Logo: to the web site of Uppsala University

LC-MS-based untargeted metabolomics is heavily dependent on algorithms for automated peak detection and data preprocessing due to the complexity and size of the raw data generated. These algorithms are generally designed to be as inclusive as possible in order to minimize the number of missed peaks. This is known to result in an abundance of false positive peaks that further complicate downstream data processing and analysis. As a consequence, considerable effort is spent identifying features of interest that might represent peak detection artifacts. Here, we present the CPC algorithm, which allows automated characterization of detected peaks with subsequent filtering of low quality peaks using quality criteria familiar to analytical chemists. We provide a thorough description of the methods in addition to applying the algorithms to authentic metabolomics data. In the example presented, the algorithm removed about 35% of the peaks detected by XCMS, a majority of which exhibited a low signal-to-noise ratio. The algorithm is made available as an R-package and can be fully integrated into a standard XCMS workflow.

The research field of untargeted metabolomics aims to determine the relative abundance of all small metabolites in a biological system in order to find biomarkers or make biological inference with regards to the internal or external stimuli. This is no trivial aim, as the small metabolites are both vast in numbers and extremely diverse in their chemical properties. As such, no single analytical method exist that is able to capture the entire metabolome on its own. In addition, the data generated from such experiments is both immense in volume and very complex. This forces researchers to use algorithmic data processing methods to extract the informative part of this data. Such algorithms are, however, both difficult to parametrize and designed to be highly inclusive, the combination of which often leads to errors. One such algorithm is the peak picking procedures used to find chromatographic peaks in liquid chromatography-mass spectrometry (LC-MS) data.

In this thesis, four papers are included that focus both on the development of new methods for sample analysis and data processing as well as the application of such, and other, methods in two interdisciplinary research projects. The first paper describes the development and application of a protocol for LC-MS based untargeted analysis of guinea pig perilymph. The focus of the study was to investigate the biochemical processes underlying the protective effect of hydrogen gas on noise-induced hearing loss (NIHL) in guinea pigs exposed to impulse noise. This study sparked two research projects based on limitations observed during the analytical work. The first limitation was that of limited chemical coverage in the analysis when sample volumes are highly limited. The second paper describes the design and validation of a novel separation method for the sequential analysis of both hydrophilic and lipophilic compounds in biological samples. The second limitation observed was the abundance of false peaks reported by peak picking software. These have a negative effect on both downstream data processing as well as data analysis and metabolite identification. The third paper describes the development of a new algorithm for comprehensive peak characterization in untargeted analytical data with the purpose of filtering such false peaks. Both methods presented in the second and third paper were applied to the analysis of guinea pigs perilymph samples in a follow-up study on the attenuating effect of hydrogen gas on NIHL in guinea pigs exposed to broad band continuous noise.

In order to increase metabolite coverage in LC-MS-based untargeted metabolomics, HILIC- and RPLC-mode separations are often combined. Unfortunately, these two techniques pose opposite requirements on sample composition, necessitating either dual sample preparations, increasing needed sample volume, or manipulation of the samples after the first analysis, potentially leading to loss of analytes. When sample material is precious, the number of analyses that can be performed is limited. To that end, an automated single-injection LC-MS method for sequential analysis of both the hydrophilic and lipophilic fractions of biological samples is described. Early eluting compounds in a HILIC separation are collected on a trap column and subsequently analyzed in the RPLC mode. The instrument configuration, composed of commercially available components, allows easy modulation of the dilution ratio of the collected effluent, with sufficient dilution to obtain peak compression in the RPLC column. Furthermore, the method is validated and shown to be fit for purpose for application in untargeted metabolomics. Repeatability in both retention times and peak areas was excellent across over 140 injections of protein-precipitated blood plasma. Finally, the method has been applied to the analysis of real perilymph samples collected in a guinea pig model. The QC sample injections clustered tightly in the PCA scores plot and showed a high repeatability in both retention times and peak areas for selected compounds.

Introduction

Noise-induced hearing loss (NIHL) is an increasing problem in society and accounts for a third of all cases of acquired hearing loss. NIHL is caused by formation of reactive oxygen species (ROS) in the cochlea causing oxidative stress. Hydrogen gas (H-2) can alleviate the damage caused by oxidative stress and can be easily administered through inhalation.

Objectives

To present a protocol for untargeted metabolomics of guinea pig perilymph and investigate the effect of H-2 administration on the perilymph metabolome of noise exposed guinea pigs.

Methods

The left ear of guinea pigs were exposed to hazardous impulse noise only (Noise, n = 10), noise and H-2 (Noise + H2, n = 10), only H-2 (H2, n = 4), or untreated (Control, n = 2). Scala tympani perilymph was sampled from the cochlea of both ears. The polar component of the perilymph metabolome was analyzed using a HILIC-UHPLC-Q-TOF-MS-based untargeted metabolomics protocol. Multivariate data analysis (MVDA) was performed separately for the exposed- and unexposed ear.

Results

MVDA allowed separation of groups Noise and Noise + H2 in both the exposed and unexposed ear and yielded 15 metabolites with differentiating relative abundances. Seven were found in both exposed and unexposed ear data and included two osmoprotectants. Eight metabolites were unique to the unexposed ear and included a number of short-chain acylcarnitines.

Conclusions

A HILIC-UHPLC-Q-TOF-MS-based protocol for untargeted metabolomics of perilymph is presented and shown to be fit-for-purpose. We found a clear difference in the perilymph metabolome of noise exposed guinea pigs with and without H-2 treatment.

Introduction: Permanent hearing loss and tinnitus as side-effects from treatment with the anticancer drug cisplatin is a clinical problem. Ototoxicity may be reduced by co-administration of an otoprotective agent, but the results in humans have so far been modest.

Aim: The present preclinical in vivo study aimed to explore the protective efficacy of hydrogen (H₂) inhalation on ototoxicity induced by intravenous cisplatin.

Materials and Methods: Albino guinea pigs were divided into four groups. The Cispt (n = 11) and Cispt+H₂ (n = 11) groups were given intravenous cisplatin (8 mg/kg b.w., injection rate 0.2 ml/min). Immediately after, the Cispt+H₂ group also received gaseous H₂ (2% in air, 60 min). The H₂ group (n = 5) received only H₂ and the Control group (n = 7) received neither cisplatin nor H₂. Ototoxicity was assessed by measuring frequency specific ABR thresholds before and 96 h after treatment, loss of inner (IHCs) and outer (OHCs) hair cells, and by performing densitometry-based immunohistochemistry analysis of cochlear synaptophysin, organic transporter 2 (OCT2), and copper transporter 1 (CTR1) at 12 and 7 mm from the round window. By utilizing metabolomics analysis of perilymph the change of metabolites in the perilymph was assessed.

Results: Cisplatin induced electrophysiological threshold shifts, hair cell loss, and reduced synaptophysin immunoreactivity in the synapse area around the IHCs and OHCs. H₂ inhalation mitigated all these effects. Cisplatin also reduced the OCT2 intensity in the inner and outer pillar cells and in the stria vascularis as well as the CTR1 intensity in the synapse area around the IHCs, the Deiters' cells, and the stria vascularis. H₂ prevented the majority of these effects.

Conclusion: H₂ inhalation can reduce cisplatin-induced ototoxicity on functional, cellular, and subcellular levels. It is proposed that synaptopathy may serve as a marker for cisplatin ototoxicity. The effect of H₂ on the antineoplastic activity of cisplatin needs to be further explored.

Noise-induced hearing loss (NIHL) is believed to be caused by oxidative stress and accounts for up to a third of all cases of acquired hearing loss worldwide. The use of hydrogen gas as a mitigating treatment for NIHL has been successfully demonstrated in the past. We here present the application of a novel LC/LC-HRMS method for sequential analysis of hydrophilic and lipophilic compounds on an untargeted metabolomics study of the attenuating effect of hydrogen gas on NIHL in guinea pigs. The study was conducted using perilymph taken from the basal turn of the cochlea. Samples were taken in the acute stage, immediately following noise exposure along with follow-up samples after 1 and 2 weeks. Data analysis using volcano plots and random forest discriminant models revealed differences in the concentration of potassium within the scala tympani indicating effects on the potassium recycling system of the inner ear. Additionally, discriminant levels of glycerophosphorylcholine, a common osmolyte and antioxidant were discovered. In addition to these two compounds, four additional unknown metabolites are described. The findings are in line with previous reports indicating that the protective effect of hydrogen gas is sustained for at least 2 weeks.