Oxygen suppression of macroscopic multicellularity

Abstract

Atmospheric oxygen is thought to have played a vital role in the evolution of large, complex multicellular organisms. Challenging the prevailing theory, we show that the transition from an anaerobic to an aerobic world can strongly suppress the evolution of macroscopic multicellularity. Here we select for increased size in multicellular 'snowflake' yeast across a range of metabolically-available O₂ levels. While yeast under anaerobic and high-O₂ conditions evolved to be considerably larger, intermediate O₂ constrained the evolution of large size. Through sequencing and synthetic strain construction, we confirm that this is due to O₂-mediated divergent selection acting on organism size. We show via mathematical modeling that our results stem from nearly universal evolutionary and biophysical trade-offs, and thus should apply broadly. These results highlight the fact that oxygen is a double-edged sword: while it provides significant metabolic advantages, selection for efficient use of this resource may paradoxically suppress the evolution of macroscopic multicellular organisms.

Fig. 1. Interior cells are O₂ limited and largely incapable of respiration.

a Images of representative snowflake yeast clusters expressing the MitoLoc construct (left: under intermediate O_2, right: under supplemental O₂). Peripheral cells are actively respiring, as shown by the dual staining of preCOX4-mCherry, which only enters mitochondria with an active proton gradient, as well as pre-SU9-GFP, which labels all mitochondria. In contrast, few internal cells are respiring. b Supplementing our batch culture with additional oxygen doubled the fraction of respiring cells per cluster, from an average of 28% to an average of 56% (Mann–Whitney test, U = 262, p = 0.0003, two-tailed; n = 29 and 37 for the intermediate and supplemental O₂ treatments, respectively). The whiskers are drawn down to the 10th and up to the 90th percentiles, and data outside of the whiskers are shown as individual points. The lines in the middle of the boxes show the median values. Source data are provided as a Source Data file.

Fig. 2. The evolution of large size in snowflake yeast is constrained under low-oxygen conditions.

a Temporal dynamics of size evolution in each treatment. b Confocal images of representative clusters for each treatment after 145 transfers. Color indicates z-axis depth. c Low oxygen constrained the evolution of large size, relative to anaerobic or highly aerobic conditions. Shown here are the final mean size of snowflake yeast clusters within each of the 20 populations shown in a, plotted against the average metabolically-available pO₂ from each experimental microcosm. Line (c) is a spline with four knots. d Larger multicellular size evolved through increased cellular aspect ratio in all treatments. The aspect ratio and cluster size values are from the three ancestral genotypes (strictly aerobic, strictly anaerobic, and mixotrophic control) and 20 evolved populations. Source data are provided as a Source Data file.

Fig. 3. Testing O₂-mediated selection on size via synthetic construction.

a Engineered (Δace2+Δgin4+Δarp5) snowflake yeast were 62.7% larger in a petite (mitochondria incapable of respiration) background (n = 1920 for small petite clusters and n = 4987 for large petite clusters; p < 0.0001, F_3,10759 = 1398, Sidak’s Method in one-way ANOVA), and 29.3% larger in a grande (functional mitochondria) background (n = 2308 for small grande clusters and n = 1548 for large grande clusters; p < 0.0001, F_3,10759 = 1398, Sidak’s Method in one-way ANOVA). Error bars show mean with 95% CI. b Engineered large snowflake yeast had higher relative fitness than small clusters when oxygen was not used for growth (anaerobic metabolism) or in a high-O₂ environment (~72% pO₂) (Tukey’s HSD in one-way ANOVA, p < 0.0001, F_3,26 = 265.1). When O₂ was more limiting, selection favored small-sized snowflake yeast. Relative fitness is reported as a daily change in proportion of the competing strains. Central lines show median values, error bars show min-to-max values, and data points show the result of each independent fitness experiment, i.e., n = 9 for anaerobes, n = 5 for control mixotrophes, n = 6 strictly aerobic (Intermediate O₂), and n = 10 for strictly aerobic (High O₂). Source data are provided as a Source Data file.

Fig. 4. Trade-offs constrain the evolution of size under intermediate O₂ conditions.

We modeled the O₂-dependent evolution of size in a simple, diffusion-limited organism. a Larger organisms can develop an anaerobic core of cells (blue shading), decreasing their average growth rate. b Large size is adaptive when oxygen is absent or abundant (100% PAL), but is maladaptive when oxygen is present but cannot efficiently reach internal cells via diffusion. Blue, purple, and red lines represent oxygen diffusion distance θ of 0 (no oxygen), 1 (low oxygen), and ∞ (high oxygen), respectively. Relative to anaerobic or highly aerobic conditions, intermediate oxygen availability strongly suppresses the evolution of large size (shown as a function of O₂ diffusion depth in c, and as a function of environmental pO₂ in d). In both c and d, dashed lines denote k = 0.1, while solid lines denote k = 0.00001. With smaller values of k, organisms need to be larger to obtain the same survival benefit of increased size. In d, black, green, and red lines model the rate of O₂ consumption, φ = 46 mg s⁻¹ l⁻¹, while the blue line has φ = 4.6 mg s⁻¹ l⁻¹. The black and blue lines model the rate of O₂ diffusion, D_e = 1.12 × 10⁻⁵ cm² s⁻¹, green line D_e = 2.24 × 10^-5 cm² s⁻¹, red line D_e = 5.56 × 10⁻⁵ cm² s⁻¹. In all figures, solid lines denote k = 0.00001 and dashed lines denote k = 0.1. MATLAB code to generate each of these figures can be found in Supplementary Code 1.