In 1987, Thomas Cavalier-Smith proposed a kingdom called Archezoa. The organisms in it — Giardia, microsporidians, trichomonads, a handful of others — were amitochondriate: no visible mitochondria. They also branched near the base of eukaryotic phylogenetic trees, suggesting they split from other eukaryotes before the mitochondrial endosymbiosis. Cavalier-Smith's claim was that these were living fossils: lineages that had diverged while eukaryotes were still premitochondrial, and had since survived unchanged in anaerobic environments where oxidative phosphorylation never mattered.
The evidence for early divergence came from small subunit ribosomal RNA — one of the few genes shared across all life. Build a tree from those sequences, and the Archezoa reliably appeared near the base. Simplest interpretation: they're old. They branched off before everyone else. Their position in the tree reflects their position in time.
But phylogenetic position and time of divergence are only the same thing under one assumption: that different lineages evolve at roughly the same rate. If that holds, distance in the tree corresponds to time. If it doesn't, a lineage can look ancient simply by evolving faster.
This is the long branch problem. In phylogenetic trees, fast-evolving sequences appear on long branches — they've accumulated more changes per unit time. In parsimony and early distance-based methods, long branches cluster with each other and with the outgroup, because they share more convergently evolved characters than they share with their actual relatives. A lineage that evolved rapidly will appear to branch off earlier than it did. It will look like a basal lineage, a primitive divergence, a living fossil — not because it is old, but because it is fast.
The Archezoa were fast-evolving. Their rRNA sequences had diverged substantially from other eukaryotes. The phylogenetic methods of the late 1980s and early 1990s placed them at the base of the tree, and the conclusion seemed obvious: primitively amitochondriate, diverged before the endosymbiosis, never had mitochondria.
In 1998, Andrew Roger and colleagues found a gene encoding a mitochondrial chaperonin — cpn60, which folds proteins inside mitochondria — in the nuclear genome of Giardia lamblia. Not in a mitochondrion. In the nucleus.
This happens routinely in mitochondrial evolution: genes transfer from the mitochondrial genome to the nuclear genome, a one-way ratchet that has moved nearly all ancestral mitochondrial genes into the nucleus over evolutionary time. Finding a gene in the nucleus with mitochondrial ancestry means the gene was once in a mitochondrion. Giardia had been carrying this testimony in its nuclear genome. The gene had survived something the organelle had not — or at least not in recognizable form.
Subsequent work confirmed it. Giardia has mitosomes: tiny, double-membrane organelles with no genome and no ATP production, doing only iron-sulfur cluster assembly. They are mitochondria reduced until almost nothing of the original function remains, except one thing that cannot be reduced: iron-sulfur clusters are essential for electron transport and DNA synthesis, and the machinery for making them traces back to the bacterial endosymbiont. Strip everything else away and that persists.
The Archezoa were not primitively amitochondriate. Their ancestors had full mitochondria, which were reduced over time as the organisms adapted to anaerobic environments, until only the irreducible minimum remained. The hypothesis was abandoned.
The interesting question is not whether it was wrong — it was — but why the evidence looked so convincing. The two errors weren't careless. The first (no visible mitochondria) required molecular biology to correct: the mitosomes are too small and too reduced to recognize as mitochondria unless you're specifically looking for vestigial organelles. The second (phylogenetic basal position) required phylogenetic methods that account for rate variation — methods that weren't computationally feasible when the hypothesis was proposed.
What makes the case interesting is the inversion at the end. The presence of a gene proved the existence of an absence. Cpn60 in the nucleus wasn't evidence that Giardia had mitochondria. It was evidence that Giardia once had mitochondria that had since been reduced. The gene's presence testified to something that was no longer there.
Never-having and having-and-lost are indistinguishable by direct observation at a single timepoint. Both produce the same result: no mitochondria. The only way to tell them apart is to ask a different question — not "do you have this structure" but "do any of your genes carry ancestry from this structure." That question requires a different tool, and a reason to ask it even when the direct observation says the structure isn't there.