A Clinical Approach


LCGC Asia Pacific

Isabelle Kohler from Leiden University, in Leiden, The Netherlands, spoke to LCGC Europe about the latest trends in clinical metabolomics using chromatography and how the field is likely to evolve in the future.

Isabelle Kohler from Leiden University, in Leiden, The Netherlands, spoke to LCGC Europe about the latest trends in clinical metabolomics using chromatography and how the field is likely to evolve in the future.

Q. What is the definition of clinical metabolomics and why is it important?

A: Metabolomics can be defined as the comprehensive analysis of all metabolites, which are intermediates and end-point products of the metabolism with a mass lower than 1 kDa, present within a biological system. In clinical metabolomics, the aim is to identify-and ideally quantify-all small molecules present in patient-derived samples. Metabolomics, together with other “-omics” techniques, such as genomics and proteomics, plays a key role in the implementation of personalized medicine in patient healthcare. Indeed, by providing a metabolite “snapshot” (phenotype) at a certain time, the information gathered at the metabolite level may contribute to a better understanding of biomolecular mechanisms involved in (patho)physiological conditions, the possibility for earlier diagnosis of diseases, or the implementation of individualized treatment therapies.

Q. What are the main challenges facing chromatographers involved in this field and how are they being addressed?

A: Even more than in other “-omics” techniques, metabolomics strongly relies on the quality of the obtained data. Indeed, the analytical variability should be lowered to its minimum to extract the relevant biological or clinical differences in the metabolome between the studied groups. It is therefore essential to opt for state‑of-the-art and reliable analytical platforms for the analysis of the samples, control the storage and sample handling conditions, and implement multiple quality controls throughout the whole workflow. Many excellent papers have been published to raise the awareness of the community on the importance of high-quality data while also providing some useful tips, including a white paper from the Metabolomics Society Initiative (1) and a recent review from Beger et al. showing the importance of quality assurance and quality control in untargeted metabolomics studies (2).

From an analytical perspective, the two challenges that chromatographers have to face are the need for (i) high‑throughput and (ii) high-resolution techniques. In metabolomics, a strong experimental design relies on the number of subjects for each group studied, which has to be high enough to ensure a sufficient statistical power of the study. Large-scale clinical studies therefore involve the analysis of thousands of samples-not including all the additional quality controls-justifying the need for development of high-throughput techniques. There is also a strong trend towards being capable of analyzing more and more metabolites to potentially enable the discovery of novel biomarkers.

The latest release of the Human Metabolome Database (HMDB 4.0), considered the standard metabolomics resource for human metabolic studies, has reported more than 110,000 fully annotated metabolites (3)! This is an incredible number of metabolites to be exploited in the clinic. This also highlights the need to use the latest analytical developments in the field of metabolomics to expand the coverage of the metabolome, and have better access to these potential metabolite biomarker candidates.


Q. Can you tell us more about the role of liquid chromatography–mass spectrometry (LC–MS) and gas chromatography mass spectrometry (GC–MS) in clinical metabolomics?

A: GC–MS was the first separation technique used in metabolomics and still remains very popular because of the excellent separation efficiency that can be obtained. However, it presents some relevant drawbacks, for example, possible loss of thermolabile analytes, cumbersome and time‑consuming sample preparation, limited metabolome coverage, and higher variability compared to LC–MS.

The innovative technological developments performed in LC over the last 15 years, are, notably, the advent of sub-2-µm porous particles for ultrahigh‑pressure liquid chromatography (UHPLC) and superficially porous particles (core–shell technology) to achieve high‑throughput and high-resolution analysis.

 LC–MS, mostly UHPLC–MS, is now considered the gold standard in metabolomics. Chromatographers who are not familiar with fast LC analysis should refer to the paper of Fekete et al. (4), which also highlights the essential role played by the instrumentation.

Most of the metabolomics applications performed in the last decade were performed using reversed‑phase LC, typically using a stationary phase based on C18 chemistry combined with an aqueous‑organic mobile phase composed of methanol or acetonitrile with 0.1% formic acid. A very simple and versatile setup, well‑adapted for a lot of simple applications but suffering from two major drawbacks, namely (i) a possible ion suppression caused by the presence of coeluting phospholipids (plasma or serum analysis), and (ii) a poor retention of polar compounds. Moreover, metabolites with very similar physicochemical properties (for example, isomers) are not well separated using reversed‑phase LC. Improving the metabolome coverage will therefore rely on the use of other approaches more powerful for the analysis of the polar metabolome or the discrimination between closely related compounds, such as supercritical fluid chromatography (SFC), capillary electrophoresis (CE), or two‑dimensional (2D-)LC.

Q. Can you discuss in more detail the role that CE, SFC, and comprehensive 2D-LC may play in clinical metabolomics?

A: CE, SFC, and 2D-LC are increasingly considered in clinical metabolomics as complementary approaches to conventional reversed-phase LC–MS. CE–MS, for example, is very interesting because it allows the highly efficient separation of polar and ionized compounds-without the need for large sample volumes. Indeed, only a few nanolitres are required for a CE injection and 2–10 µL of sample in the vial for reproducible injection. CE–MS is therefore particularly well‑suited for the analysis of limited sample amounts, for example microfluidics three‑dimensional (3D)-cell culture models (limited number of cells), microdialysates, or other samples (plasma, cerebrospinal fluid, urine) collected on small animal models (5). However, only a few applications of CE–MS in large-scale studies have been reported.

SFC has shown a spectacular comeback in the last decade in the field of pharmaceutical analysis, mostly as a result of the impressive technological developments performed in this technique. The advent of the latest generation of SFC instruments has also fostered the development of columns packed with sub-2-µm fully porous (ultrahigh-performance SFC, [UHPSFC]) and sub-3-µm superficially porous particles specially designed for SFC analysis, as well as new interface designs for hyphenating SFC with MS. This metamorphosis of SFC–MS into a powerful analytical technique is also highly beneficial for metabolomics applications. SFC–MS is not only well suited for the analysis of lipids, where it enables the separation of isomers that are difficult to analyze using other LC-based techniques, but also for polar compounds. Indeed, the state‑of-the art instruments allow for a large flexibility in tuning the experimental conditions (for example, addition of acids, bases, salts, or water to the mobile phase; use of different stationary phases). Moreover, excellent kinetic performance can be obtained with (UHP)SFC, similar to (UHP)LC, but at higher mobile phase velocity and with a lower pressure drop compared to (UHP)LC (6). SFC remains little used in metabolomics, but on the basis of recent work carried out in this field by the group of Holčapek and co-workers (7) or Guillarme and colleagues (8), I believe that this will significantly change in the near future.

Last but not least, 2D-LC has also seen a significant breakthrough in the last couple of years that will also benefit metabolomics. Indeed, adding another separation dimension on-line is an excellent strategy to improve the metabolome coverage. Comprehensive 2D-LC (LC×LC), where all the peaks are captured in the first dimension into the second dimension, is very interesting in untargeted approaches to significantly increase the number of features detected. On the other hand, multiple heart‑cutting 2D-LC (LC–LC), where one or few fractions are collected in the first dimension and sent to a high-resolution second dimension, is promising to increase the separation between closely related compounds such as isomers. Since a large diversity of chromatographic modes can be combined in 2D-LC, this opens up new perspectives in the field of metabolomics to further expand the metabolome coverage. However, method development for this technique is perceived as complicated for inexperienced users, so this technique is also in its infancy in clinical metabolomics. New users are strongly encouraged to refer to guidelines and tutorials published by experts in the fields, including Pirok and Schoenmakers (9,10), as well as Stoll and co-workers (11,12).


Q. There has also been some interesting work using hydrophilic interaction liquid chromatography (HILIC). Why is HILIC useful in clinical metabolomics?

A: If there is one field where HILIC can prove its usefulness, it is in metabolomics. Indeed, a large variety of polar or ionizable metabolites, such as small organic acids, amino acids, nucleosides, or phosphate derivatives, are very important in multiple (patho)physiological processes, but are not easily analyzed using reversed-phase LC because of poor retention. HILIC is based on a multimodal separation mechanism involving hydrophilic partitioning, dipole-dipole interaction, hydrogen bonds, and electrostatic interaction, which makes it very well-suited for the analysis of polar compounds.

 However, HILIC is not as straightforward as reversed-phase LC; reproducible and high-quality data can only be obtained with adequate procedures. Notably, it is well-known that the sample injection diluent and injected volume, the composition of the mobile phase (especially the organic solvent chosen and the composition of the buffer), and the column equilibration time have to be carefully controlled to ensure the reproducibility of the analysis. Inexperienced users are referred to the reviews published by Kohler et al. (13,14) and McCalley (15) for discussion of these important parameters and some useful practical recommendations.

Q.It has been reported that chirality can affect the results obtained in clinical metabolomics. Why is that and how is this issue being addressed?

A: Since the thalidomide crisis in the 1960s, we know that two enantiomers from a pharmaceutical racemic mixture may have different pharmaceutical activity and potency. This is also the case for chiral metabolites, which often show different biological activities as a result of, for instance, different receptor affinities. Among the reported examples, we can cite the importance of chirality in the interaction between D-serine (and not the L-form) and the NMDA receptor to modulate synaptic plasticity, playing an important role in depression and neurological diseases; the detection of trace levels of D-amino acids in blood from patients with kidney diseases; or the oncometabolite D-2-hydroxyglutarate, which is produced upon mutations of the enzyme isocitrate dehydrogenase and causes malignant transformation-while L-2-hydroxyglutarate blood levels are not affected. The latter example has been well discussed by Struys (16).

In my opinion, chirality still remains overlooked in current metabolomics applications, but this is probably about to change along with the advent of the innovative technologies mentioned earlier. Indeed, CE, SFC, 2D-LC, and especially ion mobility mass spectrometry have an important role to play in this field because they enable the discrimination between enantiomers (17). I believe that these techniques will be the core of the analytical toolbox used in modern bioanalysis-not only in clinical metabolomics, but also in forensic toxicology, environmental toxicology, and plant metabolomics.


Q. Is clinical metabolomics applied in a routine setting?

A: A lot of small molecule biomarkers are routinely used in the clinic, such as glucose, total cholesterol, lactate, acylcarnitines, urea, or creatinine. However, none of them has been discovered using metabolomics‑based approaches. As discussed by Goodacre and colleagues recently, many metabolite biomarkers have been reported in the literature while almost zero have made it to the clinic (18). Metabolomics is now well-established, but the literature is saturated with small-scale preliminary-type studies. The potential biomarker candidates discovered are rarely confirmed in replication studies or validated for clinical use.

Clinical metabolomics is extremely promising in personalized medicine, but it is only by performing large cohort multi-centre studies with adequate experimental design that possible metabolite biomarker candidates will succeed in being translated into point-of-care or rapid diagnostics. Goodacre and co-workers also mentioned that a large number of publications “claim” to have discovered a biomarker using metabolomics, despite the fact that most of this research failed to acquire sufficient statistical power because of limited sample size (<100 subjects in total) (18). Additionally, the replication of the results using independent cohorts is crucial to increase confidence in the validity of the findings and the clinical utility of the biomarkers discovered.

Q. Your research is often focused on clinical metabolomics in the brain. What are the main analytical challenges in this research, how have you overcome them, and what have been your main findings so far?

A: Studying neurological diseases, such as Alzheimer’s disease, Parkinson’s disease, or addiction, is very exciting and can have a strong impact, but it comes with a lot of challenges. First, it is difficult to study the core of the disease in affected patients, or only in post‑mortem material, which does not provide any information on prodromal phases of such diseases. Animal models are an alternative solution, but many of them are actually not fully representative of the (patho)physiological processes happening in humans. Second, analyzing brain tissues is very challenging, because lipids account for a large proportion of the total metabolite composition. The sample preparation step is therefore essential to extract the metabolites of interest, which makes the method development rather time-consuming. The analysis of cerebrospinal fluid (CSF) seems to be an interesting alternative because it has a direct connection with the brain. However, despite being a relatively simple matrix (compared to plasma–serum), a large number of metabolites of interest are present at very low concentrations, which requires extensive method development and the latest generations of MS instruments to achieve sufficient sensitivity.

We are currently very interested in the role that lipid-based signalling molecules play in neurological disorders, mostly in Alzheimer’s disease. We have therefore implemented a method that enables the simultaneous analysis of oxylipins, lysophospholipids, isoprostanes, free fatty acids, bile acids, and endocannabinoids in plasma and tissue samples. We are also currently measuring plasma samples from a large cohort of patients with dementia, mild-cognitive impairment, and Alzheimer’s diseases (N > 5000 subjects) with this method. I am very curious about the results.


Q. What else is your group focusing on at the moment?

A: Our metabolomics facility is currently measuring multiple samples obtained from large-scale studies, in which we also have access to relevant biological information on the genome, transcriptome, and proteome, which gives us the opportunity to integrate this information with metabolomics data. We are also focusing on developing novel approaches to speed up the analysis time, improve the metabolite coverage, adapt current methods to the analysis of samples from microfluidics 3D-cell cultures from on-chip format and other biomass- and volume-limited samples, and improve the separation of exosomes and lipoproteins. All our projects linked to method development actually aim to tackle the two challenges I mentioned earlier, namely, the need for high-throughput and high‑resolution analytical techniques.

Q. How will clinical metabolomics evolve in the future?

A: The value of clinical metabolomics in routine practice has not been fully demonstrated yet, but the community believes that the translation of metabolomics to clinics relies on the development of novel point-of-care tests, such as dipstick approaches, breath measurements for volatiles, and electrochemical detection for analytes such as glucose and cholesterol.

Indeed, state-of-the-art LC–MS techniques are very powerful for the discovery of new biomarker candidates, but they are too laborious to be used for large-scale screening of very large populations, or at the general practitioner’s. As an aficionada of biohacking techniques, I believe that biosensors used in wearable technologies (for example, smartphones, smart watches, health bands, contact lenses) together with the use of machine learning and artificial intelligence represent the next generation of metabolomics towards real personalized medicine.

Q. What steps need to be taken for clinical metabolomic to evolve faster?

A: Many of the techniques I mentioned earlier, such as CE, HILIC, SFC, and 2D-LC, still remain underexploited in metabolomics-and more generally in bioanalysis. The reluctance in using these techniques is, in my opinion, mostly explained by the lack of basic practical knowledge by inexperienced users. These techniques are indeed not as straightforward as reversed‑phase LC, but it is possible to reach similar performance and data quality if the operator is aware of the potential analytical pitfalls, and knows how to prevent or solve them. Therefore, and this is particularly important for me because I am involved in a lot of teaching activities, I would like to encourage all senior scientists or leading experts in such fields to helps inexperienced users by giving short courses and tutorials at conferences, publish guidelines and background review papers, give access to helpful on-line tools, and teach analytical students the state-of‑the-art techniques, ideally alongside practical experience. Today’s undergraduate students are our future colleagues; the quality of their knowledge relies on their education.


  1. R.D. Beger et al., Metabolomics 12(10), 149 (2016).
  2. R.D. Beger et al., Metabolomics15(1), 4 (2019).
  3. D.S. Wishart et al., Nucleic Acids Res.46, D608–D617 (2017).
  4. S. Fekete, I. Kohler, S. Rudaz, and D. Guillarme, J. Pharm. Biomed. Anal.87, 105–119 (2014).
  5. R. Ramautar, LCGC Europe 30(12), 658–661 (2017).
  6. A. Grand-Guillaume Perrenoud, J.L. Veuthey, and D. Guillarme, J. Chromatogr. A1266, 158–167 (2012).
  7. M. Lisa, E. Cifkova, M. Khalikova, M. Ovcacikova, and M. Holcapek, J. Chromatogr. A1525, 96–2018 (2017).
  8. V. Desfontaine, G.L. Losacco, Y. Gagnebin, J. Pezzatti, W.P. Farrell, V. Gonzalez-Ruiz, S. Rudaz, J.L. Veuthey, and D. Guillarme, J. Chromatogr. A1562, 96–107 (2018).
  9. B.W.J. Pirok, A.F.G. Gargano, and P.J. Schoenmakers, J. Sep. Sci.41(1), 68–98 (2018).
  10. B.W.J. Pirok and P.J. Schoenmakers, LCGC Europe5(31), 242–249 (2018).
  11. D.R. Stoll and P.W. Carr, Anal. Chem. 89(1), 519–531 (2017).
  12. D.R. Stoll, X. Li, X. Wang, P.W. Carr, S.E.G. Porter, and S.C. Rutan, J. Chromatogr. A1168(1–2), 1–93 (2017).
  13. I. Kohler, A. Verhoeven, R.J. Derks, and M. Giera, Bioanalysis8(14) 1509–1532 (2016).
  14. I. Kohler, R.J. Derks, and M. Giera, LCGC Europe 29(2), 60–75 (2016).
  15. D. McCalley, LCGC Europe32(3), 114–125 (2019).
  16. E. Struys, Proc. Nat. Acad. Sci. USA110(51), E4939 (2013).
  17. V. D’Atri, T. Causon, O. Hernandez-Alba, A. Mutabazi, J.L. Veuthey, S. Cianferani, and D. Guillarme, J. Sep. Sci.41(1), 20–67 (2018).
  18. D.K. Trivedi, K.A. Hollywood, and R. Goodacre, New Horiz. Transl. Med.3(6), 294–305 (2017).

Isabelle Kohler studied pharmacy at the University of Geneva, in Geneva, Switzerland. She carried out her Ph.D. at the School of Pharmaceutical Sciences at the University of Geneva, and obtained her Ph.D. in pharmaceutical sciences in 2013, focusing on the use of capillary electrophoresis hyphenated to mass spectrometry in clinical and forensic toxicology. She moved to the Netherlands for a postdoctoral fellowship at the Leiden University Medical Center, where she investigated the biomolecular mechanisms of familial hemiplegic migraine in a transgenic mouse model using untargeted and targeted metabolomics approaches. She is currently working as Assistant Professor in the group of Analytical Biosciences and Metabolomics at the Leiden Academic Center for Drug Research. Her research interests include clinical metabolomics, bioanalysis, brain metabolism, and neurological diseases, as well as capillary electrophoresis, and hydrophilic interaction chromatography coupled to mass spectrometry.

Related Videos
Robert Kennedy
John McLean | Image Credit: © Aaron Acevedo