According to the endosymbiotic theory, mitochondria are the descendents of bacteria. The theory postulates that a primitive cell engulfed some aerobic bacteria and rather than digest them, a symbiotic relationship was established, where each partner benefited from the new relationship. This relationship then set the stage for the ultimate stream-lining of the bacteria, such that they were transformed into mitochondria through the transfer of much of their gadgetry to the host nucleus.

In a nutshell, the essence of the argument for the endosymbiotic origin of mitochondria is that mitochondria look like they share a common ancestor with bacteria. The argument is quite convincing, as there are numerous mitochondrial genes whose sequences are much more similar to bacterial sequence than that which exists in the nucleus of the same cell. In fact, this is an example where no one piece of evidence carries the day, but instead it's the cumulative power of multiple lines of evidence.

There are some important lessons to be learned from the development of the endosymbiotic theory. Some have a tendency to engage in a mild form of revisionism when arguing for the neo-Darwinian model of evolution, where they portray the model as having provided more focus and output than it really has. The history behind the endosymbiotic theory shows us the picture is more complicated. The first thing to note is that the neo-Darwinian model of evolution did not predict the origin of these organelles by endosymbiosis. When the endosymbiosis hypothesis was first proposed by Wallin in the 1920s, it was essentially dismissed as something as too "fantastic" to be discussed in "polite biological society." When Margulis resurrected this hypothesis, some traditional evolutionary biologists (Uzzell and Spolsky) actually labeled it as something akin to a revival of "special creation." Evolution was supposed to proceed in small steps, not in symbiotic leaps. Just as neo-Darwinian theorists originally resisted lateral gene transfer, so too did they resist endosymbiosis.

Nevertheless, what was Wallin's evidence? According to Margulis, his evidence "was based on the size, shape, staining properties and general cytological behavior or the organelles which he claimed were comparable to bacteria." In other words, Wallin drew up a list of similarities which seems awfully much like some form of argument from analogy.

Yet it is noteworthy that Wallin did not (AFAIK) predict that mitochondria should thus possess bacterial genetic factors (at this time, no one knew that DNA was the genetic material). When Margulis revived Wallin's hypothesis, she pointed to the newly discovered facts that indicated mitochondria had their own DNA and the mitochondria replicated rather than formed de novo. This extended the bacteria-like appearance of mitochondria, making it look even more like bacteria and mitochondria are related. Yet I cannot find where Margulis actually predicted this DNA sequence would nest with bacterial sequence, let alone some specific bacterial subgroup.

The history behind the endosymbiotic hypothesis gets even more complicated. First of all, Margulis did not revive the hypothesis to account for two eukaryotic organelles, but for the origin of eukarya itself. Her first paper, "Origin of Mitosing Cells" makes it clear that the lion's share of her attention was devoted to the centrioles, where she even went so far to argue that centromeres were of bacterial origin. Suffice it to say that the main thrust of her hypothesis has been rejected by mainstream science, where endosymbiosis currently accounts for the origin of mitochondria and chloroplasts, not eukarya (although many now favor a symbiosis between archaea and eubacteria). Furthermore, classical evidence, cited by Wallin, is also suspect. For example, while mitochondria of higher organisms may appear similar in shape and size to bacteria, this is often not true among protozoa.

Margulis offered a mismash of predictions early on, some validated, some not. For example, Margulis predicted corroboration of the main part of her thesis - the centrioles/MTOC were of bacterial origin. She also erroneously predicted that we would be able to culture mitochondria and that tubulin sequence would be homologous to the flagellar sequence of bacteria. On the other hand, she successfully predicted that we would better demonstrate gene transfer from organelles to nuclei.

I mention all this not in any attempt to discredit Margulis and her hypothesis (as I accept them), but to illustrate that some have a tendency to forget the failed predictions and remember the successful ones. In other words, in terms of predictive abilities alone, both the Darwinian model and endosymbiotic theory have a mixed history of success. The simple observations of similarity have done most of the work, where with increased resolution, the similarities not only remain, but become more impressive.

While it seems rather clear that mitochondria are derived prokaryotic systems, the actual mechanism behind this transformation remains unclear. Whatever the mechanism, it would have to account for two facts.

1. Mitochondria are monophyletic. In other words, the mitochondria from various protozoa, plants, animals, and fungi are all derived from the same stem population of mitochondria that were once free-living bacteria. The significance of this is that explanations that paint with a broad brush are suspect. Citing a laundry list of common cellular and molecular events (phagocytosis, modern examples of endosymbiosis between protozoa and bacteria, mechanisms of gene transfer, etc.) left to themselves would lead us to expect a polyphyletic origin of mitochondria. The monophyletic nature of mitochondria suggest there was something unusual about the transformation.

2. Mitochondria are not just similar to bacteria, they nest with a bacterial crown group - the Rickettsia type alpha-Proteobacteria. Since alpha-proteobacteria branch late in bacterial phylogeny, and Rickettsia are one of the twigs in this late-branching group, it stands to reason that extensive bacterial evolution and divergence occurred prior to the endosymbiotic event. Since the endosymbiotic event likely occurred long after the last universal common ancestor, this raises the specter of bringing together two very different genetic control systems. How one would subsume much of the other is a challenging line of thinking.

Finally, there is another twist to the story. The standard story has some primitive eukaryote engulfing a Rickettsia-like bacteria, spawning cells with mitochondria. Since mitochondria are widespread among eukarya, this is thought to have happened first. Then some eukaryotes with mitochondria engulfed cyanobacteria to form the chloroplasts. It would be nice if this scenario matched bacterial phylogeny. But cyanobacteria were around long before alpha-proteobacteria appeared. What’s more, cyanobacteria even carry out aerobic respiration. So what’s so special about Rickettsia-like alpha-proteobacteria? The standard endosymbiotic theory would be much more strongly supported if mitochondria and chloroplasts nested together and diverged very early off the bacterial tree.