Copyright 0 1996 hy the Genetics Society of America
Structure of the
Y Chromosomal
Su(Ste)
Locus in Drosophila melanogaster
and Evidence for Localized Recombination Among Repeats
Bruce
D. McKee
and
Mark
T.
Satter
Department of Biology, University of Wisconsin, Eau Claire, Wisconsin 54702 and Departments of Zoology, and Biochemistry and Cellular and Molecular Biology, University of Tennessee, Knoxville, Tennessee 37996
Manuscript received September 6, 1994 Accepted for publication September 21, 1995
ABSTRACT
The structure of the Suppressor of Stellate [Su(Ste)] locus on the Drosophila melanogaster Y chromosome was examined by restriction analysis of both native and cloned genomic DNA. The locus consists of short subarrays of tandem repeats separated by members of other moderately repeated families. Both size variants and restriction variants proved to be common. Most repeats fell into two size classes-2.8 and 2.5 kb-but other size variants were also observed. Restriction variants showed a strong tendency to cluster, both at the gross level where some variants were present in only one of three subintervals of the locus, and at the fine level, where repeats from the same phage clone were significantly more similar than repeats from different clones. Restriction variants were shared freely among repeats of different size classes; however, size variants appeared to be randomly distributed among phage clones. These data indicate that recombination among tandem Su(Ste) repeats occurs at much higher frequencies between close neighbors than distant ones. In addition, they suggest that gene conversion rather than sister chromatid exchange may be the primary recombinational mechanism for spreading variation among repeats at the Su(Ste) locus.
M
EMBERS of multigene and simple sequence re- peat families do not evolve independently of one another. Such repeats typically show a much higher degree of sequence homogeneity than would be ex- pected if evolving independently and accumulate many of the same mutational variants. This process, known as concerted evolution (ARNHEIM 1983), is particularly easily recognized in sequences not subject to stringent selection, where repeats from one species are all very similar to one another, but repeats from even closely related species may be highly diverged. Concerted evo- lution has been documented both in functional genefamilies such as the rDNA (DOVER 1982; ARNHEIM
1983), histones (COEN et al. 1982b; MATSUO and YM-
ZAKI 1989; COLBY and WILLIAMS 1993) and chorion genes (EICKBUSH and BURKE 1985,1986), and in simple sequence repeat families such as heterochromatic satel-
lite sequences (APPELS and PEACOCK 1978; WAR~URTON
and WILLARD 1990).
Concerted evolution is thought to occur by recombi- nation between repeats. Both nonreciprocal (gene con- version) (BALTIMORE 1981; NAGYLAIU and PETES 1982) and reciprocal recombinational mechanisms such as unequal sister chromatid or interhomolog exchange
(SMITH 1976; OHTA 1980; COEN and DOVER 1983; GILL
INGS et al. 1987) have been proposed. Gene conversion involves transfer of mutational variants arising in one
Corresponding authm: Bruce D. McKee, Department of Biochemistry and Cellular and Molecular Biology, University of Tennessee, Knox- ville, TN 57996. E-mail: [email protected]
repeat to other repeats within chromatids or between sister chromatids, homologs or even nonhomologous chromosomes. Recurrent unequal exchange either be- tween homologs or sister chromatids can lead to the spread of some variants to fixation and the elimination of others, accompanied by fluctuations, sometimes dra- matic, in repeat copy number.
In tandem gene families, it is frequently difficult to determine the mechanism responsible for the observed patterns, as more than one mechanism can lead to the same outcome. Thus for example, in the Drosophila rDNA, which is present in two tandem blocks on the X
and Y chromosomes (RITOSSA 1976), although both sis-
ter chromatid (TARTOF 1974) and interhomolog ex-
changes (COEN and DOVER 1983; GILLINGS et al. 1987) have been documented, it is not clear whether either of these mechanisms or gene conversion provide the major homogenizing force (.g., see COEN 1982b). In other cases the mechanism can be inferred from the pattern of variation. For example, in the silkmoth chorion genes, a nonreciprocal exchange mechanism is implied by the observations that the repeated genes themselves are highly homogeneous but flanking sequences are much less so and that some repeats are in inverted orientation with respect to each other (EICKBUSH and BURKE 1985, 1986). In general, it is clear that unequal exchanges (both between sister chromatids and between homo- logs) and gene conversions occur in multigene families and that their relative importance in sequence homoge- nization is a function of their relative frequencies (WIL
L W S et al. 1989), which in most cases are unclear.
150 B. D. McKee and M. T. Satter
Another unresolved issue is the effect of distance on recombination frequency. Theoretical studies have shown that concerted evolution of a large gene family can occur even if recombination is limited to nearby repeats, as long as the frequency of recombination is sufficiently high (KIMURA and OHTA 1979; OHTA 1980). In some gene families, such as the Drosophila histones (COLBY and WILLIAMS 1993) and human alpha-satellite (WARBURTON and WILLARD 1990), rare variants are typi- cally found in adjacent or nearby repeats, implying that recombination occurs preferentially between closely linked repeats. In other families, however, such as the Drosophila rDNA (WILLIAMS et al. 1990) and the silk- moth chorion genes (EICKEWSH and BURKE 1985,1986),
mutational variants are distributed throughout the lo- cus in an apparently random fashion, suggesting that recombination occurs at equal frequencies between dis- tant and nearby repeats.
The purpose of the present study was to examine the distribution of restriction site variation within the
Su(Ste) gene family of Drosophila melanogaster. Su(Ste) is a tandem gene family confined to the Y chromosome; repeats have been localized to a region between the kl-
1 and kl-2 fertility factors on the long arm of the Y
(LOVETT 1983; LIVAK 1984; HARDY et al. 1984). The only
known function of Su(Ste) is a regulatory one, to repress transcription and alter the splicing of transcripts from the X-linked Stellate (Ste) locus in primary spermatocytes
(LIVAK 1984, 1990).
Ste itself is a tandem repeat family located at two sites
on the X chromosome, one euchromatic (12E1) and
one heterochromatic (LIVAK 1984, 1990; SHEVELYOV
1992). Two different Sterepeat classes have been identi- fied, 1150 and 1250 bp, both of which are capable of encoding a protein of 19,500 kD with significant homol-
ogy to casein kinase I1 (LIVAK 1990). Ste transcripts are present at high levels in testes of males lacking the
Su(Ste) region of the Y chromosome, but at low levels
in normal males (LOVETT 1983; LIVAK 1984, 1990).
Moreover, the Ste transcripts present in normal males are mostly improperly spliced and incapable of encod- ing the Ste protein, while the abundant transcripts in
X 0 males are predominantly of the “correctly” spliced class (LIVAK 1990). Derepression of Steis associated with several spermatogenic abnormalities including the pres- ence of crystals in primary spermatocytes, nondisjunc- tion of all chromosomes, meiotic drive, and, in the pres- ence of certain Ste alleles, sterility (WY 1980; HARDY et al. 1984; LIVAK 1984).
The Fchromosome-linked Su(Ste) repeats are vari- able in size but a 2800-bp repeat seems to be the most common ( BAWUREVA et al. 1992) ; 2800-bp repeats con- tain a region that is highly homologous to the Ste repeat and two regions unrelated to Ste, one of which corre- sponds to the mobile element 1360. The Ste-homolo- gous sequences contain the Ste OW, but not the poly- A signal. Moreover all five Su(Ste) repeats that have been
sequenced thus far contain one or more nonsense mu- tations and at least one splice site mutation ( B m -
m V A et al. 1992), and thus appear to be pseudogenes. It is not clear if any Su(Ste) repeats encode a functional protein.
The Su(Ste) locus is of special interest with respect to the role of recombination in the processing of muta- tional variation. Its confinement to the Y narrows the list of plausible recombinational mechanisms to intra- homolog ones. Although the X-linked Ste repeats are homologous, there are numerous, apparently fixed, dif- ferences between Ste and Su(Ste) repeats (BAIAKIREVA et
al. 1992). Moreover, the restriction of Su(Ste) sequences to male Drosophila presumably isolates them from the normal processes of meiotic exchange. Thus it is im- portant to learn whether any interrepeat recombination occurs at this locus, and if so, whether the rates and patterns are different from those described at other repetitive loci.
BALAKIREVA et al. (1992) reported evidence for recom- bination among Su(Ste) repeats. Short stretches of se- quence ranging from tens to hundreds of nucleotides including point mutations and short deletions are shared in common between various pairs of repeats. Notwithstanding these apparent recombination events, however, the sequenced repeats differed from each other at 3 4 % of nucleotide sites, a higher level of diver- gence than is typical for most tandem gene families. Moreover, the BAWUREVA et al. (1992) study was not able to shed light on the mechanism of recombination or on the effect of distance on recombination frequency.
The present study was undertaken to characterize the distribution of restriction site variation across the
Su(Ste) locus in the hope of learning more about the kinds of recombination events involved in the flux of mutational variation through the locus. Restriction vari- ant distributions were analyzed both globally, using translocation breakpoints to subdivide the locus, and locally, by comparing variant distribution patterns among repeats in the same and different phage clones. Marked evidence for clustering of restriction variants, suggestive of a distance-limited recombination mecha- nism, was uncovered. The pattern of variation is more easily explained by a nonreciprocal recombination mechanism than by unequal sister chromatid exchange.
MATERIALS AND METHODS
Fly stocks: The Drosophila genomic library was prepared from karyotypically normal males carrying a genetically marked X chromosome (y w ) and a single unmarked Y chro- mosome. These males were generated by crossing a single y w male to three virgin y w females and culturing for three generations. The y w stock was obtained from the Mid- America Drosophila Stock Center at Bowling Green State Uni- versity. The X-Y translocation stocks T(X;Y)W27, P7, E l , and
Drosophila Y Chromosome Repeats 151
matin of the X and one in the long arm of the Y (KENNISON 1981). Genetic markers and chromosomes are described in LINDSLEY and ZIMM (1992).
Genomic library preparation and screening: High molecu- lar weight genomic DNA was prepared from one gram of adult males by the method of (BENDER et al. 1983). The geno- mic library was constructed in the lambda replacement vector lambda-ZAP (Stratagene) by the partial end-filling method. Genomic DNA fragments were prepared by partial digestion with Sau3AI (GIBCO/BRL) (to maximize concentration of fragments in the 9-2-kb range), and partial end-filling with the Klenow fragment of DNA polymerase (GIBCO/BRL) in the presence of dCTP and dATP. The vector was prepared by digestion with SalI and partial end-filling in the presence of dCTP and d'ITP. Phage and genomic DNAs were ligated overnight at 15" at a 1 2 ratio of insert to vector arms in the presence of T4 DNA ligase (New England Biolabs) . Ligations were packaged in vitro using extracts from Promega Biotec.
The library was screened by the nitrocellulose lift method (BENTON and DAVIS 1977) according to the protocol in MA- NIATIS et al. (1982) using the Ste subclone pDm2L1.1 (LIVAK 1984), kindly supplied by K LIVAK, and a subcloned 2.8-kb Su(Ste) repeat as probes.
Restriction mapping: Restriction maps of the lambda phage were obtained by analysis of single and double digests using agarose gel electrophoresis and Southern blot analysis. In some cases, single or double Su(Ste) repeats were subcloned and restriction mapped more precisely. Subclones of non-Ste related DNA were also prepared and used as probes. Sub- clones were prepared in the pBluescript vector (Stratagene).
Genomic blot analysis: Male and female DNA samples ( 5 - 10 pg per lane) were digested and electrophoresed on 0.5% agarose gels, transferred to Genescreen filters (New England Nuclear) and probed with a labeled subclone containing a 2.8-kb Su(Ste) repeat. Probes were labeled by nick translation (GIBCO/BRL) in the presence of alpha "P-dCTP (New En- gland Nuclear).
Statistics and computer simulations: The degree of diver- sity among cloned Su(Ste) repeats was estimated from the re- striction site data using both the nucleotide diversity index,
r (NEI and Lr 1979; NEI 1987), and average heterozygosity, H (ENGELS 1981). Both measures estimate the average fre- quency of mismatches at a randomly chosen nucleotide site between a randomly chosen pair of repeats. The nucleotide diversity index was calculated by averaging 1 - S'", where S is the proportion of shared restriction sites between a pair of repeats and r is the length of the restriction sites (six in this case), across all possible pairwise comparisons (NEI 1987). H was calculated from ENGELS 1981, equation 11.
Clustering was assessed by comparing the diversity index for all repeats irrespective of clone (rT) with the comparable index calculated within phage clones (rw). 7rwwas calculated by restricting pairwise comparisons to repeats in the same clone. The clustering index (CI) is the proportion of total diversity present only between clones and is found by taking the difference between total and withinclone diversity and dividing by the total diversity ( x T - r w ) / r T ) .
To determine whether the clustering indices are signifi- cantly different from 0, randomization tests were carried out with the aid of the program Repeakbas running on the VAX computer at University of Tennessee computing center. For each test, the cloned repeats were first mixed in a pool, then sampled randomly without replacement to repopulate the clones. The CI was then calculated. This procedure was re- peated 10,000 times and the proportion of cases with CI values as high as or higher than the observed values was determined. This is the P value for a test of the null hypothesis that the observed CI is not significantly different from 0, because the
expected CI on the null hypothesis of no clustering is 0 (this was confirmed in the simulations).
To test whether the length polymorphism is significantly less clustered than the restriction site polymorphisms, two simulations were carried out, again with the aid of the Re- peats.bas program. In the first, the probability that the length polymorphism came from an array as strongly clustered as the array most likely to have given rise to the restriction site sample was estimated by the following procedure. First, the degree of clustering for the three most polymorphic restric- tion sites was calculated using the nucleotide diversity index. Then a series of repeat arrays polymorphic for three restric- tion sites and with varying degrees of clustering were gener- ated by three independent rounds of random sampling with- out replacement from 300 alleles divided into two subpools, the relative sizes of which were calculated from the site fre- quencies in the data. Clustering was achieved by checking each sampled repeat for identity with the previous repeat in the array and rejecting a discordant choice if a random num- ber fell below a preset clustering factor (CF). By varying CF from 0 to nearly 1, arrays with clustering levels ranging from none to very high could be generated. Each array was then sampled (four clones each containing four adjacent repeats) and the CI value for each sample calculated. Typically 100 samples were taken for each array and 10 arrays generated for each CF and the CI values averaged over the 1000 samples. The CF that gave a CI value closest to that in the real restric- tion site data was then chosen. The resulting array was then sampled again, using a single site only, and the proportion of CI values as low as or lower than the CI value for the length polymorphism was determined. This is the Pvalue for a test of the null hypothesis that there is no difference between the CI values of the three polymorphic restriction sites and that of the length polymorphism.
The second approach was to estimate the probability of obtaining a variance of CI values among four sites from a homogeneously clustered four-site array as high or higher than the variance for the CI values of the three polymorphic sites and the length difference. First, the composite CI value for the three polymorphic restriction sites plus the length polymorphism was determined and the variance of the single site CI values around that composite calculated. Next, simu- lated four-site arrays were generated as above and the CF value that maximized the likelihood of the observed composite CI selected. The resulting arrays were then sampled repeatedly and CI values for each of the four sites calculated, averaged and their variance determined. The fraction of samples with variances as high as or higher than that in the data was esti- mated. This gives the Pvalue for a test of the null hypothesis that all three restriction sites and the length polymorphism are equally clustered, so that sample variances are due only to sampling error.
RESULTS
152 B. D. McKee and M. T. Satter
RamHl Hind111 Sau3AI
? a ? a ? @
A B C D E
F
G H
F
-
T- flr-
!
o
8
: s o o x 2 5 o x ~ ~ 0 x 20x.i
154.5.5
T -
8
4 -I
2.8
1
2.5
2 . 0
goram
-
2.5
L
0 1.4“1 .l
FIGURE 1 .-Hybridization of a cloned .%@le) repeat to di- gests of male and female genomic DNA. Approximately 2pg of male or female genomic DNA were digested with the indi- cated enzymes, electrophoresed on 0.5% agarose gels, blotted to Genescreen and probed with a radiolabeled plasmid clone containing a single 2.8-kb Su(Ste) repeat. Fragment lengths are indicated in kilobases.
hybridization to male DNA on Southern blots con- taining male and female genomic DNA. The 2.8-kb frag- ments correspond to single complete Su(Ste) repeats
(BALAKIREVA et ul. 1992), which are cut from repeat arrays by enzymes with a single recognition site per unit repeat. The unit repeats from the Xchromosomal
Stellate arrays are 1250 and 1 150 bp ( LIVAK 1990). Figure 1 shows hybridization of DNA from one of the Yderived clones to equal amounts of genomic DNA from males and females digested with HindIII, BumHI and Suu3aI. It is evident that this probe hybridizes much more strongly to male-specific (Yderived) bands such as the prominent 2.8- and 2.5-kb Hind111 bands than to com- mon (Xderived) bands.
By
contrast, X chromosome- derived Stellate probes hybridize approximately equally to male-specific and common bands (data not shown). This difference results from the fact that the largerSu(Ste) repeats contain extensive regions unique to the Yrepeats (BALAKIREVA et aL 1992) whereas nearly all of the sequences in Ste repeats are also present in Su(Ste)
repeats. While most of the signal in the male Hind111 digest (Figure 1) results from hybridization to the unit 2.8-kb repeats, to a major 2.5-kb deletion class (analyzed below), and to a 5.5-kb band that likely represents di- mers of 2.8-kb repeats, there is also considerable hybrid- ization to a variety of bands of varying sizes.
Copy number of Su(Ste) repeats on a Drosophila
Y
chromosome: Hybridization of Yderived clones to male-specific bands was used to estimate the copy num- ber of Su(Ste) repeats on a normal Y chromosome (an unmarked Y from the y 7u stock). This was done by a
FIGURE 2.-Genomic reconstruction experiment. Lanes A and B contain 2 pg female and male (respectively) genomic DNA (from the y 7u stock) digested with HindIII. Lanes C-H contain aliquots of 35 ng (C), 17.5 ng (D), 7 ng (E), 5.25 ng (F), 3.5 ng (G) and 1.4 ng (H) of a 5.8-kb clone containing one 2.8-kb Su(Ste) repeat cloned into a 3.0-kb plasmid vector (pBluescript) each mixed with 2 pg salmon sperm DNA and digested with HindIII. The DNA was electrophoresed through a 0.5% agarose gel, blotted to a Genescreen membrane, and hybridized with a radiolabeled lambda phage containing
Su(Ste) repeats. The numbers at the top of the figure represent the copy number equivalent of the calibrating cloned Su(Ste)
DNA. These numbers are based on the assumption that a single copy sequence 2.8 kb in length present in 2 pg of Drosophila genomic DNA would weigh 33 pg [2 pg X 2800 bp/1.70 X 10’ per haploid genome (ASHBURNER 1989)], so
that a 20X aliquot, for example, would require 20 X 33 pg
X 5.8 kb/2.8 kb = 1.4 ng of the plasmid clone.
reconstruction experiment using dilutions of a cloned 2.8-kb repeat to calibrate the level of hybridization to the genomic DNA (Figure 2). Band intensities were compared using a scanning densitometer (BioRad Model GS670). There are
-
125- 150 2.8-kb repeats and a smaller number (60-90) 2.5-kb repeats, in addition to a large number of much more faintly hybridizing bands, most of which may be junction fragments. Thus the Su(Ste) array on one laboratory Y chromosome con- tains -200-250 repeats in the two major size classes and a substantial number of fragments of varying sizes.Restriction analysis of genomic DNA from flies con- taining subdivided Su(Ste) loci: The organization of
Drosophila Y Chromosome Repeats 153
B
Y
X D Y
-0
4@-
d
"---c700
?
-0
X
X
FIGURE 3.- Y chromosomal breakpoints of X-Y transloca- tions and karyotypes of males and females used to generate DNA for the genomic blot mapping of Su(Ste) variants. (A) Map of the proximal portion of the long arm of the Ychromo- some (cytological regions h9-hl5, G A ~ I and PIMPINEI.I.I
1983) showing the locations of the breakpoints of the four translocations used in the mapping of &@e) sequences (based on HARDY d al. 1984; LIVAK 1984,1990). The approxi- mate locations of the fertility factors kl-I and kl-2 and the approximate extent of the Su(Ste) sequences are indicated. (B) Karyotypes of males and females used to prepare genomic DNA. Circles represent centromeres, boxes represent hetero- chromatic sequences and lines represent X euchromatin. Filled and unfilled boxes and circles represent material from the Y and X chromosomes, respectively. The point is that females carryall of the Ychromosome proximal to the translo- cation breakpoint, while males carry all of the Ychromosome distal to the breakpoint.
locus (Figure 3A) (KENNISON 1981; HARDY et aL 1984; LIVAK 1984). The Y chromosomal breakpoints of the translocations T(X;Y)W27 and E15 flank the Su(Ste) se- quences distally and proximally (respectively) while the breakpoints of the translocations T(X;Y)P7 and E l fall within the locus. Collectively, these translocations de- fine three subintervals of the Su(Ste) locus (from distal to proximal): W27-P7, P7-E1, and El-E15. These subin- tervals contain -10, 70 and 20% of the Su(Ste) se- quences, respectively (LIVAK 1984). Males from each of these translocation stocks were crossed to chromosom- ally normal females and DNA prepared from the X"YP daughters and
X"p'
sons (Figure 3B). The DNA was digested with various sixcutter restriction enzymes, electrophoresed, blotted and probed with a 2.8-kb Su(Ste) repeat. In the resulting autoradiograms (Figure 4) all of the %specific signal is in the female lane forT(X;Y)W27 but in the male lane for E15, while for the translocations that subdivide the Su(Ste) locus, some sig- nal is present in both lanes.
Polymorphism for re.striction site.s: It is apparent that hy- bridization to the main bands at 2.8 and 2.5 kb relative to higher molecular weight bands varies from digest to digest. Main band hybridization is most intense in the
HindIII and BgLII digests, somewhat less intense in the RamHI digest and much less intense in the SstI digest. These differences likely reflect differences in degree of polymorphism of these sites among members of the Su(Ste) array. Most repeats evidently contain single sites for HindIII and BgLII, while fewer repeats contain a site for RamHI and fewer still contain one for SstI. Some of the higher molecular weight bands in the various di- gests likely represent multiples of 2.8 or 2.5 kb, gener- ated from arrays of repeats where an internal restriction site is absent. The 5.5-kb band evident in the BgLII, HindIII, and BamHI digests (Figures 1 and 4) is likely a dimer of 2.8-kb repeats. Other higher molecular weight bands may represent either full repeats of other size classes, or junction fragments containing both Ste and non-Ste related DNA. The XbaI digest shows main bands at 1.7 and 2.5 kb and much weaker hybridization at 2.8 kb, suggesting that most 2.8-kb repeats have two Xbd sites spaced 1.7 (or 1.1) kb apart while most 2.5-kb repeats have only one XbaI site. [Note: the predicted 1.1-kb band ran off the bottom of the gel in Figure
4,
but is present (data not shown)].CLusta'ng of variants within subintervals: The results in Figure 4 indicate that at least some of the restriction and size variants are clustered in particular subintervals of the locus. For example, the 2.5-kb repeats appear to be limited to the central P7-lCl interval (note the virtual absence of 2.5-kb repeats from the P7 male and E l female lanes in the BgLII, HindIII, and BamHI digests). A fairly prominent band at -3.5-kb in the HindIII and
BgLII digests shows a similar distribution, suggesting a similar localization to the central interval. At least three other variants are restricted to the El-El5 interval based on their restriction to the female lanes in the P7 and E l digests. Included in this class are the 2.8- and 2.5- kb repeats with SstI sites, a HindIII fragment at -1.8 kb and most of the singly cut 2.8kb repeats in the XbaI digest. Thus, the genomic blot hybridizations suggest that nearby repeats are more likely to share size or
restriction variants than are repeats from more distant regions of the locus.
Restriction analysis of cloned Su(Ste) repeats: To ad- dress the question of the organization of Su(Ste) varia- tion on a finer scale, restriction maps of the Su(Ste)
lambda clones were prepared using several sixcutter enzymes, including the five used in the genomic blot analysis and two additional enzymes. In eight of the 11 clones, 2.8-kb repeats were identified; detailed restric- tion maps for those eight appear in Figure 5. The other three clones, which were classified as Yderived on the basis of hybridizing more intensely to Yderived Su(Ste) bands than to Xderived bands on genomic Southern blots, were not analyzed in detail. All three appeared to contain extensive non-Sterelated sequences in addi- tion to the Su(Ste) sequences.
154 B. D. McKee and M. T. Satter
3.5b
2.8b
2
.SF
2 . 8 ~
2.5P
1.7,
v .
I I
5.5
2.8
2.5
FIGURE 4."Genomic blot mapping of Su(Ste)variants. Genomic DNA (2 pg) from males and females of the indicated karyo- types was digested with the indicated en- zymes, electrophoresed through 0.5% aga-
D XbuI
E
BurnHI
F HindIIIIBurnHI
rose gels, blotted to Genescreen filters and probed with a radiolabeled 2.8-kb repeat. Refer to Figure 3 for karyotypes of males and females. Fragment lengths are indi- cated in kilobases.115 E1 p7 W27 E l 5 El P7 W27 EIS E l p7 W27
5.5
c..
0
contained 2.5-kb repeats, and in two of these clones, two 2.5-kb repeats were present in tandem. Other minor size classes were also present. Some of these repre- sented complete repeats containing deletions or inser- tions, while others represented fragments of repeats at the edge of the Su(Ste) DNA bordering either other insert DNA sequences or vector sequences. All eight of the clones contained more than one repeat, and in all cases, all of the repeats were in head-to-tail orientation. The inserts in Figure 5 are drawn so that all of the
Su(Ste) repeats are in the same orientation.
The 2.5-kb repeats all share a 300-bp deletion (or perhaps several smaller deletions totaling -300 bp) lo- cated between the Hind111 and S c d sites on the left side of the repeat. Because all six cloned 2.5-kb repeats lack the leftmost XbuI site that is present in nearly all 2.&kb repeats, it is likely, but not certain, that the deletion overlaps this site. The genomic blot data (Figure 4) point to similar conclusions. First, as noted above, most 2.8-kb repeats contain two XhuI sites, while most if not all 2.5-kb repeats contain only one. Second, the deletion
r 2 8
r2.5
ul.7
1.3
q1.1
maps to the left of the BumHI site. This is based on the prominent hybridization to the expected 1.3-kb band in the BumHI/HindIII double digest. That this band is in fact a derivative of 2.5-kb repeats is confirmed by its sublocus distribution-restriction to the central P7-E1
interval, matching that of the 2.5-kb bands. Thus the data suggest that the ubiquitous 2.5-kb fragments likely represent a homogeneous size variant, all sharing a de- letion in the left half of the repeat, which most likely encompasses the leftmost XbuI site. It seems likely that this deletion arose once, early in the genesis of the present array and has since been amplified within the central 70% of the locus.
Other middle repetitive sequences interspersed with
Su(Ste) sequences: In addition to the tandem Su(Ste)
Drosophila Y Chromosome Repeats 155
8 2
c3
c4
c 5
F6
X H C G A X H C G A X H XI C , Gi $ X H XI C l A ( G X H x C G A
L L L 1 I " i I I I , I I I I j I / !
B S B S
X H x GA X H X A f X H X H
I
< >
G B
2 8 . . . . . . ... . .
X H X C G B X H X C G B X H X C G B X H C G B X H
"
lambda clones containing Su(Ste) repeats.
C
G H X C A B X
A, ApnI; B, BamHI; C , Sc& G, &&I; H: Hin-
dIII; S, M I ; and X, XbaI. 0 , Su(3~)homolo- gous sequences; I, unrelated DNA se- G
I i l l 1 ! i
2.8 quences; and B, border areas.
H1
A X H X C B X H X C G I B X H X C G I S X H S CG X H H H
H6 X H X C G B X H X C G B X H C G B X H C G y B S X
, I
2.5
"_ "~
1 kb ~
and female DNA and some male-specific (mostly the former). Each of the three probes gave a different hy- bridization pattern, two of which are shown in Figure 6. Repeated subcloning of the flanking DNA from C3 failed to separate the sequences responsible for male- specific hybridization from those responsible for hy- bridization to common bands (data not shown). This indicates that the flanking DNA consists of members of repeated DNA sequence families with representatives on the Y chromosome and elsewhere in the genome
(mostly elsewhere). The fact that the hybridization pat- terns are different for each of the probes suggests that
Su(Ste) sequences are interspersed with members of sev- eral other repeated families. As noted above, non-Ste related sequences were identified in all three phage clones that were not mapped in detail as well as in the five clones that appear in Figure 5. The ubiquity of this interspersed DNA provides an explanation for the large number of minor bands visible on Southern blots of male DNA digested with enzymes like HindIII that cut once per repeat (Figure 1 ) .
156 B. D. McKee and M. T. Satter
A
B
Hind111 EcoRI Hind111 EcoRI
FIGURE 6.-Genomic hybridization patterns of two se- quences acljacent to Su(Ste) sequences. Genomic DNA (4 pg) from y w males and females was digested with Hind111 or
EcoRI, electrophoresed through 0.5% agarose gels, hlotted to Genescreen filters and prohed with suhcloned non-Sterelated sequences from the phage clones C3 (A) and F6 (R). Arrows indicate male-specific (Y-derived) hands.
Similar polymorphisms for sites in the 2.5-kb repeats are evident as well. " h e n the size variants are also fac- tored in, it is evident that there is a great deal ofvariabil- ity among Su(Ste) repeats.
This variability was quantified using the nucleotide diversity index ( x , NEI and LI 1979; NEI 1987). This measure estimates the probability of a nucleotide mis- match at a randomly chosen site between any two ran- domly chosen alleles (or repeats in this case) based on restriction site data. For the 23 repeats in the seven clones with two or more repeats the diversity index is 0.051 (Table l ) , meaning that on average two repeats differ from each other at 5.1% of nucleotide sites.
Clustering of restriction site variants: The restriction maps of the phage clones reveal marked clustering of similar restriction variants. In general, restriction maps of same-size repeats present in the same phage clone are identical or nearly so, while repeats from different phage clones often differ substantially. For example, both of the 2.8-kb repeats in B2 have a site for SstI and
ApaI but no RnmHI sites, while all three of the 2.8-kb repeats in C5 have a RamHI site but lack both ApaI
and SslI sites. There are differences between adjacent repeats, such as the missing HindIII, BgflI, ScaI, and RamHI sites in single repeats in C3 and H1, the RamHI
TABLE 1
Nucleotide diversity within and among
clones of Su (Ste) repeats
87. 8 , y CI P
A. Restriction site variants
All repeats (7, 10)" 0.051 0.025 0.51 <0.0001
All repeats (7, 3) 0.099 0.029 0.70
A l l repeats (4, 3) 0.087 0.027 0.68 2.8-kh repeats only (7, 10) 0.051 0.018 0.65 <0.0001 2.5-kh repeats only (7, 10) 0.027 0.0087 0.68 2.5-kh repeat.. only (4, 9) 0.030 0.0097 0.68 0.083 2.5 vs. 2.Rkh repeats (4, 9) 0.031 0.01.55 0.50 0.0028
B. Size variants
2.5- 71s. 2.Rkh repeats (4, 1) 0.109 0.136 -0.25
Comparisons were based on complete repeats defined as extending from the left HindIII site to (hut not including) the right HindIII site; (refer to Figure 5 for data).
"Numhers in parentheses are (numher of phage clones, numher restriction sites) nsed in the comparison. Compari- sons involve either all seven phage clones with two or more repeats (7) or the four phage clones with both size classes (4). Restriction sites used are all ( I O ) , all sites hut the left XbnI site (9), or the three most polymorphic sites (3) (see
for formulae for R and CI values. P is the probability of o b
taining a CI value as high or higher than that reported, assum- ing that the true value is 0, determined as descrihed in MATERI-
text for further explanation). See MATERIALS AND METHODS
A I S AND METHODS.
singlet in C4, and the ApaI doublet in H5, but the overall impression is that contiguous repeats are consid- erably more similar than repeats in different phage clones. The similarity of adjacent repeats is especially striking for the ApaI and both SstI sites. Despite being present in only a minority of repeats, all three sites are present almost exclusively in contiguous arrays.
Repeats of different size classes are also more likely to be similar if present in the same clone than if in different clones. For example, the 2.5-kb repeats in clones B2, C5, H1 and H6 differ substantially from each other but are very similar, except for the deletion, to the 2.8-kb repeats in the same clones. Similarly, the repeat fragments at the edges of the Su(Ste) arrays are typically quite similar to the contiguous complete repeats.
To quantify and assess the significance of this appar- ent clustering of repeats with similar restriction maps, nucleotide diversity measures were calculated both for the data set as whole (x.,.) and for withinclone compari- sons only (xt,,). In the absence of clustering, these values should be the same. In this data set (which again consists
Drosophila Y Chromosome Repeats 157
yield substantial clustering indices, whether compari- sons involve all repeats, just 2.8-kb repeats, just 2.5-kb repeats orjust comparisons of
2.8
us. 2.5-kb repeats. The 2.5-kb repeat comparison was done over all clones with all restriction sites, and over just the four clones with both size classes using only the nine sites found in the 2.5-kb repeats, with essentially identical results. All of the clustering indices are highly significantly different from 0 for all comparisons except that involving the 2.5- kb repeats only, for which the data set is quite small. These results indicate that two repeats from the same clone (irrespective of size class) are substantially more likely to have similar restriction maps than are two re- peats from different clones. Similar results were o b tained using average heterozygosity( H )
as the diversity measure (ENGELS 1981) (data not shown).Lack of clustering of size variants: The strong tendency of restriction variants to cluster within subarrays does not seem to hold for size variants. There are 2.8- and 2.5-kb repeats present together in four of the eight mapped clones, and other size variants are also distrib uted widely. Of the eight mapped clones, only one (C3) is monomorphic for a single size class, and this may be because it has only two repeats. This lack of clustering is confirmed by the clustering index for the length poly- morphism, which is slightly negative [but not signifi- cantly different from 0 (Table 1)
1.
To assess the signficance of this apparent difference between restriction site polymorphisms and length polymorphisms, two computer simulations were con- ducted. In both, the strategy was to simulate repeated arrays that yield clone samples with clustering indices similar to those in the restriction site data, then to esti- mate the probability of drawing a sample that contains data as divergent as those for the length polymorphism. For both simulations, only the three most polymorphic restriction sites were used (BamHI, ApaI and the right- most SstI site). These three were selected as having poly- morphism levels comparable with that of the length difference. The sample was so restricted because nucle- otide diversity measures are strongly influenced by lev- els of polymorphism (when nearly all alleles are identi- cal, diversity is necessarily low and clustering can be much harder to detect). Details of the simulation meth-
ods are in MATEMALS AND METHODS.
The strategy for the first simulation was to generate arrays that maximize the likelihood of yielding samples with clustering properties like those of the three poly- morphic restriction sites. These arrays were then sam- pled repeatedly and the fraction of single-site CI values as different from the actual value for the three polymor- phic restriction sites as that of the length polymorphism were calculated. The resulting value, 0.015, is an esti- mate of P for a test of the hypothesis that the length polymorphism is derived from an array as strongly clus- tered as the array most likely to have yielded the restric- tion site data. To determine how robust this conclusion
is with respect to changes in degree of array clustering, the P value was plotted against the average CI value for a range of arrays with varying degrees of clustering (Figure 7A). For arrays with average CI values of 50.55, the length polymorphism is not statistically deviant, but for all arrays with CI values >0.55, the difference is signficant. Because the real data have a CI value s u b stantially higher than 0.55 (0.68), it seems unlikely that the length polymorphism could reflect simply an un- usual sample from a uniformly distributed array.
The strategy in the second simulation was to treat the length polymorphism as a fourth polymorphic re- striction site and to ask whether the observed variance of the four CI values falls within the 95% confidence interval generated by sampling a four-site array with similar composite clustering values (0.45). The result was that 4.8% of simulated samples from such a four- site array yielded CI variances as high as that calculated using the three polymorphic restriction sites and the length polymorphism. These results are thus statistically significant, but just barely. It is clear from Figure
7B
that this result is sensitive to changes in the average CI value of the arrays. Thus for average CI values >0.49 the probability of variances as high or higher than the observed is >0.05 (though never greater than 0.15).Taken together these results support the apparent lack of clustering of the length polymorphism and sug- gest that the restriction polymorphisms and length poly- morphisms are differentially affected by the evolution- ary forces responsible for clustering. However, due to the small sample size of the length polymorphism data, it is not possible to unambiguously test the idea that the restriction site and length polymorphisms are clus- tered to different degrees. Such a test will require a larger sample.
DISCUSSION
Structure of the Su(Ste) locus: The picture of the Su(Ste) locus that emerges from these results is that of a series of short arrays of tandem (head-to-tail) repeats interspersed with members of other (mostly dispersed) middle repetitive families. Five of the eight phage clones fully mapped in this study (and all three un-
mapped clones) contain non-Ste-related sequences in
addition to Su(Ste) sequences. The typical length of Su(Ste) arrays is unknown. Seven of the eight phage clones fully mapped in this study contained clusters of two or more rpeats, so most arrays must be longer than two. Three different arrays of four+ repeats were cloned, so the maximum length is at least four, but larger clones will have to be examined to gain an accu- rate picture of the upper limit of array length.
158 B. D. McKee and M. T. Satter
A.
Length versus Sites
P
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Clustering index
6.
Variance Test
P
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Clustering index
FIGURE 7.--Simulated comparison of length polymorphism and restriction site polymorphisms. (A) Length polymor- phism us. restriction site polymorphisms. Horizontal axis rep resents degree of clustering, measured by the clustering index (CI), of random samples of four clones of four repeats each from simulated, clustered arrays with three polymorphic re- striction sites (at levels of polymorphism identical to those of
the three most polymorphic sites in Figure 6). Each point represents an average of CI values for 100 samples each from 10 independent arrays. The vertical axis gives the fraction of single-site samples with clustering values as different from the mean as is the length polymorphism in the data in Table 1 . Dashed line indicates composite CI for the three most polymorphic restriction sites. (B) Variance test. Horizontal axis represents clustering indices for simulated four-restric- tion site arrays determined as above. Vertical axis gives frac- tion of arrays with variances of single-site clustering indices
as great or greater than the observed. Dashed line indicates composite CI of real data using the three polymorphic restric-
tion sites and the length polymorphism.
genomic distribution pattern by genomic blot analysis. All three proved to be moderately repetitive and dis- persed in the genome, a typical pattern for transposable
elements. Copies of the mobile elements MDGl (SHEV-
ELYOV et al. 1989) and hetA (DANILEVSKAYA et al. 1991) have been described adjacent to Su(Ste) or Su(Ste)-re- lated sequences.
At this level of analysis, the Su(Ste) locus is not dissimi- lar to other repeated gene families in Drosophila. The histone locus consists of 8-12 subarrays of -10 tandem repeats each separated by unrelated sequences (SAIGO et al. 1981). The rDNA arrays are long (200-250 copies of an 11.5-kb unit repeat) (RITOSSA 1976) but are inter- rupted frequently by insertions of site-specific transpos-
able elements (GLOVER and HOGNESS 1977; WELLAUER
et al. 1978), which are dispersed throughout the locus (HAWLEY and TARTOF 1983). The heterochromatic
Rsp
locus also consists of many short tandem arrays inter- spersed with other repetitive sequences (Wu et al. 1988). It is unknown whether this pattern has functional sig- nificance.Polymorphism for size variants: Another similarity between Su(Ste) and other repeated loci in Drosophila is polymorphism for repeat length. Most Su(Ste) repeats fall into two major size classes-2.8 and 2.5 kb-but both larger and smaller repeats are also present. The
histone locus shows a similar polymorphism, repeat units of 4.8 and 5.0 being the major classes (LIFI'ON et al.
1978). Stellate repeats also fall into two major length
classes-1150 and 1250 bp (LNAK 1990). rDNA repeats
exhibit numerous length variants, mostly due to varia- tions in length of an internally repetitive spacer (WEL LAUER et al. 1978; COEN and DOVER 1982), and to the
TE insertions referred to above (GLOVER and HOGNESS
1977; WELLAUER et al. 1978). It is not known whether these length polymorphisms serve any function.
Sequence variability among repeats: The level of re- striction site variability observed in this study is consis- tent with that from the report of BAWUREVA et al.
Drosophila Y Chromosome Repeats 159
repeats are used, average divergence is 5.1
%,
identical to the present results.Su(Ste) repeats appear to be more dissimilar than most tandem repeat families that have been studied. Human alpha-satellite repeats differ at -1% of sites
(WARBURTON and WILLARD 1990), the three members
of the D. pseudoobscura alpha-amylase family differ at -0.5% of sites in the coding region (BROWN et al. 1990),
Stellate repeats differ at -2.5% of sites (calculated from the data of SHEVELYOV 1992) and Drosophila rDNA in- tergenic spacer repeats differ at an average of 1.8% of sites in
D.
melanoguster and D. virilis and 1.1% of sites inD.
simulans (calculted from the data of TAUTZ et al.1987). The noncoding 360 and 500 families of tandem repeats in Drosophila appear to have levels of intraspe- cific variability comparable with those of Su(Ste) repeats
In principle, the higher level of variability of Su(Ste)
repeats could be due to a higher mutation frequency, a lower recombination rate leading to a lower rate of fixation or removal of variants, or to less stringent selec- tion. The mutation freqeuncy has not been measured, but there is no reason to expect it to be especially high at Su(Ste). It would not be surprising if recombination were reduced; not only is Su(Ste) heterochromatic but it is also present only in males which have virtually no
meiotic recombination, even in euchromatic se-
quences. Recombination is not absent, however. As de- veloped in the next section, the pattern of clustered variation observed here is consistent with recombina- tion limited to very closely linked repeats. Moreover,
the Ychromosomal rDNA locus would appear to subject
to the same general recombination conditions as
Su(Ste), yet there is considerably less variation among rDNA spacer repeats than among Su(Ste) repeats. Selec- tion may well be reduced at Su(Ste); all five repeats that have been sequenced lack significant open reading frames, despite the homology to Ste (BALAKIREVA et al.
1992). The locus as a whole has a function crucial for normal spermatogenesis, but it is not clear how this function is related to the sequence of the repeat. Thus the unusually high variability among Su(Ste) repeats could be due to low recombination frequency or to weak selection, or both.
Clustering of variants and localized homogenization:
This study provided strong evidence for clustering of restriction variants. Clustering was evident both in the clones, in which repeats in the same clone tended to be much more similar in restriction map than repeats in different clones, and in the genomic blot data, which showed that some of the variants were restricted to only one of the three regions defined by the translocation breakpoints. The observed restriction site variability is consistent with a nucleotide sequence divergence level of -5.1% for 2.8-kb repeats overall, compared to only
1.8% for repeats in the same clone. The only potentially comparable data are those from the BALAKIREVA et al.
(STRACHAN et al. 1985).
(1992) study; as noted above, the overall divergence level is similar in that study and this. However, two of the five Su(Ste) repeats for which sequences were re-
ported in BAW~REVA et al. (1992) came from the same
cosmid clone (SHEVELYOV et al. 1989) and these are
considerably more divergent (8%) than expected from
the present data. The differences included multiple point mutations and several deletions. A possible expla- nation for this discrepancy is that the two repeats from the cosmid may not have derived from a continuous block of repeats but from two discrete subarrays sepa- rated from each other by an unknown amount of unre- lated DNA. Alternatively, the multiple deletions in these two repeats may prevent recombination between them. The present results indicate that for the heterochro- matic Su(Ste) family, recombination occurs between re- peats but must be limited mainly to interactions between nearby repeats. If no recombination occured among
Su(Ste) repeats, or if recombination occurred at the same frequency irrespective of distance between repeats, adja- cent repeats should be no more similar than distant ones. This is clearly not the case, so interrepeat recombi- nation must occur but at frequencies that depend in some manner on the proximity of the repeats.
It is not clear whether distance itself is the critical factor for recombination frequency or whether the un- related DNA sequences separating each subarray act as barriers to the spread of sequence information, perhaps by providing a register for interchromatid pairing. The data support the former idea, but are somewhat incon- clusive. Where a variant is present in only two repeats out of three or four in a clone, those repeats are typ
cially adjacent rather than separated by one or two re- peats. The clearest examples of this pattern are the SstI
doublets in B2 and H1 and the ApaI doublet in H5 (Figure 5). Also consistent is the fact that in the SstI
genomic blot, clear bands are present at 2.8 and 2.5 kb but not at higher multiples such as 5.6 (Figure 4). A clear band at -5.5 kb is present in the BglII and Hind111 digests, but these sites are so abundant that the pres- ence of this band reflects the absence of a single site in some repeats and so can not be taken as evidence for transfer of information across longer distances. These observations suggest that variants spread preferentially to adjacent repeats, but do not address the question of whether interspersed foreign DNA acts as a barrier to the spread of sequence variants. These possibilities could be distinguished by examining restriction pat- terns within and across larger clones, such as cosmids, that might be expected to include more than one subar- ray and would also permit anlysis of the distribution pattern within larger subarrays than have been exam- ined in this study.
160 B. D. McKee and M. T. Satter
clusters are distributed across many megabases of DNA
(WARBURTON and WILLARD 1990). Rare restriction site
variants within the Drosophila histone gene locus are present predominantly in adjacent repeats (COLBY and WILLIAMS 1993). The clustering in this case is so pro- nounced that the authors conclude that >90% of the recombination events must occur between adjacent re- peats. Clustering of sequence variation is not a universal observation, however. Shared sequence variants appear to be distributed randomly among repeats of the silk- moth chorion late gene family (EICKBUSH and BURKE
1985, 1986). In the
D.
melanogaster 5 s ribosomal gene family, a Hind111 site is interspersed throughout the tandem array (SAMSON and WEGNEZ 1988). It is not clear what factors act to impose distance-limited recom- bination on some families but not others.The only polymorphism not clustered in the cloned repeats analyzed in this study is the length polymor- phism. Two simulations were carried out to test for a significant difference between the three most polymor- phic restriction sites and the length polymorphism, and both indicated that the difference is significant. Thus there is likely some difference in the recombinational behavior of polymorphisms due to site differences and those due to large heterologies. However, the sample size for the length polymorphism is rather small and the data should not be overinterpreted. It is certainly possible that regions of the Su(Ste) locus not repre- sented in the clone sample exhibit strong clustering of repeat length polymorphisms. Indeed, the genomic blot data in this paper point to some large-scale clustering of 2.5-kb repeats. A larger sample of cloned repeats, and especially, clones of larger fragments, will be needed to determine how general the present findings are.
Mechanism of homogenization of Su(Ste) repeats: gene conversion or unequal sister chromatid exchange?
Most theories of concerted evolution phenomena have invoked either gene conversion or unequal sister chro- matid exchange or both as explanations for the spread
or elimination of sequence variation among repeats in tandem families. Although both mechanisms have been shown to occur, in most cases it has not been possible to conclude which provides the major homogenizing force. In situations where changes in repeat copy num- ber have also been documented, such as the rDNA (TARTOF 1974; WELLAUER et al. 1978; COEN et al. 1982b; WILLIAMS et al. 1990), unequal sister chromatid ex- change has often been favored ( . g . , TAUTZ et al. 1987) because it can potentially account both for copy num-
ber changes and sequence homogeneity (SMITH 1976;
KIMURA and OHTA 1979; OHTA 1980), whereas gene conversion readily accounts only for the latter.
However, concerted evolution involving apparent
gene conversion has been documented in several
multigene families including vertebrate immunoglobu-
lins (BALTIMORE 1981) and globins (POWERS and SMITH-
IES 1986) and Drosophila noncoding tandem repeats
(STRACHAN et al. 1985). A particularly well-documented example is the silkmoth chorion late gene family which consists of 15 copies of a divergently transcribed gene pair. The gene pairs themselves are highly homologous
(<1% divergence) but the interspersed 3’ flanking
DNAs are much more divergent, thus ruling out sister chromatid exchange.
The clustering of Su(Ste) restriction site variants docu- mented in the present study can be explained either by localized conversion or by unequal sister chromatid exchange. However, if the apparent lack of clustering of repeat length variants is confirmed, that will point
toward gene conversion as the predominant mecha-
nism of sequence homogenization. If sister chromatid exchange were responsible for homogenization of se- quence variants, it should also homogenize length vari- ants. Although recombination between the sites of a sequence variant and a length variant can allow a se- quence variant to sweep to fixation in a subarray while still allowing dimorphism for the length variant, the reverse is equally likely apriori. Only if the length variant is associated with a recombination hotspot would it be expected to behave as if unlinked from all sequence variants. But in that case, the length variant should ex- hibit a higher rate of fixation than the sequence vari- ants, the reverse of what is observed here.
The conversion model can explain the lack of cluster- ing of size variants by stipulating either that conversion tracts are confined to one side or the other of any large heterology or that the conversion mechanism is blind to large heterologies. Assuming that conversion events among Su(Ste) repeats, like other conversion events, in- volve formation and repair of heteroduplexes, these ideas imply either that heteroduplexes initiated in a region of homology can not be propagated past a 300- bp deletion/insertion heterology or that hetero- duplexes containing such large heterologies are not repaired efficiently.
The conversion idea is supported by the findings of
BAW~REVA et al. (1992) who reported patches of se- quence with similar variants among otherwise dissimilar regions of different repeats. The authors did not specify a mechanism, but the patchwork pattern of sequence similarities they described is more simply explained by a conversion mechanism than by unequal exchange.
The authors thank LEONAR~ ROBBINS, WILLIAM ENGELS, and JIAN-
MING SHEN for statistical advice. Responsibility for the statistical calcu- lations and computer simulations is entirely the authors’. This study was supported by a research grant, #DCB8416813, from the National Science Foundation to B.D.M.
LITERATURE CITED
APPEIS, R., and W. J. PEACOCK, 1978 The arrangement and evolu- tion of highly repeated (satellite) DNA sequences with special reference to Drosophila. Int. Rev. Cytol. 8: 69-126.
ARNHEIM, N., 1983 Concerted evolution of multigene families, pp.