POLYMORPHISM AND CONCERTED EVOLUTION IN A TANDEMLY REPEATED GENE FAMILY - 5S RIBOSOMAL DNA IN DIPLOID AND ALLOPOLYPLOID COTTONS

5S RNA genes and their nontranscribed spacers are tandemly repeated in plant genomes at one or more chromosomal loci. To facilitate an under standing of the forces that govern 5S rDNA evolution, copy-number esti mation and DNA sequencing were conducted for a phylogenetically well-c haracterized set of 16 diploid species of cotton (Gossypium) and 4 spe cies representing allopolyploid derivatives of the diploids. Copy numb er varies over twentyfold in the genus, from approximately 1,000 to 20 ,000 copies/2C genome. When superimposed on the organismal phylogeny, these data reveal examples of both array expansion and contraction. Ac ross species, a mean of 12% of nucleotide positions are polymorphic wi thin individual arrays, for both gene and spacer sequences. This shows , in conjunction with phylo,genetic evidence for ancestral polymorphis ms that survive speciation events, that intralocus concerted evolution ary forces are relatively weak and that the rate of interrepeat homoge nization is approximately equal to the rate of speciation. Evidence pr esented also shows that duplicated 5S rDNA arrays in allopolyploids ha ve retained their subgenomic identity since polyploid formation, there by indicating that interlocus concerted evolution has not been an impo rtant factor in the evolution of these arrays. A descriptive model, on e which incorporates the opposing forces of mutation and homogenizatio n within a selective framework, is outlined to account for the empiric al data presented. Weak homogenizing forces allow equivalent levels of sequence polymorphism to accumulate in the 5S gene and spacer sequenc es, but fixation of mutations is nearly prohibited in the 5S gene. As a consequence, fixed interspecific differences are statistically under represented for 5S genes. This result explains the apparent paradox th at despite similar levels of gene and spacer diversity, phylogenetic a nalysis of spacer sequences yields highly resolved trees, whereas anal yses based on 5S gene sequences do not.