Recombination is beneficial over the long term, allowing more effective selection. Despite long-term advantages of recombination, local recombination suppression can evolve and lead to genomic degeneration, in particular on sex chromosomes. Here, we investigated the tempo of degeneration in nonrecombining regions, that is, the function curve for the accumulation of deleterious mutations over time, leveraging on 22 independent events of recombination suppression identified on mating-type chromosomes of anther-smut fungi, including newly identified ones. Using previously available and newly generated high-quality genome assemblies of alternative mating types of 13 Microbotryum species, we estimated degeneration levels in terms of accumulation of nonoptimal codons and nonsynonymous substitutions in nonrecombining regions. We found a reduced frequency of optimal codons in the nonrecombining regions compared with autosomes, which was not due to less frequent GC-biased gene conversion or lower ancestral expression levels compared with recombining regions. The frequency of optimal codons rapidly decreased following recombination suppression and reached an asymptote after ca. 3 Ma. The strength of purifying selection remained virtually constant at dN/dS = 0.55, that is, at an intermediate level between purifying selection and neutral evolution. Accordingly, nonsynonymous differences between mating-type chromosomes increased linearly with stratum age, at a rate of 0.015 per My. We thus develop a method for disentangling the effects of reduced selection efficacy from GC-biased gene conversion in the evolution of codon usage and we quantify the tempo of degeneration in nonrecombining regions, which is important for our knowledge of genomic evolution and the maintenance of regions without recombination.

Codon usage

Optimal codons where identified as the most often used synonymous codon among the 10% most highly expressed genes in two Microbotryum species grown in two conditions. The most highly expressed genes correspond to the genes with the lowest PC1 coordinates in a within-group correspondence analysis (WCA, see http://pbil.univ-lyon1.fr/datasets/charif04/). Input data were the best TPM (from kallisto pseudo alignmet of trimmed paired-end reads on M. intermedium) per gene. We checked that expression was highly correlated between single-copy orthologous genes of Microbotryum intermedium and M. lychnidis-dioicae.

Molecular dating

Single-copy genes were assigned to genome compartments based on orthology with genes in M. lagerheimi and rearrangement scenarios. Concatenated codon-based alignments (obtained per orthologous group with macse v2.04) were imported into beauti to generate beast2's XML input files. We estimated the species and genome compartment divergence times under a calibrated Yule model and a HKY substitution model (kappa = 2, estimated frequencies) with 10,000,000 or 20,000,000 mcmc generations in beast v2.5.0. We used three unlinked site models (one per codon position). The clock model and tree parameters were kept linked. We used the split of M. lychnidis-dioicae and M. silenes-dioicae as a calibration point, with a normal prior with a mean of 0.42 Ma and sigma of 0.04.

External software used

Beast v2.5.0 https://github.com/CompEvol/beast2/releases

R 3.6.3

R packages and libraries

ade4
ape
car
cowplot
dplyr
ggplot2
ggpubr
grid
interactions
plyr
pscl
reshape2
seqinr
sjmisc
sjPlot
tidyr