Figure 2

50% majority-rule bootstrap consensus tree of 304 cichlid fish specimens based on 862 bp of the mitochondrial control region (neighbor-joining, HKY85 model with gamma substitution correction, 5,000 replicates). Relevant bootstrap values are depicted on the respective branches. The branches colored in red indicate the taxa that were chosen for the phylogenetic analyses combining the control region with sequences of the NADH dehydrogenase subunit II gene (see Fig. 3; for Ctenochromis oligacanthus from GenBank no control region sequence was available). The colors of the boxes that indicate the major clades refer to Figs. 1 and 3, the labels of the clades correspond to Table 1 [see Additional file 1]. Note that Haplochromis bloyeti (marked by an asterisks) had a control region sequence identical to H. sp. 1533 of [25], which was collected in the Malagarasi River, and grouped – together with other fishes from the Malagarasi area and from the Lake Edward/George region – into their group VII. Likewise, our H. sp. Tanzania I (marked by a circle) was identical to H. sp. 1738 of [25], which was collected in Lake Chala and clustered with other taxa from Tanzania into their group VI. In addition to Haplochromis gracilior (endemic to Lake Kivu), which was recently identified as close relative of the Lake Victoria superflock [4], we found another sister group to the superflock. This lineage includes Haplochromis paludinosus that occurs in the Malagarasi, as well as undescribed species from Tanzania and Lake Edward (see also Fig. 3). Like Haplochromis gracilior from Lake Kivu, all these taxa have the diagnostic character state 'Adenine' in position 630 of the control region alignment and root to the Lake Victoria superflock through the central rift valley haplotype [4], corroborating the view that Lake Kivu is the main reservoir from which the Lake Victoria superflock evolved [4].