Finding. Madagascar's extinct hippos were not three tidy species sitting on an old island clock. Salis, Gatesy, Raxworthy, Flynn and colleagues recover the first ancient mitochondrial genomes from 37 subfossils, seven localities, and all three named taxa. Thirty-five endemic sequences form one clade beside Hippopotamus amphibius: a single colonization. Inside that clade are five deeply divergent, geographically structured mitochondrial lineages that spread across most of the island in the Pleistocene. Tip-dating puts the crown at about 1.41 million years, after a split from mainland Hippopotamus at about 3.24 million years. That is a young radiation for a large mammal, and the youngest such endemic radiation Madagascar is known to have made. The names do not survive contact with the tree. Material called H. lemerlei is scattered across clades. H. laloumena and H. madagascariensis sit inside the same broader diversity.
Why this paper matters
Island megafauna stories usually arrive as morphology plus a hope that mitochondrial fragments will agree. Here the mitogenomes are the primary instrument, and they rewrite both colonization and taxonomy. One landing from the common hippo. Five mitochondrial countries. A Pleistocene burst, not a Miocene leftover. Then a late-Holocene extinction that deleted more than a million years of endemic diversification for reasons the authors refuse to invent.
For mitochondrial biology this is phylogeography at full-genome scale in degraded bone, not OXPHOS. It still belongs here. Ancient mtDNA is how lost lineages get dates, and the dates are sharp enough to argue that Madagascar was still inventing large endemic mammals well into the Pleistocene.
What they actually measured
Thirty-seven subfossils, all three named species, seven sites. Phylogeny from mitogenomes, then morphology and a total-evidence tree. Tip-dating for colonization and crown age. Ancestral-state reconstruction of body size: early dwarfing, then a secondary size increase in H. laloumena. Morphology agrees with some lineage assignments and also points at diversity the names never captured.
How to read the score
High 80s on a mitochondria rubric because this is a real ancient mitogenome dataset with dates and a taxonomic rupture. It is not a disease paper. Confidence is high for the single-colonization, five-lineage mitochondrial claim as stated. Nuclear genomes could still rearrange species limits.
Caveats
mtDNA can split populations that nuclear DNA later joins. Coverage, damage patterns, and contamination controls are not reproduced in this brief. Extinction cause is explicitly open. Do not turn 1.41 Ma into a textbook number without the posterior intervals from the paper.
What to do with it
If you curate ancient mitogenomes or island radiations, this is the week's hippo tree. Pull the 37 genomes and the tip-dating priors. If you came for respiratory-chain disease, skip it. Do not keep teaching three clean Malagasy hippo species from a 1990s key.