SPEAKERS
Plenary Speakers
NRF2 and cysteine metabolism in cancer
Introduction: Redox regulators are emerging as critical mediators of lung tumorigenesis. Notably, NRF2 and its negative regulator KEAP1 are commonly mutated in human lung cancers. These mutations lead to NRF2 accumulation and constitutive expression of NRF2 target genes, many of which are at the interface between antioxidant function and anabolic processes that support cellular proliferation. One such metabolic process is the uptake and metabolism of the amino acid cysteine, which is required for maintaining cellular redox homeostasis in both normal and transformed cells. However, how tissues and tumors acquire cysteine in vivo is not well understood.
Methods: We generated genetically engineered mouse models of lung cancer with activating mutations in NRF2 and a liver cancer model driven by Myc overexpression. We infused healthy and tumor bearing mice with 13C-serine and 13C-cystine to track the transsulfuration pathway (de novo cysteine synthesis) and cystine uptake, respectively. Moreover, we deleted the enzymes required for cystine reduction in the lung cancer mouse model to examine the effect on lung tumor formation.
Preliminary Data: Our studies revealed that the liver and liver tumors were metabolically flexible with the ability to both synthesize cysteine and uptake it from the circulation. In contrast, healthy lung and lung tumors lacked the ability to synthesize cysteine and derived their cysteine from circulating cystine, consistent with their overexpression of the cystine transporter xCT. Interestingly, perturbation of cystine reduction only modestly impaired lung tumor formation, suggesting the presence of an alternative metabolic mechanisms of cysteine acquisition.
Novel Component: We performed the first in vivo analysis of the source of cysteine in cancer models.
A deep dive into sample normalization for improved quantitative performance in untargeted metabolomics
Introduction: Variations in starting material leads to a big challenge in the quantitative comparison of metabolites in biological samples, such as urine and saliva. For fair quantitative comparisons, sample normalization is essential during the metabolomics pipeline. Over the years, many post-acquisition sample normalization methods have been proposed. Yet, how to select the right one remains under explored. This work benchmarked six commonly used post-acquisition sample normalization methods using data simulations: sum, median, quantile, class-specific quantile, probabilistic quotient normalization (PQN), and maximal density fold change (MDFC). Using data simulations and experimental results, our work provides a mechanistic understanding of the discrepancy between sample normalization algorithms. Based on the discoveries, we proposed an evidence-based normalization workflow to facilitate accurate quantitative comparisons.
Methods: To simulate LC-MS-based metabolomics data, intensities from real experimental data were used to generate a reference sample. Next, additional samples were simulated using Gaussian noise. Each sample were then scaled by a dilution factor to represent concentration differences. To represent metabolite dysregulation from biological variation (e.g., diet and lifestyle), a set of features were dysregulated using uniform distributions. Normalization methods were benchmarked by comparing the estimated normalization factors with the true dilution factors. To examine the impact of data quality on normalization performance, we analyzed urine samples with known dilution factors. Notably, we examined how the following issues impact normalization performance: 1) non-biological features; 2) poor quantitative response; 3) fold-change bias; 4) missing values.
Preliminary Data: In our simulations, we found that most normalization methods perform poorly with unbalanced metabolite dysregulation. This is the common phenomenon where the percentage of up- and downregulated features are unequal. In these cases, normalization methods are unable to accurately estimate true dilution factors. The implications are severe as improper normalization will result in misleading quantitative comparisons. The only normalization method that was able to consistently perform well across a diverse range of data structures was MDFC. Beyond the choice of normalization method, we examined four aspects of data quality on normalization performance. Our results demonstrated that the using non-biological features and features with poor quantitative response resulted in inaccurate normalization factors. Interestingly, we observed that correcting fold-change compression resulted in the greatest improvement in normalization performance. We also show that normalization should be performed prior to missing value imputation. The reasoning is that imputation strategies can introduce artifacts that can bias normalization factors. To facilitate accurate quantitative comparisons, we propose an evidence-based normalization workflow that captures both data quality and normalization method. In this approach, a metabolomics dataset is first cleaned up by addressing the data quality issues described previously. We then introduced the idea of symmetry factors (SFs) to estimate the percentage of samples with unbalanced data. Based on this information, the normalization method can be chosen accordingly. We demonstrate this workflow using public metabolomics datasets.
Novel Component: This work deepens the understanding of sample normalization and proposes an innovative normalization workflow to achieve more reliable quantitative results.
Advances in Nanospray Desorption Electrospray Ionization (nano-DESI) Mass Spectrometry Imaging
Mass spectrometry imaging (MSI) is a powerful technique for molecular mapping of biological samples with high sensitivity and molecular specificity. Ambient ionization techniques enable imaging of biological samples with minimal sample pretreatment. We have developed an ambient MSI technique based on nanospray desorption electrospray ionization (nano-DESI). Nano-DESI is a liquid extraction-based technique, in which molecules are extracted from the sample into a dynamic liquid bridge formed between the nano-DESI probe and sample surface. The extracted analytes are transferred to a mass spectrometer inlet and ionized by electrospray ionization. The high sensitivity of nano-DESI enables imaging with high spatial resolution of 6-10 microns, which opens new directions for molecular mapping of individual cells in biological tissues. Furthermore, we have developed approaches for correlative imaging of lipids, metabolites, peptide, proteins, and glycans in biological tissues and used immunofluorescence microscopy of the same or adjacent tissue sections to extract cell-specific molecular signatures. We have also examined the effect of solvent composition on both the extraction and ionization efficiency, which provided significant enhancements in sensitivity and molecular coverage of nano-DESI MSI. These developments have established nano-DESI MSI as a powerful technique for studying biological systems.
Unlocking the metabolic secrets of Aedes aegypti mosquitoes using reverse genetics and mass spectrometry-based metabolomics.
Introduction: Despite our best efforts to control Aedes aegypti mosquitoes, these vectors of arboviruses causing infectious diseases represent a serious public health burden worldwide. The blood-sucking female mosquito must deal with a potentially life-threatening overload of nitrogen that makes up a disproportionate amount of the nutrients in a blood meal. This metabolic challenge involves most of the amino acids derived from a blood meal being oxidized, leading to a massive deamination. The application of a multidisciplinary approach including RNA interference (RNAi) and stable-label isotope tracing based metabolomics, has allowed us to discover that Ae. aegypti mosquitoes overcome the lack of a urea cycle by using unique metabolic pathways and mechanisms of regulation for ammonia detoxification and nitrogen disposal.
Methods: Currently, we are investigating how ammonia pathways closely crosstalk with the polyamine biosynthetic pathway to maintain nitrogen homeostasis in blood-fed Ae. aegypti females. To achieve this goal, we are using classical and modern techniques, including qPCR, western blots, reverse genetics and mass spectrometry-based metabolomics. Furthermore, a targeted, quantitative LC-MS/MS method – with a new mixed-mode chromatographic separation – is under development to simultaneously measure metabolites involved in both ammonia and polyamine pathways.
Preliminary Data: We identified genes encoding S-adenosylmethionine decarboxylase (SamDC, AAEL001176), spermidine synthase (SdS, AAEL010071), and spermine synthase (SmS, AAEL001378) within the Ae. aegypti genome. These proteins catalyze different steps in the polyamine biosynthetic pathway. We analyzed SamDC, SdS and SmS gene and protein expression in tissues dissected from sugar- and blood-fed females during the first gonotrophic cycle by qPCR and western blots. We found that the three genes are constitutively expressed, whereas the proteins SamDC, SdS and SmS each have a distinct profile in certain tissues. Our preliminary data also indicate an increase in polyamine levels measured in the fat body of blood-fed females dissected at 24 h post blood meal (PBM) when compared to the fat body of sugar-fed females. We have also begun to knockdown SdS and SmS by RNAi to determine the impact of their deficiencies on protein expression and uric acid (UA) excretion. Notably, we observed that RNAi-driven SdS deficiency in fat body causes a pronounced decrease in SdS protein level as expected, but also a significant reduction of SmS protein level and UA excretion at 24 h PBM. Interestingly, RNAi-mediated knockdown of SmS in fat body results in a significant decrease in SmS and SdS protein expression and a significant reduction of UA excretion at 24 h PBM. Altogether, our data not only uncover crosstalk regulations within the polyamine biosynthetic pathway but also provide strong evidence of unique crosstalk regulations between the polyamine and UA pathways. A better understanding of these metabolic processes during blood meal digestion could lead to the discovery of metabolic targets or their regulators, and ultimately provide a foundation for the development of novel metabolism-based mosquito control strategies.
Novel Component: Unique crosstalk regulations between polyamine and UA pathways maintain nitrogen homeostasis in blood-fed Ae. aegypti females.
Standardizing nontargeted metabolomics and exposomics: The LC-BinBase environment
Introduction: Current nontargeted MS/MS data do not give standardized reports and are difficult to compare. Data processing including naming and annotating metabolites is not harmonized or under quality-control. Reported compound identifications are difficult to be scrutinized by the scientific community.
Methods: We here showcase the computational infrastructure to make nontargeted analyses standardized, validated, and useful for comparisons comparable across multiple organs and studies. The LC-BinBase environment defines sample metadata via a standardizer that employs automatic user input curation through vocabularies and ontologies in MeSH, NCBI, Cellosaurus, NCIT and FDA databases. Raw data acquisition is continually monitored in the LC-BinBase control panel, including QC chart violations of upper and lower intervention limits and missed internal standards.
LC-BinBase standardizes retention times into retention index values by normalizing to 42 internal standards for the BEH-amide (HILIC) assay, and 76 internal standards for the BEH-C18 lipidomics assay. LC-BinBase then generates ‘accurate mass_MS/MS_retention time’ triplets (Bins) that are mapped to unique Spectral Hash Keys.
Preliminary Data: From over 20,000 samples that have been acquired so far, LC-BinBase has generated between 10,000-20,000 Bins per assay. MS/MS spectra are cleaned via LibGen 2.0 denoising. MS/MS spectra are automatically analyzed via Flash Entropy similarity search in MassWiki, including identity-, hybrid-, neutral loss- and open search similarity queries. The public MassWiki sites include open libraries such as MassBank.us and GNPS, while licensed libraries such as NIST23 are kept in-house. Normalized retention times are matched against machine learning predictions in Retip 2.0 software. From human clinical cohort data, plasma metabolome data with more than 1,200 unique annotated compounds per study are reported, including more than 500 exposome compounds. Chemical metadata are enriched by adduct information, InChI keys and SMILES chemical identifiers. Through links to MassIVE/GNPS and WCMC metadata, all spectra are linked to biological information such as the frequency at which they were detected in diverse species and organs. Results are presented for several studies, from plasma-based exposomics to clinical trials in small intestinal diseases.
Novel Component: A standardized, open-access database that will be developed into a community resource.