Analysis of molecular variance (AMOVA) of Y-chromosome-specific microsatellites in two closely related human populations
Analysis of molecular variance (AMOVA) of Y-chromosome-specific microsatellites in two closely related human populationsLutz Roewer1, Manfred Kayser1, Patrick Dieltjes2, Marion Nagy1, Egbert Bakker2, Michael Krawczak3,4 and Peter de Knijff2,*
1Institut für Gerichtliche Medizin, Humboldt-Universität, Charité, Hannoversche Strasse 6, D-10115 Berlin, Germany, 2MGC-Department of Human Genetics, Leiden University, P.O. Box 9503, 2300 RA Leiden, The Netherlands, 3Institut für Humangenetik, Medizinische Hochschule, D-30623 Hannover, Germany and 4Institute of Medical Genetics, University of Wales College of Medicine, Cardiff CF4 4XN, UK
Received March 19, 1996;Revised and Accepted April 19, 1996
The analysis of seven Y-chromosome-specific microsatellite loci revealed a high level of polymorphism in two closely related human populations (Dutch, n = 89, and German, n = 70). Four of these loci were found to generate at least 77 different haplotypes, only 15 of which were shared by the two populations. These results demonstrate that highly informative PCR-based DNA typing of the Y chromosome is now feasible. Assuming a stepwise mutation model, a network comprising all minimum spanning evolutionary trees connecting the haplotypes was constructed. Analysis of molecular variance based upon this network indicated that the within-population heterogeneity with respect to haplotype descent was significantly smaller than the between-population heterogeneity, suggesting that males were more closely related to males from their own population as opposed to males from the other population. These findings suggest that Y-chromosomal microsatellites might be very useful not only for forensic purposes but also in association studies of multifactorial traits, allowing the characterization of the level of genetic distinctiveness of supposedly inbred or isolated populations and discrimination even between closely related populations.
The characterization of a still rapidly increasing number of autosomal microsatellite loci in human DNA has not only facilitated the mapping of disease-associated genes (1 ), but has also proven capable of providing researchers with valuable tools for analyzing the phylogenetic relationships of human populations (2 ). Human evolutionary trees based upon autosomal microsatellite haplotype frequencies closely resemble their counterparts based upon mitochondrial (mt)DNA (3 ), although there has been at least one report of only marginal overlap (4 ). It seemed straightforward to add to the aforementioned types of methods one that is strictly based upon paternal inheritance, thus recruiting Y-chromosomal heterogeneity. The Y chromosome, however, appears to be the least polymorphic human chromosome since only a relatively small number of Y-chromosomal variations have been reported so far (for comprehensive review, see ref. 5 ). Three recent searches for sequence heterogeneity have yielded not more than six different point mutations even though 729 bp, 2.6 kb and 18.3 kb of human Y-chromosomal sequence were screened in several ethnically distinct males (6 -8 ).
To date, five microsatellite loci identifying multiple alleles on the Y chromosome have been described (9 -11 ). Since at least one of these(DYS19 or Y27H39) exhibited considerable allele frequency diversity between populations (12 ,13 ), other Y-chromosome-specific microsatellites could be equally useful in identifying male lineages. From existing databases, we obtained information on potentially useful microsatellites. After some initial testing, seven microsatellites were selected for further analysis. Haplotype analysis based upon a subset of four microsatellites revealed a significantly increased between-population heterogeneity with respect to haplotype descent, suggesting that the analyzed males were on average more closely related to males from their own population as opposed to males from the opposite population.
Dutch (n = 89) and German (n = 70) male controls were typed for seven Y-specific STRs. At loci DYS19, DYS389-II, DYS390 and DYS392 (Table 1 ), allele frequencies differed significantly between the two groups. Based upon four loci (DYS19, DYS389-I, DYS389-II and DYS390) a Y-chromosomal haplotype was constructed for each individual. In total, 77 different haplotypes were observed in the combined data set (n = 159), only 15 of which were shared by both groups. Haplotype diversity values, calculated according to Nei (14 ), were 0.962 and 0.988 for the Dutch and German male group, respectively, representing a significant difference (P<0.01).
Allele frequencies and their standard errors (SE) of seven Y-chromosome-specific microsatellite loci among German and Dutch males
Locus/allele
German (n = 70)
Dutch (n = 89)
Pa
Frequency
SE
Frequency
SE
DYS19
186 (A)
0.100
0.036
0.045
0.022
<0.001
190 (B)
0.371
0.058
0.708
0.048
194 (C)
0.257
0.052
0.191
0.042
198 (D)
0.243
0.051
0.034
0.019
202 (E)
0.029
0.020
0.022
0.016
DYS389-I
247
0.243
0.051
0.270
0.047
0.223
251
0.529
0.060
0.607
0.052
255
0.229
0.050
0.124
0.035
DYS389-II
357
0.000
0.000
0.011
0.011
0.012
361
0.214
0.049
0.303
0.049
365
0.286
0.054
0.404
0.052
369
0.329
0.056
0.247
0.046
373
0.100
0.036
0.034
0.019
377
0.071
0.031
0.000
0.000
DYS390
203
0.000
0.000
0.011
0.011
0.016
207
0.086
0.033
0.180
0.041
211
0.357
0.057
0.382
0.052
215
0.257
0.052
0.326
0.050
219
0.229
0.050
0.067
0.027
223
0.071
0.031
0.022
0.016
227
0.000
0.000
0.011
0.011
DYS391
279
0.029
0.020
0.034
0.019
0.830
283
0.643
0.057
0.580
0.053
287
0.314
0.055
0.352
0.051
291
0.014
0.014
0.034
0.019
DYS392
236
0.014
0.014
0.000
0.000
0.003
248
0.614
0.058
0.382
0.052
251
0.071
0.031
0.056
0.024
254
0.257
0.052
0.539
0.053
257
0.029
0.020
0.011
0.011
260
0.014
0.014
0.011
0.011
DYS393
120
0.043
0.024
0.112
0.033
0.310
124
0.800
0.048
0.753
0.046
128
0.100
0.036
0.112
0.033
132
0.057
0.028
0.023
0.016
Alleles are labeled by their PCR fragment length (in bp). In addition, for DYS19, the often used allelic designation A-E is given (9-11). aP value, indicating the significance of the difference of allele frequency distribution for a given locus between the Dutch and German male-groups, calculated by means of Fisher's exact test.
A comparison of allele frequencies (as above) does not take into account the structural relationships between haplotypes and thus does not enable one to infer evolutionary relatedness between populations. To be able to do this, one would have to take into account also the number of mutational events separating two observed haplotypes. This is as important as the difference between their relative frequencies. Thus, we followed the recently developed MST and AMOVA approach (see the Materials and Methods section for some detailed explanation), which was used to analyze complex mtDNA haplotypes (15 ,16 ) in this sense. As a first step, a network comprising all minimum spanning trees (MSTs) connecting the 77 observed haplotypes under a single step mutation (SSM) model was constructed (Fig. 1 ). Two mutations had to be assumed for the integration of four haplotypes, whereas one haplotype required three mutations to fit into the network.
From this network, its 77*77 Kirchhoff matrix (15 ) was next determined which has entries kij = -1 whenever haplotypes i and j are connected, and kij = 0 otherwise. Diagonal elements kii are equal to the number of nodes to which node i is connected. The determinant of any 76*76 submatrix gave the (same) upper limit, 3*1027, for the number of MSTs embedded in the network and also indicates (approximately) the number of equally parsimonious trees based on the observed 77 haplotypes.
From this, we concluded that maximum parsimony alone is not sufficient for the evaluation of evolutionary trees connecting these chromosome Y haplotypes and that other criteria are required. The molecular variance, [sigma]2, is the average measure of how far the haplotype of an individual in a given sample is away from that of any other individual in terms of mutational events (i.e. sum of branch length along the tree). By analogy to ordinary analysis of variance, [sigma]2 can be written as [sigma]2 = [sigma]a2 + [sigma]w2 where [sigma]a2 denotes the molecular variance between populations and [sigma]w2 is the molecular variance between chromosomes within a population (15 ,16 ). The ratio [Phi]ST = [sigma]a2/[sigma]2 reflects how much of the total molecular variance is explicable in terms of population difference in haplotype frequencies. Because of the high number of possible trees, comprehensive minimization of [sigma]2 and [sigma]a2 seemed impossible. Therefore, the parameters had to be characterized on the basis of a sufficiently large, representative collection of trees.
Figure 2 a illustrates the distribution of [sigma]2, [sigma]a2 and their ratio [Phi]ST as determined from 10 000 randomly chosen MSTs. Evidently, the low haplotype frequencies observed pose tight limits upon [Phi]ST, which implies that even the best MSTs with respect to [sigma]2 and [sigma]a2 would still leave a large proportion of molecular variance attributable to population difference. Nevertheless, similar to mitochondrial data, Y-chromosomal haplotypes yield a strong negative correlation between both [sigma]2 and [sigma]a2 and the cophenic correlation coefficient r(O,X) (Table 2 ). Therefore, we may conclude that molecular variance was indeed a meaningful criterion for tree quality in the Y-chromosomal situation.
In order to extract more information regarding optimal MSTs from the network in Figure 1 , we screened for potential common characteristics the 100 best trees with respect to [sigma]2, [sigma]a2 and r(O,X), out of the 10 000 chosen above. Some branches were indeed common to a number of optimal MSTs, and those found in >50% of them have been marked in bold printing.
We have demonstrated, for the first time, how sensitive PCR-based methods can be used to characterize highly informative haplotypes of Y-chromosomal microsatellite loci. With four out of the seven microsatellites presented, samples of Y chromosomes could readily be differentiated with respect to their Dutch or German origin on the basis of allele frequencies alone (Table 1 ), and as many as 77 haplotypes have been observed for these loci among the 159 males tested.
This observed chromosome Y variability seems to contradict restriction fragment length polymorphism (RFLP) studies (17 ,18 ) and sequence analyses (6 -8 ) on the human Y chromosome showing a low level of nucleotide diversity compared with autosomes. Different theories have been proposed for this phenomenon (5 ), including a simple numerical explanation (four autosomes and three X chromosomes are present for each Y chromosome in a population). The number of microsatellites appears also to be significantly reduced on the Y chromosome (19 ). Yet, as is shown by our data, Y chromosomal microsatellites are equally polymorphic as autosomal counterparts (11 ). It could be speculated that the reduced number of polymorphisms on the Y chromosome could more likely be the result of the relatively short time which has passed by since the origin of the present-day Y chromosome (19 ) and the preservation of the human Y chromosome from accumulating variation by male mating patterns and selective processes.
We next attempted to reconstruct the evolutionary relationships between the observed haplotypes. In the majority of recent publications on similar topics, evolutionary trees connecting DNA sequences from living organisms have been based upon maximum parsimony. Adopting an SSM model (2 ) and procedures previously described to construct MSTs for mtDNA data (15 ,16 ) it was possible to arrange the 77 observed Y-chromosomal haplotypes in a network comprising all MSTs connecting them. However, the upper limit for the number of equally parsimonious MSTs was 3*1027, and since the number of branches of equal unit length in the network was very large, the true figure should not be much smaller. Therefore maximum parsimony alone did not allow meaningful tree construction.
For these reasons, we employed analysis of molecular variance (AMOVA), previously described by Excoffier et al. (15 ,16 ) for mtDNA data. AMOVA is based upon the mutational difference between pairs of different haplotypes and yields analogs of variance components and F-statistics, reflecting the haplotype diversity within and between populations. When applied to four of our Y-chromosomal microsatellites, AMOVA revealed a larger mean proportion of between-population (66.8%) than within-population variance (33.2%), taken over all MSTs. Although the haplotype frequency distributions observed in our samples impose tight limits upon these ratios, optimal MSTs with respect to the total and between-population molecular variance could be shown to imply a significantly increased between-population variance component. This is in line with findings obtained for classical Y-chromosomal markers employing Southern blotting techniques (5 ), but contrasts with previous results using other Y-chromosome microsatellites (20 ).
The fact that even under an extremely diverse haplotype distribution (77 haplotypes among 159 males) enough information remained to allow discrimination even between closely related populations may seem counterintuitive at first. It could indeed be argued that if everyone is different from everyone else (or nearly so), population affiliation or common origin is likely to have become obscured. However, this argument is valid only as long as the type of polymorphism employed does not allow one to relate observed differences to an evolutionary time scale. This is, in turn, what we have undertaken, at least qualitatively, for the Y-chromosomal microsatellites, assuming that proximity in the emerging haplotype network implied evolutionary relatedness.
As explained above, the extremely high number of equally parsimonious MSTs (close to 3*1027) prevented us from using the principle of maximum parsimony to construct an evolutionary tree connecting chromosome Y haplotypes based on microsatellite loci. This is in marked contrast with, for example, the very appealing trees presented by Jobling and Tyler-Smith (5 ) based on base substitutions. With only nine such base substitutions, representative population groups of the world fall into only nine haplotypes. Trees based on these `slow' polymorphisms seem to reflect long-term evolutionary scenarios allowing us to survey major migration routes as well as the differentiation of major human populations. However, they do not allow separation of recently diverged populations. For this purpose, the rapidly evolving sequences like Y-chromosomal microsatellites seem more appropriate. They could enable the analysis of migration, settlement or mating structure of human populations in historic rather than evolutionary time spans.
For forensic purposes, it would be preferable if even more information could be obtained from a Y-chromosomal haplotype. However, expanding the four locus haplotype to, for example, a seven-locus haplotype including all loci listed in Table 1 , still did not enable us to identify all 159 males individually (results not shown). Nevertheless, we feel that the set of Y-chromosomal microsatellites presented in this manuscript, since the loci can be genotyped with PCR-based methods, will facilitate finding exclusions in, for example, rape and paternity cases in which only limited amounts of DNA are available.
In conclusion, the large within-population diversities noted for haplotypes of Y-chromosomal microsatellites will render them useful markers for forensic purposes. In addition, our results also appear to demarcate their utility in evolutionary and population genetic studies.
A logical expansion of the results presented here would be to compare more diverse and/or more distantly related populations, using more than four microsatellite loci and to directly compare these results with trees based on base substitutions. Such studies are currently under way.
The two groups of unrelated males analyzed in this study (70 Germans and 89 Dutch) comprised controls routinely used for the validation of forensic genetic markers. Care was taken that none of the males shared last names, and that all were white Caucasians.
Existing databases were searched for microsatellites fulfilling the following four criteria: (i) Y-chromosome specificity; (ii) level of polymorphism; (iii) allowing unequivocal identification of separate alleles; and (iv) robustness of use. A set of seven microsatellites was finally selected and used in the present study. Of these, DYS389 is present on the Y chromosome in at least two copies. Nevertheless, since the corresponding PCR fragments, albeit generated by the same primer pair, differed in length by ~100 bp, alleles could be unequivocally allocated to either of the two loci. All males were screened for the following tetranucleotide repeat loci (GDB locus names given):DYS19, DYS389-I and -II, DYS390, DYS391 and DYS393. In addition, DYS392, a trinucleotide repeat, was screened. Until complete characterization, we decided to label all alleles by their corresponding PCR fragment length.
Primer sequences and PCR conditions were essentially as published (11 -13 ) or as described in the Genome Database (GDB, The Johns Hopkins University, Baltimore), apart from the following modifications: (i) in the Berlin laboratory, radioactive PCR was performed with 25-50 [mu]l volumes containing 20-100 ng of genomic DNA, 50 mM KCl, 10 mM Tris-HCl (pH 8.3), 2 mM MgCl2, 0.2 mM of each dNTP (Boehringer, Mannheim, Germany), 20 pmol of each primer, 0.5 pmol of one [[gamma]-32P] ATP end-labeled primer and 1 U of Taq DNA polymerase (Promega) followed by size-separation of the PCR products on 4% denaturing polyacrylamide gels; (ii) in the Leiden laboratory (and for some loci also in the Berlin lab), fluorescent PCR assays were run in 25 [mu]l volumes containing 20-100 ng of genomic DNA, 50 mM KCl, 10 mM Tris-HCl (pH 8.3), 1.5 mM MgCl2, 0.2 mM of each dNTP (Pharmacia, Uppsala, Sweden), 100 ng of each primer with the forward primer marked by a 5' FITC label, 0.001% (w/v) gelatin and 1 U of Amplitaq polymerase (Perkin Elmer/Roche Molecular Systems, Inc., Branchburg, NJ, USA) followed by separation of the PCR products on 6% denaturing polyacrylamide gels using the ALFT automated sequencer (Pharmacia, Uppsala, Sweden) to detect and analyze the PCR products. A set of five DNA samples was tested in both laboratories in order to ensure consistent allele designation. Reference samples, allelic ladders, detailed protocols and further information on all markers are available from the authors upon request.
Differences in allele frequencies were tested for significance using the R*C Fisher-Exact method included in the StatXact software package (Cytel software corporation, Cambridge MA, USA). Haplotype diversity values were calculated and their differences tested for significance according to Nei (14 ).
Maximum parsimony is generally accepted as an intelligible criterion for the quality of candidate evolutionary trees connecting haplotypes, and a tree is regarded as sensible if it minimizes the total number of substitutions along its branches. Trees fulfilling this requirement are usually termed MSTs. In general, however, more than one MST will exist for a given set of haplotypes, and their superimposition generates a so-called `haplotype network'. In the present study, MST network construction was carried out essentially as described for mtDNA polymorphism (15 ,16 ), applying a modified Prim algorithm and a single step mutation model (SSM). The latter is tantamount to the assumption that microsatellite allele length changes occur only via the insertion or deletion of one repeat unit at a time.
It has been demonstrated for mtDNA (15 ) that MSTs which best explain the unconstrained differences between haplotypes also tend to minimize both the total ([sigma]2) and the inter-population ([sigma]a2) molecular variance in geographically stratified samples (15 ). The unconstrained fit of a given tree X to the data was measured by means of the so-called `cophenic correlation coefficient', r(O,X), which is the larger the more distant pairs of unrelated but frequent haplotypes are located in the tree, without invoking any mutation model. Since small values of [sigma]2 and [sigma]a2 have thus been found to be reasonable criteria for the credibility of MSTs with mtDNA, we decided to employ these parameters in our study of Y-chromosomal microsatellites, too.
We thank Chris Tyler-Smith and David N. Cooper for comments on the manuscript and Carmen Krüger, Claus van Leeuwen and René Mieremet for expert technical assistance. Jochen Hampe is gratefully acknowledged for data base searches. This work was supported by the Deutsche Forschungsgemeinschaft (Ro 1040/2-1 and Kr 1093/5-1).
1 Weber, J.L. (1990) Informativeness of human (dC-dA)n[middot](dG-dT)n polymorphisms. Genomics, 7, 524-530 (1990).MEDLINE Abstract
2 Di Rienzo, A., Peterson, A.C., Garza, J.C., Valdes, A.M., Slatkin, M. and Freimer, N.B. (1994) Mutational processes of simple-sequence repeat loci in human populations. Proc. Natl Acad. Sci. USA, 91, 3166-3170.MEDLINE Abstract
3 Bowcock, A.M., Ruiz-Linares, A., Tomfohrde, J., Minch, E., Kidd, J.R. and Cavalli-Sforza, L.L. (1994) High resolution of human evolutionary trees with polymorphic microsatellites. Nature, 368, 455-457.MEDLINE Abstract
4 Jorde, L.B., Bamshad, M.J., Watkins, W.S., Zenger, R., Fraley, A.E., Krakowiak, P.A., Carpenter, K.D., Soodyall, H., Jenkins, T. and Rogers, A.R. (1995) Origins and affinities of modern humans: a comparison of mitochondrial and nuclear genetic data. Am. J. Hum. Genet., 57, 523-538.MEDLINE Abstract
5 Jobling, M.A., and Tyler-Smith, C. (1995) Father and sons-the Y chromosome and human evolution. Trends Genet., 11, 449-456.MEDLINE Abstract
6 Dorit, R.L., Akashi, H. and Gilbert, W. (1995) Absence of polymorphism at the ZFY locus on the human Y chromosome. Science, 268, 1183-1185.MEDLINE Abstract
7 Hammer, M.F. (1995) A recent common ancestry for human Y chromosomes. Nature, 378, 376-378.MEDLINE Abstract
8 Whitfield, L.S., Sulston, J.E. and Goodfellow, P.N. (1995) Sequence variation of the human Y chromosome. Nature, 378, 379-380.MEDLINE Abstract
9 Chen, H., Lowther W., Avramopoulos D., and Antonarakis S.E. (1994) Homologous loci DXYS156X and DXYS156Y contain a polymorphic pentanucleotide repeat (TAAAA)n and map to human X and Y chromosomes. Hum. Mutat., 4, 208-211.MEDLINE Abstract
10 Mathias, N., Bayes, M., and Tyler-Smith, C. (1994) Highly informative compound haplotypes for the human Y chromosome. Hum. Mol. Genet., 3, 115-123.MEDLINE Abstract
11 Roewer, L., Arnemann, J., Spurr, N.K., Grzeschik, K.-H. and Epplen, J.T. (1992) Simple repeat sequences on the human Y chromosome are equally polymorphic as their autosomal counterparts. Hum. Genet., 89, 389-394.MEDLINE Abstract
12 Muller, S., Gomolka, M. and Walter, H. (1994) The Y-specific SSLP of the locus DYS19 in four different European samples. Hum. Hered., 44, 298-300.MEDLINE Abstract
13 Santos, F.R., Gerelsaikhan, B., Munkhtuja, B., Oyunsuren, T., Epplen, J.T. and Pena, S.D.J. (1996) Geographic differences in the allele frequencies of the human Y-linked tetranucleotide polymorphism DYS19. Hum. Genet., 97, 309-313.MEDLINE Abstract
14 Nei, M. (1987) Molecular Evolutionary Genetics. Columbia University Press, New York.
15 Excoffier, L. and Smouse, P.E. (1994) Using allele frequencies and geographic subdivision to reconstruct gene trees within a species: molecular variance parsimony. Genetics, 136, 343-359.MEDLINE Abstract
16 Excoffier, L., Smouse, P.E. and Quattro, J.M. (1992) Analysis of molecular variance inferred from metric distances among DNA haplotypes: application to human mitochondrial DNA restriction data. Genetics, 131, 479-491.MEDLINE Abstract
17 Jakubiczka, S., Arnemann, J., Cooke, H.J, Krawczak, M. and Schmidtke, J. (1989) A search for restriction fragment length polymorphisms on the human Y chromosome. Hum. Genet., 84, 86-88.MEDLINE Abstract
18 Malspina, P., Persichetti, F., Noveletto, C., Iodice, C., Terrenato, L. and Prantera, G. (1990) The human Y chromosome shows a low level of DNA polymorphism. Ann. Hum. Genet., 54, 297-305.
19 Spurdle, A.B. and Jenkins, T. (1992) The Y chromosome as a tool for studying human evolution. Curr. Opin. Genet. Dev., 2, 487-491.MEDLINE Abstract
20 Goldstein, D.B., Ruiz Linares, A., Cavalli-Sforza, L.L. and Feldman, M.W. (1995) Genetic absolute dating based on microsatellites and the origin of modern humans. Proc. Natl Acad. Sci. USA, 92, 6723-6727.MEDLINE Abstract
*To whom correspondence should be addressed
This page is maintained by OUP admin. Last updated Thu Oct 31 15:25:15 GMT 1996. Part of the OUP Journals World Wide Web service.Copyright Oxford University Press, 1996
