Extraordinary molecular evolution in the PRDM9 fertility gene

Abstract

Recent work indicates that allelic incompatibility in the mouse PRDM9 (Meisetz) gene can cause hybrid male sterility, contributing to genetic isolation and potentially speciation. The only phenotype of mouse PRDM9 knockouts is a meiosis I block that causes sterility in both sexes. The PRDM9 gene encodes a protein with histone H3(K4) trimethyltransferase activity, a KRAB domain, and a DNA-binding domain consisting of multiple tandem C2H2 zinc finger (ZF) domains. We have analyzed human coding polymorphism and interspecies evolutionary changes in the PRDM9 gene. The ZF domains of PRDM9 are evolving very rapidly, with compelling evidence of positive selection in primates. Positively selected amino acids are predominantly those known to make nucleotide specific contacts in C2H2 zinc fingers. These results suggest that PRDM9 is subject to recurrent selection to change DNA-binding specificity. The human PRDM9 protein is highly polymorphic in its ZF domains and nearly all polymorphisms affect the same nucleotide contact residues that are subject to positive selection. ZF domain nucleotide sequences are strongly homogenized within species, indicating that interfinger recombination contributes to their evolution. PRDM9 has previously been assumed to be a transcription factor required to induce meiosis specific genes, a role that is inconsistent with its molecular evolution. We suggest instead that PRDM9 is involved in some aspect of centromere segregation conflict and that rapidly evolving centromeric DNA drives changes in PRDM9 DNA-binding domains.

Figure 1. Sites of positive selection in primate PRDM9 genes.

Schematic of PRDM9 protein showing KRAB, SSXRD (SSX repression domain), SET, and ZF domains. The ZF domain region of five aligned primate proteins is shown expanded and split into three sections, with blue shading proportional to amino acid conservation. The 12 human zinc fingers are boxed with thickened lines showing the DNA-binding turn-helix and its three major nucleotide contact residues. Below the alignment is a histogram of the Bayes-Empirical-Bayes estimate of d_N/d_S ratio for each codon as computed by codeml. Filled red squares above the histogram indicate codons where the P-value for positive selection was 0.95 or more. Filled green circles indicate the positions of all known nonsynonymous human polymorphisms; positions with more than one mark indicate more than one distinct nucleotide change. Wavy red lines between the chimp and human indicate positions at which the reference protein sequence for the two species differ. Regions of the alignment with only two sequences are paled to indicate that the d_N/d_S estimate is poorly informed.

Figure 2. Human nonsynonymous polymorphisms according to position in their zinc finger.

The amino acid sequence below shows the defining C2H2 zinc coordinating residues and their invariant spacing. The thickened line shows the DNA-binding turn-helix and its three major nucleotide contact residues. The open circles indicate the positions of each distinct nonsynonymous polymorphism, which are strongly associated with DNA-binding residues.

Figure 3. Zinc finger sequences are homogenized within genes and divergent between genes.

Panel A. Self dot plot of all tandem 21 amino acid zinc fingers from PRDM9 orthologs in all 15 species with 5 or more zinc fingers. Dot shading corresponds to sequence similarity using the BLOSUM62 matrix, scaled so that the lowest scoring match is white and the highest scoring is black. Lines separate fingers from each species. Sequences are arranged by their taxonomic name and within each gene fingers are in order by position (N-terminal zinc finger first). Panel B. The same PRDM9 fingers with their order randomized, showing that the appearance of self-similarity in panel A is significant. Multiple randomizations were inspected; this one is typical. Panel C. A maximum-likelihood tree of the PRDM orthologs based on the protein encoded by their entire ZF encoding exon. The protein tree approximates the species phylogenetic tree, supporting orthology of the PRDM9 genes. Panel D. Self dot plot of all 21 amino acid zinc fingers from 13 ZNF773 orthologs (representative of other KRAB zinc finger genes, data not shown). All species except one have 9 zinc fingers (the additional divergent zinc finger in armadillo (species 4) is N-terminal and not tandem with the 9 orthologous fingers). Each zinc finger is closely related to its orthologous fingers from the other species and divergent from the other zinc fingers in the same gene. The species in figure order are cow (species 1), dog (species 2), marmoset (etc.), armadillo, horse, cat, human, macaque, rabbit, baboon, orangutan, chimpanzee, and ground squirrel. Because many of the PRDM9 genes come from low-coverage assemblies it is difficult to find a representative zinc finger gene from all of the same species as shown in panel A. The species are domestic cow (Bos taurus), common marmoset (Callithrix jacchus), domestic dog (Canis familiaris), guinea pig (Cavia porcellus), armadillo (Dasypus novemcinctus), lesser tenrec (Echinops telfairi), domestic horse (Equus cabalus), domestic cat (Felis catus), human (Homo sapiens), African elephant (Loxodonta africana), Rhesus macaque (Macaca mulatta), mouse lemur (Otolemur garnettii), domestic mouse (Mus musculus), domestic rabbit (Oryctolagus cuniculus), Hamadryas baboon (Papio hamadryas), Sumatran orangutan (Pongo pygmaeus abelii), chimpanzee (Pan troglodytes), flying fox bat (Pteropus vampyrus), Norway rat (Rattus norvegicus), thirteen-lined ground squirrel (Spermophilus tridecemlineatus), and bottlenose dolphin (Tursiops truncatus).

Figure 4. Lack of synonymous site differences among fingers.

Codon alignment of the 12 human (top) and 11 mouse (bottom) zinc fingers including the upstream 7-codon linker region (reference genome sequences). Blue shading is proportional to the frequency of the residue in its aligned column. For all codons, except the three major nucleotide contacts, the predominant encoded amino acid is shown below the codon, with the zinc-coordinating residues underlined. The next two lines indicate the number of possible codons for the predominant amino acid (degeneracy) and the number of synonymous changes observed among any of the fingers. The striking lack of synonymous divergence between fingers strongly supports recombination among fingers. At the major nucleotide contact residues (arrows) all amino acids encoded by at least two fingers are listed.