Endosymbiosis can be viewed as a means of genetically linking two or more species that are otherwise found in a symbiotic relationship, plus can potentially provide greater biochemical integration between the two species. In their origin they can be viewed as extensions of existing symbioses, including microbial consortia, or may instead be a consequence of an accidental capture of one organism by another. In addition to being a form of horizontal gene transfer that has the effect of perhaps instantaneously creating a new form of life (e.g., mitochondrial eukaryotes), endosymbiosis also forms the basis of a type of horizontal gene transfer that is nearly as profound: The migration of endosymbiont genes from endosymbiont to the host chromosome or chromosomes. Endosymbiotic relationships are impressively numerous, both quantitatively (effectively all eukaryotes) and qualitatively. They can readily form in single-celled organisms, and also are prevalent among multicellular organisms. The latter goes well beyond the mitochondria and plastids that represent the most common endosymbiotic bacteria, though the origin of these two organelles represent key steps in eukaryote evolution. Quoting Hoffmeister and Martin (2003) (p. 647):
…the concept of endosymbiosis in evolution… has fought a long uphill battle for acceptance. The main reason for reluctance among biologists to embrace the notion of endosymbiosis is probably because it runs contrary to Darwin's principle: endosymbiotic origins of organelles entail the occasional merger of two highly disparate cells into a single, bipartite genetic unit, simultaneously giving rise to novel and distinct taxa at higher levels (for example among the algae…). Darwin envisaged nothing of the sort (but he was also not primarily concerned with microbes). Endosymbiotic models have always drawn support from modern, observable examples of symbioses between free-living cells; extrapolating back in time yields models of interspecific evolution, which can and must accommodate lateral gene transfer. Accepting the premise that cell-cell interactions similar to those observable today should also have occurred in the past (uniformitarianism) and drawing upon molecular data, biologists have gradually become accustomed to the view that chloroplasts and mitochondria were in fact once free-living prokaryotes…
For endosymbiosis to occur, one organism, e.g., a bacterium, must be able to gain access to another organism, e.g., a eukaryote. There exist four general mechanisms by which such access may occur: (i) Gradual increase in the physical tightness of the association between microbes found in a consortium, until the association transitions from partial surrounding of one member by the other to full surrounding, i.e., engulfment. (ii) Acquisition of a bacterium as food, as in a food vacuole, which then escapes to the cytoplasm of the would-be host and takes up long-term residence. (iii) Intracellular parasitism by a bacterium which transitions to a more cooperative, i.e., mutually beneficial relationship. Or (iv), an otherwise temporary enslavement of one organism within the cells of another, i.e., which is effected by the larger organism and that then transitions into a more permanent arrangement. There also exists a fifth, grab-all explanation, and that is that a cell somehow finds itself within a second organism, as a consequence of some kind of accident, but then persists there. For a given capture event, presumably only one of the above mechanisms operates, though there is no reason not to believe that all of these mechanisms could have played a role in the generation of one or more endosymbiotic relationships.
Once captured, an endosymbiont must persist. For persistence to occur within replicating host cells, as for any horizontal gene transfer event, there must be a mechanism of post-acquisition replication and segregation to the daughter cells of the host. For endosymbionts, as formerly free-living and autonomously replicating organisms, it is not unreasonable to expect that such replication might be possible following capture, especially if sufficient nutrients are supplied by the host cell. Given sufficient replication, then segregation may occur by simple chance, just as high copy number plasmids can assure their segregation to daughter bacteria via high numbers. A bigger question, therefore, is what assures the survival, in the face of genetic drift, of the newly formed endosymbiotic relationship once it has formed. The answer, other than simply luck, is presumably the same as it is for any horizontal gene transfer event. That is, unless the endosymbiont provides a selective advantage to the larger organism, it is unlikely to persist. This is one appeal of the idea that at least some endosymbioses might have originated as microbial consortia, since in this case the selective advantage of the symbiosis would be preexisting.
Note regardless that an endosymbiont consists of linked genes that have been transferred as a single unit. As a consequence it is not unreasonable to imagine that useful biochemical pathways may be readily transferred into another organism via an endosymbiotic event. This suggests that this issue, that of the completeness of biochemical pathways, should not be of great concern with regard to the initial utility of initially acquired endosymbionts. Exactly how the new host organisms can utilize these newly acquired pathways, however, is uncertain, and particularly so to the extent that metabolic products may remain inside of endosymbionts rather than being directly accessible to a host cell. Thus, for example, what was the initial utility of mitochondria, i.e., if ATP was not initially leaked out of the proto-mitochondrion into the host's cytoplasm? Similarly, what was the initial utility of plastids? On the other hand, for microbial consortia that transition into being endosymbioses, this problem of access to useful metabolic products presumably was solved, as well, well prior to the formation of the endosymbiosis, that is, to the extent that microbial consortia involve organisms whose cytoplasms are fully separated by membranes. In this case of microbial consortia transitioning into endosymbioses, in fact there potentially would be one less membrane for metabolites to cross should the resulting endosymbiont be found directly within the cytoplasm of the new host rather than surrounded by a host membrane.
Perhaps the original endosymbiont, or at least the earliest which is unambiguously still in existence, are the mitochondria. The mitochondria were acquired by either eukaryotic or proto-eukaryotic cells 1.5 billion or more years ago. Their modern role in eukaryotic metabolism consists primarily though not exclusively of supplying products of oxidative phosphorylation to their hosts, just as algae supply products of photosynthesis to their fungal hosts in lichens. Oxidative phosphorylation involves numerous components – including those associated with glycolysis, the tricarboxylic acid cycle, electron transport, and ADP phosphorylation, etc. – that probably are not acquirable by more standard mechanisms of horizontal gene transfer, particularly as found between bacteria (i.e., transduction, conjugation, or transformation). As a result, selection need not have favored the linking together of the genes involved into operons or genomic islands within aerobic bacteria (that is, unless all of these functions could be linked together into a single genomic island that could then be transferred en masse, then there could be no selection on such linkage). The consequence is a difficulty for one organism to acquire oxidative phosphorylation in toto without acquiring an entire oxidative phosphorylation-encoding organism.
As with any horizontal gene transfer event, we have to consider the initial utility of the acquired genes rather than their ultimate utility. While acquisition of oxidative phosphorylation is an obvious ultimate utility, one has to wonder how the endosymbiont-acquiring organism might have been able to reliably access products of oxidative phosphorylation. Instead, it is possible that some other aspect of oxidative phosphorylation supplied the original utility of mitochondria or, alternatively, that the original motivation for existing as an endosymbiont was associated instead with the endosymbiont. For the latter, if the endosymbiosis were derived from a microbial consortium, then perhaps the utility simply was one of each organism living off of each other's waste products, but with access to those waste products simply more assured with one cell living inside the other. In particular, the endosymbiont could perhaps acquire those waste products while those products were still residing in the cytoplasm of the larger cell (making the endosymbiont a "parasite" of intracellular rather than extracellular wastes) while the endosymbiont's wastes would be efficiently captured by the larger cell without requiring transport into the latter's cell.
A possible other aspect of oxidative phosphorylation that could have supplied the proto-mitochondrion with an initial selective advantage is oxygen scavenging. In particular, the proto-mitochondrion might have provided their original hosts with protection from molecular oxygen, then found in increasing densities within the cyanobacterium-transformed atmosphere. Thus, just as especially facultatively anaerobic bacteria today sequester oxygen, and thereby contribute to the generation of anaerobic environments, so too could facultatively aerobic bacteria potentially protect individual cells with which they are associated by sequestering relatively rare oxygen molecules. By doing so, otherwise anaerobic organisms potentially could have invaded niches that are less anaerobic, and do so as whole organisms rather than as disparate parts of consortia. Meanwhile, the newly formed endosymbiont might have subsisted at least in part on the fermentative wastes of the anaerobe (i.e., products of the strict anaerobe's glycolytic pathways). Only with time, then, need the aerobe have come to directly supplying ATP to the larger organism. Indeed, even today we can view the mitochondrion-eukaryote symbiosis as consisting, at least in part, of mitochondria living off of the larger cell's glycolytic waste (i.e., pyruvate and NADH) while supplying ATP in return.
The acquisition of mitochondria appears to have happened only once. This is not to say that eukaryote acquisition of an aerobic endosymbiont did not happen more than once but, instead, that of those acquisition events, only one seems to have survived to the present. Another way of stating this point is that mitochondria appear to be monophyletic, that is, they represent a single lineage that dates back, presumably, to the original mitochondrion acquisition event. We can ascertain this monophyly in part by comparing among mitochondrial DNA. That is, mitochondria, as once free-living bacteria, initially had and to a large extent still have a genome that is independent of that found in the nucleus of their hosts. Some seemingly mitochondria-like organelles, in certain eukaryotic lineages, appear to lack even a remnant genome, however. This complicates claims as to the monophyletic nature of mitochondria and their ilk. In other words, without DNA to sequence, one must resort to old-fashioned comparative phenomics. The result of these latter efforts is a claim that mitochondria, as well as hydrogenosomes and mitosomes, together represent mitochondria along with mitochondria-derived organelles (van der Giezen and Tovar, 2005) .
Hydrogenosomes, as their name suggests, are hydrogen-generating organelles while mitosomes are hydrogenosome-like but nevertheless do not generate molecular hydrogen. In either case, these organelles are not involved in the aerobic respiration that one typically associates with mitochondria. There also exist mitochondria which, when under anaerobic conditions, employ organic compounds, rather than molecular oxygen, as final electron acceptors. Quoting van der Giezen and Tovar (2005) (p. 526):
…many eukaryotic organisms contain mitochondria that can produce ATP without consuming oxygen. Instead, these facultative anaerobic organelles use terminal oxidases that are capable of using endogenous organic compounds, such as fumarate or inorganic nitrate, as final electron acceptors… Facultative anaerobic mitochondria are found not only in some obscure unicellular protists that live in microaerophilic or anaerobic environments, but also in 'higher' multicellular eukaryotes, such as mussels, snails and parasitic helminths. The biochemical diversity of anaerobic mitochondria and the biochemical range of aerobic mitochondria illustrate the biochemical heterogeneity of this organelle.
Thus, the "truism" that mitochondria are associated with aerobic respiration in eukaryotes presumably is true, or at best is only true for eukaryotes that in fact are respiring aerobically.