Essays on Science and SocietyEPPENDORF 2003 PRIZE-WINNING ESSAY

Ubiquitin and the Deconstruction of Synapses

See allHide authors and affiliations

Science  31 Oct 2003:
Vol. 302, Issue 5646, pp. 800-801
DOI: 10.1126/science.1092546

The remodeling of synapses is a fundamental mechanism for information storage and processing in the brain (1, 2). Much of this remodeling occurs at the postsynaptic density (PSD), a specialized biochemical apparatus containing neurotransmitter receptors and associated scaffolding proteins that organizes signal transduction pathways at the postsynaptic membrane (35). Easily visible in the electron microscope, the PSD is a disk-like structure ∼30 nm thick and a few hundred nanometers wide, positioned opposite presynaptic terminals (Fig. 1).

Fig. 1.

Electron micrograph of an excitatory synapse showing the PSD (red arrow). [Adapted from figure 1 in (3)]

During synapse maturation and in response to synaptic activity, the PSD undergoes remarkable structural changes (2), including growth (6), complexification, and perforation (7, 8). Such structural plasticity provides a physical basis for enduring changes in neural circuits and is thought to be mediated by alterations in the molecular composition of the PSD (2, 5). Indeed, stabilization or removal of neurotransmitter receptors and signaling proteins from the PSD has been proposed to account for long-term changes in synaptic strength (2, 914). Nevertheless, with few exceptions, the specific (and likely numerous) molecular changes that occur in the PSD in response to synaptic activity remain unknown. More important, understanding the patterns of molecular changes in large sets of PSD proteins, which ultimately encode the history of activation at the synapse, represents a level of analysis not previously attempted. It has been our goal to obtain a more complete understanding of the molecular underpinnings of neural circuit modification through a multiprotein analysis of activity-dependent changes in the PSD.

To examine activity-dependent changes in the PSD, we used quantitative protein profiling approaches, together with a modified procedure for isolating PSDs from cultured cortical neurons (15, 16). This experimental system allowed for strong and reproducible manipulation of activity levels across a relatively homogeneous population of synapses using pharmacological tools. We found that changes in PSD composition are bidirectional, saturable, reversible, and involve multiple classes of PSD proteins (16). Surprisingly, the time course and magnitude of bidirectional change in PSD composition were remarkably similar across all protein classes (Fig. 2). The striking resemblance among the patterns of protein accumulation and loss from the PSD suggests that large sets of postsynaptic proteins exist as functional or physical ensembles, and it implies that control over PSD composition may be governed by the incorporation or removal of certain key “master organizing molecules” that preserve stoichiometric relationships between PSD proteins.

Fig. 2.

Postsynaptic protein ensembles coregulated by activity. Quantitative analysis of selected PSD proteins at various times after adding and then removing TTX to block neuronal activity (left) or bicuculline to enhance excitatory synaptic activity (right) is shown. The gray bars indicate the duration of drug treatment. Data represent means ± SEM of band intensities normalized to control values from untreated control neuron PSDs (*P < 0.05 for all pairwise comparisons between red and green proteins; **P < 0.01 for all pairwise comparisons between red, green, and black proteins; t test; n = 3 to 6 for each time point). This segregation of proteins into stoichiometrically preserved ensembles requires the ubiquitin-proteasome system. See (16) for details.

The long-lasting changes in the molecular content of synapses we observed could arise by two general mechanisms: the incorporation of new proteins or the selective removal of existing synaptic proteins. For much of the past two decades, the prevailing model for enduring changes in synapse function and structure has been stimulus-dependent gene expression and protein synthesis (17, 18). Indeed, substantial evidence indicates that transcriptional events are critical for long-term activity-dependent plasticity (17, 19). In addition, several studies support a role for local translation of dendritic mRNAs in orchestrating long-lasting forms of learning-related synaptic plasticity (18). On the other hand, considerably less attention has been given to the contribution of protein turnover to long-term structural and functional changes at synapses.

To examine the effect of activity on dynamic PSD turnover, we measured the half-life or turnover rate of synaptic proteins using metabolic labeling pulse-chase analysis. Remarkably, we found a robust ongoing turnover of total PSD protein in control neurons that was accelerated by neuronal activity and slowed in inactive cultures (16). Much to our surprise, the turnover rate of total PSD protein occurred on the order of only a few hours, suggesting that the entire complement of synaptic proteins in mature neural circuits are replaced multiple times a day!

What causes this robust and regulated turnover of synaptic proteins? Using further biochemical approaches, we demonstrated that ubiquitin conjugation and proteasome-mediated degradation are the primary mechanisms for activity-dependent remodeling of the PSD. Specifically, we found that activity regulates de novo ubiquitin conjugation and turnover of postsynaptic proteins generally, and ubiquitination of certain scaffolding molecules specifically (16). Moreover, activity-dependent changes in PSD composition were completely dependent on the degradation of ubiquitinated proteins by the proteasome. Our findings therefore indicate that activity-dependent protein turnover joins stimulus-dependent gene expression and protein synthesis (17, 18) in engineering durable changes in the cellular machinery that governs synaptic morphology and function.

The modular multiprotein nature of these durable changes may be due to removal or turnover of master organizing molecules. Consistent with this notion, we have shown that the postsynaptic scaffolds Shank, GKAP, and AKAP79/150 undergo selective activity-dependent ubiquitination (16). Shank and GKAP are multivalent adaptors that bind to each other and to numerous additional proteins in the PSD (5, 20). In addition, AKAP79/150 anchors PKA and PP2B to complexes containing AMPA receptors or NMDA receptors and PSD-95 family members (21). The ubiquitin-dependent removal of one or more of these scaffolds could provide a mechanism for selectively destabilizing numerous associated proteins in the PSD complex, thereby accounting for the clusters of proteins coregulated by activity.

The PSD is, in essence, a proteinaceous signal-processing machine, with scaffolds, receptors, and enzymes comprising the various gears (3). Our findings indicate that the molecular components of this machine accumulate or disperse in reproducible patterns influenced by activity level and ubiquitination, raising the possibility that the signaling properties of the PSD machine are similarly plastic. To address this question, we examined the effect of alterations in activity on downstream signaling to CREB and ERK-MAPK: two prominent signal transduction cascades organized by proteins in the PSD and involved in synaptic plasticity (4, 19, 22). Our results indicated that NMDA receptors at active synapses elicit augmented activation of CREB, whereas their counterparts at inactive synapses are selectively coupled to ERK-MAPK (16). This reciprocal regulation of CREB and ERK-MAPK pathways provides clear evidence that activity-dependent reorganization of the postsynaptic apparatus regulates multiple facets of synaptic signaling.

Thus, far from being an immutable structure, the PSD contains hidden dimensions of interconnected protein networks within which reside the molecular traces of experience. By demonstrating that activity controls the global composition of the synapse through ubiquitin-dependent turnover, our research provides a new conceptual framework for understanding and ultimately predicting functional changes in neural circuits.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.

Navigate This Article