A diverse proteome is present and enzymatically active in metabolite extracts

Metabolism, a complex network of chemical reactions, is pivotal in various physiological and pathological processes. Mass spectrometry-based metabolomics offers a powerful approach to study metabolism, but it faces challenges, including inconsistent extraction protocols and potential for inaccurate phenotype detection. A critical step in metabolomics is metabolite extraction from biological matrices, which commonly involves precipitation of proteins using organic solvents. However, the extent of protein removal is often unclear. Prior research suggested that some proteins remain in the soluble fraction, raising concerns about potential post-extraction enzymatic activity altering metabolite composition. This study aimed to evaluate and optimize metabolite extraction approaches for metabolomics, specifically focusing on the impact of residual proteins and their enzymatic activity on metabolite profiles.

The literature extensively documents various metabolite extraction protocols, each with its strengths and limitations. Optimization efforts often lead to variations across labs, hindering data reproducibility and comparison. The influence of procedural details (solvent composition, additives, pH, etc.) on metabolomic data quality remains poorly defined. While the insoluble fraction is sometimes utilized for other analyses (proteomics, RNAseq), evidence supporting complete protein precipitation in metabolite extracts is lacking. One study showed that 2-6% of serum proteins persist in the soluble fraction, depending on the extraction solvent used. This highlights the need for improved extraction methods to minimize the influence of residual proteins on the accuracy of metabolomics experiments. The potential for enzymatic activity and subsequent metabolite interconversion after extraction remains largely unexplored.

This study used cryopulverized mouse liver tissue as a model system to optimize polar metabolite coverage by varying extraction water content (25-60%) using a common 40% acetonitrile, 40% methanol, 20% water (AMW20) extraction approach. The impact of water content on metabolite abundance was assessed using untargeted metabolomics. Proteomic analysis was performed on extracts to identify proteins present in the soluble fraction. Gene set enrichment analysis was used to characterize the functional pathways of these proteins. Post-extraction enzymatic activity was investigated using stable isotope-labeled glutamate and transaminase inhibitors. The effects of 3 kDa filtration for protein removal on both the metabolome and proteome were evaluated across multiple extraction methods (AMW, 80% methanol, Bligh-Dyer) and sample types (mouse brain, skeletal muscle, adipose tissue, plasma, human HEK cells). Untargeted metabolomics with repeated injections was used to assess time-dependent metabolite changes in extracts with and without protein removal. Targeted metabolomics using three orthogonal LC-MS methods was applied to assess the impact of filtration on polar metabolite recovery. Statistical analysis methods including ANOVA, t-tests, PCA, GSEA, Spearman's correlation, and multiple testing corrections (Benjamini-Hochberg) were used throughout the study.

This study identified over 1000 proteins in metabolite extracts, predominantly enriched for metabolic pathways, across various extraction methods and sample types. Extraction water content significantly affected metabolite detection, with a strong effect on nucleotides and other phosphate-containing compounds. A water-dependent loss of D5-glutamate (internal standard) was observed, accompanied by a gain of D4-glutamate, suggesting enzymatic activity post-extraction. The presence of eleven transaminases in metabolite extracts, including aspartate-glutamate transaminase (GOT1), further supported this hypothesis. The D5→D4-glutamate conversion was causally linked to transaminase activity, as it was prevented by both 3 kDa filtration and the addition of a transaminase inhibitor (aminooxyacetic acid). Experiments with doubly labeled [13C515N] glutamate demonstrated transaminase-mediated futile cycling. Untargeted metabolomics revealed extensive time-dependent metabolite changes in unfiltered extracts, including de novo formation of glutamate dipeptide and depletion of total glutathione. These changes were significantly mitigated by 3 kDa filtration. Combining high-water AMW extraction with 3 kDa filtration improved polar metabolite recovery while preventing enzyme-mediated interconversions, demonstrating a superior extraction methodology. The presence of proteins was ubiquitous across various sample types and extraction techniques, highlighting the potential for protein-mediated observer effects in metabolomics.

The findings challenge the conventional understanding of metabolite extraction, demonstrating the presence of a diverse and enzymatically active proteome in extracts. This has broad implications for metabolomics, as post-extraction enzymatic activity and protein-metabolite interactions can lead to inaccurate biological phenotype detection and characterization. The development of a 3 kDa filtration-based workflow helps to address these challenges by removing proteins while preserving the metabolome, improving polar metabolite coverage. The study highlights the need to carefully consider the potential for protein-mediated artifacts in metabolomics experiments and emphasizes the importance of optimized extraction methods to ensure data accuracy and reproducibility. This work also underscores the importance of careful experimental design in stable isotope tracing studies to avoid misinterpretation due to post-extraction enzymatic activities.

This study reveals a previously unrecognized source of error in metabolomics, demonstrating that residual proteins in extracts can maintain enzymatic activity and alter metabolite profiles. The proposed method incorporating 3 kDa filtration offers a practical solution to mitigate these protein-mediated effects, improving the accuracy and reproducibility of metabolomics experiments. Future research could focus on further optimizing filtration parameters, exploring additional protein removal techniques, and investigating the broader implications of protein-metabolite interactions in various biological systems.

The study primarily focused on mouse liver tissue, although other tissues and cell types were examined. The generalizability of the findings to all biological matrices and metabolomics platforms needs further investigation. While the 3 kDa filtration method is effective, it adds an extra step to the workflow, potentially introducing additional variability and sample loss. The identities of some metabolites affected by protein-mediated changes remain unknown, warranting further investigation.