teemi: An open-source literate programming approach for iterative design-build-test-learn cycles in bioengineering.

Synthetic biology dictates the data-driven engineering of biocatalysis, cellular functions, and organism behavior. Integral to synthetic biology is the aspiration to efficiently find, access, interoperate, and reuse high-quality data on genotype-phenotype relationships of native and engineered biosystems under FAIR principles, and from this facilitate forward-engineering strategies. However, biology is complex at the regulatory level, and noisy at the operational level, thus necessitating systematic and diligent data handling at all levels of the design, build, and test phases in order to maximize learning in the iterative design-build-test-learn engineering cycle. To enable user-friendly simulation, organization, and guidance for the engineering of biosystems, we have developed an open-source python-based computer-aided design and analysis platform operating under a literate programming user-interface hosted on Github. The platform is called teemi and is fully compliant with FAIR principles. In this study we apply teemi for i) designing and simulating bioengineering, ii) integrating and analyzing multivariate datasets, and iii) machine-learning for predictive engineering of metabolic pathway designs for production of a key precursor to medicinal alkaloids in yeast. The teemi platform is publicly available at PyPi and GitHub.

CRI-SPA: a high-throughput method for systematic genetic editing of yeast libraries.

Biological functions are orchestrated by intricate networks of interacting genetic elements. Predicting the interaction landscape remains a challenge for systems biology and new research tools allowing simple and rapid mapping of sequence to function are desirable. Here, we describe CRI-SPA, a method allowing the transfer of chromosomal genetic features from a CRI-SPA Donor strain to arrayed strains in large libraries of Saccharomyces cerevisiae. CRI-SPA is based on mating, CRISPR-Cas9-induced gene conversion, and Selective Ploidy Ablation. CRI-SPA can be massively parallelized with automation and can be executed within a week. We demonstrate the power of CRI-SPA by transferring four genes that enable betaxanthin production into each strain of the yeast knockout collection (≈4800 strains). Using this setup, we show that CRI-SPA is highly efficient and reproducible, and even allows marker-free transfer of genetic features. Moreover, we validate a set of CRI-SPA hits by showing that their phenotypes correlate strongly with the phenotypes of the corresponding mutant strains recreated by reverse genetic engineering. Hence, our results provide a genome-wide overview of the genetic requirements for betaxanthin production. We envision that the simplicity, speed, and reliability offered by CRI-SPA will make it a versatile tool to forward systems-level understanding of biological processes.

Engineering of Pseudomonas putida for accelerated co-utilization of glucose and cellobiose yields aerobic overproduction of pyruvate explained by an upgraded metabolic model.

Pseudomonas putida KT2440 is an attractive bacterial host for biotechnological production of valuable chemicals from renewable lignocellulosic feedstocks as it can valorize lignin-derived aromatics or glucose obtainable from cellulose. P. putida EM42, a genome-reduced variant of strain KT2440 endowed with advantageous physiological properties, was recently engineered for growth on cellobiose, a major cellooligosaccharide product of enzymatic cellulose hydrolysis. Co-utilization of cellobiose and glucose was achieved in a mutant lacking periplasmic glucose dehydrogenase Gcd (PP_1444). However, the cause of the co-utilization phenotype remained to be understood and the Δgcd strain had a significant growth defect. In this study, we investigated the basis of the simultaneous uptake of the two sugars and accelerated the growth of P. putida EM42 Δgcd mutant for the bioproduction of valuable compounds from glucose and cellobiose. We show that the gcd deletion lifted the inhibition of the exogenous ß-glucosidase BglC from Thermobifida fusca exerted by the intermediates of the periplasmic glucose oxidation pathway. The additional deletion of hexR gene, which encodes a repressor of the upper glycolysis genes, failed to restore rapid growth on glucose. The reduced growth rate of the Δgcd mutant was partially compensated by the implantation of heterologous glucose and cellobiose transporters (Glf from Zymomonas mobilis and LacY from Escherichia coli, respectively). Remarkably, this intervention resulted in the accumulation of pyruvate in aerobic P. putida cultures. We demonstrated that the excess of this key metabolic intermediate can be redirected to the enhanced biosynthesis of ethanol and lactate. The pyruvate overproduction phenotype was then unveiled by an upgraded genome-scale metabolic model constrained with proteomic and kinetic data. The model pointed to the saturation of glucose catabolism enzymes due to unregulated substrate uptake and it predicted improved bioproduction of pyruvate-derived chemicals by the engineered strain. This work sheds light on the co-metabolism of cellulosic sugars in an attractive biotechnological host and introduces a novel strategy for pyruvate overproduction in bacterial cultures under aerobic conditions.

Laboratory evolution reveals general and specific tolerance mechanisms for commodity chemicals.

Although strain tolerance to high product concentrations is a barrier to the economically viable biomanufacturing of industrial chemicals, chemical tolerance mechanisms are often unknown. To reveal tolerance mechanisms, an automated platform was utilized to evolve Escherichia coli to grow optimally in the presence of 11 industrial chemicals (1,2-propanediol, 2,3-butanediol, glutarate, adipate, putrescine, hexamethylenediamine, butanol, isobutyrate, coumarate, octanoate, hexanoate), reaching tolerance at concentrations 60%-400% higher than initial toxic levels. Sequencing genomes of 223 isolates from 89 populations, reverse engineering, and cross-compound tolerance profiling were employed to uncover tolerance mechanisms. We show that: 1) cells are tolerized via frequent mutation of membrane transporters or cell wall-associated proteins (e.g., ProV, KgtP, SapB, NagA, NagC, MreB), transcription and translation machineries (e.g., RpoA, RpoB, RpoC, RpsA, RpsG, NusA, Rho), stress signaling proteins (e.g., RelA, SspA, SpoT, YobF), and for certain chemicals, regulators and enzymes in metabolism (e.g., MetJ, NadR, GudD, PurT); 2) osmotic stress plays a significant role in tolerance when chemical concentrations exceed a general threshold and mutated genes frequently overlap with those enabling chemical tolerance in membrane transporters and cell wall-associated proteins; 3) tolerization to a specific chemical generally improves tolerance to structurally similar compounds whereas a tradeoff can occur on dissimilar chemicals, and 4) using pre-tolerized starting isolates can hugely enhance the subsequent production of chemicals when a production pathway is inserted in many, but not all, evolved tolerized host strains, underpinning the need for evolving multiple parallel populations. Taken as a whole, this study provides a comprehensive genotype-phenotype map based on identified mutations and growth phenotypes for 223 chemical tolerant isolates.

Engineering precursor supply for the high-level production of ergothioneine in Saccharomyces cerevisiae.

Ergothioneine (ERG) is an unusual sulfur-containing amino acid. It is a potent antioxidant, which shows great potential for ameliorating neurodegenerative and cardiovascular diseases. L-ergothioneine is rare in nature, with mushrooms being the primary dietary source. The chemical synthesis process is complex and expensive. Alternatively, ERG can be produced by fermentation of recombinant microorganisms engineered for ERG overproduction. Here, we describe the engineering of S. cerevisiae for high-level ergothioneine production on minimal medium with glucose as the only carbon source. To this end, metabolic engineering targets in different layers of the amino acid metabolism were selected based on literature and tested. Out of 28 targets, nine were found to improve ERG production significantly by 10%-51%. These targets were then sequentially implemented to generate an ergothioneine-overproducing yeast strain capable of producing 106.2 ± 2.6 mg/L ERG in small-scale cultivations. Transporter engineering identified that the native Aqr1 transporter was capable of increasing the ERG production in a yeast strain with two copies of the ERG biosynthesis pathway, but not in the strain that was further engineered for improved precursor supply. Medium optimization indicated that additional supplementation of pantothenate improved the strain's productivity further and that no supplementation of amino acid precursors was necessary. Finally, the engineered strain produced 2.39 ± 0.08 g/L ERG in 160 h (productivity of 14.95 ± 0.49 mg/L/h) in a controlled fed-batch fermentation without supplementation of amino acids. This study paves the way for the low-cost fermentation-based production of ergothioneine.

Emergence of Phenotypically Distinct Subpopulations Is a Factor in Adaptation of Recombinant Saccharomyces cerevisiae under Glucose-Limited Conditions.

Cells cultured in a nutrient-limited environment can undergo adaptation, which confers improved fitness under long-term energy limitation. We have shown previously how a recombinant Saccharomyces cerevisiae strain, producing a heterologous insulin product, under glucose-limited conditions adapts over time at the average population level. Here, we investigated this adaptation at the single-cell level by application of fluorescence-activated cell sorting (FACS) and showed that the following three apparent phenotypes underlie the adaptive response observed at the bulk level: (i) cells that drastically reduced insulin production (23%), (ii) cells with reduced enzymatic capacity in central carbon metabolism (46%), and (iii) cells that exhibited pseudohyphal growth (31%). We speculate that the phenotypic heterogeneity is a result of different mechanisms to increase fitness. Cells with reduced insulin productivity have increased fitness by reducing the burden of the heterologous insulin production, and the populations with reduced enzymatic capacity of the central carbon metabolism and pseudohyphal growth have increased fitness toward the glucose-limited conditions. The results highlight the importance of considering population heterogeneity when studying adaptation and evolution. IMPORTANCE The yeast Saccharomyces cerevisiae is an attractive microbial host for industrial production and is used widely for manufacturing, e.g., pharmaceuticals. Chemostat cultivation mode is an efficient cultivation strategy for industrial production processes as it ensures a constant, well-controlled cultivation environment. Nevertheless, both the production of a heterologous product and the constant cultivation environment in the chemostat impose a selective pressure on the production organism, which may result in adaptation and loss of productivity. The exact mechanisms behind the observed adaptation and loss of performance are often unidentified. We used a recombinant S. cerevisiae strain producing heterologous insulin and investigated the adaptation occurring during chemostat growth at the single-cell level. We showed that three apparent phenotypes underlie the adaptive response observed at the bulk level in the chemostat. These findings highlight the importance of considering population heterogeneity when studying adaptation in industrial bioprocesses.

Genome-scale metabolic modeling of P. thermoglucosidasius NCIMB 11955 reveals metabolic bottlenecks in anaerobic metabolism.

Parageobacillus thermoglucosidasius represents a thermophilic, facultative anaerobic bacterial chassis, with several desirable traits for metabolic engineering and industrial production. To further optimize strain productivity, a systems level understanding of its metabolism is needed, which can be facilitated by a genome-scale metabolic model. Here, we present p-thermo, the most complete, curated and validated genome-scale model (to date) of Parageobacillus thermoglucosidasius NCIMB 11955. It spans a total of 890 metabolites, 1175 reactions and 917 metabolic genes, forming an extensive knowledge base for P. thermoglucosidasius NCIMB 11955 metabolism. The model accurately predicts aerobic utilization of 22 carbon sources, and the predictive quality of internal fluxes was validated with previously published 13C-fluxomics data. In an application case, p-thermo was used to facilitate more in-depth analysis of reported metabolic engineering efforts, giving additional insight into fermentative metabolism. Finally, p-thermo was used to resolve a previously uncharacterised bottleneck in anaerobic metabolism, by identifying the minimal required supplemented nutrients (thiamin, biotin and iron(III)) needed to sustain anaerobic growth. This highlights the usefulness of p-thermo for guiding the generation of experimental hypotheses and for facilitating data-driven metabolic engineering, expanding the use of P. thermoglucosidasius as a high yield production platform.

Fluctuations in glucose availability prevent global proteome changes and physiological transition during prolonged chemostat cultivations of Saccharomyces cerevisiae.

Chemostat cultivation mode imposes selective pressure on the cells, which may result in slow adaptation in the physiological state over time. We applied a two-compartment scale-down chemostat system imposing feast-famine conditions to characterize the long-term (100 s of hours) response of Saccharomyces cerevisiae to fluctuating glucose availability. A wild-type strain and a recombinant strain, expressing an insulin precursor, were cultured in the scale-down system, and analyzed at the physiological and proteomic level. Phenotypes of both strains were compared with those observed in a well-mixed chemostat. Our results show that S. cerevisiae subjected to long-term chemostat conditions undergoes a global reproducible shift in its cellular state and that this transition occurs faster and is larger in magnitude for the recombinant strain including a significant decrease in the expression of the insulin product. We find that the transition can be completely avoided in the presence of fluctuations in glucose availability as the strains subjected to feast-famine conditions under otherwise constant culture conditions exhibited constant levels of the measured proteome for over 250 hr. We hypothesize possible mechanisms responsible for the observed phenotypes and suggest experiments that could be used to test these mechanisms.

Combined In-Cell NMR and Simulation Approach to Probe Redox-Dependent Pathway Control.

Dynamic response of intracellular reaction cascades to changing environments is a hallmark of living systems. As metabolism is complex, mechanistic models have gained popularity for describing the dynamic response of cellular metabolism and for identifying target genes for engineering. At the same time, the detailed tracking of transient metabolism in living cells on the subminute time scale has become amenable using dynamic nuclear polarization-enhanced 13C NMR. Here, we suggest an approach combining in-cell NMR spectroscopy with perturbation experiments and modeling to obtain evidence that the bottlenecks of yeast glycolysis depend on intracellular redox state. In pre-steady-state glycolysis, pathway bottlenecks shift from downstream to upstream reactions within a few seconds, consistent with a rapid decline in the NAD+/NADH ratio. Simulations using mechanistic models reproduce the experimentally observed response and help identify unforeseen biochemical events. Remaining inaccuracies in the computational models can be identified experimentally. The combined use of rapid injection NMR spectroscopy and in silico simulations provides a promising method for characterizing cellular reactions with increasing mechanistic detail.

MARSI: metabolite analogues for rational strain improvement.

Summary: Metabolite analogues (MAs) mimic the structure of native metabolites, can competitively inhibit their utilization in enzymatic reactions, and are commonly used as selection tools for isolating desirable mutants of industrial microorganisms. Genome-scale metabolic models representing all biochemical reactions in an organism can be used to predict effects of MAs on cellular phenotypes. Here, we present the metabolite analogues for rational strain improvement (MARSI) framework. MARSI provides a rational approach to strain improvement by searching for metabolites as targets instead of genes or reactions. The designs found by MARSI can be implemented by supplying MAs in the culture media, enabling metabolic rewiring without the use of recombinant DNA technologies that cannot always be used due to regulations. To facilitate experimental implementation, MARSI provides tools to identify candidate MAs to a target metabolite from a database of known drugs and analogues. Availability and implementation: The code is freely available at https://github.com/biosustain/marsi under the Apache License V2. MARSI is implemented in Python. Supplementary information: Supplementary data are available at Bioinformatics online.

Integration of enzymatic data in Bacillus subtilis genome-scale metabolic model improves phenotype predictions and enables in silico design of poly-γ-glutamic acid production strains.

BACKGROUND: Genome-scale metabolic models (GEMs) allow predicting metabolic phenotypes from limited data on uptake and secretion fluxes by defining the space of all the feasible solutions and excluding physio-chemically and biologically unfeasible behaviors. The integration of additional biological information in genome-scale models, e.g., transcriptomic or proteomic profiles, has the potential to improve phenotype prediction accuracy. This is particularly important for metabolic engineering applications where more accurate model predictions can translate to more reliable model-based strain design. RESULTS: Here we present a GEM with Enzymatic Constraints using Kinetic and Omics data (GECKO) model of Bacillus subtilis, which uses publicly available proteomic data and enzyme kinetic parameters for central carbon (CC) metabolic reactions to constrain the flux solution space. This model allows more accurate prediction of the flux distribution and growth rate of wild-type and single-gene/operon deletion strains compared to a standard genome-scale metabolic model. The flux prediction error decreased by 43% and 36% for wild-type and mutants respectively. The model additionally increased the number of correctly predicted essential genes in CC pathways by 2.5-fold and significantly decreased flux variability in more than 80% of the reactions with variable flux. Finally, the model was used to find new gene deletion targets to optimize the flux toward the biosynthesis of poly-γ-glutamic acid (γ-PGA) polymer in engineered B. subtilis. We implemented the single-reaction deletion targets identified by the model experimentally and showed that the new strains have a twofold higher γ-PGA concentration and production rate compared to the ancestral strain. CONCLUSIONS: This work confirms that integration of enzyme constraints is a powerful tool to improve existing genome-scale models, and demonstrates the successful use of enzyme-constrained models in B. subtilis metabolic engineering. We expect that the new model can be used to guide future metabolic engineering efforts in the important industrial production host B. subtilis.

Escher: A Web Application for Building, Sharing, and Embedding Data-Rich Visualizations of Biological Pathways.

Escher is a web application for visualizing data on biological pathways. Three key features make Escher a uniquely effective tool for pathway visualization. First, users can rapidly design new pathway maps. Escher provides pathway suggestions based on user data and genome-scale models, so users can draw pathways in a semi-automated way. Second, users can visualize data related to genes or proteins on the associated reactions and pathways, using rules that define which enzymes catalyze each reaction. Thus, users can identify trends in common genomic data types (e.g. RNA-Seq, proteomics, ChIP)--in conjunction with metabolite- and reaction-oriented data types (e.g. metabolomics, fluxomics). Third, Escher harnesses the strengths of web technologies (SVG, D3, developer tools) so that visualizations can be rapidly adapted, extended, shared, and embedded. This paper provides examples of each of these features and explains how the development approach used for Escher can be used to guide the development of future visualization tools.

Evolution reveals a glutathione-dependent mechanism of 3-hydroxypropionic acid tolerance.

Biologically produced 3-hydroxypropionic acid (3 HP) is a potential source for sustainable acrylates and can also find direct use as monomer in the production of biodegradable polymers. For industrial-scale production there is a need for robust cell factories tolerant to high concentration of 3 HP, preferably at low pH. Through adaptive laboratory evolution we selected S. cerevisiae strains with improved tolerance to 3 HP at pH 3.5. Genome sequencing followed by functional analysis identified the causal mutation in SFA1 gene encoding S-(hydroxymethyl)glutathione dehydrogenase. Based on our findings, we propose that 3 HP toxicity is mediated by 3-hydroxypropionic aldehyde (reuterin) and that glutathione-dependent reactions are used for reuterin detoxification. The identified molecular response to 3 HP and reuterin may well be a general mechanism for handling resistance to organic acid and aldehydes by living cells.

Probing efficient microbial CO2 utilisation through metabolic and process modelling.

Acetogenic gas fermentation is increasingly studied as a promising technology to upcycle carbon-rich waste gasses. Currently the product range is limited, and production yields, rates and titres for a number of interesting products do not allow for economically viable processes. By pairing process modelling and host-agnostic metabolic modelling, we compare fermentation conditions and various products to optimise the processes. The models were then used in a simulation of an industrial-scale bubble column reactor. We find that increased temperatures favour gas transfer rates, particularly for the valuable and limiting H2 , while furthermore predicting an optimal feed composition of 9:1 mol H2 to mol CO2 . Metabolically, the increased non-growth associated maintenance requirements of thermophiles favours the formation of catabolic products. To assess the expansion of the product portfolio beyond acetate, both a product volatility analysis and a metabolic pathway model were implemented. In-situ recovery of volatile products is shown to be within range for acetone but challenging due to the extensive evaporation of water, while the direct production of more valuable compounds by acetogens is metabolically unfavourable compared to acetate and ethanol. We discuss alternative approaches to overcome these challenges to utilise acetogenic CO2 fixation to produce a wider range of carbon negative chemicals.

Bottom-up parameterization of enzyme rate constants: Reconciling inconsistent data.

Kinetic models of metabolism are promising platforms for studying complex metabolic systems and designing production strains. Given the availability of enzyme kinetic data from historical experiments and machine learning estimation tools, a straightforward modeling approach is to assemble kinetic data enzyme by enzyme until a desired scale is reached. However, this type of 'bottom up' parameterization of kinetic models has been difficult due to a number of issues including gaps in kinetic parameters, the complexity of enzyme mechanisms, inconsistencies between parameters obtained from different sources, and in vitro-in vivo differences. Here, we present a computational workflow for the robust estimation of kinetic parameters for detailed mass action enzyme models while taking into account parameter uncertainty. The resulting software package, termed MASSef (the Mass Action Stoichiometry Simulation Enzyme Fitting package), can handle standard 'macroscopic' kinetic parameters, including Km, kcat, Ki, Keq, and nh, as well as diverse reaction mechanisms defined in terms of mass action reactions and 'microscopic' rate constants. We provide three enzyme case studies demonstrating that this approach can identify and reconcile inconsistent data either within in vitro experiments or between in vitro and in vivo enzyme function. We further demonstrate how parameterized enzyme modules can be used to assemble pathway-scale kinetic models consistent with in vivo behavior. This work builds on the legacy of knowledge on kinetic behavior of enzymes by enabling robust parameterization of enzyme kinetic models at scale utilizing the abundance of historical literature data and machine learning parameter estimates.

Simultaneous application of enzyme and thermodynamic constraints to metabolic models using an updated Python implementation of GECKO.

IMPORTANCE: The metabolism of biological cells is an intricate network of reactions that interconvert chemical compounds, gathering energy, and using that energy to grow. The static analysis of these metabolic networks can be turned into a computational model that can efficiently output the distribution of fluxes in the network. With the inclusion of enzymes in the network, we can also interpret the role and concentrations of the metabolic proteins. However, the models and the experimental data often clash, resulting in a network that cannot grow. Here, we tackle this situation with a suite of relaxation algorithms in a package called geckopy. Geckopy also integrates with other software to allow for adding thermodynamic and metabolomic constraints. In addition, to ensure that enzyme-constrained models follow the community standards, a format for the proteins is postulated. We hope that the package and algorithms presented here will be useful for the constraint-based modeling community.

Reconstruction of a catalogue of genome-scale metabolic models with enzymatic constraints using GECKO 2.0.

Genome-scale metabolic models (GEMs) have been widely used for quantitative exploration of the relation between genotype and phenotype. Streamlined integration of enzyme constraints and proteomics data into such models was first enabled by the GECKO toolbox, allowing the study of phenotypes constrained by protein limitations. Here, we upgrade the toolbox in order to enhance models with enzyme and proteomics constraints for any organism with a compatible GEM reconstruction. With this, enzyme-constrained models for the budding yeasts Saccharomyces cerevisiae, Yarrowia lipolytica and Kluyveromyces marxianus are generated to study their long-term adaptation to several stress factors by incorporation of proteomics data. Predictions reveal that upregulation and high saturation of enzymes in amino acid metabolism are common across organisms and conditions, suggesting the relevance of metabolic robustness in contrast to optimal protein utilization as a cellular objective for microbial growth under stress and nutrient-limited conditions. The functionality of GECKO is expanded with an automated framework for continuous and version-controlled update of enzyme-constrained GEMs, also producing such models for Escherichia coli and Homo sapiens. In this work, we facilitate the utilization of enzyme-constrained GEMs in basic science, metabolic engineering and synthetic biology purposes.

A Genome-Scale Metabolic Model of Methanoperedens nitroreducens: Assessing Bioenergetics and Thermodynamic Feasibility.

Methane is an abundant low-carbon fuel that provides a valuable energy resource, but it is also a potent greenhouse gas. Therefore, anaerobic oxidation of methane (AOM) is an essential process with central features in controlling the carbon cycle. Candidatus 'Methanoperedens nitroreducens' (M. nitroreducens) is a recently discovered methanotrophic archaeon capable of performing AOM via a reverse methanogenesis pathway utilizing nitrate as the terminal electron acceptor. Recently, reverse methanogenic pathways and energy metabolism among anaerobic methane-oxidizing archaea (ANME) have gained significant interest. However, the energetics and the mechanism for electron transport in nitrate-dependent AOM performed by M. nitroreducens is unclear. This paper presents a genome-scale metabolic model of M. nitroreducens, iMN22HE, which contains 813 reactions and 684 metabolites. The model describes its cellular metabolism and can quantitatively predict its growth phenotypes. The essentiality of the cytoplasmic heterodisulfide reductase HdrABC in the reverse methanogenesis pathway is examined by modeling the electron transfer direction and the specific energy-coupling mechanism. Furthermore, based on better understanding electron transport by modeling, a new energy transfer mechanism is suggested. The new mechanism involves reactions capable of driving the endergonic reactions in nitrate-dependent AOM, including the step reactions in reverse canonical methanogenesis and the novel electron-confurcating reaction HdrABC. The genome metabolic model not only provides an in silico tool for understanding the fundamental metabolism of ANME but also helps to better understand the reverse methanogenesis energetics and its thermodynamic feasibility.

Whole-cell modeling in yeast predicts compartment-specific proteome constraints that drive metabolic strategies.

When conditions change, unicellular organisms rewire their metabolism to sustain cell maintenance and cellular growth. Such rewiring may be understood as resource re-allocation under cellular constraints. Eukaryal cells contain metabolically active organelles such as mitochondria, competing for cytosolic space and resources, and the nature of the relevant cellular constraints remain to be determined for such cells. Here, we present a comprehensive metabolic model of the yeast cell, based on its full metabolic reaction network extended with protein synthesis and degradation reactions. The model predicts metabolic fluxes and corresponding protein expression by constraining compartment-specific protein pools and maximising growth rate. Comparing model predictions with quantitative experimental data suggests that under glucose limitation, a mitochondrial constraint limits growth at the onset of ethanol formation-known as the Crabtree effect. Under sugar excess, however, a constraint on total cytosolic volume dictates overflow metabolism. Our comprehensive model thus identifies condition-dependent and compartment-specific constraints that can explain metabolic strategies and protein expression profiles from growth rate optimisation, providing a framework to understand metabolic adaptation in eukaryal cells.