Centrioles look highly complex, but so do snowflakes. Extremely simple self-organizing chemical and physical processes can generate structures of extraordinary complexity, and one must be careful to distinguish between two different types of complexity. One type of complexity is called "informational complexity" which can be measured as the number of bits of information required to explicitly describe the structure in question. For instance, one could recognize image features such as corners and edges and record their positions within the object. Structures that look visually complicated, such as a Persian rug, will have much higher informational complexity than a simple-looking pattern such as a checkerboard. An alternative measure of complexity stems from the fact that simple computer programs can generate complex fractal patterns. Depending on the pattern, a very simple program may suffice, while for other structures, it may take a longer program to generate the pattern. This type of consideration has led to a newer way to describe structures in terms of their "algorithmic complexity", also known as Kolmogorov complexity, which is the size of the smallest computer program (usually measured in bits) sufficient to generate the pattern (34
). Evolution and genomics is concerned with algorithmic complexity and not informational complexity. No matter how visually complicated a centriole is, its evolvability depends only on how complicated the genomic "program" must be that generates the structure.
We must therefore consider how many different genes are really needed to build a structure that is minimally functional as a centriole. This question has two parts: (a) how much of the structure of the centriole is really critical for its function, and (b) how many genes are necessary to generate the critical core structure. We will tackle the first question first - how much could the structure of a centriole be altered and still work? In this case, "work" would be defined as being able to provide at least some fitness benefit to a cell. Since the main job of centrioles is to form cilia, we can focus on how centriole structure contributes to assembly of cilia. Templating of cilia by centrioles requires is a set of preexisting microtubules, but the canonical arrangement of nine triplets in the centriole and nine doublets in the axoneme is by no means absolutely required for ciliary function, as organisms are known that have other numbers of centriolar or axonemal microtubules (36
). These cilia and flagella are motile, despite deviating from the canonical ninefold symmetry, hence we can only conclude that there is nothing magical about the number nine. Moreover, it is possible for centrioles having nine triplets to generate cilia having more than nine doublets, indicating that cilia have a degree of self-organization independent of the influence of centrioles (40
). Mutants in proteins of the centriole cartwheel can produce centrioles with variable numbers of triplets instead of nine (41
), and these modified centrioles can still form cilia, demonstrating that ninefold symmetry is not an essential feature of centrioles. It is also important to point out that the existence of a symmetrical array of triplets (be it nine-fold symmetric or with some other symmetry) is not required for the triplets themselves to form. Mutants also exist in which centrioles form asymmetric arrangements of triplet-containing units in variable orientations (43
). These studies demonstrate that single triplet subunit-sized chunks of centrioles can form without the overall rotational symmetry being present, so that if a single chunk could perform some useful function, this could be selected for prior to the development of the final nine-fold symmetrical structure.
Although ciliary motility is a highly coordinated process that might be quite sensitive to deviation away from nine-fold symmetry, cilia also play important sensory functions that would require little more than the microtubules and a membrane into which receptors could be localized. Cilia can also drive gliding motility which is independent of normal ciliary motility and might not require strong nine-fold symmetry (44
). In one interesting protist, the cilium consists almost entirely of an elongated central pair of microtubules, lacking the nine doublets over most of its length, yet this is able to drive swimming (45
). It thus seems reasonable that even a very rudimentary centriole-like precursor could allow formation of a proto-cilium that would give cells a tremendous advantage in terms of either sensory or gliding functions.
The second key question is how many genes would have to evolve in order to generate a centriole. One way to get at this question is to ask how many genes are essential to maintain centrioles. A recent genome-wide RNAi screen in Drosophila has argued that only nine genes are required for centriole duplication (46
). The apparent complexity of the centriole proteome, which likely consists of at least 50–100 proteins (47
), does not contradict the idea that only a small core set of genes are essential for centriole formation, provided the majority of the centriole proteins constitute add-ons, for example fibers that attach centrioles to different cytoskeletal elements.
Combining these considerations, it is apparent that only a small number of molecular innovations would be needed to produce some reduced, fragmentary version of a centriole that could in turn nucleate some sort of microtubule-based cellular extension that would be useful for gliding or sensation.
I propose that the original evolutionary precursor to the centriole may have resembled a single triplet blade subunit of the present day centriole, and consisted of a microtubule doublet or triplet structure that was able to extend a rigid proto-cilium consisting of a single doublet surrounded by plasma membrane out into the extracellular environment (). Extension of microtubules from the end of the doublet would not require additional evolutionary novelty since it is known that centriole triplets can serve as templates from assembly of purified tubulin (12
). The resulting structure, while simple, would have been able to provide basic sensory and gliding motility functions. The molecular requirements for formation of such a structure could be quite minimal: one or more proteins involved in forming the doublet microtubule structure plus a protein capable of linking the base of the structure to the cellular cortex. It is not currently known how microtubule doublets and triplets form, but the tektin family of proteins is thought to be involved in the process (49
). Evolution of one or more tektins, plus one or more proteins that can bind microtubule doublets and attach to the cortex, such as ODF2/cenexin (50
), might have been sufficient to produce a centriole precursor with basic function.
Figure 2 Proposed evolution of a proto-centriole capable of nucleating a primitive cilium-like structure. (A) Evolution of tektins allows formation of stable microtubule doublet and triplet structures. (B) Acquisition of one or more appendage proteins allows docking (more ...)
Once the original proto-centriole was established, it eventually was modified by addition of further gene products to produce the characteristic nine-fold symmetric array of triplets found today. Genes required for this transition would be recognizable as those which when mutated lead to aberrations in symmetry without affecting assembly of the triplet blade subunits themselves. This phenotype has been seen for mutants in the centriole cartwheel proteins SAS-6 and BLD10 (41
) suggesting that these genes may have evolved after the proto-centriole in order to bring about the modern cylindrical symmetric structure. The acquisition of a symmetric cylindrical arrangement might provide a structure with greater mechanical strength to support more powerful motility, and might allow a cilium to project at a right angle to the cell surface by presenting a rotationally symmetric array of attachment appendages. Although as discussed above variations in structure are seen in some lineages, the cylinder of nine triplets is by far the most common arrangement. This relatively invariant centriole architectural plan seen among eukaryotes, combined with the fact that the ninefold triplet architecture is found even in the earliest branching eukaryotic lineages, such as Giardia, suggests that the structure evolved just once and was then inherited throughout the eukaryotes. Modifications to centriole structure, including the reduction of centrioles to discs of singlets in nematodes would thus have occurred by secondary loss events.