Nitrogen limitation causes a seismic shift in redox state and phosphorylation of proteins implicated in carbon flux and lipidome remodeling in Rhodotorula toruloides

Abstract

Background: Oleaginous yeast are prodigious producers of oleochemicals, offering alternative and secure sources for applications in foodstuff, skincare, biofuels, and bioplastics. Nitrogen starvation is the primary strategy used to induce oil accumulation in oleaginous yeast as part of a global stress response. While research has demonstrated that post-translational modifications (PTMs), including phosphorylation and protein cysteine thiol oxidation (redox PTMs), are involved in signaling pathways that regulate stress responses in metazoa and algae, their role in oleaginous yeast remain understudied and unexplored.

Results: Towards linking the yeast oleaginous phenotype to protein function, we integrated lipidomics, redox proteomics, and phosphoproteomics to investigate Rhodotorula toruloides under nitrogen-rich and starved conditions over time. Our lipidomics results unearthed interactions involving sphingolipids and cardiolipins with ER stress and mitophagy. Our redox and phosphoproteomics data highlighted the roles of the AMPK, TOR, and calcium signaling pathways in regulation of lipogenesis, autophagy, and oxidative stress response. As a first, we also demonstrated that lipogenic enzymes including fatty acid synthase are modified as a consequence of shifts in cellular redox states due to nutrient availability.

Conclusions: We conclude that lipid accumulation is largely a consequence of carbon rerouting and autophagy governed by changes to PTMs, and not increases in the abundance of enzymes involved in central carbon metabolism and fatty acid biosynthesis. Our systems-level approach sets the stage for acquiring multidimensional data sets for protein structural modeling and predicting the functional relevance of PTMs using Artificial Intelligence/Machine Learning (AI/ML). Coupled to those bioinformatics approaches, the putative PTM switches that we delineate will enable advanced metabolic engineering strategies to decouple lipid accumulation from nitrogen limitation.

Fig. 1

Growth and lipid kinetics for the multi-omics experimental approach. A Experimental approach for integrated multi-omics analysis. Briefly, R. toruloides was cultivated in nitrogen-rich (C:N of 5:1) and limited (C:N of 90:1) conditions with sampling at 24, 48, and 72 h. For growth and lipid kinetics, a 0 h timepoint was included to analyze media and the inoculum. Lipidomics and multiplexed quantitative proteomics/PTMomics analyses were conducted to attain an integrative perspective of protein regulation, cellular PTM states, and metabolism. B Line plots demonstrating biomass production (CDWs; cell dry weights) and substrate consumption (glucose and ammonium). The overlayed bar chart shows lipid contents for the different conditions. Error bars represent one standard deviation from biological triplicates (except for 0 h in which technical replicates were evaluated). For the bar chart, significance levels from two-sample t-tests are included: ** = p-value ≤ 0.01, *** = p-value ≤ 0.001

Fig. 2

Lipidome dynamics according to nitrogen availability. A Nested pie chart presenting the distribution of unique lipid IDs in the major classes: glycerophospholipids (GC), glycerolipids (GL), fatty acyls (FA), sphingolipids (SL), and prenol lipids (PR). Lipid subclass abbreviations are as follows: coenzyme Q (CoQ), ceramide (Cer), cardiolipin (CL), prenol lipid (PR), triacylglyceride (TG), diacylglyceride (DG), fatty acid (FA), fatty acid ester of hydroxy fatty acid (FAHFA), glycosphingolipid (HexCer), phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylglycerol (PG), phosphatidylinositol (PI), phosphatidylserine (PS), phosphatidic acid (PA), and the corresponding lyso-derivatives (e.g., LPC). B Principal component analysis of normalized lipidomics data. The 0 h condition (inoculum) was evaluated in technical replicates, but all other conditions were investigated using biological triplicates. C Heatmap showing summed peak heights for lipids in the observed subclasses. The data was transposed and scaled to a mean of zero and standard deviation of 1 prior to visualization (see the legend on the far right)

Fig. 3

Protein thiol oxidation and phosphorylation patterns are distinguished by nitrogen availability and cultivation time. A Workflow for multiplexed quantification of global protein abundance, cysteine thiol oxidation, and phosphorylation. During homogenization, protein free thiols in “Thiol Oxidation” samples were blocked with NEM, whereas NEM was omitted for “Total Thiol” samples. Automated single-pot, solid-phase-enhanced sample preparation (SP3) was then used for sample cleanup, followed by on-bead digestion and TMT labeling. After desalting, samples were split for protein abundance analysis and enrichment of PTMs. Reversibly oxidized cysteine thiols (represented by “-SOx”) were reduced and enriched by resin-assisted capture (RAC), while phosphorylated peptides were captured using immobilized metal affinity chromatography (IMAC). B Faceted bar chart presenting unique peptide, PTM site, and protein identifications for each sample type. C PCA plots with ellipses delineating the 95% confidence intervals for the 24 h replicates. D Distributions of percent cysteine thiol oxidation (the “% Occupancy” of thiol oxidation abundances/total thiol abundances). The population mean value is labeled on each distribution. E Heatmap showing Pearson correlation between the summed lipid subclass intensities and normalized thiol oxidation abundances

Fig. 4

Summary of multi-PTM proteomics results for central carbon metabolism and lipogenesis. Metabolic pathways constructed based on annotations aggregated from JGI [81], eggNOG [82], as well as BLASTp results against S. cerevisiae S288C [87] (http://www.yeastgenome.org/) and R. toruloides NP11 [83]. Labels include common gene nomenclature abbreviations and the corresponding JGI accession ID. Details for gene and metabolite abbreviations are provided in Additional File 6 along with several additional annotations for desaturases, elongases, lipooxygenases, and enzymes involved in carotenoid and sterol synthesis. The gene label box is split into four sections describing relative protein abundances for four comparisons. The first is 72 vs. 24 h of high nitrogen, and the remaining three are low vs. high nitrogen at 24, 48, and 72 h. Circle (phosphorylation) and star (thiol oxidation) labels for each of the four comparisons provide a snapshot of PTM data for one particular residue (details for all observed residues found in Additional File 6). Grey and white labels define proteins and/or PTM sites that were not observed or did not exhibit significant changes, respectively. Multiple arrows denote multiple reaction steps, while dotted arrows indicate transport. Enzyme annotations for cardiolipin metabolism are not included, because they were not observed

Fig. 5

Significant proteome changes in signaling pathways that regulate stress response and metabolism according to nitrogen availability. Labels include common gene nomenclature abbreviations referenced herein and defined in Additional File 7. Circle (phosphorylation) and star (thiol oxidation) labels provide a snapshot of PTM data for one residue. In many cases (e.g., PAH1), multiple residues exhibited significant changes, but for the sake of simplicity, proteins with significant changes in at least one of the comparisons are summarized in the pathway portion of this figure. Grey labels define proteins that were not observed. Dotted arrows represent multiple steps