Discovered living in the solfatane fumaroles around Naples, Italy, Cyanidioschyzon merolae is a very small red algal cell. At just two micrometers in length, most of its interior is occupied by its single plastid and small nucleus. A single mitochondrion is wedged between them. Its division is inextricably yoked to its host’s cell cycle. When such intimate dependence is found, it is easy to presume that it is the host cell’s replication program that imposes requirements on the organelles, but in nearly all cells, the mitochondria replicate before the nucleus. The question of who is controlling whom remains.
A recent article in Nature magazine’s Communications Biology by a group of Japanese researchers has exploited the minimalist cell architecture and mitochondrial dynamics of C. merolae to shed new light on life at its most primitive stage of endosymbiosis.1 In their paper, Shoichi Kato et al. trace patterns of critical phosphorylation points in division cycle regulator proteins, and show that they were conserved throughout the course of evolution. In particular, Kato et al. looked at a serine/threonine kinase known as Aurora. While humans have an A, B, and C form of Aurora kinase, and lesser invertebrates have two, C. merolae has just one, which has been designated CmAUR.
Building upon earlier studies showing that Aurora kinase was imported into mammalian mitochondria, Kato et al. show that in humans, Aurora phosphorylates are a key component of the mitochondrial division ring known as Drp1 (dynamin-related peptide). They also demonstrate that in C. merolae, Aurora (CmAUR) phosphorylated Drp1 (CmDnm1) at precisely the conserved locations controlling mitochondrial division. These observations reveal the surprisingly large and sometimes opposing effects of adding phosphate groups at each of the many accessible serines or threonines in proteins.
The actuation of division cycles through linked chains of phosphate switches is the bread and butter of propagating life. Kato et al. show that the remarkable, frozen-in-time preservation of ancient yet extant organisms translates what was formerly a process of evolutionary guesswork into one largely comprised of phosphate accounting.
The importation of Aurora into mitochondria is only the beginning of a long enzymatic journey, one that begins with two rounds of N-terminal protease cleavage. Before phosophyloration can commence, Aurora must first undertake its own auto-phosphorylation. Downstream targets like Drp1 are just one of many for Aurora, while Aurora itself is only one among many phosphorylation patrons of Drp1. The mitochondrial hose clamp made from homomeric Drp1 has been engineered with only a limited dynamic range of compression. In the cytoplasm, that range is good enough to pinch off endocytos clathrin-coated vesicles from the cell membrane. In double-membraned mitochondria, a host of other proteins are actually needed to finish the job.
In the face of these constraints, nature forked the ancestral, bacterially derived dynamin precursor into independently targeted variants. In the last eukaryotic common ancestor, the presumed bifunctional Drp1 was duplicated into specialized mitochondrial and vesicle forms at least three times. These moments marked the evolution of the alveolates, green algae, and the fungi and metazoans. In mammals, there are several more additional rings of complexity required for full division. It is now believed that the final pinch requires the efforts of a completely new and essential dynamin, Drp2.2
A few formal terms are needed in order to complete an informal model of mitochondrial division. At a minimum, we have an outer domain, or OD, and an inner domain, or ID, the cell membrane making the boundary between them. For all practical purposes, the critical ID to which the membrane must be compressed in order for spontaneous division to occur approaches the thickness of four membranes. Depending on the species and the source, Drp1 charged with its essential guanosine triphosphate (GTP) cofactor preferentially assembles into rings of 13 to 18 monomers, a diameter of around 50nm. In yeast, a related dynamin-like protein known as Dnm1 forms double-start helices under a wide variety of experimental conditions, while other closely related isoforms generate parallel sets of fixed, closed rings. Our own Drp1 seems to prefer polymerization into a single-starting spiral thread whose ends are able to slide past each other during compression.3
Many protists, including red algae, retain the essential components of bacterial binary fission to crack their organelles. Foremost among them is the protein Filamentous Temperature-sensitive mutant Z (FtsZ), the forerunner to GTP-hydrolyzing tubulin itself. The acquisition of additional new dynamins from the host was accompanied over the course of evolution by a parallel downgrading of FtsZ. In protists, mitochondrial division typically involves painting the inside ID with a contractile FtsZ ring. This marks the scission site for a tandem dynamin-mediated effort on the outside. To this day, all dividing plastids simultaneously pull from the inside with FtsZ while squeezing from the outside with dynamins. Each dynamin can act in different kinds of membrane. Some prefer the cardiolipin-containing outer mitochondrial membranes; others, such as the isoforms, the region of rapid-fire synapses that mediate short latency vesicle reabsorption. These subtle preferences are handled by various post-translational modifications, primarily phosphorylation/dephosphorylation events linked into banks of miniature timer relays.
In addition to dynamins, a role for cytoplasmic actin polymerization has come to light in controlling division. In mammals, there are also fingerlike tentacles of specialized endoplasmic reticulum that envelope free mitochondria, or portions of a budding network syncytium, and break them up like a cephalopod on a clamshell. This is a process that typically occurs at the immediate periphery of the nucleus. In brain cells, mitochondria serve to push far reaching neurites out from the cell. When traversing axons, mitochondria often find harbor in target tissues. In dendrites, there are elusive outpockets of ER known as the Nissl substance. Although it has been suggested that the dendritic ER could be involved in protein translation, it is tempting to speculate that the Nissl substance plays a role in mediating autophagic mitochondrial selection.4
A major theme of the paper by Kato et al. is the conserved elements of mitochondria division. But it is also important to take a look at what is not conserved at all. Just as a cell cannot divide until it has replicated its nuclear DNA and organelles for its daughter cells, the mitochondrion has no business dividing until it has replicated its nucleoid.
This raises several questions. First, mitochondria are always fissioning and fusing, with or without nucleoid replication. Fusion can be thought of as a special case of fission. Unique dedicated proteins are involved in the fusion of the respective inner and outer membranes. There is no evidence that nucleoids replicating in the matrix will lead to later mitochondrial division. Nucleoids can be cleaved off in tiny vesicles and transmitted to other cells, and perhaps even beyond. Second, the most interesting cells tend to be highly differentiated, post-mitotic cells that are hard-pressed ever to divide again. There is no obvious need for them to double up on organelles. Third, replication of nucleoid DNA is not independent of the transcription of nucleoid DNA. Transcription is the only way to print the RNA primers that are needed for replication. Nucleoid replication has not been shown to be distinct from mitochondrial DNA recombination and repair. Fourth, the mechanism of nucleoid replication is not conserved. Different species use alternative strategies for nucleoid replication, and different organs—heart muscle, skeletal muscle, neuron, and fibroblast—use different mechanisms within the same organism. It is likely that different kinds of nucleoid replication are employed in different parts of the same cell, depending on whether the mitochondria are in their individual or network phase.
Exactly how are nucleoids replicated, and why are different mechanisms used? Experiments have defined three ways in which to produce new nucleoids:5
- In the strand displacement model, mtDNA replication is unidirectional, asymmetric, and asynchronous, and does not involve the formation of Okazaki fragments.
- In the synchronous strand-coupled model, replication proceeds bidirectionally from various origins with a subsequent synchronous movement of two forks along the mtDNA.
- In the RITOLS (RNA incorporated throughout the lagging strand) model, replication incorporates ribonucleotides to form RNA-DNA hybrids and shares features of the other to two mechanisms.
Each method appears to be optimized for different needs, among them speed, error correction, and resistance to single-strand breakage. Nucleoids usually consist of one or more circles of DNA (mitochondrial chromosome M), often found concatenated along with the appropriate binding proteins. In the brain, mitochondria are frequently found to have nucleoids that contain complex three- and four-way junctions (G-quadruplexes). These complexes are now believed to be possible sites of recombination. By contrast, mitochondria in the adult human heart muscle often rearrange their nucleoids into a network of rapidly replicating linear genomes.6
It is an axiom of nature that what holds for the macroscopic world holds for mitochondria. Fish form schools for safety in numbers, slime mold cells aggregate to pool resources in times of stress, geese flock to increase the reliability of their migratory decisions, and ants and locusts swarm to tackle what they cannot do alone. Just as the diffusion of oxygen sets a hard upper bound on the size, shape, and deformability of red blood cells navigating the microvasculature, various unseen factors restrict mitochondria in traversing the brain. Mitochondria seeking passage through the tightly packed, crystalline, parallel-fiber axon arrays face a hard limit on the size of their ODs. The cytoskeletal structure in the axon might form a barrier to motion as well as a scaffold.
It is possible to pin down the role played by Aurora kinase at the molecular level, at least in theory. But its macroscopic effects still require experimental exploration. Genetic techniques can be used to under- or overexpress the Aurora transcript; its function can be silenced using interference RNA; or it can be directly blocked with other inhibitors. Varying the concentration of active Aurora has recently been found to give a U-shaped curve for the size of the network. When Aurora is completely silenced, fission is inhibited. When it is overexpressed, fusion is enhanced. When energy production is investigated, a monotonically increasing concentration of adenosine triphosphate (ATP) was found as Aurora was increased.7
While advances in microscopy are now revealing what dividing mitochondrial profiles look like from the outside, much remains to be discovered regarding how their internal matrix is laid out in three dimensions. Where are the ribosomes, and how many are there at division time? Is the matrix largely random insofar as there are no obvious rotational symmetries about the long axis? Are the cristae and therefore their crista junctions tubular and round, or are they oval? The post-translational protein modifications, particularly phosphorylations, that control so many macroscopic observables in cell morphology undoubtedly play a role here, too. The handbook of ingenious mitochondrial mechanisms is currently being written. The number of controlling phosphorylations may be a large index, but it is finite.
