The relative-rate tests, in which three monophyletic lineages were defined to represent the taxa of a lake, revealed constant evolutionary rates among fishes from the three different lakes (P,0.05). The branch length for the Tropheus clade was 0.072, that for the Malawi mbunawas 0.046, and that for the Lake Victoria cichlids was 0.062. Using Tropheus as the reference group, aZ value of 0.906 was obtained; using Lake Malawimbuna, theZvalue was 0.509; and using Lake Victoria cichlids, the Zvalue was 1.376; in all cases rate constancy was not rejected. The minimum spanning tree for the Tanganyikan genus Tropheus identified six phylogeographically informative clusters of individuals having extremely small genetic differences from each other (average genetic distance51.33 base substitutions; SD5 0.63; 36 pairwise comparisons; see fig. 1AandB). The clusters were found in three regions of the lake, and taxa often were collected at opposite shores. The most striking similarities among populations were found in the central and southern regions of the lake, where individuals of one side of the lake had mitochondrial genotypes identical to those found at the opposite shoreline (fig. 1Aand Band table 2). For Eretmodus cyanostictus, closely related individuals were also found in four localities in the northern basin of Lake Tanganyika (one base substitution between Bemba and Ngombe, two base substitutions between Luhanga and Cape Kabogo as well as Luhanga and Ngombe; represented by triangles in fig. 1A; see also fig. 1C). Closely related individuals ofO. nasutuswere found in four localities of the southern basin of Lake Tanganyika (one to three base substitutions; represented by squares in fig. 1A, see also fig. 1D). Average genetic distances were not calculated for the latter two species due to the small sample size. Quinn 1992; Brown, Beckenbach, and Smith 1993; Stewart and Baker 1994). Since no calibration of the evolutionary rate was available for cichlid fishes, we used new evidence for the history of Lake Malawi (Delvaux 1995) and the average genetic distance among two ancestral lineages of Lake Malawi cichlids to derive a rate estimate. Among 25 mbunaand 14 utakahaplotypes (350 pairwise comparisons), an average Kimura distance of 6.54% (SD50.98%) was found. Our rate estimate for the most variable section of the control region thus amounts to 6.5%–8.8% per Myr, depending on which age is assumed for the Lake Malawi lacustrine ecosystem (fig. 4). The highly similar average genetic distances within clusters of closely related genotypes in cichlids of all three lakes (Lake Tanganyika, 1.33 mutations, 0.37%; Lake Malawi, 1.27 mutations, 0.35%; Lake Victoria, 1.33 mutations, 0.37%) would translate into an age range of between 57,000 and 40,000 years. This estimate would fit to a period of moderately low lake level (2160 m) in Lake Tanganyika 40,000–35,000 years ago (Lezzar et al. 1996; Cohen et al. 1997) but is older than the dating of the most recent period of very
The analyzed DNA sequences of rock-dwelling cichlids from Lake Malawi of 13 populations from both the central steeply sloping core region of the lake and the southernmost shallow sections of the lake contained four distinct haplotypes which were found in more than one species and in populations at the steeply sloping central region, to where the lake would retreat at a lake stand 500 m below its present level, as well as at localities in shallow water at the southern end of the lake (fig. 2 and table 2). These haplotypes also correspond to four phylogeographically informative clusters (average genetic distance51.27 base substitutions; SD5 0.96; 171 pairwise comparisons; see fig. 2).
The DNA sequences of Lake Victoria haplochromines from seven localities in the lake and four localities from surrounding rivers and lakes contained 4 out of 38 haplotypes that were shared by up to eight species originating from opposite sides of the lake, again defining geographically informative clusters of genotypes (average genetic distance51.33 base substitutions; SD5 0.97; 212 pairwise comparisons; see fig. 3 and table 2).
The presence of individuals having identical or very closely related mitochondrial haplotypes at distant localities in all three lakes, together with the remarkable degree of genetic heterogeneity found in many populations, consistently indicates an event of secondary admixis of previously isolated populations in the very recent past, followed by a range expansion. In each of the three lakes, this event was triggered by a major fluctuation of the lake level. Even though base substitutions occur stochastically, they can be taken as measures of divergence time when three criteria are met. First, all taxa have similar rates of base substitution in the analyzed gene segment. Second, the stochastic variation of mutation rates among individuals is homogenized by averaging multiple pairwise comparisons of taxa. Third, incomplete lineage sorting must be taken into account by comparing those individuals from a given population only to their counterparts from another population that are direct descendants of the same ancestral mitochondrial haplotype. All haplotype clusters that have diverged after the most recent admixis thus exhibit similar levels of genetic variation.
low lake level in all three lakes (18,000–12,000 years ago). This discrepancy may be due to the still relatively imprecise age estimate for Lake Malawi. Alternatively, it was repeatedly shown that mtDNA mutation rates seem faster when studied over relatively few generations, due to rapid accumulation of mutations at hypervariable sites. These mutation hot spots rapidly saturate after short periods of divergence, or genotypes may be quickly removed from populations due to selection against slightly deleterious mutations (Parsons et al. 1997; Gibbons 1998). We thus feel that the time estimate for the latest major dispersal event between 40,000 and 57,000 years ago is likely to be too old. To us, the most likely time for an almost synchronous spread in all three lakes would be their rise dated about 11,000 years ago by geological and sedimentological evidence, because the drop in the lake level 40,000–35,000 years ago was not sufficiently large to allow for crossing of the lake at the central basin by individuals of Tropheus (Gasse et al. 1989; Owen et al. 1990; Finney and Johnson 1991; Johnson et al. 1996; Lezzar et al. 1996; Cohen et al. 1997). The fluctuation of Lake Malawi over, at most, 150 m betweenA.D. 1500 and A.D. 1840 (Owen et al. 1990; but see Nicholson 1998) is not likely to have resulted in the observed pattern of genotype distribution, because this would have not fused all studied populations. The observed phylogeographic patterns of genetic divergences not only corroborate the suggested synchrony of the latest most severe dessication events in the three major East African lakes (Broecker et al. 1998), but also demonstrate equally severe consequences for their cichlid faunas. The same climatic phenomenon caused synchronization of genetic divergences of lineages within and among distinct species flocks and thus links the most recent evolutionary history and the stability of the fish communities of all three lakes to the same external modulators of adaptive radiation.
Our findings show that patterns of genetic divergences of stenotopic organisms provide valuable feedback on geology-based time estimates for events affecting lacustrine ecosystems. They have important implications for future works, since they open the possibility to compare simultaneous processes of species diversification among lakes that differ widely in the absolute ages of their radiating species communities (Sturmbauer 1998). Over the same period, a multitude of species originated in Lake Victoria and Lake Malawi with an impressive degree of ecological differentiation, whereas the Tanganyikan taxa that were exposed to the same habitat changes have hardly diverged ecologically and morphologically at all in the recent past. Future studies may allow identification of the characteristics and the driving modulators dominating an early stage of adaptive radiation as prevailing in Lake Victoria, in comparison with the intermediate stage in Lake Malawi and the highly advanced stage in Lake Tanganyika.
The strikingly similar average numbers of base substitutions in all phylogeographically informative haplotype clusters suggest that the latest periods of low lake level in Lakes Tanganyika, Malawi, and Victoria happened, or at least ended, roughly at the same time. The finding of identical genotypes of Tropheus at opposite shores in the central region of Lake Tanganyika (fig. 1A andB) suggests a retreat of the lake level by a minimum of 550 m, which would be sufficient to shift a continuous band of rock bottom into the depth limit of Tropheus (about 50 m). The distribution of identical or very closely related genotypes in Lake Malawi, and particularly their occurrence at the edge of the deep basin at locality 1 (fig. 2), suggests a retreat of about 500 m, but certainly more than 400 m, below its present level.
Since all data sets were found to have similar evolutionary rates in the analyzed segment of the control region, the inference of relative age estimates seems justified. The published rate estimates for the control region on various organisms differ widely (Vigilant et al. 1991;