High-throughput Saccharomyces cerevisiae cultivation method for credentialing-based untargeted metabolomics

Abstract

Identifying metabolites in model organisms is critical for many areas of biology, including unravelling disease aetiology or elucidating functions of putative enzymes. Even now, hundreds of predicted metabolic genes in Saccharomyces cerevisiae remain uncharacterized, indicating that our understanding of metabolism is far from complete even in well-characterized organisms. While untargeted high-resolution mass spectrometry (HRMS) enables the detection of thousands of features per analysis, many of these have a non-biological origin. Stable isotope labelling (SIL) approaches can serve as credentialing strategies to distinguish biologically relevant features from background signals, but implementing these experiments at large scale remains challenging. Here, we developed a SIL-based approach for high-throughput untargeted metabolomics in S. cerevisiae, including deep-48 well format-based cultivation and metabolite extraction, building on the peak annotation and verification engine (PAVE) tool. Aqueous and nonpolar extracts were analysed using HILIC and RP liquid chromatography, respectively, coupled to Orbitrap Q Exactive HF mass spectrometry. Of the approximately 37,000 total detected features, only 3-7% of the features were credentialed and used for data analysis with open-source software such as MS-DIAL, MetFrag, Shinyscreen, SIRIUS CSI:FingerID, and MetaboAnalyst, leading to the successful annotation of 198 metabolites using MS² database matching. Comparable metabolic profiles were observed for wild-type and sdh1Δ yeast strains grown in deep-48 well plates versus the classical shake flask format, including the expected increase in intracellular succinate concentration in the sdh1Δ strain. The described approach enables high-throughput yeast cultivation and credentialing-based untargeted metabolomics, providing a means to efficiently perform molecular phenotypic screens and help complete metabolic networks.

Keywords: High-throughput sample generation; Liquid chromatography; Metabolomics; Saccharomyces cerevisiae; Stable isotope labelling; Untargeted high-resolution mass spectrometry.

Fig. 1

High-throughput sample generation using a D48-well plate. (1) Cultivation: yeast strains are cultivated in a D48 well plate in the presence of unlabelled and/or stable isotope-labelled substrates. (2) Normalization: before pelleting the cells, 200 μL aliquots of the cell cultures are transferred into a new D48 plate containing ISOTONE II (Beckman Coulter) solution for biovolume (μL/mL) determination using the Multisizer Z3 (Beckman Coulter). The latter is used to calculate the resuspension volume for the dried metabolite extracts prior to LC-MS analysis. (3) Cell pelleting: fast centrifugation treatment. (4) Quenching: the bottom of the plate is immerged in liquid nitrogen. (5) Metabolite extraction: the cell pellets are resuspended in pre-cooled MeOH:H₂O (− 20 °C). After resuspension,  − 20 °C pre-cooled MTBE is added for metabolite extraction. After the incubation step, the cell extracts of each well are transferred into 2-mL Eppendorf tubes. (6) Phase separation: the cell extracts are centrifuged in order to achieve a phase separation between the upper organic phase (MTBE) and the lower aqueous phase (MeOH:H₂O). (7) Sample evaporation: the collected phases are evaporated overnight using a SpeedVac vacuum device (Labconco). (8) Sample reconstitution: samples are reconstituted, adapting the resuspension volume to the biovolume (μL/mL) for the purpose of normalization. (9) Untargeted LC-MS analysis: the biovolume-normalized samples are analysed by HILIC-MS (metabolomics) or RP-LC-MS (lipidomics). Abbreviations are reported in the abbreviation list

Fig. 2

Experimental and computational pipelines used in this study. Yeast WT (ho:kanMX) and KO (sdh1Δ) strains were cultivated in the presence of unlabelled and/or labelled substrates in a D48 well plate and SF, followed by extraction of polar and nonpolar metabolites for LC-HRMS analysis. The raw data obtained was processed with the MS-DIAL peak-picker to obtain a feature list for the PAVE workflow. Credentialed features underwent a mass shift quality control using Shinyscreen. Features with confirmed mass shift were imported into MS-DIAL for Level 2A annotation. The MS² spectra of the remaining, non-annotated features were extracted with Shinyscreen and annotated as Levels 2A–3 with MetFrag combined with the PubChemLite (PCLite) chemical database, or as Levels 3–4 with Sirius CSI:FingerID. The non-annotated features are reported as Level 5 with their relative carbon and nitrogen number. Abbreviations are reported in the abbreviation list

Fig. 3

Quantification of intracellular ¹³C-Succinate in yeast WT and KO strains following classical SF or D48 cultivation. A The ¹³C uniformly labelled cell extracts (black, MS¹ signal) were spiked with unlabelled succinate (light blue, MS¹ signal). B Calculated intracellular concentration of ¹³C succinate for the yeast WT and KO strains after applying the D48 (D48-KO/WT) or SF (SF-KO/WT) approaches. The bar plot values refer to means ± SDs for three biological replicates. The ¹³C succinate concentrations calculated in μM amount to 24.5 ± 2.0 (D48-WT), 16.2 ± 6.3 (SF-WT), 166.0 ± 29.2 (D48-KO), and 141.3 ± 19.0 ( SF-KO). The resulting fold changes (KO/WT) were 7.2 ± 0.2 (D48) and 9.0 ± 0.9 (SF). Abbreviations are reported in the abbreviations list

Fig. 4

PCA score plot of the annotated credentialed features from the HILIC analysis of the D48 cultivation. Green dots: extraction blank samples (BLANK-GLU, n = 8). Red squares: procedural blank samples (BLANK, n = 3). The dashed zone shows the section of the PCA plot, where extraction and procedural blanks overlap. Purple triangles: KO samples from the D48 method (D48-KO, n = 3). Light blue diamonds: WT samples from the D48 method (D48-WT, n = 3)

Fig. 5

CV distribution for annotated credentialed features measured through the D48 (A) and SF (B) approaches. The bar plot represents the number of features inside 5% CV bin intervals, whereas the red line shows the cumulative frequency

Fig. 6

Box plot of the selected differential metabolites in D48 and SF conditions. Statistical significance was evaluated using a one-way ANOVA followed by Tukey’s HSD post hoc test (p < 0.01) from the HILIC-HRMS analyses. Metabolites were grouped by signals showing the same (succinate, xanthurenate, and kynurenate) or opposite (NAD⁺, gluconate, histidine) trends. The identities of succinate (A), xanthurenate (B), kynurenate (C), NAD⁺ (D), gluconate (E), and histidine (F) were confirmed (Level 1)