DNA sequence comparisons of a 1200-base pair (bp) region in 14 human fetal globin genes in seven linked pairs reveal 31 nucleotide substitutions at positions where the fetal globin genes, G, and A,, usually differ. In each case, the newly substituted nucleotide is identical to the one found at the same position in the linked nonallelic gene. Most of these nucleotide substitutions are clearly the result of gene conversions, but 11 could be the result of either very short gene conversions or of point mutations. T h e unexpectedly frequent occurrence of these short gene conversions suggests that they may be the relics of some normal interaction between homologous but nonallelic DNA sequences, and we discuss the possibility that they result from interactions occurring between homologous sequences during the process of meiotic chromosome pairing. ENE conversion, the nonreciprocal transfer of DNA sequence information G from one DNA duplex to another, was first observed as the non-Mendelian segregation of the products of a single meiosis in fungi. Since the initial observations, genetic and molecular analyses in fungi have demonstrated conversions of the whole or portions of genes by allelic genes and by related but nonallelic genes (JACKSON and FINK 1981; KLEIN and PETES 1981; PETES and FINK 1982; KLAR and STRATHERN 1984; KLEIN 1984). The first molecular evidence for gene conversions in higher eukaryotes was provided in 1980 by SLIGHTOM, BLECHL and SMITHIES (1980), who analyzed the DNA sequences of three human fetal globin genes from one individual and showed that part of the DNA sequence in one of the fetal globin genes had probably been replaced by the corresponding region of the linked but nonallelic fetal globin gene. All of the features of this unusual DNA sequence arrangement were consistent with the molecular models of gene conversion that had already been proposed to account for the genetic observations of gene conversion in fungi (HOLLIDAY 1964; MESELSON and RADDING 1975). Additional molecular evidence for gene conversion in higher eukaryotes has since been reported. These additional examples of gene conversion occur between the duplicated DNA sequences of small gene families, including members of the aand &type globin gene clusters (LIEBHABER, GOOSSENS and KAN Genetics 1 1 2 343-358 February, 1986. 344 P. A. POWERS AND 0. SMITHIES 1981; MICHELSON and ORKIN 1983; Hardison and MARGOT 1984), the immunoglobulin gene family (BENTLEY and RABBITTS 1983), the human haptoglobin gene family (N. MAEDA, personal communication) and the gene family encoding the mouse histocompatibility genes (SCHULZE et al. 1983; WEISS et al. 1983; MELLOR et al. 1983). Several other gene conversions between fetal globin genes in human (STOECKERT, COLLINS and WEISSMAN 1984), gorilla (SCOTT et al. 1984) and chimpanzee (J. SLIGHTOM, personal communication) have also been observed. The human fetal globin genes are an excellent system in which to look for additional examples of gene conversion because of the nature of these two genes. The two human fetal globin genes, G, and A,, are the result of a 5kilobase pair (kbp) tandem duplication estimated to have occurred about 35 million years ago (SHEN, SLIGHTOM and SMITHIES 198 1). The genes themselves consist of three exons interrupted by a small 5’ intervening sequence (IVS1) and a larger 3’ intervening sequence (IVS2). The coding regions of the two genes differ by only a single nucleotide located in the third exon within the triplet coding for a glycine residue in the 5’ gene (G,) or an alanine residue in the 3’ gene (Ay). The nucleotide sequences of the two genes differ by about 3% within their large intervening sequences, whereas over most of the nontranscribed regions, the present-day copies of the duplicated region differ by an average of 14% (SHEN, SLIGHTOM and SMITHIES 1981). The structures of the two genes are so similar that it appears unlikely that any strong evolutionary selection would exist either for or against gene conversion between them, except, perhaps, in the 5’ flanking regions. In addition, because of the high degree of homology between these two genes, gene conversion will be unhindered by regions containing extensive sequence differences. Consequently, gene conversions may be detected in this particular gene pair that might be lost by unfavorable selection or be prevented by extensive sequence differences in other genes. In order to explore further the incidence of gene conversions between the human fetal globin genes, we chose to compare regions including the large intervening sequence, the third exon and some 3’ flanking DNA in the 14 paired genes of seven chromosomes. This particular region of the fetal globin genes was chosen for detailed study because previous work has shown that the large intervening sequences and the 3’ flanking regions of globin genes are poorly conserved during evolution (EFSTRATIADIS et al. 1980), so that selective forces are not likely to play a great role in determining acceptable mutations within much of the region we have studied. Some of the chromosome pairs were selected for study because we already had evidence indicating that they might have been involved in gene conversions. Other pairs were from randomly selected chromosomes. Our comparison has revealed several new examples of gene conversions between the human fetal globin genes. The relative ease of detecting these gene conversions, and their lengths (most are short), suggests to us that they may be the result of some general cellular process; and we discuss the possibility that these short gene conversions may be byCONVERSION IN HUMAN "I-GLOBIN GENES 345 products of the normal interactions occurring between homologous sequences during the process of meiotic chromosome pairing. MATERIALS AND METHODS DNA preparation: Large molecular weight DNA was prepared from fibroblasts by the method of BLATTNER et al. (1978) or from peripheral blood leukocytes by the procedure of PONCZ et al. (1982). Cloning the linked fetal globin genes: Cloning into bacteriophage vectors, and subcloning into plasmid vectors, was performed essentially as described by POWERS et al. (1 984). DNA sequencing and analysis: DNA sequence analysis was performed as described by MAXAM and GILBERT (1977), using the modifications of SLIGHTOM, BLECHL and SMITHIES (1980). The nucleotide sequences of all of the regions that differ between the genes were determined at least twice. Nucleotide sequences were analyzed using software provided by the University of Wisconsin Genetics Computer Group (DEVEREUX, HABERLI and SMITHIES 1984).