Metabolomics Needs Precision: The Case for Mass Spectrometry
Because of Metabolon’s long history with the science and technology of metabolomics, they’re often asked about the utility of various technologies. Martin Hornshaw, PhD compares two popular metabolomics technology platforms: NMR versus LC-MS.
Every chemist and biochemist is familiar with nuclear magnetic resonance (NMR). It is a long established, powerful technique dating back to the 1940s used to determine molecular identity and structure. I suspect that NMR is actually quite broadly known about. Perhaps you will have experienced NMR in a different way via magnetic resonance imaging (MRI), which is basically NMR applied to imaging. Indeed, MRI was first known as NMR Imaging, but at least partly due to the use of the word “nuclear” and patients’ concerns raised about that word, the nuclear in NMR Imaging was mostly dropped by the 1980s.
At this point I am going to admit to, and proudly express, a bias. During and since my PhD I have worked with mass spectrometry (MS; mass spec). It is a technique I know and love, and in my opinion it is the most powerful analytical technique currently in the armoury of the analytical scientist, particularly when coupled to liquid chromatography (LC).
It is very broadly used to analyze a whole range of samples such as water and food to check for chemical impurities such as pesticides; in the clinical lab for therapeutic drug monitoring; in sports anti-doping to catch sporting cheats who are enhancing their performance with drugs; in the clinical microbiology lab to identify infectious bacteria; in the pharmaceutical industry to look at how a drug is metabolized; and, in academia, particularly in life science and clinical research, to broadly analyze proteins with proteomics and metabolites with metabolomics. This is actually a very short and incomplete list of examples. The uses of mass spec are far greater than this.
There are areas where both NMR and MS may be applied. One such area is metabolomics, the comprehensive analysis of metabolites in a biological sample such as a tissue or cell culture. Most research conducted with metabolomics is carried out using a hybrid technology called liquid chromatography-mass spectrometry (LC-MS), but gas chromatography-mass spectrometry (GC-MS) and NMR are also used.
Here, I’ll limit my comparison to NMR and LC-MS. Both have their strengths and weaknesses, but I will give the game away somewhat and tell you now that I strongly favour LC-MS as the analytical tool of choice to be used for successful metabolomics studies.
Why do I prefer mass spec for metabolomics?
I can boil the reason down to two very key, very critical areas of analytical performance. Firstly, mass spec is sensitive while NMR is not. It’s that simple, and the difference between the two is as night and day. Indeed, mass spec is several orders of magnitude more sensitive than NMR. What is the consequence of that?
We come now to my second reason: in biological samples NMR detects only a handful of abundant metabolites, such as glucose and pyruvate, whereas an LC-MS approach, properly developed, as at Metabolon, will identify on the order of 1,000 metabolites in a sample such as blood plasma. If you also choose to study complex lipids, then more than 2,000 small molecule metabolites and lipids in total are typically identified. The approach developed by Metabolon based on LC-MS detects metabolites at a much lower concentration than an NMR-based approach. In addition, LC-MS-based approaches continue to improve, and it is foreseeable that in the next several years this number will be greatly exceeded.
To discover what is changing in metabolism, for example the differences between diseased samples relative to ‘normal’ (control) samples, often requires the maximum sensitivity that can be achieved. Even if a scientist thinks they know what they are looking for and that they already know what is of importance, they will often receive unexpected surprises that may elucidate the mechanism of disease or possibly identify biomarkers that may be of future clinical relevance diagnostically or prognostically. In addition, our current understanding of the human metabolome constituents, of say plasma, is incomplete. Mass spec is the only approach with a realistic possibility of identifying unknown metabolites present at a low level.
Examples of NMR and LC-MS in genome wide association studies
In genetics, a genome wide association study (GWAS) is an examination of a genome-wide set of genetic variants such as single-nucleotide polymorphisms (SNPs) to see if any variant (allele) is associated with a specific phenotypic trait, for example high blood pressure, or the aetiology of a disease. If one type of variant is particularly frequent in people with a disease, then the variant is said to be associated with that disease. These associated SNPs are then considered to mark a region of the human genome that may influence the risk for that disease. As usual, Wikipedia contains a useful summary.
A typical GWAS performed today might well involve tens of thousands of individuals, sometimes more, to increase the probability of finding associations. The reason for this large size is the drive toward detecting ‘risk-SNPs’ that have smaller odds ratios (a statistical measure of the strength of an association in a population) and lower allele frequency. In other words, these associations are difficult to identify with a typical GWAS unless a very large population cohort is studied, or so goes the general perception.
What if we add metabolomics to GWAS?
Could metabolomics enhance our ability to identify associations that might inform us about disease, for example, and enable the use of a smaller population cohort to identify significant associations? This could be very relevant to the study of rare or complex diseases or to reduce the cost associated with population health studies.
A number of studies have been performed looking for association of genes with metabolism, for example, to identify the metabolic function of genes and/or to uncover metabolic pathophysiology underlying established disease variants. For the purposes of comparison between NMR and LC-MS metabolomics approaches, two recent studies are worth describing, one using NMR, the other using LC-MS.
How do they do?
In 2016, Kettunen et al1 wrote, ”We conduct an extended genome-wide association study of genetic influences on 123 circulating metabolic traits quantified by nuclear magnetic resonance metabolomics from up to 24,925 individuals and identify eight novel loci for amino acids, pyruvate and fatty acids. The LPA locus link with cardiovascular risk exemplifies how detailed metabolic profiling may inform underlying aetiology via extensive associations with very-low-density lipoprotein and triglyceride metabolism.”
Picking that apart a little, 123 is actually a small number of ‘metabolic traits’ to study, and of that total, many were not actual small molecule metabolites but were, for example, lipoprotein particles. Amino acids, pyruvate and fatty acids are rather abundant species, and eight novel loci is a small number for nearly 25,000 individuals studied, as you will see by comparison with the paper published in 2017 in Nature Genetics by Long et al.2
“Many blood metabolites are intermediate phenotypes linking cellular functions to diseases,” wrote Long et al. “We conducted a whole genome sequencing study of common, low-frequency, and rare variants to map causal genetic variation for blood metabolite levels using comprehensive metabolite profiling in 1,960 individuals. We focused the analysis on 644 metabolites with consistent levels across three separate longitudinal data collections exhibiting high heritability (median h2 = 49%). Levels of 246 (38%) metabolites were associated with genetic sequence variation at 101 loci (p-value < 1.9×10-11). The genetic information supported the identification of unknown metabolites. In addition, we identified 175 individuals (113 unrelated, 10.7% of all unrelated individuals in the cohort) with rare variants likely influencing function of 17 genes, among which 35 individuals (26 unrelated, 2.5% of all unrelated) had metabolite levels beyond four standard deviations from the mean. Thirteen of the 17 genes are associated with inborn errors of metabolism or known pediatric genetic conditions. This study extends the map of loci influencing the metabolome and highlights the importance of heterozygous rare variants in determining abnormal blood metabolome phenotypes in adults.”
Far more metabolites were studied using Metabolon’s advanced Precision Metabolomics™ approach. Over the course of three measurements that were separated in time by years, LC-MS metabolomics measurements were stable. In addition, far more were associated with genetic sequence variation, some of which variation was rare.
A simple but telling measurement of the success of the two approaches is this: the NMR-based approach identified eight novel loci, while the LC-MS-based approach identified 48 loci associated to metabolites which were completely novel. There were even previously unknown metabolites identified using the LC-MS approach. Not all of the “metabolic traits” studied with NMR were actually measurements of metabolites. A last point worth strongly emphasizing is that the number of individuals necessary for the LC-MS metabolomics-GWAS was more than ten times fewer than in the NMR association study.
NMR has real benefits. The data generated by NMR is relatively straightforward; it quickly creates a single data file without any up-front complex separation, producing insight for some applications, including metabolomics. LC-MS requires some sample preparation, and for comprehensive, precision metabolomics, different types of chromatographic separation coupled with mass spec detection to identify as many metabolites as possible are needed to cover as much of the metabolome as possible. Each of the x4 LC-MS experiments performed by Metabolon take several minutes.
However, the benefits of NMR are far outweighed by those of LC-MS in the context of metabolomics, because of NMR weaknesses that are currently impossible to overcome, namely a lack of sensitivity and inability to detect the great majority of metabolite components in a complex sample. These are the strengths of the LC-MS-based Precision Metabolomics approach taken by Metabolon. NMR’s deficits limit its usefulness. While LC-MS certainly requires more effort, the benefits of superior coverage of the total metabolome make the extra effort worthwhile.
The conclusion from the comparison of GWAS coupled with metabolomics for the two metabolomics platforms seems straightforward. We see more associations of gene to metabolite identified with far fewer individuals studied using an LC-MS metabolomics approach. This implies very strongly that the LC-MS metabolomics approach from Metabolon was superior analytically at identifying metabolic associations than the NMR-based approach.
A final word
The ability to identify many hundreds of metabolites from all the major pathways of metabolism in a complex, rich sample such as plasma, the sample where biomarkers lie, and in tissues, diseased and otherwise, is compelling. The uses for this robust data across many areas of life sciences research, population health and precision medicine are many-fold. The combination of rich LC-MS data with the advanced bioinformatics and QC tools of Precision Metabolomics are simply too potent, too powerful, too useful not to use.
To learn more about Precision Metabolomics, visit www.metabolon.com.
1. Kettunen J, Demirkan A, Würtz P, Draisma HH, Haller T, Rawal R, Vaarhorst A, Kangas AJ, Lyytikäinen LP, Pirinen M, Pool R. Genome-wide study for circulating metabolites identifies 62 loci and reveals novel systemic effects of LPA. Nature communications. 2016 Mar 23;7.
2. Long T, Hicks M, Yu HC, Biggs WH, Kirkness EF, Menni C, Zierer J, Small KS, Mangino M, Messier H, Brewerton S, Turpaz Y, Perkins BA, Evans AM, Miller LA, Guo L, Caskey CT, Schork NJ, Garner C, Spector TD, Venter JC, Telenti A. Whole-genome sequencing identifies common-to-rare variants associated with human blood metabolites. Nat Genet. 2017 Mar 6. doi: 10.1038/ng.3809. [Epub ahead of print]