Use of nuclear magnetic resonance spectroscopy in diagnosis of inborn errors of metabolism

Abstract

Nuclear Magnetic Resonance (NMR) spectroscopy has been applied in many fields of science and is increasingly being considered as a tool in the clinical setting. This review examines its application for diagnosis of inborn errors of metabolism (IEMs). IEMs, whether involving deficiency in the synthesis and degradation of metabolites, or in lipoprotein metabolism, affect nearly 3% of the global population. NMR is a preferred method for comprehensive evaluation of complex biofluids such as blood or urine, as it can provide a relatively unbiased overview of all compounds that are present and does not destroy or otherwise chemically alter the sample. While current newborn screening programs take advantage of other more sensitive methods, such as mass spectrometry, NMR has advantages especially for urine analysis with respect to ease of sample preparation and the reproducibility of results. NMR spectroscopy is particularly compatible with analysis of lipoproteins because it provides information about their size and density, not easily attained by other methods, that can help the clinician to better manage patients with dyslipidemia. We believe that NMR holds great potential for expanding clinical diagnosis in the future, in the field of IEMs and beyond.

Introduction

Inborn errors of metabolism (IEMs) are a group of conditions typically caused by a single gene disorder yielding a disruption to a metabolic pathway. This heterogeneous group of disorders reflects a deficiency in the conversion of one molecule into another and includes organic acidemias, urea cycle disorders, aminoacidopathies, fatty acid oxidation disorders, and mitochondrial disorders [1]. A second group involving lipoprotein metabolism (IELMs) is a failure in proper interconversion of lipoproteins, such as low-density lipoprotein (LDL), leading to high serum cholesterol. There are over 1000 known IEMs, each with a low prevalence on its own, but as a group are estimated to occur in 1 out of 800 live births [2,3]. Specific countries have higher rates; Turkey, for example, sees a prevalence of IEMs closer to 1 out of 400 newborns, partially due to consanguineous marriages [4]. Most untreated IEMs present with neurologic symptoms and seizures in newborns shortly after birth, within hours to a few days [5]. If left untreated, many of these can quickly lead to intellectual or developmental disability, and eventually to death [6]. Many IEMs are treatable if diagnosed early and may be detected prior to the infant showing symptoms, thus countries around the world have implemented newborn screening programs. The inborn errors of lipid metabolism (IELM) do not present with neurological symptoms in infants, but with cardiovascular disease later in life. Heterozygous familial hypercholesterolemia (FH), which causes an accumulation of cellular damage over time yielding an increased risk of early atherosclerotic disease, is more common than any IEM as it presents in approximately one in 250 people [3,7]. Although the molecular defects are heterogeneous, combined familial hyperlipidemia is even more prevalent, affecting 1–3% of the population [8]. Dyslipidemias do not severely impact the infant’s health nor are they treated until the child is older [9–11]. Given dyslipidemias do not necessitate intervention in the period immediately after birth, IELMs are not included in general newborn screening programs and are additional to the IEMs of infants [12].

Inborn errors of metabolism and metabolomics

Each inborn error is caused by a genetic defect, most commonly in a gene encoding an enzyme, receptor or transport protein. The disruption of the biochemical pathway yields either an excess or deficiency of a specific metabolite, which may serve as a biomarker [13]. For example, phenylketonuria (PKU), is caused by a defect in one of two enzymes necessary for the hydroxylation of phenylalanine to tyrosine in the liver. A defect in either phenylalanine hydroxylase or dihydrobiopterin reductase, which recycles the reaction’s cofactor, causes a buildup of phenylalanine and may cause a deficiency of tyrosine [14]. Excess phenylalanine will via a different enzymatic pathway be converted to phenylpyruvic acid and other related compounds, which in excess are associated with cognitive impairment. Broadly, abnormalities in the concentration of a specific metabolite can be indicative of a specific inborn error. For the past few decades, newborn screening programs have analyzed dried blood spots collected from infants (heel pricks) using tandem mass spectrometry (MS/MS) to screen for abnormal levels of specific metabolites [6]. Metabolomics, defined as the identification and quantification of all the metabolites in a system, is an ideal method for studying IEMs.

In more recent years, metabolomics based on Nuclear Magnetic Resonance Spectroscopy (NMR) has emerged as a method for analyzing metabolites in body fluids [15]. The most common form based on the ¹H (proton) nucleus (¹H-NMR), can be used to identify and quantify small molecule components of the complex mixtures associated with clinical samples, such as blood or urine. More recently, body fluids such as saliva and sweat have shown great potential as well [16–19]. Although other nuclei have been studied, most notably ¹³C and ³¹P, ¹H provides the highest sensitivity. The NMR technique utilizes transitions between different states of the ¹H nucleus when placed in an external magnetic field. The ¹H nuclei within a molecule have different resonance frequencies due to the local environmental differences such as electron density within the molecule. A one-dimensional (1D) NMR spectrum consists of a graph in which each ¹H nucleus of the molecule provides a peak that may be singular (a singlet) or complex in shape (multiplet). The characteristic splitting relates to the fine structure of the molecule. Protons attached to electronegative atoms may also possess a broadened width as well, typically relating to exchange processes with the bulk solvent or relaxation properties of the attached heteroatom [20].The x-axis shows the frequency of resonance of each nucleus. The resonance frequency is expressed in ppm (parts-per-million) of the overall magnetic field, to allow for comparison between spectra collected at different external magnetic field strengths on different instruments. The height of each peak on the y-axis is directly proportional to the concentration of that nucleus in the sample. A molecule will give a characteristic spectrum; alanine, ${\mathrm{NH}}_3^{+}\text{-}\mathrm{CH}(\text{-}{\mathrm{CH}}_3)\text{-}{\mathrm{COO}}^{-}$, for example, shows no peaks in aqueous solution for the amino group due to rapid exchange, a quartet for the alpha CH group with heights of 1 : 3 : 3 : 1 at 3.77 ppm, and a doublet at 1.47 ppm that reflects contributions by all three hydrogen atoms of the methyl group. The spectra are very reproducible from laboratory to laboratory given the same pH and ionic strength of the medium (typically buffered with 50 mM phosphate at pH 7). Thus, the spectrum of an authentic compound collected in one laboratory may be used as a reference spectrum in another. Lists of metabolites and their concentrations are obtained from either chemometric or deconvolution analysis of the ¹H-NMR spectrum (Figure 1) [21]. In the chemometric approach, the spectra are aligned to account to small changes in peak position due to changes in pH or ionic strength and the spectral intensities are summed in bins typically of 0.01 to 0.05 ppm in width [22]. The spectral vectors that change as a result of the biological perturbation are identified, then the metabolite or metabolites that contribute to the bins are identified a posteriori. In the deconvolution method, which is objectively superior but more time consuming, directly delivers quantitative metabolite data relative to the concentration of a standard (such as Sodium trimethylsilylpropanesulfonate, DSS), that is spiked into the sample at a known concentration [23].

Figure 1. Workflow for the use of NMR in the diagnosis of inborn errors of metabolism.

A biofluid, such as urine, is obtained from the human subject. The sample is prepared for NMR spectroscopy and the data obtained. The NMR spectra are analyzed for metabolite composition and concentration. A graph is generated on the basis of healthy subjects to understand the normal distribution of any particular metabolite (metabolite x). In the case of the urine, the concentration is normalized to the creatinine level, to control for dilution effects. An individual with an abnormal measurement that deviates significantly from the median of the standard curve would be flagged for follow-up.

Currently, newborn screening programs for IEMs utilize MS/MS, however there are a number of advantages provided by use of ¹H-NMR spectroscopy over MS methods in metabolomics. The most important advantage is that NMR methods are less biased and particularly useful for identification of unknown compounds. While MS metabolite identification of unknowns has in recent years improved with developments in techniques, labeling schemes and extensiveness of databases, NMR methods provide an overview of all metabolites present and allow for the nearly complete identification of unknown metabolites [22,24,25]. NMR is also quantitative without the use of multiple standards, requires very little sample preparation, is non-destructive and highly reproducible [13,24,26]. ¹H-NMR spectroscopy is particularly well-suited for use in newborn screening programs as it is the preferred method for urine analysis as compared with other methods in terms of the number of metabolites covered and quantified [27]. Collecting urine from newborns is non-invasive and repeated measures can easily be taken over a short period of time to monitor increasing concentrations of a toxic metabolite or response to treatment. Thus, many ¹H-NMR studies assessing IEMs in newborns have measured metabolites in urine as opposed to blood.

¹H-NMR-based metabolomics and diagnosis of IEMs

¹H-NMR spectroscopy of body fluids has long been used as a complement to MS-based methods in the diagnosis of IEMs [4,15,28]. For example, PKU, maple syrup urine disease, propionic aciduria, and methylmalonic aciduria, among many others, have been identified via ¹H-NMR spectroscopy of blood and urine [5,29,30]. More recently, various groups have looked into how NMR spectroscopy could be applied on a larger scale to screen for IEMs. A 2014 study by Aygen et al. looked at ¹H-NMR spectra of urine samples from nearly 1000 healthy neonates in Turkey [4]. Accounting for the metabolic fingerprints of gender, birth mode, day of life, and other variables known to influence a person’s metabolome, they characterized the urine spectra of healthy neonates [13]. They proposed that the established metabolite distributions could be used to devise a standardized procedure for IEM screening in Turkey; with a baseline for normal, diseased neonates that could easily be identified from analysis of their ¹H-NMR spectra (Figure 2).

Figure 2. Atypical ¹H NMR spectrum of human urine with identified metabolites.

The numbers refer to metabolites as follows: 1, valine; 2, 3-hydroxyisobutyrate; 3, 4-deoxyerythronic acid (4-DEA); 4, 3-aminoisobutyrate; 5, 4-deoxythreonic acid (4-DTA); 6, 3-hydroxyisovalerate; 7, threonine or lactate; 8, alpha-hydroxyisobutyrate; 9, alanine; 10, acetate; 11, phenylacetylglutamine; 12, p-cresol sulfate; 13, citrate; 14, dimethylamine; 15, creatine; 16, creatinine; 17, proline or betaine; 18, carnitine; 19, trimethylamine-N-oxide (TMAO); 20, taurine; 21, glycine; 22, hippurate; 23, trigonelline; 24, ascorbate; 25, xylose; 26, allantoin; 27, urea; 28, 4-hydroxyphenylacetate; 29, tyrosine; 30, 3-indoxyl sulfate; 31, histidine; 32, formate; 33, trigonelline. Figure and legend adapted from Giskeødegård et al. [56] under a Creative Commons Attribution 4.0 International License https://creativecommons.org/licenses/by/4.0/.

This NMR method was then applied to derive the normal metabolic profile of a different group of newborns (in Basque Country, Spain) [5]. They used ¹H-NMR spectra of urine samples to create probabilistic models for concentrations of 150 different metabolites relative to creatinine. Creatinine, which is proportional to muscle mass, is produced and excreted at a relatively constant rate so it is commonly used as a urinary biomarker for normalization. Normalization to creatinine accounts for dilution of the sample due to different states of hydration across subjects. Significant deviations from the model of a given metabolite indicates strong suspicion of an inborn error of metabolism; these experimentally derived models could be used to implement an automated newborn screening routine.

Despite ample evidence of the efficacy of ¹H-NMR in screening for IEMs, it is still predominantly used as a research tool. The most significant drawback of ¹H-NMR as compared with MS/MS is its lower sensitivity, thus it is less suited to measurement of relatively low-abundance compounds that would be obtained from a blood spot from the heel prick used in newborn screening [31]. In addition, NMR instruments are usually more expensive than mass spectrometers (Table 1). ¹H-NMR is superior to MS/MS for untargeted analyses and the discovery of new biomarkers while MS/MS is the preferred method for conducting targeted analyses looking for specific metabolites [32]. The persistent usage of MS/MS and targeted analyses can be understood when considering the overarching goal of newborn screening for IEMs. The screening program recommendations around the world vary greatly. Many states in the U.S. screen for around 54 inborn errors while countries within the European Union screen for anywhere between 2–30 IEMs; there is a lack of consensus around the world [6]. Turkey, for example, only screens for 3 IEMs despite the higher prevalence of metabolic diseases [4]. If the purpose of screening all newborns for IEMs is to detect the treatable disorders before they cause permanent damage, not to discover novel diseases, it follows that MS/MS is the preferred clinical method. While there is certainly potential for the use of ¹H-NMR in screening programs, currently, the relatively lower sensitivity and higher cost of NMR instrumentation precludes widespread use at this time [33]. In vitro ¹H-NMR is, however, currently being used clinically to help understand cardiovascular disease (CVD) risk through lipoprotein profiling, a methodology with implications for management of certain IELM such as familial hypercholesterolemia (FH).

Table 1.

Advantages and limitations of NMR spectroscopy as compared with mass spectrometry

	NMR	MS
Detected metabolites	Universal (less biased)	Specific (more biased)
Metabolite Identification	Nearly complete	Incomplete
Sensitivity	Less sensitive	More sensitive
# of compounds quantified	Dozens to 200	Several hundred
Reproducibility	Higher	Lower
Quantitation standard	Only one needed	Standard for each metabolite
Sample preparation	Minimal	Involved
Sample integrity	Non-destructive	Destructive
Cost	More expensive	Less expensive

Clinical diagnostic Use of ¹H-NMR: lipoprotein analysis

¹H-NMR is currently in use clinically to help identify and manage patients at high risk of developing CVD [34]. LipoScience has developed Vantera, an FDA-approved automated clinical analyzer utilizing NMR technology, and a diagnostic LipoProfile test, which measures the amount of LDL particles (LDL-p) in a blood sample [34]. Similar platforms, such as Liposcale and Nightingale, have also been commercialized [35,36]. The Vantera clinical analyzer is being used around the United States to improve CVD risk assessments [34]. Traditional assessments of LDL and associated CVD risk measure the mass of LDL cholesterol (LDL-c) carried by the lipoproteins [37,38]. Traditionally, these detailed determinations of lipoprotein particle distributions rely on time-consuming (hours to days) separation-based methods such as ultracentrifugation or high-performance liquid chromatography (HPLC); the NMR method takes only minutes (Figure 3) [39]. Furthermore, LDL-c measurement does not capture variations in size among LDL particles, an important risk determinant, as the smaller LDL particles are more atherogenic than larger ones [40,41]. When CVD risk assessments utilizing LDL-c differ from those using LDL-p, LDL-attributable atherosclerotic risk and CVD events are better predicted by the latter [37,42,43]. As ¹H-NMR is non-destructive and preserves lipoprotein structure, it is an ideal method for evaluating the size and density of intact lipoproteins, LDL-p, as opposed to other platforms which generally only measure the molecular composition [34,39,44]. ¹H-NMR could also be used to inform clinical decision making about specific diseases yielding poor CVD outcomes, like FH and other cardiovascular diseases.

Figure 3. A typical ¹H-NMR spectrum of human serum (measured at 600 MHz).

The lower part of the figure shows the full spectrum whereas the top part shows an expansion between 1.55 and 0.6 ppm. Characteristic metabolites of low molecular mass, which show narrow signals, are labeled as well as the broader signals that arise from either the protein or lipids within lipoproteins. The ¹H nuclei corresponding to the signals are shown in red. The ¹H signals for lipids within lipoprotein tend to increase in chemical shift value with size of the lipoprotein [57].

Heterozygous familial hypercholesterolemia (FH) is present in approximately one out of 250 children and is a leading cause of accelerated atherosclerosis and premature CVD [7,40,45]. The most common cause of FH is a mutation in the LDL receptor (LDL-R) gene leading to reduced clearance of LDL and higher levels of serum cholesterol [3]. FH is considered an inborn error of metabolism, but its lack of impact early in life differentiates it from the diseases discussed previously and it is not included in newborn screening programs [3]. Elevated LDL from a young age can lead to premature atherosclerosis and greater risk of CVD [3,46]. FH is characterized by high LDL-c levels from a young age and is currently diagnosed either in the context of family history or by genetic testing [47]. Rodriguez-Borjabad et al. investigated the clinical value of lipoprotein particle counts and subclass breakdown, provided by ¹H-NMR assay, in CVD risk prediction in children with FH compared with controls [40]. They found total LDL-p count and small LDL particle numbers to be independent predictors of subclinical atherosclerosis in children with FH, indicating ¹H-NMR can provide useful clinical information [40]. The National Heart, Lung and Blood Institute recommends all children ages 9–11 undergo cholesterol screening, however this has not been the norm in practice widely [48,49]. More in-depth risk assessment can help clinicians decide when to initiate treatment for FH and assess which therapeutic options are indicated [38,46].

The use of in vitro NMR for metabolic analysis could be translated to a in vivo measurements using non-invasive magnetic resonance spectroscopy (MRS) analysis to monitor disease and treatment response [50]. MRS is an add-on to conventional Magnetic Resonance Imaging (MRI). While well-developed in the field of cancer, particularly in the brain [51], the method has also been applied in the setting of IEM [50]. Lactate has been considered a consequence of tissue injury, and elevated lactate is observed in several models of metabolic dysfunction [52].

NMR technology is rapidly developing. Recently, cryogenically cooled probes have boosted sensitivity two to four-fold and are now in widespread use. Although the biofluids covered in this review, urine and blood sera, are not normally limiting in volume, new microprobe designs have enabled the measurement of ten or more less-fold volume of a conventional 5 mm probe that requires a volume of ~0.5 ml [53]. Metabolic data are typically collected on instrument with a field strength that results in a ¹H resonance frequency of ~600 MHz (Figures 2 and 3). Higher field strengths provide several-fold increases in sensitivity, concomitant with greatly increased cost of instrumentation. Another noteworthy development is the use of dynamic nuclear polarization (DNP) technique which can boost the NMR signal several hundred-fold, particularly for ¹³C nuclei [54]. However, the technology is currently far from the accessibility needed for routine NMR analysis. A challenge that remains for both NMR and MS-based studies is the lack of standardization of protocols among laboratories, such as sample processing parameters, which would allow for a more accurate comparison between studies [22,55].

Conclusions

¹H-NMR spectroscopy is a research tool with great potential for clinical diagnosis of IEMs. It is a preferred method for comprehensive evaluation of complex biofluids such as blood and urine, as it can provide a relatively unbiased overview of all compounds that are present in a sample. Nearly all of the 1000 known IEMs are associated with increased or decreased levels of a specific metabolite as a result of the disruption to the metabolic pathway. ¹H-NMR can identify and quantify the relative concentration of any given metabolite present in a biofluid, and numerous studies have shown it can be used to identify and help diagnose IEMs. This methodology could be standardized, automated, and implemented in newborn screening programs; probability density functions for specific metabolites could be derived by measuring blood or urine concentrations across a population of newborns. Newborn biofluid samples could be processed using a standardized ¹H-NMR platform to identify the presence and relative concentration of metabolites of interest. If a metabolite indicative of an inborn error is seen in a concentration that deviates significantly from the population median, as determined by the probability density function, the sample is flagged and the newborn would undergo further testing (Figure 1). In the future, a standardized ¹H-NMR platform for IEMs may complement current mass spectrometry methods that are used for newborn screening programs.

A standardized, clinically available platform is available for the assessment of lipoproteins in a blood sample. ¹H-NMR spectroscopy is particularly compatible with lipoprotein analysis because it is non-destructive and provides information not easily attained by other methods about the size and density of lipoproteins. This information is used clinically to improve risk predictions of cardiovascular disease, and thus also has implications for evaluating certain IEMs which yield poor CVD outcomes. Patients with FH, for example, have increased LDL levels from birth and often have premature atherosclerosis and more severe CVD outcomes earlier in life. The ¹H-NMR lipoprotein assessment provides information that can be used to guide appropriate treatment for patients with a specific subset of inborn errors related to lipoprotein metabolism.

While the ¹H-NMR platform is not yet used diagnostically to evaluate IEMs in the clinic, it provides crucial information to understanding human health. We believe it holds great potential for clinical diagnosis in the future, in the field of IEMs and beyond.

Summary.

• ¹H-NMR spectroscopy has great potential for clinical diagnosis of IEMs.

• Newborn biofluid samples processed using a standardized ¹H-NMR platform to identify the presence and relative concentrations of metabolites of interest may provide crucial information to understanding human health.

• Adult blood samples can be assessed for lipoprotein content by standardized, clinically available ¹H-NMR platforms that provide data underlying cardiovascular disease.

Abbreviations

CVD

cardiovascular disease

FH

familial hypercholesterolemia

IELM

inborn errors of lipid metabolism

IEMs

inborn errors of metabolism

LDL

low-density lipoprotein

PKU

phenylketonuria