Bioenergetics and the Evolution of Cellular Traits
Evolutionary processes have transformed simple cellular life into a great diversity of forms, ranging from the ubiquitous eukaryotic cell design to the more specific cellular forms of spirochetes, cyanobacteria, ciliates, heliozoans, amoeba, and many others. The cellular traits that constitute these forms require an evolutionary explanation. Ultimately, the persistence of a cellular trait depends on its net contribution to fitness, a quantitative measure. Independent of any positive effects, a cellular trait exhibits a baseline energetic cost that needs to be accounted for when quantitatively examining its net fitness effect. Here, we explore how the energetic burden introduced by a cellular trait quantitatively affects cellular fitness, describe methods for determining cell energy budgets, summarize the costs of cellular traits across the tree of life, and examine how the fitness impacts of these energetic costs compare to other evolutionary forces and trait benefits.
From the dawn of life on Earth, evolutionary forces have wrought great transformations in the cellular fabric, resulting in the great diversity of cellular forms around the planet in the present day. Ciliates, amoebozoans, choanoflagellates, heliozoans, and chlorophytes are just a small sample of this diversity within eukaryotes alone. Scenarios have been put forth to explain many cellular traits: eukaryotic flagella (60, 61), ciliature in ciliates (20), endomembranes (6, 8, 16, 17, 27, 28, 34, 35, 59), phagocytosis (11, 12), and the nucleus (36). Most of these scenarios for the evolution of cellular traits lack a quantitative accounting of the net fitness consequences of the traits and their intermediates. These net fitness changes depend not only on the downstream effects on survival and reproduction but also on a baseline fitness effect that derives from the energetic cost of a trait (48, 51).
Cellular traits are burdens as well as boons. Nothing is for free. This is as true for cells as it is in human affairs. Cellular traits are generally composed of proteins, lipids, carbohydrates, and nucleotides. When a new membrane-enveloped compartment, appendage, or metabolic pathway is introduced, resources need to be devoted to the synthesis of the constituent parts. There are two options: (a) The resources are added on top of the preexisting cellular budget, extending the time required for the cell to transport and assimilate these resources and prolonging the cell division time. (b) The resources are taken away from preexisting traits, negatively affecting survival, growth rate, or both. Under either scenario, when the contribution of a trait to net fitness is evaluated, the resource cost must be accounted for alongside the benefit.
One convenient way of accounting for the resource burden is to define the energetic cost of the trait in units of ATP hydrolyses (48, 52, 98), the primary energy currency of the cell. This brings proteins, lipids, nucleotides, and other organic molecules under the same accounting system and unifies the definitions of construction and operating costs. The construction cost consists of both direct and opportunity costs (54), with direct cost referring to chemical conversion of ATP molecules and opportunity cost referring to the potential energy associated with biomass.
Once an accounting mechanism for trait costs is established, the costs of individual traits can be examined. The composition of protein complexes can be determined by X-ray crystallography or electron microscopy, and by fluorescence microscopy for larger, more loosely organized assemblies. Protein lengths can be obtained from genome sequences and their abundances by proteomics or fluorescence microscopy. Membrane areas can be obtained from electron microscopy. Specific examples are examined in Section 5.
The impact of a trait cost on fitness depends on the total cell energy budget. A trait that costs 0.01% of the cell budget has a different impact on fitness than a trait that costs 10% of the cell budget. The cell budget depends strongly on the cell volume (48). Various methods for estimating the cell budget, with different strengths and weaknesses, are available and are discussed in Section 4. Once both a trait cost and a cell budget are known, a relative cost can be determined. The relative cost defines the negative baseline fitness impact of the newly introduced cellular trait, independent of any downstream benefits.
The fate of a new trait can be understood when the fitness cost, associated with the resource burden, is combined with the fitness benefit that the trait bestows upon the cell. This yields a net fitness change upon introducing or modifying a trait. If this net fitness change is larger than the power of genetic drift, and if the change is not opposed by a strong mutational pressure (45), the trait can sweep through the population under the power of natural selection (47, 52). An exact fitness benefit is difficult to determine given uncertainties about the ecology of single-celled species. Plausible estimates can be arrived at in some cases, and these are discussed and compared to the fitness impacts of the energetic costs.
Evolutionary change occurs gradually, often one mutation and fixation event after another. In time, this leads to the development of distinct cellular traits. For both the (immediate) small steps and the (longer term) distinct trait differences, it is important to understand the fitness impact of a change between an ancestral state and a derived state. A new trait comes with both positive and negative impacts on fitness, and the combination of these effects determines the fate of the cellular modification. Negative impacts can be chemical cross-reactivity (23), chaperone burden (40), surface area occupation (87, 88, 94), or blocking of transport in the cytoplasm (68), but here we primarily examine the resource burden in terms of the energetic costs of construction and operation. In this scenario, the fitness of a derived state is given by (51) 1. Fitness=1−Ctrait+sbenefit.
This fitness is a function of the fitness of the ancestral state normalized to 1, the relative energetic cost of the trait as a fraction of the cell budget, Ctrait, and the selection coefficient of the downstream effect of this trait on fitness, sbenefit. If the absolute value of this fitness differential exceeds the power of genetic drift, the trait can be advanced or purged by natural selection (47).
The impact of the energetic cost of a cellular trait on fitness, in Equation 1, is linearly proportional to the relative cost. This was surmised by Orgel & Crick (70), used by Lynch & Marinov (48) and Wagner (98) to understand gene evolution, and put on a more solid footing by Ilker & Hinczewski (32) (see also 51). Ideally, the impact should be tested experimentally by introducing nonfunctional proteins and examining the effect on cell fitness. An attempt at such an analysis involved a cost-benefit analysis of the lac operon (18; for earlier work, see 41). In this experiment, the lac operon expression level was altered artificially in the absence of lactose, such that the cell experienced the burden of expressing the operon but did not benefit from metabolizing lactose. Examination of the growth rate at different lac operon expression levels revealed a decrease in growth rate larger than that predicted by a linear model (18). The growth defect may have resulted in part from an increased chaperone burden or cross-reactivity of lac operon proteins with other cellular traits rather than entirely from the direct consequence of an energetic burden.
The impact of the energetic cost of a cellular trait on fitness is crucial for the comparison with trait benefits, and it is therefore worthwhile to explore this matter more thoroughly, especially in the context of the cell. The introduction of a new trait into a cell can happen by addition or replacement, and it can be put in the cytoplasm, the plasma membrane, or the external medium (Figure 1). Here, additions to and replacements of organelle membranes and lumens are treated as an addition to the cytoplasm but these may require separate treatment in the future. How and where a trait is introduced have consequences for the accounting of resource costs. For conceptual clarity, first let us consider a cell that consists of only cytoplasm and no plasma membrane or external proteins. Introducing a new trait, say, a set of enzymes that constitutes a metabolic pathway, can then have one of three effects (Figure 1): (a) The cell grows in volume (37), (b) the cell increases its density, or (c) the resource investment in preexisting traits is reduced. These effects can occur simultaneously.
The first two of these effects increase the overall cell budget and therefore the assimilation requirement for completion of the cell cycle. If the assimilation machinery is saturated and does not change in abundance, the increased cell budget causes a delay in the completion of the cell cycle and thereby contributes negatively to fitness (32, 48, 51). The delay in the completion of the cell cycle is due to a change in the ratio between, for example, ribosome abundance and protein production requirement. Such ratio effects are more general and exist for chaperones, protein-modifying enzymes, metabolic pathways for amino acids, and the production of osmoprotectants, among others. An increase in volume may also have ecological fitness-associated effects. An increase in density may slow down diffusion, causing an additional reduction in growth rate.
The third effect has simpler consequences for the cell. It lacks fitness consequences for changes in cell volume and cytoplasmic density, but it suffers from the same ratio effect that plagues the other two options. The preexisting traits lose investment and therefore lose proteins (and other components), reducing the ratio of the protein abundance of each preexisting trait to that of the entire cell.
Next, consider a cell composed of a cytoplasm enveloped by a plasma membrane (Figure 1). As above, a new cellular trait can be added, increasing the cell budget, or it can substitute (parts of) other traits, keeping the cell budget constant. The addition or substitution can happen in either the cytoplasm or the plasma membrane. The addition of a plasma membrane–localized trait would change the cell shape, increase the cell volume, or increase membrane density—depending on how the trait was added. These changes affect fitness independent of the increase in the cell budget. The substitution of preexisting plasma membrane–localized traits would cause no such baseline fitness impacts and keep the cell budget constant (Figure 1). Adding a new cellular trait to the cytoplasm increases the cytoplasmic density or requires a larger investment in the cell membrane. Again, the substitution of traits keeps all of these parameters and the cell budget constant.
Finally, consider a cell that has a cytoplasm, a plasma membrane, and external features such as appendages or excreted proteins (Figure 1). Adding a trait externally increases the cell budget but does not affect any of the other cellular parameters considered above. A substitution of (parts of) external traits with a new external trait has no effect on any of the discussed cellular parameters. Substitution of cytoplasm- and plasma membrane–localized traits with externally localized traits has no effect on the cell budget but does affect cell volume, membrane area, or compartment density, all of which may have effects on fitness unrelated to energetic costs (Figure 1).
For analytical simplicity, it is best to consider substitution scenarios, as the effects are confined to the diminishment of the fitness of the preexisting traits. Trait-addition scenarios typically feature additional fitness effects, except for the addition of an external trait (e.g., a flagellum). However, whether a new cellular trait is added to the cell budget or substituted depends on the behavior of actual cells and may differ between traits or species. It may also depend on the timescale considered. If a new trait is the consequence of a single mutation, the choice between addition or substitution depends on the regulatory system that is already in place. If the new trait emerges from an accumulation of many mutations over a long period of time, there is an opportunity to reorganize the cell by regulatory mutations and the cell will be optimized with respect to the new trait. This may involve a choice between addition or substitution as well.
Energetic costs of cellular traits have been discussed in relation to an energy limitation (1, 51, 53). However, a unicellular organism has a particular cell volume required by its ecological niche, and a particular macromolecular density because this optimizes transport rates. These factors, independently of any energy limitation, set the cell budget and the impact of the energetic cost of cellular traits on fitness. In this view, the energy limitation does not determine the cell energy budget but affects the cell division time.
The evolutionary impact of the energetic cost of a trait depends on the relative cost of this trait with respect to the cell budget. Methods such as measuring nutrient uptake rate, thermodynamic calculations, knowledge of metabolic pathways, and metabolic models have estimated aspects of cell budgets for various species (5, 22, 25, 56, 62, 66, 71, 93, 97, 100, 101). Some studies have focused on just the metabolic rate (19, 21, 24, 65), extreme energy limitation (31), how proteome cost is minimized (39), or how eukaryogenesis has affected the energy budget (15, 43, 48, 49, 63, 64). Here, we discuss four methods for estimating the energy budget of a cell: from the Pirt plot approach using chemostat data, from the heat of combustion and elemental composition of cells, from metabolic rate estimates, and from macromolecule volume fraction and membrane surface area. Each of these methods has advantages and limitations.
In the Pirt plot approach (48, 50, 73), cells are grown on glucose (or other simple substrates) in a chemostat at different dilution rates. The rate of consumption of the limiting substrate (e.g., glucose) is measured at the different dilution rates (equivalent to growth rates), and knowledge of metabolic pathways is then used to convert the glucose consumption into units of ATP, the primary energy currency of the cell. The implications of different energy metabolisms, for example, respiration versus fermentation, on the calculation of the energy budget require further examination. The conversion of glucose into units of ATP allows for the determination of a cell duplication (growth) cost and a maintenance cost, with the growth costs being incurred just once per cell cycle and the maintenance cost increasing linearly with the cell cycle time. This method requires that cells can be grown in a chemostat on a simple energy source. When cells are consuming other cells, the conversion of the organic matter to ATP equivalents becomes prohibitively complex. Fortunately, prior results from applying this approach to a diverse collection of species have been used to derive the relation between the energy budget of a cell and the cell volume (48) (Figure 2a), 2. Ccell=27Vcell0.97+0.39tdVcell0.88.
Here, the cell energy budget Ccell is in units of 109 ATP, td is the cell division time in hours, and Vcell is the cell volume in cubic microns.
The second method for estimating cell energy budgets uses the elemental composition of cells. A measure called the degree of reduction can be calculated from the number of carbon, hydrogen, and oxygen atoms in a chemical compound (50). A variant of the degree of reduction also includes phosphorus atoms (52). This degree of reduction correlates strongly with the heat of combustion and the ATP requirement for synthesizing those compounds (52). Using the cell dry weight, the mass fraction of carbon in that dry weight, and estimates of the C:H:O ratio allowed for the determination of a scaling law for ciliates (single-celled eukaryotes) that relates the cell volume to the cell budget in units of ATP (50), 3. Ccell=52Vcell0.92, where, as above, Ccell is in units of 109 ATP and Vcell is in cubic microns. Using elemental data for two bacterial species and two yeast (including phosphorus) yields a cell budget estimate of 4.7 × 1010 ATP for a 1 μm3 cell, compared to 2.7 × 1010 ATP from Equation 2, when excluding the maintenance cost (52).
A third estimate, also derived for ciliates, uses cell growth rates and oxygen consumption rates (metabolic rates). These are combined with an estimate of the ratio of energy in carbon skeletons to the metabolic rate to arrive at an estimate of (50) 4. Ccell=115Vcell0.77, with units as above. It is assumed that in these ciliates, energy metabolism is limited mostly to oxidative phosphorylation.
A fourth estimate of the cell budget derives from knowing what compartments constitute a cell and how many lipids, proteins (57, 58), and other constituents are in each compartment. This approach has been used to estimate the cost of introducing a vacuole into a cell in which cytoplasm, plasma, membrane, vacuole membrane, and cell wall are treated as separate compartments (79). Consider a bacterium, modeled after Escherichia coli, with five compartments: cytoplasm, inner membrane, periplasm, outer membrane, and external filaments. The areas and volumes of the first four compartments can be obtained from micrographs. Using a cell volume of 1 μm3, a periplasm width of 0.02 μm, and a macromolecule volume fraction of 0.16 (13, 14, 42) and including four flagella as external filaments (86), we calculate the total cell cost as 4.85 × 1010 ATP (Figure 3). Here, it is assumed that all macromolecules in the cytoplasm and periplasm are proteins. Protein costs (48) and membrane costs (49, 86) are derived from knowledge of biosynthetic pathways of amino acids and lipids. This cell budget estimate is in reasonable agreement with that of the other methods, but because it accounts for specific features associated with the cell of interest and not the average features of a diverse set of organisms, we should not expect exact agreement.
There are two advantages to this fourth method. First, it can account for species-specific differences in cytoplasmic density and abundance of external biomass, which the regression-based methods average out. Second, cytoplasmic density (91) and abundance of external biomass (e.g., filaments and excreted proteins) can be tuned in an evolutionary model and optimized together with other traits of interest that make a demand on cellular resources.
Over evolutionary time, cells can grow or shrink in volume, increase or decrease in density, modify the number of external appendages, or adjust the amount of excreted biomass—and thereby change the cell budget. They can also crank the operating cost up or down (e.g., by swimming more or less vigorously). In addition to this, the flux of nutrients into the cell, be it for energy or specific chemical elements, can change and shift the resource limitations. This can happen by a change in cell shape, nutrient transporters, swimming speed, or hunting behavior. The impact of these cell budget changes on cellular evolution remains largely unexamined.
Energy is not the only resource on which the cell depends, and it is worthwhile to consider the impact of resources. Cost-benefit analyses have been performed with respect to chemical elements such as carbon, nitrogen, and phosphorus (79, 81, 83). An explicit calculation of the fitness consequences of an elemental resource burden has not been carried out but would presumably work in the same way as for energy, except that it cannot account for maintenance cost. How fitness is affected when accounting for multiple elements and energy at the same time has yet to be examined.
When cellular parts contain the same energy density [i.e., the same amount of proteins and lipids (which differ somewhat in energy density) per unit volume], a volume ratio can be used as a proxy for cost. This has been used to determine costs of dinoflagellate flagella (83) and proto-mitochondrial symbionts (88). In a similar vein, one could examine the impact on another limiting resource, the cell surface area. This has been used to explain why bacteria enter overflow metabolism at fast growth (94, 102). It was also used, together with energetic costs, to calculate the fitness effects of primitive vesicle systems that occupy the plasma membrane during their construction (87).
The energetic costs of cellular traits are central to determining the net fitness consequences of introducing these traits. Here, we start with a summary of the energetic costs for various traits and then explain the importance of these costs in relation to effective population sizes (Ne), trait benefits, and physical models of the evolution of cellular traits. This discussion focuses on the construction costs of cellular traits. Operating costs can be a significant contributor to trait costs in some cases (86).
Understanding the cost of a cellular trait involves three questions (78): (a) How fast does it perform its task? (b) How big is it? (c) How many are there? Here, “it” is any unit that performs a task. This unit could be an individual metabolic enzyme, catalyzing a single reaction—but it could also be a motor protein, a nuclear pore complex (NPC), an endoplasmic reticulum to Golgi complex transfer vesicle, or a mitotic spindle.
The introduction of a trait burdens the cell by way of energy (48, 79, 82), elemental requirement (e.g., C, N, or P) (79, 81, 83), biomass (77), volume (83, 88), or a combination thereof. Here, we focus on energetic costs, expressed in units of ATP hydrolyses, which include both direct costs and opportunity costs (52, 54), defined in Section 1. The other methods of accounting for the cellular burden of a trait are mentioned when necessary. To make trait costs evolutionarily relevant, they must be converted into relative costs by normalizing the cell budget. Energetic costs are often reported as watts or joules, but for comparison these are converted into units of ATP, assuming 55 kJ mol–1 ATP (58), in the following examples.