- Email: [email protected]
The Flip Side of Synapse Elimination
Neuron, Vol. 37, 1–6, January 9, 2003, Copyright 2003 by Cell Press
Previews
The Flip Side of Synapse Elimination The fine tuning of synaptic circuits often requires an activity-dependent phase in which appropriate connections are strengthened and inappropriate connections are eliminated. In this issue of Neuron, Walsh and Lichtman propose a novel “synaptic takeover” mechanism for synapse elimination at the vertebrate NMJ, where withdrawal of one axon is accompanied by expansion of a competing axon into the newly vacated territory. One tremendous advantage of the advent of green fluorescent protein (GFP) as a marker in living cells has been the ability to observe the behavior of cells and proteins in living organisms, often providing dramatic new insights into their in vivo functions. In this issue of Neuron, Walsh and Lichtman studied synapse elimination at the vertebrate neuromuscular junction using transgenic mice expressing spectral variants of GFP in different subsets of motor neurons (Walsh and Lichtman, 2003). Through time-lapse imaging of two axons innervating the same muscle fiber, they were able to observe the competition between these axons for synaptic space in vivo. The authors propose a novel mechanism for synapse elimination—“synaptic takeover”—in which the advancing axon invades the area recently vacated by the retreating axon. Surprisingly, the axon initially with the largest area did not always have the competitive advantage, and the process was very dynamic and unpredictable, with axons occasionally flipping from advance to retreat. During embryogenesis, skeletal muscle fibers are initially innervated by multiple motor axons, sometimes up to five motor neurons per fiber. Over the first several weeks after birth, most of the connections are withdrawn, until only one axon remains per fiber. The battle between axons for survival is thought to be a competitive process and to rely upon differential activity patterns (Lichtman and Colman, 2000). In 1994, inspired by ideas from an elegant set of experiments probing the role of synaptic transmission in synapse elimination (BaliceGordon and Lichtman, 1994), Jennings drew a schematic model that nicely summarized the current state of the field (Jennings, 1994; see Figure, panel A). In this model, synaptic transmission produces two signals, a local, “protective” signal and a longer reaching, “elimination” signal. An inactive synapse does not produce these signals, and thus, in the absence of protection, is susceptible to an elimination signal from a neighboring active axon. The elimination signal was thought to decrease the density of postsynaptic acetylcholine receptors (AChRs) under the inactive axon, which was then followed by axon withdrawal. To a large extent, this model still holds true today; however, the experiments by Walsh and Lichtman suggest a fundamental modification. The authors utilized doubly transgenic mice that expressed either CFP or
YFP in all motor axons and GFP or CFP in a subset of motor axons. Multiply innervated, two-color junctions were imaged beginning at the second postnatal week for various time-lapse intervals of 24 hr to a few days. In addition, AChRs were labeled with a red fluorescent ␣-bungarotoxin, enabling simultaneous observation of postsynaptic dynamics. Contrary to the accepted view that synapse elimination is preceded or at least accompanied by a loss of the underlying AChRs, in the majority of cases, the authors observed no loss of AChR areas. Remarkably, instead what they observed was that the postsynaptic area vacated by the withdrawing axon was invaded by the expanding axon, in a process they dubbed “synaptic takeover” (see Figure, panel B). What is the mechanism of synaptic takeover? Does the more powerful axon actively push the weaker axon out of the way? Or does the weaker axon retreat first? Because they observe in some cases withdraw of an axon without subsequent takeover, Walsh and Lichtman argue that the first case is unlikely. Instead, they propose that the site vacated by the withdrawing axon provides an opportunity for takeover by the neighboring axon. Possibly the sudden abundance of a secreted growth factor, no longer necessary to support the withdrawing axon, attracts the neighboring axon. Or perhaps axons are constantly sampling their surrounding environment for vacant postsynaptic territory on which to expand. The latter proposal suggests that axons should intrinsically be highly dynamic. Examining the motility of axons in the absence of competition at singly innervated junctions, Walsh and Lichtman found that indeed axons in the second postnatal week were all highly dynamic, extending and retracting over 24 hr intervals, in contrast to 2 weeks later, when axons were stable. It will be exciting to see a quantitative examination of the dynamics at higher temporal resolution. Such observations are reminiscent of experiments at central synapses in the Xenopus tectum in which the dynamic behavior of axons imaged in vivo is proposed to contribute to the formation of a topographically ordered retinotectal projection (O’Rourke et al., 1994). What determines which axon wins and which loses? One simple prediction would be that the axon with greater synaptic strength should have a competitive advantage. As synapse strength often correlates with synapse size at the NMJ (Costanzo et al., 1999), the expected outcome would be that the axon initially occupying greater area would win. Surprisingly, in one quarter of the cases examined, Walsh and Lichtman found the input that was ultimately maintained occupied less than 30% of the terminal area at some earlier time point. In addition, they sometimes observed a “flip-flop” in which axons that were initially retreating began advancing, or vice versa. These results are not consistent with the idea that relative synaptic strength is the sole determinant of the outcome. More complicated models suggest that it could be the pattern of synaptic activity that may be important. Experiments which combine electrophysiology and dual color imaging will be important to answer these questions, both through monitoring
Neuron 2
Synapse Elimination at the Vertebrate NMJ (A) Synaptic transmission produces two signals in the postsynaptic muscle, a protective signal (blue cloud) and an elimination signal (red arrow). Synapses with relatively low activity (middle panel, left axon) are no longer protected from the elimination signal of a neighboring active axon, resulting in disappearance of postsynaptic AChRs and withdrawal of the axon. Modified from Jennings, 1994. (B) As in (A), synaptic transmission produces two signals in the postsynaptic muscle, a protective signal (blue cloud) and an elimination signal (red arrow). Protrusions represent intrinsic dynamic behavior of axons, whereas postsynaptic AChR areas are mostly stable. Withdrawal of one synapse provides an opportunity for invasion by a competing synapse, which expands into the space vacated by the retreating axon in a process Walsh and Lichtman term “synaptic takeover.”
what happens during normal synapse elimination, and also through examining the outcomes of controlled stimulation paradigms. In addition, the implementation of genetic regimes for reducing or silencing transmission in subsets of motor neurons would provide insight into the role of activity in competition (Zemelman and Miesenbock, 2001). The process of synaptic takeover requires the rapid redistribution of presynaptic machinery, removed from withdrawing nerve terminals and inserted into invading ones. The withdrawing axon is capped by a retraction bulb, which is likely packed full of presynaptic materials. Do these materials get recycled and redistributed to other axonal branches? Keller-Peck et al. demonstrated that the terminals of a single motor neuron are all at different stages of synapse elimination, implying that the process is under local control at the site on the muscle fiber (Keller-Peck et al., 2001). However, could retracting axon branches influence the outcome of other branches by supplying them with many of the materials necessary for rapid synapse assembly? From observations of GFP-tagged synaptic vesicle proteins, there is evidence that presynaptic proteins can be transported together in packets to nascent synapses (Ahmari et al., 2000). One theory is that such a process could be responsible for some of the unpredictable “flip-flop” behavior observed by Walsh and Lichtman or for tipping the balance in favor of the weaker axon. But why does an axon decide to retreat? What are the molecular mechanisms that initiate withdrawal? Returning to the model (see Figure), the molecular identities of the “protective” signal and the “elimination” signal are still a mystery. There has been some speculation of roles for growth factors and proteases but no substantial
evidence as of yet (Sanes and Lichtman, 1999). Many of the molecules and signaling mechanisms responsible for the initial steps of synapse formation are known (Buonanno and Fischbach, 2001; Burden, 2002). Could some of these molecules play a role in synaptic competition and synapse elimination? Testing roles for candidate molecules will require the use of spatially restricted dominant negatives or knockouts in a subset of motor neurons, and it will be necessary also to have precise temporal control in order to examine effects of candidate genes only after synapses have properly formed. Now the question is can we still call this synapse elimination? The synapse, in fact, is essentially still there after axon withdrawal, albeit with a different presynaptic partner. Perhaps “synaptic takeover” will not only describe a novel mechanism for synaptic refinement, but will refer to a new field of study. Karen Zito Cold Spring Harbor Laboratory Cold Spring Harbor, New York 11724 Selected Reading Ahmari, S.E., Buchanan, J., and Smith, S.J. (2000). Nat. Neurosci. 3, 445–451. Balice-Gordon, R.J., and Lichtman, J.W. (1994). Nature 372, 519–524. Buonanno, A., and Fischbach, G.D. (2001). Curr. Opin. Neurobiol. 11, 287–296. Burden, S.J. (2002). J. Neurobiol. 53, 501–511. Costanzo, E.M., Barry, J.A., and Ribchester, R.R. (1999). J. Physiol. 521, 365–374. Jennings, C. (1994). Nature 372, 498–499. Keller-Peck, C.R., Walsh, M.K., Gan, W.-B., Feng, G., Sanes, J., and Lichtman, J.W. (2001). Neuron 31, 381–394. Lichtman, J.W., and Colman, H. (2000). Neuron 25, 269–278. O’Rourke, N.A., Cline, H.T., and Fraser, S.E. (1994). Neuron 12, 921–934. Sanes, J.R., and Lichtman, J.W. (1999). Annu. Rev. Neurosci. 22, 389–442. Walsh, M.K., and Lichtman, J.W. (2003). Neuron 37, this issue, 67–73. Zemelman, B.V., and Miesenbock, G. (2001). Curr. Opin. Neurobiol. 11, 409–414.
Current Compensation in Neuronal Homeostasis How do neurons maintain stable intrinsic properties over long periods of time as the channels that govern excitability turn over in the membrane? In this issue of Neuron, MacLean et al. argue that homeostatic regulation of intrinsic activity can occur by an activityindependent mechanism. Human neurons live many decades, yet the ion channels and receptors that give neurons their distinctive electrical properties turn over in the membrane in hours, days, or weeks. What processes govern the number and distri-
Previews
The Flip Side of Synapse Elimination The fine tuning of synaptic circuits often requires an activity-dependent phase in which appropriate connections are strengthened and inappropriate connections are eliminated. In this issue of Neuron, Walsh and Lichtman propose a novel “synaptic takeover” mechanism for synapse elimination at the vertebrate NMJ, where withdrawal of one axon is accompanied by expansion of a competing axon into the newly vacated territory. One tremendous advantage of the advent of green fluorescent protein (GFP) as a marker in living cells has been the ability to observe the behavior of cells and proteins in living organisms, often providing dramatic new insights into their in vivo functions. In this issue of Neuron, Walsh and Lichtman studied synapse elimination at the vertebrate neuromuscular junction using transgenic mice expressing spectral variants of GFP in different subsets of motor neurons (Walsh and Lichtman, 2003). Through time-lapse imaging of two axons innervating the same muscle fiber, they were able to observe the competition between these axons for synaptic space in vivo. The authors propose a novel mechanism for synapse elimination—“synaptic takeover”—in which the advancing axon invades the area recently vacated by the retreating axon. Surprisingly, the axon initially with the largest area did not always have the competitive advantage, and the process was very dynamic and unpredictable, with axons occasionally flipping from advance to retreat. During embryogenesis, skeletal muscle fibers are initially innervated by multiple motor axons, sometimes up to five motor neurons per fiber. Over the first several weeks after birth, most of the connections are withdrawn, until only one axon remains per fiber. The battle between axons for survival is thought to be a competitive process and to rely upon differential activity patterns (Lichtman and Colman, 2000). In 1994, inspired by ideas from an elegant set of experiments probing the role of synaptic transmission in synapse elimination (BaliceGordon and Lichtman, 1994), Jennings drew a schematic model that nicely summarized the current state of the field (Jennings, 1994; see Figure, panel A). In this model, synaptic transmission produces two signals, a local, “protective” signal and a longer reaching, “elimination” signal. An inactive synapse does not produce these signals, and thus, in the absence of protection, is susceptible to an elimination signal from a neighboring active axon. The elimination signal was thought to decrease the density of postsynaptic acetylcholine receptors (AChRs) under the inactive axon, which was then followed by axon withdrawal. To a large extent, this model still holds true today; however, the experiments by Walsh and Lichtman suggest a fundamental modification. The authors utilized doubly transgenic mice that expressed either CFP or
YFP in all motor axons and GFP or CFP in a subset of motor axons. Multiply innervated, two-color junctions were imaged beginning at the second postnatal week for various time-lapse intervals of 24 hr to a few days. In addition, AChRs were labeled with a red fluorescent ␣-bungarotoxin, enabling simultaneous observation of postsynaptic dynamics. Contrary to the accepted view that synapse elimination is preceded or at least accompanied by a loss of the underlying AChRs, in the majority of cases, the authors observed no loss of AChR areas. Remarkably, instead what they observed was that the postsynaptic area vacated by the withdrawing axon was invaded by the expanding axon, in a process they dubbed “synaptic takeover” (see Figure, panel B). What is the mechanism of synaptic takeover? Does the more powerful axon actively push the weaker axon out of the way? Or does the weaker axon retreat first? Because they observe in some cases withdraw of an axon without subsequent takeover, Walsh and Lichtman argue that the first case is unlikely. Instead, they propose that the site vacated by the withdrawing axon provides an opportunity for takeover by the neighboring axon. Possibly the sudden abundance of a secreted growth factor, no longer necessary to support the withdrawing axon, attracts the neighboring axon. Or perhaps axons are constantly sampling their surrounding environment for vacant postsynaptic territory on which to expand. The latter proposal suggests that axons should intrinsically be highly dynamic. Examining the motility of axons in the absence of competition at singly innervated junctions, Walsh and Lichtman found that indeed axons in the second postnatal week were all highly dynamic, extending and retracting over 24 hr intervals, in contrast to 2 weeks later, when axons were stable. It will be exciting to see a quantitative examination of the dynamics at higher temporal resolution. Such observations are reminiscent of experiments at central synapses in the Xenopus tectum in which the dynamic behavior of axons imaged in vivo is proposed to contribute to the formation of a topographically ordered retinotectal projection (O’Rourke et al., 1994). What determines which axon wins and which loses? One simple prediction would be that the axon with greater synaptic strength should have a competitive advantage. As synapse strength often correlates with synapse size at the NMJ (Costanzo et al., 1999), the expected outcome would be that the axon initially occupying greater area would win. Surprisingly, in one quarter of the cases examined, Walsh and Lichtman found the input that was ultimately maintained occupied less than 30% of the terminal area at some earlier time point. In addition, they sometimes observed a “flip-flop” in which axons that were initially retreating began advancing, or vice versa. These results are not consistent with the idea that relative synaptic strength is the sole determinant of the outcome. More complicated models suggest that it could be the pattern of synaptic activity that may be important. Experiments which combine electrophysiology and dual color imaging will be important to answer these questions, both through monitoring
Neuron 2
Synapse Elimination at the Vertebrate NMJ (A) Synaptic transmission produces two signals in the postsynaptic muscle, a protective signal (blue cloud) and an elimination signal (red arrow). Synapses with relatively low activity (middle panel, left axon) are no longer protected from the elimination signal of a neighboring active axon, resulting in disappearance of postsynaptic AChRs and withdrawal of the axon. Modified from Jennings, 1994. (B) As in (A), synaptic transmission produces two signals in the postsynaptic muscle, a protective signal (blue cloud) and an elimination signal (red arrow). Protrusions represent intrinsic dynamic behavior of axons, whereas postsynaptic AChR areas are mostly stable. Withdrawal of one synapse provides an opportunity for invasion by a competing synapse, which expands into the space vacated by the retreating axon in a process Walsh and Lichtman term “synaptic takeover.”
what happens during normal synapse elimination, and also through examining the outcomes of controlled stimulation paradigms. In addition, the implementation of genetic regimes for reducing or silencing transmission in subsets of motor neurons would provide insight into the role of activity in competition (Zemelman and Miesenbock, 2001). The process of synaptic takeover requires the rapid redistribution of presynaptic machinery, removed from withdrawing nerve terminals and inserted into invading ones. The withdrawing axon is capped by a retraction bulb, which is likely packed full of presynaptic materials. Do these materials get recycled and redistributed to other axonal branches? Keller-Peck et al. demonstrated that the terminals of a single motor neuron are all at different stages of synapse elimination, implying that the process is under local control at the site on the muscle fiber (Keller-Peck et al., 2001). However, could retracting axon branches influence the outcome of other branches by supplying them with many of the materials necessary for rapid synapse assembly? From observations of GFP-tagged synaptic vesicle proteins, there is evidence that presynaptic proteins can be transported together in packets to nascent synapses (Ahmari et al., 2000). One theory is that such a process could be responsible for some of the unpredictable “flip-flop” behavior observed by Walsh and Lichtman or for tipping the balance in favor of the weaker axon. But why does an axon decide to retreat? What are the molecular mechanisms that initiate withdrawal? Returning to the model (see Figure), the molecular identities of the “protective” signal and the “elimination” signal are still a mystery. There has been some speculation of roles for growth factors and proteases but no substantial
evidence as of yet (Sanes and Lichtman, 1999). Many of the molecules and signaling mechanisms responsible for the initial steps of synapse formation are known (Buonanno and Fischbach, 2001; Burden, 2002). Could some of these molecules play a role in synaptic competition and synapse elimination? Testing roles for candidate molecules will require the use of spatially restricted dominant negatives or knockouts in a subset of motor neurons, and it will be necessary also to have precise temporal control in order to examine effects of candidate genes only after synapses have properly formed. Now the question is can we still call this synapse elimination? The synapse, in fact, is essentially still there after axon withdrawal, albeit with a different presynaptic partner. Perhaps “synaptic takeover” will not only describe a novel mechanism for synaptic refinement, but will refer to a new field of study. Karen Zito Cold Spring Harbor Laboratory Cold Spring Harbor, New York 11724 Selected Reading Ahmari, S.E., Buchanan, J., and Smith, S.J. (2000). Nat. Neurosci. 3, 445–451. Balice-Gordon, R.J., and Lichtman, J.W. (1994). Nature 372, 519–524. Buonanno, A., and Fischbach, G.D. (2001). Curr. Opin. Neurobiol. 11, 287–296. Burden, S.J. (2002). J. Neurobiol. 53, 501–511. Costanzo, E.M., Barry, J.A., and Ribchester, R.R. (1999). J. Physiol. 521, 365–374. Jennings, C. (1994). Nature 372, 498–499. Keller-Peck, C.R., Walsh, M.K., Gan, W.-B., Feng, G., Sanes, J., and Lichtman, J.W. (2001). Neuron 31, 381–394. Lichtman, J.W., and Colman, H. (2000). Neuron 25, 269–278. O’Rourke, N.A., Cline, H.T., and Fraser, S.E. (1994). Neuron 12, 921–934. Sanes, J.R., and Lichtman, J.W. (1999). Annu. Rev. Neurosci. 22, 389–442. Walsh, M.K., and Lichtman, J.W. (2003). Neuron 37, this issue, 67–73. Zemelman, B.V., and Miesenbock, G. (2001). Curr. Opin. Neurobiol. 11, 409–414.
Current Compensation in Neuronal Homeostasis How do neurons maintain stable intrinsic properties over long periods of time as the channels that govern excitability turn over in the membrane? In this issue of Neuron, MacLean et al. argue that homeostatic regulation of intrinsic activity can occur by an activityindependent mechanism. Human neurons live many decades, yet the ion channels and receptors that give neurons their distinctive electrical properties turn over in the membrane in hours, days, or weeks. What processes govern the number and distri-