Phase separation-mediated actin bundling by the postsynaptic density condensates

  1. Xudong Chen
  2. Bowen Jia
  3. Shihan Zhu
  4. Mingjie Zhang

  • Curated by eLife

    eLife logo

    eLife assessment

    This manuscript presents an exciting set of experiments on the mechanisms through which PSD proteins induce actin bundle formation. The work included deep mechanistic analyses which determine the necessity of upper vs. lower levels of PSD proteins for actin bundle formation, identify the domains and interactions of these proteins that are necessary and sufficient to induce actin bundles, and provide a first assessment in neurons of potential roles of the newly discovered mechanisms.

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

The volume and the electric strength of an excitatory synapse is near linearly correlated with the area of its postsynaptic density (PSD). Extensive research in the past has revealed that the PSD assembly directly communicates with actin cytoskeleton in the spine to coordinate activity-induced spine volume enlargement as well as long-term stable spine structure maintenance. However, the molecular mechanism underlying the communication between the PSD assembly and spine actin cytoskeleton is poorly understood. In this study, we discover that in vitro reconstituted PSD condensates can promote actin polymerization and F-actin bundling without help of any actin regulatory proteins. The Homer scaffold protein within the PSD condensates and a positively charged actin-binding surface of the Homer EVH1 domain are essential for the PSD condensate-induced actin bundle formation in vitro and for spine growth in neurons. Homer-induced actin bundling can only occur when Homer forms condensate with other PSD scaffold proteins such as Shank and SAPAP. The PSD-induced actin bundle formation is sensitively regulated by CaMKII or by the product of the immediate early gene Homer1a . Thus, the communication between PSD and spine cytoskeleton may be modulated by targeting the phase separation of the PSD condensates.

Article activity feed

  1. Version published to 10.7554/elife.84446 on eLife
  2. eLife

    eLife assessment

    This manuscript presents an exciting set of experiments on the mechanisms through which PSD proteins induce actin bundle formation. The work included deep mechanistic analyses which determine the necessity of upper vs. lower levels of PSD proteins for actin bundle formation, identify the domains and interactions of these proteins that are necessary and sufficient to induce actin bundles, and provide a first assessment in neurons of potential roles of the newly discovered mechanisms.

  3. eLife

    Reviewer #1 (Public Review):

    This manuscript presents an exciting set of experiments on the mechanisms through which PSD proteins induce actin bundle formation. The study builds on a previous observation from the Zhang laboratory that phase condensates of six PSD proteins lead to the formation of actin bundles. Here, deep mechanistic analyses determine the necessity of upper vs. lower level PSD proteins for actin bundle formation, identify the domains and interactions of these proteins that are necessary and sufficient to induce actin bundles, and provide a first assessment in neurons of potential roles of the newly discovered mechanisms. The authors find that a patch of arginines in the Homer EVH1 domain plays a central role. Strikingly, no adaptors are needed for PSD condensates to induce actin bundles. This work is important for the understanding of roles and mechanisms of interactions between postsynaptic receptor scaffolds and cytoskeletal elements in dendritic spines. The mechanisms that are uncovered are likely mediators of structural and functional synaptic plasticity.

    Overall, the data are rigorously acquired and convincing, the presentation of the findings is logical and clear, and the manuscript is well-written. In my view, a few adjustments in data presentation (quantitative assessment of in vitro experiments, statistical analyses) and additional analyses of existing data (on the localization and roles of transfected Homer proteins in neurons) will improve the paper, but new experiments are not necessary.

  4. eLife

    Reviewer #2 (Public Review):

    In the manuscript, Chen and colleagues reconstituted the minimal system that indicates the coupling of PSD condensates with actin polymerization. While the functional connection between the assembly and dynamics of PSD and actin was known, the molecular mechanism remained elusive. Using a series of elegant biochemical reconstitutions and in-vitro assays complemented with analysis in living cells and primary neurons, the authors characterized whether PSD condensates of Homer-1, Shank-3 and SAPAP/GKAP are sufficient to induce F-actin bundling. Furthermore, they dissected the positively-charged Arg patch within EVH1 domain of Homer to be crucial for the F-actin bundling. Postsynaptic CaMKII and a short isoform of Homer, Homer1a, can both attenuate this process, suggesting various mechanisms neurons can regulate this process. Overall, the topic is timely, the study is well-designed, and the assays are clearly executed. However, several aspects need to be experimentally addressed, including some important controls:

    1. It is well established that molecular crowding plays a crucial role in F-actin bundling. For example, in the reconstitution assays in Fig.1, the authors use 10 µM of each component of PSD (total of 60 µM), to which 5 µM actin is added. Yet, in their control assays (Supp. Fig. 1), only 10 µM of each protein was checked with the same amount of actin. A control is missing where the total protein crowding would be preserved, for example, by adding BSA or protein to mimic non-specific protein crowding.
    2. Is the F-bunding observed under these physiological ratios of PSD proteins and actin? For instance, a recent quantitative study (PMID: 34168338) suggests actin:Homer-1 is 200:1 or 100:1, which is in stark difference from the 1:2 molar ratio used in the study. The protein concentrations (molar ratios) need to match the physiological.
    3. In the cell migration assays, it is somewhat unclear to what extent the interaction is direct. For instance, co-sedimentation at ultra-speed (100,000 g) was used to suggest a direct binding of EVH1-GNC4 fusions (Homer1, Enah) with F-actin. The control that needs to be included is a protein known not to bind to F-actin incubated under the same conditions (salt concentration, duration of incubation) and spun down at 100,000xg. This is important to exclude that the tested proteins non-specifically entangle into F-actin without specifically binding to it, particularly at such high speed.
    4. The imaging assay in hippocampal neurons uses an increased spine head size as a proxy for F-actin bundling. However, one needs to be careful as the baseline includes soluble mCherry, which is both much smaller in size and does not specifically enrich the spines. The image of Homer 1 R3E shows overall lower localization at the spines. Thus, one cannot exclude that the spine enlargement upon overexpression of Homer 1 wt and R3E+EN is not primarily driven by their overall enrichment in the PSD phase. A suitable control for this assay would be mCherry-tagged PSD95, which would localize to the spines yet is not directly involved in F-actin bundling.

  5. Version published to 10.1101/2022.12.06.519305v1 on bioRxiv