Mitochondria and Neurotransmission: Evacuating the Synapse

Abstract

An abundance of mitochondria has been the hallmark of synapses since their first ultrastructural description 50 years ago. Mitochondria have been shown to be essential for synaptic form and function in many systems, but until recently it has not been clear exactly what role(s) they play in neurotransmission. Now, evidence from the nervous system of Drosophila identifies the specific subcellular events that are most dependent upon nearby mitochondria.

The synapse signifies a special type of polarized apposition or junction of two nerve cells which has the particular property of transmitting nervous impulses from one cell to the other. Therefore, it is reasonable to seek in electron micrographs for structures along the surfaces of perikarya and dendrites which have a peculiar, characteristic, polarized internal organization. An invaluable clue provided by light microscopy is that the synaptic terminals of axons contain a remarkable concentration of mitochondria.

—Sanford L. Palay, 1956

Thus began the modern ultrastructural study of synapses. In the earliest EM explorations of the nervous system, synapses were indeed located based upon their high density of mitochondria (Palay, 1956; and see references within), and decades later, this feature of pre- and postsynaptic regions is known to be extremely common. Since synapses share with other mitochondrion-rich regions of the neuron an acute need for ATP production and/or calcium buffering, there is an obvious biophysical and cell biological common sense to the arrangement (Hollenbeck, 1996). But how do mitochondria become concentrated in presynaptic regions, and for exactly what purpose?

The question of how mitochondria arrive at the synapse has become clearer in recent years. Fast anterograde axonal transport along cytoskeletal tracks conveys organelles and their proteins from their major sites of biosynthesis in the cell body to their sites of use and residence in the axon and terminal (Vallee and Bloom, 1991). Mitochondria are part of that flow, but unlike most axonal organelles, mitochondria move bidirectionally, through the use of several motor proteins, giving rise to a more complex pattern of axonal transport that can specifically position and reposition them along the axon (Hollenbeck, 1996). But, although important features of their transport remain to be worked out, their means of arrival at the synapse is no mystery. How they are then retained in the synapse is less well-understood, although several plausible mechanisms have been suggested by which cell signaling could halt and retain organelles in a particular region of the neuron (Bloom et al., 1993; Ratner et al., 1998; Morfini et al., 2002, 2004; Chada and Hollenbeck, 2003, 2004).

Once large numbers of mitochondria arrive in the presynaptic region, what essential functions do they carry out? A large body of data from various vertebrate and invertebrate systems has shown that mitochondria—and their Ca,²⁺ uptake and release in particular— are necessary for the events of neurotransmission. Sequestration of Ca²⁺ by presynaptic mitochondria regulates the cytosolic [Ca²⁺] during normal neurotransmission (e.g., David and Barrett, 2000). In addition, high mitochondrial transmembrane potential and/or mitochondrial Ca²⁺ sequestration are necessary to resist or recover from synaptic depression(Nguyen et al., 1997; Billups and Forsythe, 2002; David and Barrett, 2003; Talbot et al., 2003), to support posttetanic potentiation (Tang and Zucker, 1997), and to prevent asynchronous neurotransmitter release (David and Barrett, 2003; and references therein). While some synapses show evidence for cooperative Ca²⁺ buffering between ER and mitochondria (Ohnuma et al., 1999; see refs in Barrett, 2001), in other presynaptic regions, such as in the calyx of Held (Billups and Forsythe, 2002) and the mouse neuromuscular junction (NMJ) (David and Barrett, 2003), Ca²⁺ uptake by the ER cannot be detected, and mitochondria are thought to be the major agent of Ca²⁺ buffering.

Most of the data cited above derive from pharmacological disruption of mitochondrial function in normal synapses. But two papers in this issue of Neuron (Guo et al., 2005; Verstreken et al., 2005) describe the use of Drosophila mutants to analyze synapses with few or no mitochondria present and thus ask: what happens to synaptic structure and function if mitochondria are not available? Guo et al. carried out a genetic screen for mutants in synaptic function and identified dMiro, a member of a family of atypical monomeric GTPases associated with mitochondria (Fransson et al., 2003). In the fly nervous system, as elsewhere, dMiro is located on mitochondria, and homozygous mutant larvae show abnormal locomotion and die by the early pupal stage. Although the mitochondria of dmiro mutants are sparse in distal axons and absent from presynaptic regions, other organelles show only modest disruption of their normal distribution, and synaptic vesicle staining at the NMJ was essentially normal. Analysis of activity at NMJs showed a likely cause for the locomotory effects and lethality of dmiro mutations: in the absence of mitochondria, the mutant NMJs showed activity-dependent synaptic depression and more frequent asynchronous release of neurotransmitter than in wild-type. In other systems, this type of phenotypic abnormality would point to Ca²⁺. However, the effects on presynaptic Ca²⁺ levels were surprisingly modest: although mitochondria do sequester Ca²⁺ in wild-type NMJs, in the mutants only prolonged stimulation resulted in cytoplasmic Ca²⁺ levels higher than wild-type, and even this did not exceed physiological levels. Posttetanic clearance of cytoplasmic Ca²⁺ did not differ between dmiro mutants and wild-type, suggesting that little mitochondrially sequestered Ca²⁺ is released in the normal fly NMJ. Based on these data, the authors argue that the lack of mitochondrial Ca²⁺ sequestration is unlikely to be the cause of the abnormalities at dmiro mutant NMJs.

In work also reported in this issue, Verstreken et al. have taken advantage of a mutation of the gene for Drp1, the fly homolog of dynamin-related protein, which has been implicated in mitochondrial outer membrane fission (Praefcke and McMahon, 2004). In the mutant animals, mitochondria are much longer than normal in axons and are greatly reduced in number in terminals relative to wild-type. These authors also find that mutant NMJs show significant but unexpectedly modest differences in Ca²⁺ handling relative to wild-type. As in dmiro mutants, drp1 mutant NMJs have elevated resting Ca²⁺ levels and exhibit frequency-dependent synaptic depression in response to prolonged stimulation. But even prolonged stimulation did not produce Ca²⁺ levels above the physiological range. Again, these data argue that impaired Ca²⁺ handling in the absence of mitochondria is not the major cause of the physiological defects at mutant NMJs. If not Ca²⁺, the other major defect that ought to be expected in mitochondrion-deficient synapses is inadequate ATP production. In support of this, Verstreken et al. found that the synaptic depression of drp1 mutant NMJs could be partially rescued by front-filling the neurons with ATP.

The energy-intensive endocytosis, exocytosis, and vesicle recycling of the presynaptic region seem likely to be compromised by inadequate ATP levels, and Verstreken et al. pursued this possibility. Using an FM1-43 dye-loading and -unloading method under different stimulation regimes, they probed the ability of wild-type and drp1 mutant NMJs to form and release their rapidly recycling versus reserve vesicle pools. These labeling experiments suggested that mutant NMJs could load and unload their recycling pools as easily as wild-type, but that they had a smaller reserve vesicle pool. Further experiments in wild-type NMJs using inhibitors of oxidative phosphorylation or mitochondrial Ca²⁺ uptake recapitulated the results seen in mutant NMJs and indicated that mitochondrial ATP production, but not Ca²⁺ uptake, was required to fully unload dye from the reserve vesicle pool. The authors thus propose that the event in these NMJs that requires local mitochondrial function most acutely is the provision of ATP for the mobilization of reserve pool vesicles. Finally, Verstreken et al. further provide evidence that the activity of myosin and its regulation by myosin light chain kinase and protein kinase A are among the specific foci of ATP consumption for vesicle mobilization.

These mutants affect synaptic function, but what about structure? NMJs of the drp1 mutants actually showed similar bouton and synapse densities to wild-type (Verstreken et al., 2005). However, dmiro mutants, which had a more complete absence of mitochondria from the presynaptic region or even the distal axon, also displayed greater defects in synaptic structure. In dmiro mutants, the normalized NMJ length and number of boutons were actually greater than in wild-type, possibly a developmental adaptation to reduced synaptic strength. At the same time, the bouton volume and bouton-to-bouton distance were reduced, and the MT cytoskeleton at NMJs was attenuated and failed to extend into terminal boutons in mutants (Guo et al., 2005). It is not yet clear whether the mitochondrial deficit compromises initial synaptic development, ongoing plasticity (as in dendrites, Li et al., 2004), or both. And the structural differences in dmiro mutants could reflect the role of mitochondria in supporting development of synaptic regions, or pleiotropic effects of the mutations.

So, how do we compare these new data from Drosophila NMJs to those cited above from frog, lizard, crayfish, lobster, and mouse? Do some presynaptic regions require more—or different—mitochondrial support than others? Without a doubt! Different synapses have different properties, and junctional synapse physiology and mitochondrial properties can be very different even between neurons in the same nervous system. Note for example the intraspecies differences between phasic and tonic motor neurons: tonic synapses have more mitochondria (Brodin et al., 1999), with higher transmembrane potentials, and greater resistance to synaptic depression (Nguyen et al., 1997). (It is worth noting here that the NMJ is a phasic synapse.) On the other side of the synapse, things can be wilder still, as shown by the dramatic effects of mitochondrial content on the structure and function of dendritic spines and synapses in hippocampal neurons (Li et al., 2004). In addition, while it is clear that the Drosophila NMJs studied by Verstreken et al. and Guo et al. have a different, less pronounced requirement for mitochondrial Ca²⁺ handling than previously studied synapses, this issue bears careful analysis, lest our measurements and interpretations prove too simple. The mechanisms and purpose of ER versus mitochondrial Ca²⁺ sequestration in the synapse may be quite different, and it is possible that neither is there simply to “buffer” cytosolic Ca²⁺ (Barrett, 2001). In addition, the presence of mitochondria may fundamentally alter Ca²⁺ handling by the ER, complicating this picture further (Barrett, 2001). Accurate, dynamic [Ca²⁺] measurements and detailed modeling may be required to resolve the exact roles of Ca²⁺-handling organelles in the synaptic region (see Ohnuma et al., 1999). However, it does seem beyond question that many events in the presynaptic region require abundant ATP, the vast majority of which is supplied by the mitochondria. Perhaps what the recent Drosophila studies have done most clearly is to pinpoint which events in fly NMJs are most dependent upon high ATP levels. Can this be generalized to other synapses? Although the critical mitochondrion-requiring event(s) may differ among different synapses, the fly NMJ may nonetheless be guiding us toward key components of interest. Perhaps when the fly NMJ has been subjected to pharmacological and ultrastructural analyses similar to those that have been applied to other synapses, the picture will become clearer.

Indeed, much additional information is yet to come in the study of the fly NMJ. For example, most of the work reported thus far has analyzed dynamic physiological parameters but static images of mitochondria. Mitochondria redistribute willy-nilly in response to physiological events (Hollenbeck, 1996; Li et al., 2004), and we cannot know what pattern of mitochondrial movement underlies a fixed image. Postsynaptic mitochondria move quickly and specifically into regions of a dendrite where synaptic activity has been stimulated (Li et al., 2004). High-resolution real-time imaging with functional probes, combined with the available manipulations of fly NMJ synaptic activity, is likely to yield a similar bounty on the axonal side of the synapse.

A coda to this recent work is that both the drp1 and dmiro mutants are of considerable interest as regulators of the mitochondrial life cycle in neurons. The likelihood that Drp1 mediates mitochondrial fission raises several interesting questions. Axonal mitochondria in drp1 mutants are longer than normal, an expected result of the failure to divide. So, are the NMJs lacking mitochondria because longer mitochondria are transported less efficiently? Or because the fission of small, transport-friendly mitochondria from larger ones in the cell body has been inhibited? Or because mitochondrial growth and fission in the distal axon itself is necessary to replenish mitochondria at the NMJ? Direct observations of mitochondrial movement in these mutants could answer this question and provide insight into the relationship between mitochondrial biogenesis and transport in the axon. In the case of dMiro, the decrease in the density of mitochondria with increasing distance from the cell body is consistent with a gross defect in their anterograde transport. This calls to mind the work of Bloom et al. (1993), who originally suggested that inhibition of monomeric GTPase activity could inhibit axonal organelle motility. It is telling that Guo et al. find that presynaptic expression of normal dMiro not only rescues the phenotype by moving mitochondria into the NMJ, but actually drives them preferentially out to the most distal boutons. This implies that dMiro plays a regulatory role in the transport and/or positioning of axonal mitochondria, and so it joins the recently discovered Milton protein as a potential regulator of mitochondrial traffic in the distal axon (Stowers et al., 2002; Gorska-Andrzejak et al., 2003).