Landmark Literature: Part II

The Shape of Things to Come

By Koen Sandra, Scientific Director, Research Institute for Chromatography, Kortrijk, Belgium; co-founder and R&D Director anaRIC biologics, Ghent, Belgium.

It will come as no surprise that my selected paper is related to biopharmaceutical analysis. Biopharmaceuticals are becoming a core aspect of the pharmaceutical industry and, in recent years, we have observed many advances in the analysis of these huge and complex structures. 2017 saw a number of exciting papers and lectures in the field – a few that spring to mind are:

• Four-dimensional liquid chromatography (4D-LC) in hyphenation to mass spectrometry (MS) for the characterization of monoclonal antibodies (mAbs) (1);
• Detailed characterization of the glycosylation profile of the fusion protein etanercept by high-resolution native MS (2);
• Hydrophilic interaction chromatography (HILIC) for the separation of mAbs at middle-up level (3);
• Direct coupling of cation exchange chromatography (CEX) to MS to directly assign a molecular weight (MW) value to acidic and basic variants of mAbs (4).

And though all of these developments are landmarks in their own right, the paper I ultimately selected is related to higher-order structural analysis of proteins using mass spectrometry. The higher-order structure (secondary, tertiary, quaternary) defines the function of a protein and as such contributes to the quality, safety and efficacy of biopharmaceuticals. At protein level, higher-order structures can be assessed using native MS and ion mobility MS (IM-MS). Higher-order structures can also be assessed at peptide level, despite the fact that all structural information is lost upon generating peptides. Limited proteolysis can provide insight into higher-order structure but nowadays we typically rely on more advanced tools, such as cross linking, footprinting and hydrogen deuterium exchange (HDX). HDX is especially promising and is quickly being implemented in biopharmaceutical analysis. The principle is simple but clever. Upon placing a protein in a D₂O solution, amide hydrogens can exchange with deuterium based on their solvent accessibility. Exposed and dynamic regions exchange rapidly, while rigid and protected regions exchange more slowly. Peptide measurements following labeling and digestion allow discrimination of these regions (due to the mass shift induced by deuterium) and so provide an indirect read-out of the higher-order structure.

In my chosen paper, the authors demonstrate the applicability of HDX-MS to monitor structural changes caused by chemical modifications. Though earlier HDX-MS studies in a biopharmaceutical context typically focused on major structural differences, what makes this paper stand out is the sensitivity of the workflow with its ability to detect tiny local structural differences. Indeed, an increase in oxidation of just one percent in the conserved region of a mAb (methionine at position 254) resulted in statistically relevant deuterium uptake differences. Furthermore, the article demonstrates the importance of studying biopharmaceuticals using different complimentary techniques. Indeed, chromatographic and spectroscopic techniques – size exclusion chromatography (SEC), CEX, circular dichroism (CD) and Fourier-transform infrared (FT-IR) – were used to monitor global structural changes, indicating preserved structural integrity.

HDX-MS is extremely sensitive to back-exchange (D → H), which results in the loss of structural information and so demands substantial user expertise. All post-labeling handling involves compromises; digestions are typically performed at low pH using pepsin (as opposed to neutral pH using trypsin in traditional peptide mapping experiments) and separations are performed at very low temperatures (as opposed to elevated temperatures). The authors perfectly understand the weaknesses of the technique and took sensible precautions to limit back-exchange, dictating the success of their work. The set-up involves a nice interplay of valves and columns, similar to that in a 2D-LC set-up. Indeed, the deuterated mAb is injected on a pepsin column for digestion (low pH, 15°C), before the resulting peptides are directed to a trapping column and subsequently eluted from the trap onto the analytical column operated at 0°C for high-resolution MS and MS/MS measurements.

In the coming years we expect this methodology to be more widely applied during the development of biopharmaceuticals. In the development of biosimilars, a comprehensive comparability exercise to the innovator product is required to demonstrate similarity in terms of physicochemical characteristics, efficacy and safety. In that respect, an enormous weight is placed on analytics, and the analytical package for a biosimilar submission is considerably larger than that of an innovator product. In my opinion, the value of HDX-MS data in such dossiers is indispensable. Further developments are needed to take this tool out of the hands of the experts and into routine use, but with instrument vendors now offering HDX-MS workflows, this seems only a matter of time.

To my great satisfaction, my landmark paper demonstrates that cutting-edge research is not the sole preserve of academic laboratories, but is also being performed within industry.

Attogram-Scale LIBS

By Vassilia Zorba, Group Leader, Lawrence Berkeley National Laboratory, Berkeley, California, USA.

Nanoparticle analysis is extremely useful in many disciplines and applications. Optical excitation techniques are limited in resolution by the diffraction limit, and imaging below the diffraction limit is only possible with near-field optics in most cases. Single nanoparticle analysis was previously unattainable with laser induced breakdown spectroscopy (LIBS), but the authors of my 2017 landmark paper were able to achieve attogram-scale absolute limits of detection by using optical trapping and levitation to isolate single nanoparticles. It represents an entirely new paradigm for high-sensitivity elemental analysis (in an all-optical fashion), opening up new possibilities in single particle analysis.

I hope to see future demonstrations of this technology in different nanoparticle systems to assess absolute limits of detection, the use of different laser wavelengths (for example, UV) for LIBS excitation, and the effects of laser pulse duration on analytical figures of merit.

Shifting Gears

By Liam Heaney, Research Associate, Department of Cardiovascular Sciences, University of Leicester, Leicester, UK

As a researcher with an interest in both targeted and non-targeted metabolomics, it is encouraging to see the constant evolution of technical and computational advances for the analysis of metabolites in complex biological samples such as urine and plasma. Mass spectrometry is undoubtedly the most powerful analytical tool for these experiments, with the ability to measure thousands of analytes in a single bio-fluid screen. However, this power does not come without drawbacks. Constant and careful attention to the instrumentation and samples must be maintained to minimize crossover, contamination, competition for ionization and fluctuations in detector responses.

Nuclear magnetic resonance (NMR) spectroscopy could be described as a cruder tool for metabolomics analyses. Its excellent analytical capability comes with a caveat: reduced number of detectable analytes, which is in turn offset by its ability to analyze samples with minimal clean-up in a high-throughput manner, with quantitation possible without extensive calibration experiments. That said, automatic detection and quantitation of analytes has been hampered by the variability of metabolite chemical shifts in NMR spectra, which are attributable to multifaceted interactions of analytes within the complex bio-fluid matrices.

In 2017, Claudio Luchinat and his colleagues published the findings from extensive experiments to clarify the relationships between variable chemical shifts and analyte concentrations, with the goal of improving automatic identification and subsequent quantitation of metabolites in urine. They identified five signals from three main urine metabolites (citrate, creatinine and glycine) and labeled them as navigator signals.

Subsequent analysis of close to 4,000 artificial urine samples made up from variable concentrations of the principal 40 metabolites and 11 “NMR invisible” inorganic ions was performed and used for modeling purposes. They later demonstrated that the chemical shift of all signals from participant urine samples, along with their corresponding concentrations, could be predicted following analysis of the navigator signals.

There was impressive consistency between predicted and actual chemical shift values (R values of ≥ 0.98). It was particularly exciting that their prediction modeling could very accurately quantitate the 11 inorganic ions that could not be directly analyzed by NMR. Furthermore, they showed improved identification of analytes that present with one singlet, which are often obscured through the presence of crowded regions of NMR spectra (for example, trimethylamine N-oxide, TMAO).

The paper demonstrates the fast and accurate prediction of many clinically relevant metabolite biomarkers, measured in a high-throughput and multiplex manner. To find out whether this tool is applicable in the clinic, we will need to see inclusion of additional metabolites and upscaling to more extensive real-world sample sets. However, I’m very hopeful that these initial results will stand the test of translation to more demanding scenarios.

A beta version of the team’s NMR modeling tool can be found at 150.217.146.252.

Arsenic: an Old Trace

By Julian Tyson, Emeritus Professor Analytical Chemistry, University of Massachusetts, Amherst, USA.

For several years, my research group has been developing methods to measure potentially hazardous arsenic compounds in our diet. However, establishing an unambiguous link between the consumption of particular foodstuffs and ill health is an extremely challenging proposition. No single research group working over a limited timescale could produce the necessary data to answer important societal questions, such as: “To what extent are we being poisoned by what we eat?” Researchers need to examine the results from multiple studies over relatively long time periods. Review articles by recognized experts are, therefore, of critical importance.

The Landmark article I have chosen is one such a review article. It addresses the question: “Is there evidence that eating rice increases arsenic exposure?” The information needed to answer this question illustrates very nicely the role of analytical chemistry in supporting ongoing studies of major humanitarian health problems.

We now know that the foodstuff that contains the highest concentrations (triple digit µg per kg) of inorganic arsenic (a non-threshold human carcinogen) is rice. We also know that rice and tuna contains similar concentrations of arsenic in the form of dimethylarsinate (less toxic to humans, but probably not innocuous).

Although scientists make such statements on a regular basis, it is important to remember that inorganic arsenic and dimethylarsinate are the species that appear in solution as a result of the sample preparation procedure. So we know that rice-eaters are probably eating more inorganic arsenic than non-rice eaters. It is necessary to specify the chemical species, because fish-eaters are consuming much higher amounts of arsenic per serving, as fish contains high concentrations (up to 5 mg per kg) of arsenobetaine, a tetramethylarsonium derivative, which is thought to be harmless.

To assess exposure, it is necessary to measure the internal dose. As the internal dose cannot be measured directly, biomarkers are used; the concentrations of arsenic species in urine is the most widely studied biomarker, but nails, hair, and blood have also been analyzed. The reviewers identified 16 studies whose designs ensured that the arsenic species in urine could be correlated with the consumption of rice, specifically. Most of the studies involved the determination, by HPLC-ICP-MS, of inorganic arsenic and dimethylarsinate (whose concentration is derived from the metabolism of inorganic arsenic – via monomethylarsonate – and from the ingestion of dietary dimethylarsinate) and compared the values obtained with those obtained from the urine of non-rice eaters.

The reviewers scrutinized the study design and statistical tests applied, to ascertain whether the differences observed were significant. They concluded that, despite the variations in study design, and ethnicity and age of participants, the results showed a consistent positive association between rice intake and arsenic exposure.

Latent Talents

By Christopher C. Mulligan, Professor of Analytical Chemistry, Department of Chemistry, Illinois State University, USA.

The field of ambient mass spectrometry continues to flourish, with several reports of novel methodologies and applications published in the top analytical journals during 2017. One emerging trend was combining simplistic ambient ionization methods with chemometric processing, allowing rapid, yet impactful, profiling of target samples; for example, Rabi Musah and co-workers reported the use of direct analysis in real time to determine the species of necrophagous insect eggs by statistical examination of amino acid profiles (5).

One paper that caught my eye was the work of Richard Zare and Zhenpeng Zhou, who combined desorption electrospray ionization (DESI) with machine learning techniques to glean personal information (e.g., gender, ethnicity, and age) from latent fingerprints. Building upon the pioneering work by Cooks and co-workers (6), the Zare group performed DESI imaging of lipid distributions in fingerprints, yielding both spatial patterns of the print (e.g., ridges, furrows) and broad chemical mapping. Using a modest population of study participants (around 200 people), lipid distribution and demographic information was processed using a classification algorithm (gradient-boosting tree ensemble) in an effort to classify samples by personal characteristics.

One of the most intriguing aspects of this work is the reported accuracy of their classification model, which was able to identify samples from a diverse test set in terms of gender, ethnicity and age (within 10 years) with accuracies ranging from 82.4 to 89.2 percent. The methodology even worked for overlapping prints from different individuals, suggesting that identifying information can still be obtained even when complex latent fingerprint evidence is present. Furthermore, smeared fingerprints that have lost the spatial patterns needed for dactyloscopy can still yield information of probative value in criminal investigations. In terms of forensic processing of fingerprint evidence, this method could take Locard’s exchange principle that “every contact leaves a trace” to a whole new level.