantitation of 15 different lipid classes including fatty acids, triglycerides, phospholipids, and cholesterols in serum. A semiautomatic software, LipSpin, converts raw NMR data based on mathematical and reference spectral models, and provides quantitative information on lipids (Barrilero et al. 2018). Detailed protocols for extraction and quantitative analysis of lipids in biological specimens such as serum, tissue and cells are provided, which serve as a practical guide for beginners in the field (Gil et al. 2019).

2.5. Quantitation Methods Using Stable Isotope Labeling

Stable isotope incorporation in vivo or ex vivo offers opportunities to quantitate metabolites using NMR with improved resolution and sensitivity. In vivo analysis of metabolites in live systems enables monitoring of dynamic changes, measuring fluxes and monitoring metabolism in real time. The use of heteronuclear 2D (two-dimensional) NMR pulse techniques involving stable isotopes offers a combination of selectivity, sensitivity and resolution, and alleviates major challenges in NMR experiments involving nuclei with low natural abundance. To date, stable isotopes including ¹³C, ¹⁵N, ²H and/or ³¹P have been employed for analysis of metabolites in biological mixtures and investigation of metabolic pathways.

2.5.1. Isotope Labeling Focused on Metabolic Fluxes and Pathways.

Isotope labeling in vivo enables measurement of fluxes and tracing of metabolic pathways. Using this approach, the same metabolite that flows through multiple pathways can be quantified. A growing number of pathways, including glycolysis, pentose phosphate pathway, glutaminolysis, fatty acid oxidation, and TCA cycle can be investigated using the combination of NMR and selective or uniformly isotope labeled substrates such as ¹³C-glucose and ¹³C/¹⁵N-glutamine (Lin et al. 2019). Quantitative analysis of in vivo isotope labeled metabolites can be measured either ex vivo, after extraction of metabolites, or in live systems in vivo. While analysis after extraction provides a snapshot of metabolite levels at a particular time point, in situ analysis using live systems enables the measurement of the dynamic changes in metabolite levels and monitoring of metabolism in real time. Analysis after extraction of metabolites has been widely used in the metabolomics field. However, the growing technological and methodological advances in NMR are witnessing increasing number of in vivo investigations using live systems such as C. elegans, cells and isolated mitochondria (Nguyen et al. 2020; Wen et al. 2015; Xu et al. 2018). Isotope labeled studies using cells and subcellular organelles enable understanding of metabolic pathways under controlled conditions. And the use of organisms, animal models or humans can translate the findings from studies of cells and subcellular organelles to investigate the pathogenesis of human diseases (Fan et al. 2009; 2011; Locasale et al. 2011; Lane et al 2011).

2.5.2. Isotope Labeling Focused on Metabolite Analysis.

Isotope labeling in vivo in plants and organisms such as bacteria and yeast offer significant enhancement to spectral resolution and the detection sensitivity (Zhang et al. 2012; Chikayama et al. 2008; Bingol et al. 2013; Bingol et al. 2012). In particular, it alleviates the challenges invariably met with the analysis involving low natural abundance heteronuclei and enables analysis of a large number of metabolites using conventional high-resolution 2D NMR experiments such as HSQC and HMBC. The uniform labeling using nuclei such as ¹³C also enables characterization of metabolites based on homonuclear 2D ¹³C NMR experiments. Carbon-bond topology networks obtainable from such homonuclear 2D ¹³C experiments provide additional avenues for metabolite identification (Chikayama et al. 2008; Bingol et al. 2012).

An altogether different approach is to label different classes of metabolites based on the specific functional group (Shanaiah et al. 2007; Desilva et al 2009; Ye et al. 2009). Chemical derivatization of metabolites using a substrate that contains isotope such ¹³C, ¹⁵N or ³¹P offers both sensitivity and resolution enhancement, owing to the high isotopic abundance and wide chemical shift dispersion imparted by the incorporated isotope. The ¹H decoupled 1D or 2D NMR spectrum involving the isotope labeled heteronuclei provides a single peak for each metabolite, which further adds to the sensitivity and resolution. Metabolite classes including amines, carboxylic acids and hydroxyls have thus been tagged with isotopes and analyzed using 1D or 2D NMR (Shanaiah et al. 2007; Desilva et al 2009; Ye et al. 2009). Owing to its high natural abundance, ³¹P signals from metabolites, however, can show up as strong background peaks in the ³¹P enriched experiments, unlike the other nuclei. Incorporation of a “smart isotope tag” such as ¹⁵N-cholamine, enables analysis of carboxylic acid class of metabolites using both NMR and MS methods (Tayyari et al. 2013). The smart isotope tag possesses an NMR sensitive isotope (¹⁵N) that offers good chemical shift dispersion, and a permanent positive charge that improves MS sensitivity and enables quantitation of metabolites more accurately by both NMR and MS. Such analysis allows direct comparison of NMR and MS data, which is an important characteristic for biomarker discovery and biological interpretation in the metabolomics field.

3. Conclusion

The ability to identify unknown metabolites, absolute quantitation and analysis of intact bio-specimens including live cells and subcellular organelles, is expanding the application of NMR to new and exciting areas. Technological advances have provided significant improvements to sensitivity and resolution, which have led to the identification and quantitation of an expanded pool of metabolites. NMR spectroscopy offers opportunities to gain mechanistic insights into biochemical pathways in health and diseases, to discover biomarkers and potential therapy targets, and to translate laboratory findings to clinical applications. Continuing, multifaceted efforts to boost sensitivity, resolution and the speed of data acquisition and to improve quantitative accuracy promise to alleviate the increasingly realized complexity of biological mixtures and large-scale metabolomics studies. Moreover, ongoing technical and methodological advances contribute to further expanding the routinely quantifiable metabolites in biological specimens and hence the NMR-based metabolomics is anticipated to greatly improve and impact our understanding of systems biology and to help make progress in the treatment and management of human diseases.