Soukaina Timouma

Lipoprotein composition is altered in sepsis, and supplementation with high-density lipoproteins has been reported to improve outcomes in experimental settings. In this study, we aimed to investigate the nature and inter-individual variability in the lipoprotein proteome to inform risk stratification and opportunities for precision medicine approaches. In a large proteomic dataset including 1134 patients (1781 samples) with sepsis and 149 healthy volunteers, we analysed 18 protein components of lipoproteins. We characterise heterogeneity of the lipoprotein proteome, defining three step-wise sub-phenotypes associated with increasing disease severity, one close to health, then an early phase patient group showing increased abundance of proteins that integrate HDL under inflammatory conditions (SAA1 and SAA2), then a group with decreased abundance of proteins that are components of HDL under healthy conditions that was associated with higher organ failure intensity (SOFA score) and increased mortality. We developed and externally validated a quantitative score reflective of lipoproteins alterations in sepsis, and machine learning predictive models to predict the LP class, advancing future individualised lipoproteins-based therapeutics in sepsis. Purpose: During sepsis, the lipoprotein proteome is altered, impacting function and potentially driving severity. Many drugs targeting lipoprotein metabolism are available, but sepsis is a highly heterogenous disease, and there is no large-scaled exploration of lipoprotein alterations in sepsis. In this study, we aimed to explore lipoprotein proteome based sub-phenotypes and their implications in terms of immune dysfunction, organ failure, and mortality in sepsis. The secondary objective was to develop classifiers enabling risk stratification and personalised intervention for future trials. Methods: In a large proteomic dataset, we identified 18 proteins that are components of lipoproteins. We conducted unsupervised clustering and compared those patient clusters in terms of immune function (SRS scores, differential gene expression analysis), organ dysfunction, and mortality. We then built a continuous score for lipoprotein alteration in sepsis. In the third part, we developed and externally validated two predictive machine learning models that can be used as classifiers (for sub-phenotypes assignment and continuous score calculation). Results: We analysed data from 1134 patients (1781 samples) with sepsis and 149 healthy volunteers. On day 1 of ICU admission, we identified 3 clusters based on lipoproteins’ proteome profile (named LP3, LP2 and LP1). Transition between the cluster of lowest severity (LP3) to LP2 was characterised by an increased abundance in proteins that integrate HDL under inflammatory conditions (SAA1 and SAA2) and transition between LP2 and LP1 by a decreased abundance of a group of proteins that are component of HDL under healthy conditions. Those clusters were associated with sepsis response signature scores, organ failure, and mortality. Next, we developed a continuous score reflective of lipoproteins alterations in sepsis (LPq) and two machine learning classifiers to predict LP cluster membership and LPq. Analysis on an external dataset composed of 265 patients (353 samples) with COVID-19, sepsis and healthy volunteers enabled us to validate the robustness of our models and the external validity of our score. Conclusions: We described the heterogeneity of lipoproteins’ proteome in sepsis and defined three sub-phenotypes of increasing severity. Those sub-phenotypes were mostly driven by the HDL proteome. Our results suggest that lipoprotein proteome alteration occurs as a continuum in patients with sepsis. The first step was characterised by increased abundance of proteins that integrate HDL composition under inflammatory conditions (SAA1 and SAA2), while the second step was characterised by decreased abundance of proteins that are components of HDL under healthy conditions. This second step was associated with higher organ failure intensity (SOFA score) and increased mortality. We developed and externally validated a quantitative score reflective of lipoproteins alterations in sepsis, and machine learning predictive models to predict the LP class, paving the way for individualised lipoproteins-based therapeutics in sepsis.

Advanced transcriptomic technology has identified a great number of non-coding RNAs (ncRNAs) that are pervasively transcribed in the yeast genome. ncRNAs can be classified into short ncRNAs (200 nt). Those transcripts are strictly regulated through transcription and degradation mechanisms to maintain proper cellular homeostasis and prevent aberrant expression. It has been revealed that ncRNAs can play roles in various regulatory processes, particularly in transcriptional regulation. While short ncRNAs are well characterised, the function of lncRNAs remains poorly understood. Both functional and transcriptional profiling have been applied to fill the gap in the lncRNA functions landscape. It has been proven by functional profiling that these long transcripts can serve important cellular roles in gene regulation, RNA metabolism, sexual differentiation and telomeric overhang homeostasis. In addition, transcriptional profiling allowed the characterisation of ncRNAs involved in the cell cycle, colony subpopulation dynamics, virulence and regulatory networks. In this review, we introduce the classification, the cellular fate, the evolution and conservation, the mechanisms of action, and the profiling of yeast ncRNAs.

The Starmerella clade is known for displaying osmotolerant and acidophilic traits from their association with bees. Several species in this genus can produce sophorolipids, which are commercially produced as biosurfactants. Here, we isolated a yeast contaminant from the laboratory environment, identified as Starmerella batistae , able to thrive under low pH, including high concentrations of lactic acid, and relative high temperatures. We have sequenced and conducted a de novo genome assembly in three chromosomes and a mitochondrial genome for S. batistae (ca 9.3 Mb). Based on this reference genome we functionally annotated 29 different Starmerella species, using the publicly available sequences. Phylogenetic analysis across different yeast clades revealed a close relationship of Starmerella species with Schizosaccharomyces yeasts. Fifteen genes were uniquely shared between Sz. pombe and S. batistae , of which twelve were involved in cell morphology. Interestingly, the shape of S. batistae cells is elongated rather than round, similar to the Sz. pombe . Additionally, we found that all the Starmerella strains capable of producing sophorolipids shared a last common ancestor. Such clustering can help identify other sophorolipid-producing Starmerella yeasts that have not yet been characterised. We did not find the one-to-one orthologs of S. bombicola sophorolipid pathway in any of the Starmerella sp. with the exception of S. kuoi , S. powellii and S. floricola . In S. etchelsii , the antisense and telomeric pair UGTA1/CYP52M1 was found to be structurally conserved although not functionally. These findings support the notion that alternative pathways for the production of sophorolipids have evolved in different Starmerella lineages.

In silico tools such as genome-scale metabolic models have shown to be powerful for metabolic engineering of microorganisms. Saccharomyces pastorianus is a complex aneuploid hybrid between the mesophilic Saccharomyces cerevisiae and the cold-tolerant Saccharomyces eubayanus . This species is of biotechnological importance because it is the primary yeast used in lager beer fermentation and is also a key model for studying the evolution of hybrid genomes, including expression pattern of ortholog genes, composition of protein complexes, and phenotypic plasticity. Here, we created the iSP_1513 GSMM for S. pastorianus CBS1513 to allow top-down computational approaches to predict the evolution of metabolic pathways and to aid strain optimization in production processes. The iSP_1513 comprises 4,062 reactions, 1,808 alleles, and 2,747 metabolites, and takes into account the functional redundancy in the gene-protein-reaction rule caused by the presence of orthologous genes. Moreover, a universal algorithm to constrain GSMM reactions using transcriptome data was developed as a python library and enabled the integration of temperature as parameter. Essentiality data sets, growth data on various carbohydrates and volatile metabolites secretion were used to validate the model and showed the potential of media engineering to improve specific flavor compounds. The iSP_1513 also highlighted the different contributions of the parental sub-genomes to the oxidative and non-oxidative parts of the pentose phosphate pathway. Overall, the iSP_1513 GSMM represent an important step toward understanding the metabolic capabilities, evolutionary trajectories, and adaptation potential of S. pastorianus in different industrial settings. IMPORTANCE: Genome-scale metabolic models (GSMM) have been successfully applied to predict cellular behavior and design cell factories in several model organisms, but no models to date are currently available for hybrid species due to their more complex genetics and general lack of molecular data. In this study, we generated a bespoke GSMM, iSP_1513, for this industrial aneuploid hybrid Saccharomyces pastorianus , which takes into account the aneuploidy and functional redundancy from orthologous parental alleles. This model will (i) help understand the metabolic capabilities and adaptive potential of S. pastorianus (domestication processes), (ii) aid top-down predictions for strain development (industrial biotechnology), and (iii) allow predictions of evolutionary trajectories of metabolic pathways in aneuploid hybrids (evolutionary genetics).

In silico tools such as genome-scale metabolic models (GSMM) have shown to be powerful for metabolic engineering of microorganisms. Here, we created the iSP_1513 GSMM for the aneuploid hybrid S. pastorianus CBS1513 to allow top-down computational approaches to predict the evolution of metabolic pathways and to aid strain optimisation and media engineering in production processes. The iSP_1513 comprises 4062 reactions, 1808 alleles and 2747 metabolites, and takes into account the functional redundancy in the gene-protein-reaction rule caused by the presence of orthologous genes. Moreover, a universal algorithm to constrain GSMM reactions using transcriptome data was developed as a python library and enabled the integration of temperature as parameter. Essentiality datasets, growth data on various carbohydrates and volatile metabolites secretion were used to validate the model. Overall, the iSP_1513 GSMM represent an important step towards understanding the metabolic capabilities, evolutionary trajectories and adaptation potential of S. pastorianus in different industrial settings.

Kazachstania bulderi is a non-conventional yeast species able to grow efficiently on glucose and δ-gluconolactone at low pH. These unique traits make K. bulderi an ideal candidate for use in sustainable biotechnology processes including low pH fermentations and the production of green chemicals including organic acids. To accelerate strain development with this species, detailed information of its genetics is needed. Here, by employing long read sequencing we report a high-quality phased genome assembly for three strains of K. bulderi species, including the type strain. The sequences were assembled into 12 chromosomes with a total length of 14 Mb, and the genome was fully annotated at structural and functional levels, including allelic and structural variants, ribosomal array and mating type locus. This high-quality reference genome provides a resource to advance our fundamental knowledge of biotechnologically relevant non-conventional yeasts and to support the development of genetic tools for manipulating such strains towards their use as production hosts in biotechnological processes.

The lager yeasts, Saccharomyces pastorianus, are hybrids of Saccharomyces cerevisiae and Saccharomyces eubayanus and are divided into two broad groups, Group I and II. The two groups evolved from at least one common hybridisation event but have subsequently diverged with Group I strains losing many S. cerevisiae chromosomes while the Group II strains retain both sub-genomes. The complex genomes, containing orthologous alleles from the parental chromosomes, pose interesting questions regarding gene regulation and its impact on the fermentation properties of the strains. Superimposed on the presence of orthologous alleles are complexities of gene dosage due to the aneuploid nature of the genomes. We examined the contribution of the S. cerevisiae and S. eubayanus alleles to the gene expression patterns of representative Group I and II strains during fermentation. We show that the relative expression of S. cerevisiae and S. eubayanus orthologues is positively correlated with gene copy number. Despite the reduced S. cerevisiae content in the Group I strain, S. cerevisiae orthologues contribute to biochemical pathways upregulated during fermentation which may explain the retention of specific chromosomes in the strain. Conversely, S. eubayanus genes are significantly overrepresented in the upregulated gene pool in the Group II strain. Comparison of the transcription profiles of the strains during fermentation identified both common and unique gene expression patterns, with gene copy number being a dominant contributory factor. Thus, the aneuploid genomes create complex patterns of gene expression during fermentation with gene dosage playing a crucial role both within and between strains.

It is now accepted that non-coding RNAs (ncRNAs) play important functional roles in the cell. Recent studies in Saccharomyces cerevisiae have identified a diverse catalogue of ncRNAs, where the majority of which a function has yet to be ascribed. Large scale phenotypic screening and transcriptional analysis using the yeast ncRNAs deletion collection have identified ncRNAs responsible for major fitness changes in rich and minimal media. Here, to understand the interplay between these non-coding RNAs, the synthetic genetic array (SGA) methodology was employed to shed light on the epistasis and plasticity of key ncRNAs in different environmental conditions. Thirty-eight ncRNA query strains were crossed with a library of 411 ncRNAs deletion mutant strains, encompassing CUTs, SUTs, snoRNAs and tRNAs. Over 1000 significant genetic interactions were observed. Unlike the protein SGA genetic interaction network, for which the majority of interactions are negative, the ncRNAs mostly displayed positive epistasis (~90.2%). Interestingly, in the presence of stressors, a significant increase in negative interactions was observed, suggesting that these ncRNAs may have environmentally dependent functions or may differentially regulate neighbouring genes in different conditions. The vast majority of ncRNA interactions (>90%) did not correlate with the epistasis of their neighbouring protein coding genes, suggesting that the ncRNA and protein interaction networks are mainly independent. The two U3 paralogues, SNR17A and SNR17B, responsible for 18S rRNA processing, share the majority of genetic interactions in rich medium as expected. However, a surprising large number of unique epistatic interactions were observed for these two snoRNAs when grown in stressful conditions, suggesting that SNR17A and SNR17B may have other roles in the cell under different environmental pressures and may have sub-functionalised after genome duplication, likely driven by environmental adaptation. Additionally, SUT480 was found to interact with SNR17A and its function was linked to the processing of 18S rRNA.

Non-coding RNAs (ncRNAs), including the more recently identified Stable Unannotated Transcripts (SUTs) and Cryptic Unstable Transcripts (CUTs), are increasingly being shown to play pivotal roles in the transcriptional and post-transcriptional regulation of genes in eukaryotes. Here, we carried out a large-scale screening of ncRNAs in Saccharomyces cerevisiae, and provide evidence for SUT and CUT function. Phenotypic data on 372 ncRNA deletion strains in 23 different growth conditions were collected, identifying ncRNAs responsible for significant cellular fitness changes. Transcriptome profiles were assembled for 18 haploid ncRNA deletion mutants and 2 essential ncRNA heterozygous deletants. Guided by the resulting RNA-seq data we analysed the genome-wide dysregulation of protein coding genes and non-coding transcripts. Novel functional ncRNAs, SUT125, SUT126, SUT035 and SUT532 that act in trans by modulating transcription factors were identified. Furthermore, we described the impact of SUTs and CUTs in modulating coding gene expression in response to different environmental conditions, regulating important biological process such as respiration (SUT125, SUT126, SUT035, SUT432), steroid biosynthesis (CUT494, SUT053, SUT468) or rRNA processing (SUT075 and snR30). Overall, these data capture and integrate the regulatory and phenotypic network of ncRNAs and protein-coding genes, providing genome-wide evidence of the impact of ncRNAs on cellular homeostasis.

Genome-scale computational approaches are opening opportunities to model and predict favorable combination of traits for strain development. However, mining the genome of complex hybrids is not currently an easy task, due to the high level of redundancy and presence of homologous. For example, Saccharomyces pastorianus is an allopolyploid sterile yeast hybrid used in brewing to produce lager-style beers. The development of new yeast strains with valuable industrial traits such as improved maltose utilization or balanced flavor profiles are now a major ambition and challenge in craft brewing and distilling industries. Moreover, no genome annotation for most of these industrial strains have been published. Here, we developed HybridMine, a new user-friendly, open-source tool for functional annotation of hybrid aneuploid genomes of any species by predicting parental alleles including paralogs. Our benchmark studies showed that HybridMine produced biologically accurate results for hybrid genomes compared to other well-established software. As proof of principle, we carried out a comprehensive structural and functional annotation of complex yeast hybrids to enable system biology prediction studies. HybridMine is developed in Python, Perl, and Bash programming languages and is available in GitHub.