Resource allocation in yeasts

One of the fascinating things about microbial life is how swiftly they can adapt themselves to very different environments and conditions - likely one of the key reasons why microbes have colonized even the most unwelcoming patches of our planet Earth. A part of this can be attributed, of course, to the extremely diverse biochemistry even a single bacterial cell can contain, but even more importantly, metabolism itself is plastic. Metabolic shifts - as we call them - are key ingredient of the microbial adaptation to changes in environmental conditions. A canonical, and, personally, close-to-heart example is yeasts starting to ferment glucose into ethanol even in the presence of oxygen in the environment, so-called Crabtree effect.

If we start computing balances, this shift looks, pardon my French, ridiculous. Saccharomyces cerevisiae is rather inefficient when it comes to using respiration (complete oxidation of carbon source to CO₂ and water) for energy generation, since it can convert 16 molecules of ADP (“spent fuel”) into ATP (“new fuel”) out of one molecule of glucose - compared to around 32-36 ATP per glucose in mammalian cells! How low can you go, you might ask? Eight-fold less, to 2 ATP per glucose, if S. cerevisiae breaks it down to ethanol. If you (as a cell) were interested in using your food efficiently, fermentation is really not the way to go.

Even though, why do such shifts happen? It turns out that the cells do the math, and do that a little bit more comprehensively than just comparing the ATP yields on glucose. The chemistry of life is highly dependent on protein catalysts (enzymes), and producing proteins is a very costly business. If we compute the protein investments to generate 1 ATP, we get to know that energy generation via fermentation requires less protein per ATP than respiration. Such behavior is what we describe using an umbrella term optimal resource allocation: the winning strategy is the one that uses available resources in the best way. Here is, therefore, the key for understanding the shift: when glucose is scarce in the environment, it is the higher ATP yield on glucose (=respiration) is favored, but as soon as glucose levels are high enough, the ATP yield on protein becomes the determining factor.

To investigate the “protein economy”, as we often call this constellation, we need to do some advanced bookkeping - ideally, as fine-grained as possible. In systems biology, the most successful type of large-scale modeling are so-called stoichiometric, or genome-scale metabolic models (GEMs). However, the conventional GEMs cannot predict the “substrate-inefficient” metabolic phenotypes unless these models are extended to capture the demand of enzymes and other macromolecular machinery to operaste the metabolic networks. My doctoral research was mostly on (i) construction and (ii) applications to learn new biology from these large-scale, fine-grained models of resource allocation - with the primary focus on yeasts as the biological agents. Nowadays, I tend more and more to get involved in discussions with colleagues on the topic of yeast cell economy, rather than do modeling myself.

Some history

Our lab has worked out the formalism of so-called “proteome-constrained”, or, in short, pc-models, with the Jens Nielsen’s lab in Chalmers University of Technology (SE). In fact, the pc-models share a lot of similarities in the formalism and implementation with the Models of Metabolism and Macromolecular Expression (ME-models), developed in early 2010s in Bernhard Palsson’s lab, and Resource Balance Analysis (RBA) models, crafted by Anne Goelzer and colleagues. I have co-authored a chapter on large-scale resource allocation models in the open textbook Economical Principles of Cell Biology.

Two tandems of PhD students/postdocs were busy with the pc-models: Yu Chen and Eunice van Pelt-KleinJan worked on the pc-model of Lactococcus lactis, pcLactis (paper). In the meantime, I and Ibrahim Elsemman were busy with developing the pc-model of Saccharomyces cerevisiae, which we named pcYeast (paper). At the time of publishing the preprint of the pcYeast paper, it was the first resource allocation model for an eukaryotic organism. The paper reviewers acknowledged the model to be very well-calibrated, and highlighted for having high predictive power, which, in turn, allowed to generate biologically meaningful predictions.

Together with the pcYeast model, our colleagues in Technical University of Delft (Angelica Rodriguez Prado, lab of Pascale Daran-Lapujade) have generated a multi-layer dataset of S. cerevisiae growth. They have cultivated yeasts in glucose-limited chemostat cultures under different dilution rates, as well as batch cultures in controlled bioreactors. For every cell culture, they have measured physiological parameters, extracellular fluxes (substrate, oxygen, CO₂, ethanol, and small organic acids), and took samples for quantitative proteomics.

Additional information

You can read more about my journey taming the giant beast - that pcYeast turned out to be - in the accompanying article of the pcYeast7.6 paper, or see the papers for the yeast pc-models, listed below, for more details.