Memories are encoded by efficient signalling between neurons. The myosin V proteins help this process by shuttling receptors and membranes to make synaptic junctions better detectors of incoming signals.
Synaptic junctions transmit information between neurons. The efficiency with which they do this is affected by how frequently they are activated, a cellular equivalent of experience. For example, repeated activation yields a lasting increase in the efficiency of synaptic transmission — a process called long-term potentiation (LTP) — which is thought to underlie memory formation. LTP depends both on enhanced insertion of receptors for the neurotransmitter glutamate at spines (the sites of synapses) and on spine growth. How neurons coordinate these two processes to trigger LTP has remained a mystery. Writing in Cell, Wang, Ehlers and colleagues1 demonstrate that the motor protein myosin Vb is one missing link.
Dedicated to receiving signals, spines are tiny protrusions on the dendritic processes of neurons. They contain two types of glutamate receptor: AMPA receptors, which mediate most glutamate-dependent synaptic transmission; and NMDA receptors. On induction of LTP, activated NMDA receptors mediate a rapid influx of calcium ions (Ca2+) into spines, which initiates a cascade of signalling events, culminating in the insertion of extra AMPA receptors into the spines, and spine growth.
Previously, Ehlers and colleagues had shown2,3 that, during LTP, organelles known as recycling endosomes, which are normally found in dendritic regions near the base of spines, provide both AMPA receptors and membrane material for spine growth. In their latest work, the authors asked1 what molecule directs the transport of recycling endosomes from dendrites into spines in response to Ca2+ signals.
Within the cell, motor proteins move organelles along tracks of cytoskeletal structures called microtubules and actin filaments. Dendrites are rich in microtubules, whereas spines are mostly filled with actin. So the ideal candidate for Wang and colleagues' search would be a Ca2+-dependent protein that could hijack recycling endosomes from microtubules in dendrites and move them along actin within spines.
Myosin V proteins seem to fit the bill. In many cell types, these actin-based motor proteins specialize in organelle transport, their structures are regulated by Ca2+, and they can counteract organelle movements along microtubules4. Furthermore, one of the three myosin V isoforms, myosin Vb, binds to recycling endosomes through the Rab11 protein on these organelles and another protein known as Rab11-FIP2, which interacts with Rab11 (ref. 4). In neurons, myosin-Vb activity affects surface expression and clustering of AMPA receptors5.
Wang et al.1 find that myosin Vb is indeed the protein mediating AMPA-receptor and membrane trafficking to spines. Visualizing the cellular movement of this protein by tagging it with a fluorescent molecule, they show that, under resting conditions, it is mostly present in spines. However, on induction of LTP and activation of NMDA receptors, it moves to dendrites, binds to recycling endosomes through Rab11/Rab11-FIP2 and then returns to spines with its cargo containing the AMPA-receptor subunit GluR1 (Fig. 1).
Myosin Vb can adopt two conformations: a folded, inactive state with a high affinity for actin, and an extended state, which binds to Rab11/Rab11-FIP2. The LTP-triggered increase in Ca2+ concentrations seems to affect myosin Vb by inducing a conformational change in this protein from the inactive to the extended state, allowing it to connect to recycling endosomes1.
But is myosin-Vb activity required for induction of LTP? To propel their cargo, myosin V proteins use ATP molecules as fuel. The ATP-binding pocket of these proteins can be subtly mutated to preserve their normal function, but to allow the protein to be quickly locked by chemical means into the inactive actin-bound form6. This approach avoids potential complications due to compensatory mechanisms that could arise during long-term manipulations of myosin V activity. Using engineered mice with such mutations, Wang et al.1 measured synaptic efficiency in mouse brain slices. They show that acute inhibition of myosin Vb blocks LTP within minutes.
Another recent study7 has also demonstrated a role for myosin Va — another myosin-V isoform — in LTP. With the help of active Rab11, this protein binds directly to the cytoplasmic tail of GluR1, supporting the connection between myosin isoforms, AMPA receptors and recycling endosomes. Nevertheless, brain slices of a mouse mutant lacking myosin Va show normal LTP (ref. 8), suggesting that the role of myosin Va in LTP is subtle. It would be interesting to test whether, using Wang and colleagues' approach, acute inhibition of myosin Va over a timescale of minutes would be sufficient to block LTP.
A key feature of LTP is its input specificity — that is, an increase in synaptic strength is confined to spines in which LTP has been induced. Because spines are present at high density, how does myosin Vb deliver its cargo to where it is needed? The protein could use local Ca2+ gradients, specific to active spines, to adjust and direct its motility. But LTP is not alone in its requirement for Ca2+ influx through NMDA receptors. Another form of lasting change called long-term depression (LTD), which weakens synaptic strength by removing AMPA receptors, also requires NMDA-receptor activation and is input specific. So myosin Vb may be tuned to preferentially detect LTP-associated Ca2+ dynamics in specific spines and/or local dendrites, and work in concert with other signals associated only with LTP or LTD. For instance, spines undergoing LTD could generate signals that block the entry of recycling endosomes. A detailed look at how Ca2+ dynamics at specific spines controls the local distribution of myosin V and its GluR1-containing cargo would be valuable.
Identification of an actin motor protein as the delivery vehicle for membranes and AMPA receptors to spines during LTP highlights the significance of actin polymerization in spine growth. Actin is highly enriched in spines, in which at least two distinct pools of it are found: stable actin filaments within the spine core and a dynamic pool at the spine 'head', where it could organize glutamate receptors by interacting with other scaffold proteins9. LTP-associated spine growth — often an enlargement of the spine head — probably involves polymerizing a new set of actin filaments9.
How does myosin Vb cope with different pools of actin? When entering spines, this protein could carry recycling endosomes along the stable actin at the core. On reaching the head region, actin polymerization itself could help to propel the recycling endosomes (or part of them) towards the spine periphery by elongating filaments behind the cargo. This would ensure that membrane insertion is synchronized with the growth of the actin cytoskeleton, thus preventing the inserted membranes from collapsing. Alternatively, a distinct motor protein or trafficking machine could catalyse the final cell-surface delivery of membranes, with additional mechanisms in place for coupling membrane insertion and actin polymerization.
The abundance of GluR1 at the spine surface must also be tightly controlled with respect to spine growth10,11. Curiously, Wang et al.1 find that, on induction of LTP, the extent of GluR1 insertion and that of bulk endosomal membrane do not always go hand in hand. Compared with bulk endosomal traffic, therefore, GluR1 may be sorted into a subpopulation of recycling endosomes that take a different course to reach the cell surface. Alternatively, the extent of GluR1 accumulation on the cell surface could be restricted.
Other questions also remain. How do myosin V proteins select their cargo? What are the specific functions of their three isoforms in AMPA-receptor and membrane trafficking, and are their activities coupled to the membrane-fusion machinery? In addition, myosin V may have other LTP-related functions at spines. For example, that myosin Va is involved in localizing messenger RNA to spines12 suggests its involvement in translation-dependent forms of LTP. It is likely that synapses will remain a gold mine for revealing the molecular repertoire of memory mechanisms for years to come.