Beneficial variants enriched in industrial strains span both Ras/PKA & TOR/Sch9, affecting 23/24 genes. Three genes—CYR1, SCH9, and GPB1—stand out, pointing to specific alleles under selection in industrial environments.

To tease this apart, we examined clade-restricted variants. Strikingly, variants in clades with more industrial strains were far more likely to be beneficial

However, disentangling the role of ecology from shared ancestry is challenging because of the phylogenetic structure of the 1,011-strain panel: closely related strains often share both habitats and variants.

However, the enrichment was even stronger for beneficial variants, with no enrichment for deleterious ones—consistent with measuring fitness in conditions known to select for pathway variants impacting fermentation, respiration, and starvation tolerance: www.cell.com/current-biol...

We first focused on 2,851 variants found exclusively in industrial strains (purple), and compared them to the 3,250 variants solely found in natural strains (orange). We find that industrial variants are modestly enriched for non-neutral effects.

However, surprisingly, many variants that are common in nature still have large fitness effects.
About 24% of non-neutral variants are found in >10 strains, and 13% of these show effects >2× the average across all conditions.
(top-right quadrant of the figure)

Taking into account the number of assayed variants per gene, we found beneficial variants were particularly enriched in the terminal effector kinases: Sch9 (SCH9) in TOR/Sch9 and PKA (TPK3) in Ras/PKA.

Non-neutral variants are widespread across both the Ras/PKA and TOR/Sch9 pathways, with no entire gene devoid of them.

In total, 3,673 (39%) variants show significant fitness effects. Of these, 659 were beneficial in ≥1 conditions (green), while the rest were exclusively deleterious when non-neutral (red).

We then used these frequency trajectories to estimate fitness effects of each variant employing BarBay, a Bayesian approach developed by the amazing @mrazo.bsky.social:
journals.plos.org/ploscompbiol...

Example from the 2D condition:
Ras2 G18D (beneficial, next to human RAS G12X oncogenic hotspot) rises in frequency, while Tor1 G1088E (deleterious) declines. Non-editing neutral controls (orange) drop slightly as beneficial variants drive up mean population fitness.

These conditions span key physiological states:
1D → active fermentation
2D → fermentation + respiration
3D/5D → fermentation + respiration + stationary
Such conditions are known to expose fitness effects of adaptive mutations in these pathways.

To probe Ras/PKA & TOR/Sch9 function, we competed our edited pool in minimal medium with limiting glucose. We ran growth competitions under 4 conditions—1D, 2D, 3D, 5D—each differing in how long cells grew before transfer. 3 replicates per condition.

To our knowledge this effort represents the first comprehensive pathway-wide, nucleotide-level fitness landscape of natural variation in any signaling pathway.

With CRISPEY-BAR+, we assayed nearly two-thirds of all polymorphisms known to be segregating in Ras/PKA & TOR/Sch9, drawn from 1,011 globally diverse yeast strains, including both wild (yellow) and domesticated lineages (purple).

We focused on two ancient, conserved signaling pathways—Ras/PKA (blue) and TOR/Sch9 (green)—that control cell growth and proliferation in response to environmental cues across eukaryotes. These pathways are frequent targets of adaptation, and mutations in them drive ~1/3 of human cancers.

The alternative—balance theory—argues that some high-frequency variants are maintained because they matter. They can be advantageous in some environments but costly in others, staying in populations through mechanisms like local adaptation or frequency-dependent selection.

The nearly neutral theory says that most common variants within a species are effectively neutral. Strongly deleterious mutations are quickly removed by selection, so anything common is expected to have little or no effect on fitness—a view that has long served as the field’s null model.