1. Developmental Biology
  2. Neuroscience
Download icon

Synapse maintenance is impacted by ATAT-2 tubulin acetyltransferase activity and the RPM-1 signaling hub

  1. Melissa A Borgen
  2. Andrew C Giles
  3. Dandan Wang
  4. Brock Grill  Is a corresponding author
  1. The Scripps Research Institute, United States
Research Article
  • Cited 0
  • Views 610
  • Annotations
Cite this article as: eLife 2019;8:e44040 doi: 10.7554/eLife.44040

Abstract

Synapse formation is comprised of target cell recognition, synapse assembly, and synapse maintenance. Maintaining established synaptic connections is essential for generating functional circuitry and synapse instability is a hallmark of neurodegenerative disease. While many molecules impact synapse formation generally, we know little about molecules that affect synapse maintenance in vivo. Using genetics and developmental time course analysis in C.elegans, we show that the α-tubulin acetyltransferase ATAT-2 and the signaling hub RPM-1 are required presynaptically to maintain stable synapses. Importantly, the enzymatic acetyltransferase activity of ATAT-2 is required for synapse maintenance. Our analysis revealed that RPM-1 is a hub in a genetic network composed of ATAT-2, PTRN-1 and DLK-1. In this network, ATAT-2 functions independent of the DLK-1 MAPK and likely acts downstream of RPM-1. Thus, our study reveals an important role for tubulin acetyltransferase activity in presynaptic maintenance, which occurs via the RPM-1/ATAT-2 pathway.

https://doi.org/10.7554/eLife.44040.001

Introduction

Synapse formation is comprised of several steps including target recognition, synapse assembly and synapse maintenance (Chia et al., 2013; Jin and Garner, 2008). Synapse maintenance, also referred to as synapse stability, is required to complete the synapse formation process, and is also important for maintaining circuitry and allowing plasticity throughout an animal’s life (Lin and Koleske, 2010). Indeed, increasing evidence indicates synapse instability is a hallmark of many neurodegenerative diseases, including Alzheimer’s disease (Lin and Koleske, 2010; Selkoe, 2002; Spires-Jones and Hyman, 2014). Understanding how synapse maintenance influences nervous system development, plasticity and disease will require far greater knowledge of the molecules and signaling networks that regulate this process.

Previous genetic studies have been invaluable for informing our understanding of how synapse maintenance is regulated in vivo. At the fly neuromuscular junction (NMJ), genes encoding regulators of the microtubule cytoskeleton, such as Dynactin and Ankyrin, are crucial for maintaining the presynaptic terminal (Eaton et al., 2002; Pielage et al., 2008). Spectrin, a scaffold that links cell adhesion with the microtubule cytoskeleton, is also important for NMJ synapse maintenance (Massaro et al., 2009; Pielage et al., 2005).

In C. elegans, mechanosensory neurons that form glutamatergic neuron-neuron synapses (reminiscent of mammalian central synapses) have proven particularly valuable for understanding the molecular and genetic underpinnings of synapse maintenance. For instance, pharmacological and genetic perturbation of microtubules impairs presynaptic bouton maintenance in these cells (Chen et al., 2014). Genetic screens using mechanosensory neurons revealed that the microtubule minus-end binding protein PTRN-1/CAMSAP and the actin binding protein ZYX-1 are required for synapse maintenance (Luo et al., 2014; Marcette et al., 2014; Richardson et al., 2014). Thus, studies from both flies and worms emphasize the power genetic model systems wield in identifying molecules, and potentially unraveling entire signaling networks, that are required for synapse maintenance.

While increasing evidence has linked genetic perturbation of the microtubule cytoskeleton with synapse deterioration, it remains unknown whether mutants that affect post-translational modification of microtubules, such as acetylation, affect synapse maintenance. Two α-tubulin acetyltransferases, MEC-17 and ATAT-2, were identified in C. elegans. MEC-17 and ATAT-2 function via enzymatic acetyltransferase activity and non-enzymatic mechanisms to regulate microtubule structure, touch sensation, axon polarity and axon degeneration in mechanosensory neurons (Akella et al., 2010; Neumann and Hilliard, 2014; Shida et al., 2010; Topalidou et al., 2012). Despite this prior work, it remains unknown if α-tubulin acetyltransferases affect the synapse, and in particular synapse maintenance, in any system. The importance of addressing this question is highlighted by evidence that altered α-tubulin acetylation is associated with neurodegenerative diseases, such as Alzheimer’s and Parkinson’s disease (Godena et al., 2014; Govindarajan et al., 2013; Hempen and Brion, 1996; Pellegrini et al., 2017).

The Pam/Highwire/RPM-1 (PHR) proteins, including C. elegans RPM-1, are enormous signaling hubs that also have ubiquitin ligase activity (Grill et al., 2016). PHR proteins are important regulators of neuronal development with conserved roles in synapse formation (Bloom et al., 2007; Schaefer et al., 2000; Wan et al., 2000; Zhen et al., 2000), axon guidance (Bloom et al., 2007; Lewcock et al., 2007; Park and Rongo, 2018) and axon termination (Borgen et al., 2017b; Feoktistov and Herman, 2016; Schaefer et al., 2000). PHR proteins regulate synapse formation at NMJs (Bloom et al., 2007; Wan et al., 2000; Zhen et al., 2000) and glutamatergic neuron-neuron synapses formed by C. elegans mechanosensory neurons (Schaefer et al., 2000). At present, it is unclear whether PHR proteins impact synapse formation by regulating synapse assembly or maintenance. Furthermore, PHR protein signaling can influence microtubules in the context of axon guidance and termination (Borgen et al., 2017b; Hendricks and Jesuthasan, 2009; Lewcock et al., 2007). However, whether there is a functional genetic relationship between PHR proteins and tubulin acetyltransferases and, if so, how this influences synapse formation and maintenance remains unknown.

Here, we use developmental time course analysis, genetics and pharmacology to show that the α-tubulin acetyltransferase ATAT-2 regulates synapse maintenance in mechanosensory neurons, and does so via its enzymatic acetyltransferase activity. Moreover, ATAT-2 functions in a pathway with RPM-1 to regulate presynaptic maintenance in mechanosensory neurons, as well as behavioral habituation to repeated gentle touch. Genetic analysis indicates that RPM-1 is a hub in a network containing ATAT-2, PTRN-1 and DLK-1. Importantly, the RPM-1/ATAT-2 pathway represents a mechanism that functions independent of DLK-1 to regulate synapse maintenance. Overall, our findings not only reveal a novel, functional role for ATAT-2, but place it within an RPM-1 signaling network that is required for synapse maintenance.

Results

RPM-1 regulates presynaptic bouton maintenance during development

C. elegans has two PLM mechanosensory neurons, each of which has a single primary axon that extends a collateral branch to form chemical synapses (Figure 1A,B). Similar to neurons in the mammalian central nervous system, C. elegans mechanosensory neurons form glutamatergic, neuron-neuron connections (Chalfie et al., 1985; Lee et al., 1999). Previous work showed RPM-1, the C. elegans PHR protein, is an important regulator of PLM neuron synapse formation (Grill et al., 2016; Schaefer et al., 2000). In adult animals, the PLM neurons of rpm-1 mutants lack both the collateral branch and chemical synapses (Figure 1B). At present, it is uncertain whether this defect arises from failed synapse assembly or impaired synapse maintenance. To address this, we began by revisiting time-course analysis of PLM presynaptic bouton development in wild-type (wt) animals and rpm-1 mutants. When L1 larvae hatch, the primary PLM axon has extended, but the synaptic branch is absent (Figure 1B,C). By 6 hr post-hatch (PH), wt animals have formed morphological presynaptic boutons (Figure 1B,C). In contrast, synaptic bouton development is delayed in rpm-1 mutants at 6 hours PH. Despite this initial delay, rpm-1 mutants eventually form synaptic boutons, which are readily observable at 12 hr PH (Figure 1B,C). Between 12 and 48 hr PH, boutons destabilize and the synaptic branch retracts in rpm-1 mutants to phenotypic levels characteristic of adult rpm-1 mutants (Figure 1B,C). Transgenic rescue showed expression of RPM-1 using the native promoter or a mechanosensory neuron promoter rescued these defects (Figure 1D). These results indicate that RPM-1 primarily regulates presynaptic bouton maintenance by functioning cell autonomously in mechanosensory neurons. Notably, our results differ from a prior study that suggested synaptic branch defects in rpm-1 mutants arise primarily from impaired initial bouton formation with a relatively small frequency of bouton loss over time (Schaefer et al., 2000).

RPM-1 functions cell autonomously in mechanosensory neurons to regulate presynaptic bouton maintenance.

(a) Schematic highlighting location of collateral synaptic branch and chemical synapses in PLM mechanosensory neurons. (b) Confocal images at different developmental time points showing presynaptic boutons of PLM neurons are delayed in formation and destabilize in rpm-1 mutants. Note brackets denote PLM presynaptic boutons (PLML and PLMR) and arrows highlight synaptic branch retraction. Note that in confocal images at 16 hr and 60 hr PH, PLML synaptic branch is out of focal plane but bouton is visible. (c) Developmental time course showing synaptic boutons in PLM neurons of rpm-1 mutants are delayed in formation but reach normal levels by 16 hr PH (blue). Subsequently, rpm-1 mutant boutons are progressively lost over time (orange). (d) Quantitation showing bouton maintenance defects in adult rpm-1 mutants, and rescue with transgenic RPM-1 expressed using native rpm-1 promoter or mechanosensory neuron promoter. Significance tested using Fisher’s exact test for c, and Student’s t-test with Bonferroni correction for d. ***p<0.001, **p<0.01 and ns = not significant (p>0.05).

https://doi.org/10.7554/eLife.44040.002

RPM-1 is localized to presynaptic terminals during development

To assess RPM-1 localization during development, we transgenically expressed RPM-1::GFP in mechanosensory neurons along with tdTOMATO as a cell fill (Figure 2). RPM-1 was observed at presynaptic boutons during synapse development (Figure 2A). RPM-1 localization to presynaptic boutons occurred as early as 5 hr PH, when boutons are first consistently present, and was observed through adulthood (Figure 2B,C).

RPM-1 localizes to presynaptic terminals of developing and adult mechanosensory neurons.

(a) Confocal images showing RPM-1::GFP localized at presynaptic boutons of PLM neuron at 16 hr PH. tdTOMATO shows PLM axon and presynaptic terminal morphology. (b) Quantitation of RPM-1::GFP presynaptic localization in PLM neurons at different times in development. (c) Confocal image showing RPM-1 localized to periactive zones adjacent to active zone marker UNC-10::tdTOMATO in adults.

https://doi.org/10.7554/eLife.44040.003

To confirm RPM-1 is localized at presynaptic terminals, we used confocal microscopy to evaluate localization of RPM-1 and the active zone component UNC-10/RIM (Koushika et al., 2001). In adult animals, RPM-1 localized directly adjacent to UNC-10 at presynaptic terminals (Figure 2C). This indicates RPM-1 localizes to the periactive zone of presynaptic terminals in mechanosensory neurons. Our observation is consistent with prior studies that examined RPM-1 localization in motor neurons (Abrams et al., 2008).

Collectively, these results support several conclusions. 1) RPM-1 is localized to presynaptic terminals, which is consistent with its role in presynaptic maintenance. 2) Our observation that RPM-1 is present early in the synapse formation process and persists into adulthood suggests signaling that affects presynaptic maintenance could be initiated relatively early in the synapse formation process. 3) Localization of RPM-1 to presynaptic terminals early in development is consistent with delayed presynaptic bouton formation in rpm-1 mutants.

rpm-1 mutants assemble synapses prior to synapse destabilization

Defects in presynaptic bouton morphology and retraction of the synaptic branch suggested rpm-1 mutants have synapse maintenance defects (Figure 1). Failed synapse maintenance could arise because synapses deteriorate or because the synaptic branch successfully extends to postsynaptic target neurons, but fails to properly assemble presynaptic components.

To initially test this, we evaluated a synaptic vesicle marker, RAB-3 (Luo et al., 2014; Nonet et al., 1997). Using integrated transgenes expressing RAB-3::GFP and RFP cell fill in mechanosensory neurons, we evaluated several developmental time points. In wt animals, synaptic boutons began to form by 3 hr PH, and we observed RAB-3::GFP accumulation at presynaptic terminals even at this early time point in synapse assembly (Figure 3A). We also noted that RAB-3 accumulated at the axon point where the synaptic branch initially descends (Figure 3A). At 12 and 24 hr PH, RAB-3 was enriched at presynaptic terminals (Figure 3A). In rpm-1 mutants, we observed accumulation of RAB-3 at presynaptic terminals by 12 hours PH (Figure 3B). In contrast to wt animals, synapses destabilize in rpm-1 mutants with synaptic branches retracting by 24 hours PH and RAB-3 present only in the primary axon (Figure 3B).

Figure 3 with 1 supplement see all
RPM-1 regulates synapse maintenance.

(a) Confocal images showing synaptic vesicle marker RAB-3 (green) at presynaptic terminals of PLM mechanosensory neurons during development. RFP shows PLM morphology (magenta). Yellow brackets highlight presynaptic terminals of PLML and PLMR neurons. Note one synaptic branch is out of focal plane. (b) rpm-1 mutants accumulate RAB-3 (green) at presynaptic terminals by 12 hours PH, but presynaptic terminals fail to be maintained leading to synaptic branch retraction by 24 hr PH. (c) Developmental time course of presynaptic boutons and RAB-3::GFP accumulation in wt animals. Full assembly of presynaptic terminals with RAB-3 occurs by 12 hr PH. (d) Developmental time course of presynaptic boutons and RAB-3::GFP in rpm-1 mutants. Presynaptic assembly with RAB-3 is complete by 16 hours PH, but is not maintained and presynaptic terminals are lost at later time points. (e, f) UNC-10::tdTOMATO marks the active zone and assembles at presynaptic terminals in e) wt and (f) rpm-1 mutants at critical synapse assembly time points of 12 and 16 hr PH. (g) Quantitation of boutons containing UNC-10. At 12 hr PH, all boutons contain UNC-10 in wt and rpm-1 mutants. At 16 hr PH, there is a small defect in UNC-10 accumulation at presynaptic terminals of rpm-1 mutants. (h) Quantitation of bouton area. rpm-1 boutons are initially the same size as wt boutons (5 and 7 hours PH). rpm-1 mutants show small decreases in bouton size just prior to synapse loss (12 and 16 hr PH). Significance tested using Fisher’s exact test. ***p<0.001, *p<0.05 and ns = not significant.

https://doi.org/10.7554/eLife.44040.004

Quantitation across a wider range of developmental time points showed that in wt animals presynaptic bouton frequency is maximal at 12 hr PH, and RAB-3 accumulates in every bouton (Figure 3C). RAB-3 accumulation at presynaptic terminals was maintained across all time points examined (Figure 3C). For rpm-1 mutants, quantitation indicated presynaptic bouton formation is slightly delayed, but all PLM neurons in rpm-1 mutants have presynaptic boutons containing RAB-3 by 16 hr PH (Figure 3D). These results are consistent with rpm-1 mutants having largely normal, although slightly delayed, synapse assembly. In contrast, presynaptic maintenance is strongly impaired in rpm-1 mutants with significant, rapid loss of presynaptic boutons and RAB-3 by 21 hr PH, and further reductions by 24 hours PH (Figure 3D). Notably, presynaptic terminals that do not destabilize in rpm-1 mutants retain RAB-3 (Figure 3D).

Next, we tested two presynaptic active zone markers, UNC-10/RIM and SYD-2/Liprin. We focused on 12 and 16 hr PH, as RAB-3 analysis demonstrated these are key time points for assessing completion of synapse assembly (Figure 3A,C). In wt animals, UNC-10::tdTOMATO labeled presynaptic terminals at both 12 and 16 hr PH (Figure 3E). Likewise, presynaptic terminals of rpm-1 mutants contained UNC-10 at 12 and 16 hr PH (Figure 3F). Quantitation indicated that UNC-10 is at all presynaptic terminals in rpm-1 mutants at 12 hours PH, and the majority of terminals at 16 hours PH (Figure 3G). Similar results occurred with mScarlet::SYD-2 (Figure 3—figure supplement 1). Interestingly, a small but significant decrease in the number of terminals with UNC-10 occurred in rpm-1 mutants at 16 hr PH compared to wt animals (Figure 3G). While a subtle observation, 16 hr PH is a critical time point just prior to synapse destabilization and branch retraction in rpm-1 mutants. This observation prompted us to also assess presynaptic bouton size over development in rpm-1 mutants. Bouton size was normal in rpm-1 mutants at 5 and 7 hr PH (Figure 3H). A small, but significant, decrease in presynaptic bouton size emerged in rpm-1 mutants at 12 and 16 hr PH (Figure 3H).

These results with multiple presynaptic markers support several important points. First, our results indicate that synapses assemble in rpm-1 mutants, and while delayed, this process is largely normal in rpm-1 mutants. Second, analysis of presynaptic bouton morphology, RAB-3 and UNC-10 indicate synapses rapidly destabilize in the absence of RPM-1. Finally, we observed subtle changes in bouton size and UNC-10 accumulation just prior to synapse destabilization, which suggests these mild presynaptic changes are likely to signal the onset of failed synapse maintenance. Taken as a whole, these results indicate RPM-1 is an important regulator of synapse maintenance in mechanosensory neurons.

Loss of RPM-1 in combination with pharmacological manipulation of microtubule stability enhances synapse destabilization

We previously showed that RPM-1 signaling affects microtubule stability during growth cone collapse and axon termination (Borgen et al., 2017b). Therefore, we wanted to assess how synapse maintenance defects in rpm-1 mutants are affected by pharmacological manipulation of microtubule stability. Consistent with prior work (Chen et al., 2014; Richardson et al., 2014), treating wt animals with colchicine, a microtubule-destabilizing drug, resulted in loss of PLM synapses (Figure 4A). Treating rpm-1 mutants with colchicine significantly enhanced synapse maintenance defects (Figure 4A). Conversely, treatment of rpm-1 mutants with the microtubule-stabilizing drug taxol suppressed synapse maintenance defects (Figure 4B). These results are consistent with destabilized synapses in rpm-1 mutants resulting from less stable microtubules.

Drugs that alter microtubule stability affect synapse maintenance defects in rpm-1 mutants.

(a) Decreasing microtubule stability with colchicine enhances synapse maintenance defects in rpm-1 mutants. (b) Increasing microtubule stability with taxol suppresses synapse maintenance defects in rpm-1 mutants. Significance tested using Student’s t-test with Bonferroni correction. ***p<0.001.

https://doi.org/10.7554/eLife.44040.006

ATAT-2 and RPM-1 function in a pathway to regulate synapse maintenance

Next, we wanted to test the genetic relationship between RPM-1 and molecules that affect microtubules. We started by evaluating how null mutants for different microtubule binding proteins and tubulin acetyltransferases affect synapse maintenance. Interestingly, loss of function in atat-2, an α-tubulin acetyltransferase, resulted in synapse maintenance defects similar to rpm-1 mutants in which presynaptic boutons are lost and the synaptic branch is absent (Figure 5A). Quantitation indicated theses defects occurred at a moderate but significant frequency in atat-2 mutants (Figure 5B). This observation was confirmed using a second transgenic background (Figure 5—figure supplement 1). Destabilized synapses were also observed in mutants for mec-17, another α-tubulin acetyltransferase isoform, and the minus-end binding protein ptrn-1 (Figure 5B). However, defects in these mutants occurred at lower frequency than in atat-2 mutants. Our observation that ptrn-1 affects PLM synapse maintenance is consistent with a prior study (Marcette et al., 2014). Synapse maintenance defects were not observed in ptl-1/Tau mutants (Figure 5B).

Figure 5 with 1 supplement see all
Several mutants that affect microtubules interact with rpm-1 to affect synapse maintenance.

(a) Confocal images of presynaptic boutons and synaptic branches in adult PLM neurons. In wt animal, presynaptic boutons from PLML and PLMR are visible. Note one synaptic branch is shown and the other is out of the focal plane. atat-2 and rpm-1 mutants lack a synaptic branch and only show PLMR bouton (note loss of PLML, arrow). (b) Quantitation of synapse maintenance defects for indicated genotypes. Note atat-2 shows higher frequency defects than ptrn-1 or mec-17. (c) Quantitation showing synapse maintenance defects are similar in rpm-1; atat-2 and rpm-1; ptrn-1 double mutants compared to rpm-1 single mutants. In contrast, rpm-1; mec-17 and rpm-1; ptl-1 double mutants show enhanced defects compared to rpm-1 single mutants. (d) Quantitation indicates synapse maintenance defects are enhanced in atat-2; ptrn-1 double mutants compared to atat-2 single mutants. (e) Quantitation showing synapse maintenance defects are suppressed in rpm-1; dlk-1 double mutants, but not atat-2; dlk-1 double mutants. Significance tested using Student’s t-test with Bonferroni correction. ***p<0.001 and ns = not significant.

https://doi.org/10.7554/eLife.44040.007

Having tested how different mutants for microtubule binding proteins and tubulin acetyltransferases impact synapse maintenance in PLM mechanosensory neurons, we constructed double mutants with rpm-1. Interestingly, rpm-1; atat-2 double mutants showed a similar frequency of synapse maintenance defects as rpm-1 single mutants (Figure 5C). We validated this result using a second transgenic background (Figure 5—figure supplement 1). These results demonstrate that ATAT-2 functions in the same pathway as RPM-1. Similarly, the frequency of synapse maintenance defects was not increased in rpm-1; ptrn-1 double mutants compared to single mutants, which suggests PTRN-1 and RPM-1 function in the same pathway (Figure 5C). We note that this result differs with a prior study that suggested RPM-1 and PTRN-1 function in parallel pathways to regulate synapse formation (Marcette et al., 2014). In contrast to outcomes with atat-2 and ptrn-1, the frequency of destabilized synapses was enhanced in rpm-1; mec-17 and rpm-1; ptl-1 double mutants (Figure 5C). These results indicate that MEC-17 and PTL-1/Tau function in parallel pathways with RPM-1 to regulate synapse maintenance.

Given our observations indicating that RPM-1 functions in the same pathway as both ATAT-2 and PTRN-1, we tested the genetic relationship between ptrn-1 and atat-2. To do so, we evaluated atat-2; ptrn-1 double mutants. These animals showed strong, significant enhancement of synaptic maintenance defects compared to single mutants (Figure 5D). Thus, ptrn-1 and atat-2 function in parallel genetic pathways to regulate synapse maintenance.

Previous studies showed one mechanism by which RPM-1 regulates synapse formation is ubiquitination and inhibition of the DLK-1 MAP kinase (Grill et al., 2007; Nakata et al., 2005; Yan et al., 2009). Therefore, we tested the relationship between dlk-1 and atat-2. Consistent with this prior work, synapse maintenance defects were strongly suppressed in dlk-1; rpm-1 double mutants compared to rpm-1 single mutants (Figure 5E). In contrast, we did not observe suppression of synapse maintenance defects in dlk-1; atat-2 double mutants compared to atat-2 single mutants (Figure 5E).

To our knowledge, these results show for the first time that the tubulin acetyltransferases ATAT-2 and MEC-17 are required for presynaptic bouton maintenance, with ATAT-2 playing a particularly prominent role. Furthermore, our results demonstrate that ATAT-2 and RPM-1 function in a novel pathway to regulate synapse maintenance, while ATAT-2 functions in parallel to PTRN-1 and independently of DLK-1. The simplest model that explains our findings is that ATAT-2 and DLK-1 are part of a signaling network that is differentially regulated downstream of RPM-1. Because dlk-1 suppresses rpm-1 but not atat-2, it is particularly likely that ATAT-2 functions downstream of RPM-1. If ATAT-2 functioned upstream of RPM-1, one would expect suppression of both rpm-1 and atat-2 by dlk-1, which did not occur.

ATAT-2 acetyltransferase activity is required for synapse maintenance

Genetic interactions with rpm-1, and the loss of presynaptic boutons and synaptic branches in PLM neurons of atat-2 mutants prompted us to further evaluate if synapse maintenance was impaired in these animals. Indeed, we observed that presynaptic terminals form and accumulate UNC-10/RIM and RAB-3 in atat-2 mutants at 16 hr PH (Figure 6—figure supplement 1A,C). Quantitation indicated that a small defect in UNC-10/RIM accumulation occurred at 16 hours PH in atat-2 mutants (Figure 6—figure supplement 1B). Thus, similar to rpm-1 mutants, presynaptic assembly is largely normal in atat-2 mutants with subtle defects in accumulation of UNC-10 occurring at the critical 16-hr PH developmental time point. These results indicate ATAT-2 is affecting synapse maintenance, which is consistent with ATAT-2 functioning in the same pathway as RPM-1.

To further validate synapse maintenance defects in atat-2 mutants, we performed transgenic rescue experiments. Synapse destabilization in atat-2 mutants was rescued by expression of ATAT-2 using either the native atat-2 promoter or a mechanosensory neuron promoter (Figure 6A,B). These results demonstrate that ATAT-2 functions cell autonomously in mechanosensory neurons to regulate presynaptic maintenance.

Figure 6 with 1 supplement see all
ATAT-2 tubulin acetyltransferase activity functions in mechanosensory neurons to regulate presynaptic maintenance.

(a) Confocal images of presynaptic boutons and synaptic branches in adult PLM neurons. In the wt animal, presynaptic boutons are shown for both PLML and PLMR neurons (brackets). Note one synaptic branch is shown and other is out of focal plane. atat-2 mutant lacks presynaptic boutons from both PLML and PLMR, and synaptic branch absent (arrow). Expression of ATAT-2 in mechanosensory neurons rescues defects. ATAT-2 lacking acetyltransferase activity (ATAT-2 dead) fails to rescue (b) Quantitation of synapse maintenance defects for indicated genotypes. Defects in atat-2 mutants are rescued by using native atat-2 or mechanosensory neuron promoters to transgenically express ATAT-2. No significant rescue occurs with ATAT-2 lacking acetyltransferase activity. (c) Microtubule destabilizing drug colchicine enhances synapse maintenance defects in atat-2 mutants. Significance tested using Student’s t-test with Bonferroni correction. ***p<0.001 and ns = not significant.

https://doi.org/10.7554/eLife.44040.009

Importantly, rescue experiments allowed us to test whether the enzymatic acetyltransferase activity of ATAT-2, or non-enzymatic mechanisms affect synapse maintenance. To do so, we performed rescue with ATAT-2 (G125W, G127W) which lacks acetyltransferase activity (Topalidou et al., 2012). Unlike wt ATAT-2, acetyltransferase dead ATAT-2 failed to rescue synapse maintenance defects (Figure 6A,B). This indicates that ATAT-2 functions via its acetyltransferase activity to regulate synapse maintenance. This finding is consistent with the nature of the atat-2 allele we used, ok2415, which contains a large deletion in the acetyltransferase domain (Shida et al., 2010).

To provide further evidence that ATAT-2 influences microtubule stability to affect synapse maintenance, we evaluated how the microtubule destabilizing drug, colchicine, affects defects in atat-2 mutants. Consistent with ATAT-2 function increasing microtubule stability, synapse maintenance defects caused by atat-2 (lf) were enhanced by colchicine (Figure 6C).

Taken together, these results demonstrate that ATAT-2 α-tubulin acetyltransferase activity regulates synapse maintenance by functioning in mechanosensory neurons. Importantly, this is the first evidence in any system that tubulin acetyltransferase activity is required to maintain synapse stability.

RPM-1 and ATAT-2 function in a pathway to regulate short-term learning

Our results showed that ATAT-2 and RPM-1 function in a linear pathway to regulate maintenance of chemical synapses in PLM mechanosensory neurons. To test the genetic relationship between rpm-1 and atat-2 in a behavioral context, we evaluated habituation to repeated tap stimuli, a form of gentle touch, that is sensed by the mechanosensory neurons.

C. elegans respond to tapping the plate they are grown on by reversing their direction of movement. Initial tap sensation is thought to be primarily mediated by electrical gap junction synapses along the primary axon (Figure 7A) (Chalfie et al., 1985; Wicks and Rankin, 1995). Repeated tap stimulus leads to habituation, a simple form of short-term learning in which responses progressively decrease. Tap habituation is influenced by the glutamatergic chemical synapses of mechanosensory neurons (Crawley et al., 2017; Giles et al., 2015; Rankin and Wicks, 2000). RPM-1 functions in mechanosensory neurons to regulate habituation, and habituation defects in rpm-1 mutants likely result, at least in part, from defects in chemical synapse formation (Crawley et al., 2017; Giles et al., 2015). Because electrical synapse formation is not impaired in rpm-1 mutants, they respond normally to initial tap (Borgen et al., 2017b; Giles et al., 2015; Meng et al., 2016).

Habituation to repeated mechanical stimulation is affected by ATAT-2 and RPM-1.

(a) Chemical and electrical synapses in PLM mechanosensory neurons primarily affect habituation to repeated tap stimuli and initial tap sensation, respectively (adapted from Crawley et al., 2017). (b) Multi-worm tracker was used to quantitate tap habituation for indicated adult genotypes. Habituation is defective in atat-2 mutants (red) and rpm-1 mutants (blue) compared to wt animals. rpm-1; atat-2 double mutants (magenta) are not significantly different than rpm-1 single mutants indicating atat-2 and rpm-1 function in a linear pathway to regulate habituation. Habituation level (HL) is shaded in grey. Significance assessed by Student’s t-test with Bonferroni correction. *p<0.05, ***p<0.001 and ns = not significant.

https://doi.org/10.7554/eLife.44040.011

Consistent with prior studies, wt adult animals habituated to repeated tap with decreased responses over time, while habituation was strongly impaired in rpm-1 mutants (Figure 7B). Consistent with defects in PLM synapse maintenance, habituation was also defective in atat-2 animals, although defects were less severe than those observed in rpm-1 animals (Figure 7B). rpm-1; atat-2 double mutants phenocopied rpm-1 single mutants (Figure 7B). These results with whole animal behavior provide further support that RPM-1 and ATAT-2 function in the same pathway.

Prior studies (Crawley et al., 2017; Giles et al., 2015; Rankin and Wicks, 2000) and our results here are all consistent with impaired chemical synapses in mechanosensory neurons affecting habituation to repeated tap, a simple form of short-term learning. However, it is notable that the frequency of synapse maintenance defects in rpm-1 mutants is only slightly stronger than atat-2 mutants (Figure 5B), while habituation defects are much stronger in rpm-1 mutants (Figure 7B). This might occur because RPM-1 is a signaling hub that regulates several downstream pathways (Grill et al., 2016). In contrast, atat-2 is only known to affect microtubules. Because synapses in rpm-1 mutants face several insults to signaling compared to atat-2 mutants, it is possible remaining synapses that do not destabilize could be functionally weaker in rpm-1 mutants than atat-2 mutants. This idea is supported by previous studies in Drosophila which have shown that loss of function in the RPM-1 ortholog Highwire results in synapse formation defects and synaptic transmission defects that are mediated by distinct molecular mechanisms (Borgen et al., 2017a; Collins et al., 2006).

Discussion

This study breaks new ground on several fronts regarding the molecular and genetic mechanisms that regulate synapse maintenance (Figure 8). 1) We provide new evidence that PHR proteins, such as RPM-1, impact synapse formation during development primarily via effects on synapse maintenance. 2) We show for the first time that ATAT-2 regulates presynaptic maintenance, and does so via its α-tubulin acetyltransferase activity. 3) ATAT-2 acts in a novel pathway with RPM-1 that functions cell autonomously to regulate presynaptic maintenance. 4) Extensive genetic analysis revealed RPM-1 is a hub in a signaling network consisting of ATAT-2, PTRN-1 and DLK-1. 5) Finally, our results indicate that ATAT-2 functions independently of DLK-1, and is therefore likely to function downstream of RPM-1 to affect synapse maintenance.

ATAT-2 tubulin acetyltransferase activity functions in a pathway with RPM-1 to regulate synapse maintenance.

Model summarizing genetic and pharmacological results suggesting RPM-1 functions upstream of ATAT-2 acetyltransferase activity to regulate microtubule stability and synapse maintenance. Outcomes indicate the RPM-1/ATAT-2 pathway functions independently of DLK-1 to regulate synapse maintenance.

https://doi.org/10.7554/eLife.44040.012

RPM-1 regulates synapse maintenance during development

Experiments in worms, flies and mice showed that PHR proteins, such as RPM-1, regulate synapse formation (Bloom et al., 2007; Grill et al., 2016; Schaefer et al., 2000; Wan et al., 2000; Zhen et al., 2000). Important progress has been made in understanding signaling networks regulated by RPM-1 and PHR proteins, which act as both signaling hubs and ubiquitin ligases (Grill et al., 2016). Nonetheless, we still lack a clear cellular explanation for why synapse formation is abnormal in the absence of PHR proteins. One prior study examined the presence of presynaptic boutons, and hinted that synapse assembly might be the principal defect in rpm-1 mutants with possible minor defects in synapse maintenance (Schaefer et al., 2000). We now expand significantly on this prior work with more extensive developmental time course analysis of bouton morphology, and several presynaptic markers. Our results indicate that synapse formation defects in rpm-1 mutants result primarily from a failure to maintain synapses (Figures 1, 3 and Figures 3—figure supplement 1). Consistent with RPM-1 regulating synapse maintenance during development, we observed RPM-1 accumulation at presynaptic terminals during synapse formation through adulthood (Figure 2).

In mice, the RPM-1 ortholog Phr1 regulates synapse formation at the NMJ. While it remains unknown if this is due to defects in synapse assembly or maintenance, the Diantonio group observed orphan presynaptic terminals in Phr1 knockout mice (Bloom et al., 2007). Our results suggest these orphan presynaptic terminals could reflect failed synapse maintenance. Consistent with this, studies in flies and mice indicate orphan terminals are a hallmark of destabilizing synapses (Bhattacharya et al., 2016; Eaton et al., 2002; Graf et al., 2011; Pielage et al., 2008).

Our findings demonstrate that RPM-1 regulates developmental synapse maintenance, which we consider the final step in synapse formation. This differs from long-term synapse maintenance, which facilitates synapse integrity for months and is regulated by molecules like αLaminin and p190Rho at central synapses, and LRP4 at NMJs (Barik et al., 2014; Kerrisk et al., 2013; Lin and Koleske, 2010; Omar et al., 2017). The relationship between regulators of developmental synapse maintenance and long-term synapse maintenance remains unclear and awaits future studies.

ATAT-2 functions via acetyltransferase activity to regulate synapse maintenance

In C. elegans, there are two α-tubulin acetyltransferases, ATAT-2 and MEC-17 (Akella et al., 2010; Shida et al., 2010). We show here that both molecules affect synapse formation in mechanosensory neurons (Figure 5). ATAT-2 is a more prominent player and functions via enzymatic acetyltransferase activity to regulate synapse maintenance (Figures 5 and 6). The implications of these findings potentially extend beyond C. elegans, as the mammalian acetyltransferase ortholog called Atat1 affects hippocampal development and touch sensation (Kalebic et al., 2013; Kim et al., 2013; Morley et al., 2016). Whether these phenotypes arise from defects in synapse maintenance now becomes an intriguing question.

Our genetic analysis also revealed that ATAT-2 functions in a linear pathway with RPM-1 (Figure 8). This was observed in two functional contexts: synapse maintenance in mechanosensory neurons (Figure 5) and behavioral habituation mediated by these neurons (Figure 7). Consistent with RPM-1 and ATAT-2 acting in a linear pathway, both RPM-1 and ATAT-2 function cell autonomously in mechanosensory neurons to regulate presynaptic maintenance (Figures 1D and 6A,B).

Several observations indicate the RPM-1/ATAT-2 pathway influences microtubule stability to impact synapse maintenance. Treating either rpm-1 or atat-2 mutants with colchicine enhanced synapse destabilization defects (Figures 4A and 6C). RPM-1 functions in parallel to other molecules that can stabilize microtubules, such as PLT-1/Tau and MEC-17 (Figure 5). Finally, the enzymatic α-tubulin acetyltransferase activity of ATAT-2 was necessary for synapse maintenance (Figure 6B).

Our results do not provide definitive evidence for the order of ATAT-2 and RPM-1 within this novel pathway. We attempted to address this with transgenic bypass experiments, but results were inconclusive. Nonetheless, there are several reasons why RPM-1 most likely functions upstream of ATAT-2 (Figure 8). First, RPM-1 is a signaling hub and ubiquitin ligase that positively and negatively regulates at least six different downstream signaling pathways (Grill et al., 2016). In contrast, ATAT-2 regulates microtubules directly and is not known to regulate signaling events (Akella et al., 2010; Shida et al., 2010; Topalidou et al., 2012). Second, RPM-1 functions in parallel to other molecules that can stabilize microtubules, including MEC-17 and PTL-1/Tau. If RPM-1 were to function downstream of microtubule stability in general, we might expect the same genetic relationship between RPM-1 and all mutants that affect microtubule stability, which did not occur. The final argument is perhaps the most convincing. Our results indicate that ATAT-2, PTRN-1 and DLK-1 function within the RPM-1 signaling network (Figure 5). While dlk-1 can suppress rpm-1, it failed to suppress atat-2. The simplest explanation for this is a model in which ATAT-2 and DLK-1 have opposing functions with both molecules acting downstream of RPM-1 (Figure 8). If ATAT-2 were to function upstream of RPM-1, we would expect dlk-1 to suppress both atat-2 and rpm-1, which did not occur. Despite cumulative reasons for favoring the model that RPM-1 functions upstream of ATAT-2, we cannot entirely rule out the alternative possibility.

It is intriguing that the functional genetic relationship between RPM-1 and molecules that affect microtubule stability differs between synapse maintenance in mechanosensory neurons described here (Figure 8), and axon termination of the same neurons that occurs in a different anatomical location (Borgen et al., 2017b). For example, RPM-1 functions in the same pathway as both ATAT-2 and PTRN-1 to regulate synapse maintenance (Figure 5C), but acts in parallel opposing pathways to these molecules to regulate axon termination (Borgen et al., 2017b). Further, RPM-1 functions in parallel to PTL-1/Tau to regulate synapse maintenance (Figure 5C), but Tau likely inhibits RPM-1 during axon termination (Borgen et al., 2017b). Thus, the functional genetic relationship between RPM-1 and regulators of microtubule stability varies with subcellular location and the developmental process in question. These findings in C. elegans might explain why studies in fish and mice that analyzed different types of neurons arrived at opposing conclusions about how PHR protein signaling influences microtubule stability (Hendricks and Jesuthasan, 2009; Lewcock et al., 2007).

Another worthwhile consideration emerges from our findings here, and prior observations about axon termination (Borgen et al., 2017b). We now uncover a second example of a functional genetic relationship between RPM-1 and PTL-1/Tau, which has interesting implications given the prominence of Tau in neurodegenerative disease (Brunden et al., 2009; Wang and Mandelkow, 2016). Likewise, our discovery that RPM-1 functions in a pathway with ATAT-2 acetyltransferase activity to regulate synapse maintenance could have important disease implications, as alterations in α-tubulin acetylation and synapse instability are associated with neurodegenerative diseases (Godena et al., 2014; Govindarajan et al., 2013; Hempen and Brion, 1996; Lin and Koleske, 2010; Pellegrini et al., 2017). Thus, converging themes from several studies, including this one, suggest it could be informative to test whether the RPM-1/ATAT-2 pathway impacts neurodegenerative disease models.

Materials and methods

Genetics and transgenics

C. elegans strains were maintained using standard procedures. Alleles used included: rpm-1 (ju44), mec-17 (ok2109), ptrn-1 (tm5597), ptl-1 (ok621), dlk-1 (ju476), and atat-2 (ok2415). Integrated transgenes and extrachromosomal arrays used in this study are as follows: muIs32 (Pmec-7GFP), zdIs5 (Pmec-4GFP), bggEx8 (Prpm-1RPM-1), bggEx127 (Pmec-3RPM-1::GFP), bggEx141 (Pmec-7SYD-2::mScarlet), jsIs973 (Pmec-7mRFP), jsIs821 (Pmec-7RAB-3::GFP), bggIs28 (Pmec-7UNC-10::tdTOMATO), and bggIs34 (Pmec-3RPM-1::GFP). jsIs973 and jsIs821 were kind gifts from Dr. Michael Nonet (Washington University). All alleles were outcrossed a minimum of four times. All double mutants were constructed following standard mating procedures. Genotypes were confirmed by PCR, or sequencing as needed. Primers and PCR conditions are available upon request.

For rescue experiments, transgenic extrachromosomal arrays were constructed by injecting DNA of interest with a coinjection marker, Pmyo-2RFP (2 ng/µl) or Pttx-3RFP (50 ng/µL) and pBluescript to reach a total DNA concentration of 100 ng/µL. For all rescue experiments, two or more independently derived transgenic lines were analyzed for a given genotype. Supplementary file 1 details transgenic extrachromosomal arrays and injection conditions.

Cloning

For atat-2 rescues with the native promoter, the atat-2 locus (including promoter, open reading frame, and 3’ UTR) was PCR amplified from N2 genomic DNA. Primer sequences used for cloning are available upon request. For atat-2 rescue with mechanosensory neuron promoters, atat-2 cDNA was amplified from C. elegans RNA and TOPO cloned into pCR8 Gateway entry vector (Invitrogen) to generate pBG-GY896. pBG-GY896 was recombined into a destination vector containing the mec-7 promoter, pBG-GY119, to generate pBG-GY897 (Pmec-7ATAT-2). Site-directed mutagenesis was performed on pBG-GY896 to change two glycine residues into tryptophan (G125W and G127W) resulting in pBG-GY898. Mutation of these conserved glycine residues in the ATAT-2 paralog MEC-17 (G121W and G123W) was previously shown to render MEC-17 catalytically inactive (Topalidou et al., 2012). After mutagenesis, pBG-GY898 was recombined with the Pmec-7 destination vector to yield pBG-GY899 (Pmec-7ATAT-2 dead). For expression of SYD-2::mScarlet, syd-2 genomic DNA was cloned from C. elegans and TOPO cloned into pCR8 Gateway entry vector to generate pBG-GY699. pBG-GY699 was recombined into a destination vector containing the mec-7 promoter and a C-terminal mScarlet tag, pBG-GY880, to generate pBG-GY936 (Pmec-7SYD-2::mScarlet).

Developmental analysis of synapses

The transgenic strain muIs32 (Pmec-7GFP) was used to label PLM neurons for analysis of synaptic branch and boutons during development. Where indicated, bggIs28 was used to coexpress the active zone marker UNC-10::tdTOMATO. Transgenic arrays were used to express SYD-2::mScarlet in muIs32. For RAB-3::GFP analysis (jsIs821), jsIs973 (Pmec-7RFP) was used as a cell fill to visualize PLM axon and presynaptic bouton morphology. Synchronized developmental time course analysis was done by collecting freshly hatched L1 larvae and aging animals in hour-long intervals at room temperature (22°C). At indicated time points between 1 and 48 hr post-hatch (PH), animals were mounted in 5 mM levamisole (M9 buffer) on agar pads to score phenotypes. A minimum of 18 PLM neurons were scored at each time point for all markers with the exception of SYD-2::mScarlet which was scored in at least 14 PLM neurons. Phenotypes were scored using epifluorescent microscopy, and all images were acquired using confocal microscopy. Epifluorescent microscopy was done using 100x magnification on a Leica CTR6500 microscope with Leica Application Suite software. Confocal microscopy was done under 63x magnification and 2x zoom factor on a Leica SP8 confocal microscope. Z-stacks were collected (0.5 µm slices for larvae and 1 µm for adults) and maximum intensity projections are shown for each genotype. Image analysis was done using ImageJ software from NIH image (http://rsb.info.nih.gov/ij/).

Synaptic bouton defects were scored by tracking the PLM synaptic branch to the ventral nerve cord. Branches with presynaptic boutons/varicosities were scored as synapses. If no bouton was discernable, even if the branch was present, it was scored as a destabilized synapse. Similar logic was used to evaluate RAB-3::GFP, UNC-10::tdTOMATO and SYD-2::mScarlet at presynaptic terminals.

RPM-1 localization in PLM neurons

Transgenic ju44 mutants expressing RPM-1::GFP using the mec-3 promoter (Pmec-3RPM-1::GFP) and tdTOMATO using the mec-7 promoter (Pmec-7tdTOMATO) were anesthetized with levamisole. Freshly hatched L1 larvae were allowed to age for 5–6 hr at 22°C before imaging. RPM-1::GFP and tdTOMATO were assessed using confocal microscopy and 63x magnification with 2x zoom factor. Adult animals were imaged using 40x magnification and 2x zoom. Increased fluorescence of tdTOMATO was essential for visualizing PLM morphology in L1 larvae because of their small size, and relatively weak expression of the mec-7 promoter at this age (data not shown). Periactive localization of RPM-1 at presynaptic terminals was assessed using both RPM-1::GFP and UNC-10::tdTOMATO as an active zone marker. Because UNC-10::tdTOMATO is difficult to observe in early L1 animals due to low expression and small boutons, we performed this analysis using young adult PLM synapses.

Microtubule pharmacology

For pharmacological manipulation of microtubule stability, taxol (2 µm), colchicine (0.25 mM) or the vehicle DMSO were spread on NGM plates and left overnight. Plates were seeded with OP50 E. coli, and 16 hr later 3–5 P0 adults were placed on plates with drugs. F1 progeny developed normally on colchicine or taxol, and synapse maintenance was scored in young adults. Opposing effects of taxol and colchicine on rpm-1 mutants indicates these drugs are at appropriate concentrations for use in C. elegans, which was also shown previously (Borgen et al., 2017b).

Habituation

Tap habituation experiments were performed as described previously with minor modifications (Giles et al., 2015). Briefly, age-synchronized animals (~50–100) were cultivated from egg until gravid adult (72–75 hours PH) at 23°C, and assayed on 5 cm NGM plates with 50 μL of E. coli. (OP50). Using Multi-Worm Tracker (Swierczek et al., 2011), animal behavior was recorded for 550 s. After the first 100 s, 45 tap stimuli were given with a 10 s inter-stimulus interval. Response to tap was measured by reversal probability (the fraction of animals that reversed their locomotion within 2 s of the tap). For each plate, exponential curves were fit to responses across stimuli, and habituation level (HL) was measured as the value of the fit at the final stimulus. All strains analyzed contained the muIs32 transgene.

Statistics

For developmental time course analysis, we used the Fisher’s exact test to compare the percentage of PLM neurons lacking synapses in single populations of rpm-1 and wt at each time point. Data points presented represent the mean of the population. A minimum of 18 PLM neurons were scored at each time point for all markers, except SYD-2::mScarlet which was scored in at least 14 PLM neurons.

For analysis of synapse maintenance defects (% destabilized synapses), we scored a minimum of five independent sets of 20–50 PLM neurons from adult animals for each genotype. For rescue experiments, data was acquired from a minimum of two independent transgenic lines for each genotype. Data shown represents the mean and error bars represent the SEM. Comparisons between genotypes were done using the Student’s t-test, with Bonferroni correction for the number of comparisons in each experiment.

For habituation, data presented represents the mean and error bars represent the SEM across 12 plates (50–100 animals per plate) tested on three independent days per genotype. Differences were assessed by comparing habituation level using Student’s t-tests with Bonferroni correction for multiple comparisons.

References

  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
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57

Decision letter

  1. Graeme W Davis
    Reviewing Editor; University of California, San Francisco, United States
  2. K VijayRaghavan
    Senior Editor; National Centre for Biological Sciences, Tata Institute of Fundamental Research, India

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

[Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]

Thank you for submitting your work entitled "Synapse maintenance is impacted by the RPM-1 signaling hub and the tubulin acetyltransferase ATAT-2" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

There is general agreement that the topic of synapse stability is important and of general interest. After decades of studying how synaptic connections are driven to form, there is new interest in what keeps them stable. As such, the topic should have a wide appeal in the field of neuroscience. Further, the reviewers generally agree that the genetic data, as presented, are solid and clear. Specifically, this refers to the identification of the new effects of atat-2. Further, it is acknowledged that the discovery of atat-2 is important in the context of other microtubule regulatory genes that did not have a similar phenotype. Finally, the anti-correlation between growth cone and synapse was appreciated, but it was also felt that this was a side observation that, while interesting, did not much advance the core of the work.

With this enthusiasm in mind, there was also general agreement that the study relies heavily on a limited number of assays and would benefit greatly from a more in depth analysis of the microtubule cytoskeleton and the synapse. There are a number of studies that have pursued direct visualization of the microtubule cytoskeleton in the worm, either live or by EM analyses, and these have in some cases been performed in the context of synapse remodeling, which closely resembles the phenomenon being studied here. By directly addressing the phenotypes associated with the central argument of your work, it is expected that the interpretation of the genetics will be simplified and more powerful. There was general agreement that the behavioral analysis was difficult to interpret, both as a phenomenon and in terms of 'regulatory control' by the genes you are studying. Finally, there were concerns raised about prior work on the topic that seem directly relevant but not acknowledged.

Reviewer #1:

This manuscript (Borgen et al.) identifies the RPM-1 signaling protein and the tubulin acetyltransferase as key requirements for synapse maintenance and touch habituation.

The team first establishes that RPM-1 is required both for synapse stabilization and for the temporal coordination of synaptic bouton formation and axon termination in PLM neurons. RPM-1 localizes to presynaptic terminals just adjacent the active zone and mature active zone-bearing terminals are observed in rpm-1 mutants prior to their destabilization. Next the team demonstrates that treatments that increase destabilization of microtubules appear to destabilize synapses. In alignment with these observations, the group shows that the tubulin acetyltransferase atat-2 is required for synapse maintenance and that it interacts functionally with rpm-1 via epistasis interactions, but other MT stabilizing proteins (e.g. Tau) do not appear to be involved. Finally, the group shows that disruption of the rpm-1 decreases habituation to touch and that the effects of this mutation are much larger than the effects of atat-2.

Overall the data are clean and convincing. That said, this work is very descriptive – it defines new pieces involved in synapse maintenance but it does not provide new mechanistic insights into how these new players regulate synapse maintenance. As such, it would be much more appropriate for solid journal focusing on genetic mechanisms (e.g. Genetics).

1) The work describes new players but does not establish how these new players impact synapse stability. The lack of such a mechanistic understanding limits the impact of the discovery. One possibly productive area would be to measure microtubule dynamics in the various genetic backgrounds to test whether and how these treatments impact these. Alternatively, identifying key players or cargoes whose trafficking to synapses requires RPM-1 and/or ATAT-2 could provide mechanistic insights.

2) rpm-1 and atat-2 mutants have similar defects in PML synapse destabilization. The finding that habituation is compromised in rpm-1 much more than in atat-2 mutants would suggest that this phenotype is not simply explained by loss of these synapses. How do the authors reconcile this.

3) The finding that axon elongation and synapse formation are inversely coupled is very interesting observation but would be even more interesting if the authors could identify the underlying mechanisms.

Reviewer #2:

In this manuscript Borgen and colleagues describe a role for the signaling molecule Rpm-1 in microtubule-dependent synapse maintenance in C. elegans. Using the mechanosensory neuron PLM as a model system they demonstrate that the signaling molecule Rpm-1 is not required for synapse formation but for the control of synapse maintenance. Synaptic retraction phenotypes are shared by mutations in microtubule stabilizing factors including the α-tubulin acetyltransferase ATAT-2 and Mec-17 and can be modulated by pharmacological alterations of microtubule stability. Using genetic interaction assays the authors then try to establish a signaling pathway and propose that Rpm-1 acts via Atat-2 and in parallel to other microtubule stabilizing proteins to control synapse maintenance.

While this topic is in principle of great interest for the neuroscience community the current manuscript does not extend significantly enough beyond prior published work. It fails to sufficiently acknowledge prior work and does not incorporate or extend on these published results that addressed at least parts of this important question.

Three papers previously identified microtubule stability as a key factor for synapse maintenance using the same PLM synapse as a model system (Richardson et al., 2014 eLife; Marcette, Chen and Nonet, 2014 eLife; Chen et al., 2014). All three studies demonstrated that impairing MT stability by manipulating either Tubulin, Tubulin-binding proteins or Tubulin-modifying proteins results in defects in synapse maintenance. Importantly, these studies also demonstrated that pharmacological alterations mimic and/or modulate these effects, that synapse formation occurs normally and is followed by synapse retractions (e.g. Marcette Figure 5), that these effects are coupled to defects in the neurite extension and that rpm-1, ptrn-1, mec-17 and atat-2 are part of the signaling machinery controlling synapse maintenance. Furthermore, these studies demonstrated that defects in microtubule stability result in dlk-1 activation that then in turn induces neuronal remodeling (neurite extension and synapse retraction). Mutations in dlk-1 effectively suppress synaptic retraction phenotypes caused by impairments of microtubule stability. As Dlk-1 is the main effector of Rpm-1 (Nataka et al., 2005; rpm-1 inhibits dlk-1 activation) these studies already place rpm-1 upstream of dlk-1 in synapse maintenance (see Marcette, Chen and Nonet, 2014).

Thus the majority of phenotypes characterized in the Borgen et al. study have been previously described:

1) Rpm-1 required for synapse maintenance;

2) Coupling of neurite extension and retraction;

3) Peri-active zone localization of the rpm-1/Dlk-1 complex;

4) Time course of retraction phenotype for MT regulators (but not for rpm-1);

5) Pharmacological manipulation of MTs causing synaptic retractions;

6) Identification of mec-17, ptrn-1 and atat-2 as synapse stability genes).

In addition, there is convincing evidence that dlk-1upregulation is the main mediator of neuronal remodeling as a consequence of the MT perturbations. Rpm-1 as the main dlk-1 inhibitor can thus clearly be integrated into this model (see Marcette, Chen and Nonet, 2014).

Unfortunately, Borgen et al., do not test and extend the current model but use (weak) genetic interaction data to suggest a reverse order of the signaling complex with rpm-1 being upstream or parallel to the control of MT stability. In addition, the current manuscript does not provide sufficient evidence to support their proposed conclusions.

For these reasons I cannot recommend publication of this manuscript in eLife.

Additional comments:

1) Figure 1 there is no clear distinction between synapse retraction and axon branch degeneration. Throughout the paper it is not clear which comes first and all phenotype could be explained by axon degeneration (at 60h the axon is gone but remnants of the synapse remain).

In addition, there is a clear synapse formation phenotype. This should be quantified. Does the synaptic bouton region ever reach wild type dimensions? Impaired synapse formation might thus contribute to synaptic retraction.

2) Synapse retraction should be demonstrated using unc-10 as a marker. Currently the authors only show that unc-10 assembles at the synapse by 16 hr but do not demonstrate that unc-10 (and thus the presynaptic active zone) is indeed disassembled in rpm-1 mutants. This would also enable to differentiate between induced axonal degeneration and/or synaptic disassembly.

3) A large number of results have been previously published by Marcette, Chen and Nonet, 2014/Chen et al., 2014 – these findings are not acknowledged in this study (see general comment above).

4) The behavioral data is relatively meaningless – if these synapses are no longer present it is not surprising that the stimulus cannot evoke behavioral responses. It remains unclear why there are differences between rpm-1 and atat-2 mutants that display the same frequency of synaptic retractions.

5) The model is only based on genetic interactions and at least partly contradict findings from Marcette, Chen and Nonet, 2014, Borgen et al., 2017 and Chen et al., 2014. This should have been addressed experimentally.

https://doi.org/10.7554/eLife.44040.016

Author response

[Editors’ note: the author responses to the first round of peer review follow.]

We appreciate the reviewers emphasizing the significance of our study’s topic, synapse maintenance. We also appreciate the helpful and thorough reviews we received. These comments have allowed us to dramatically overhaul and improve our paper, and have prompted numerous informative new experiments that address the reviewers’ concerns. We now highlight our novel finding that ATAT-2, acting via its enzymatic acetyltransferase activity, is required for synapse maintenance. As well as the novel discovery that ATAT-2 functions in a pathway with RPM-1 to regulate synapse maintenance.

Unfortunately, our prior paper did a poor job of highlighting these and other novel findings in our study. Our new manuscript was heavily rewritten to address this issue, and we now present new figures (Figure 6, and Figure 6—figure supplement 1) that show ATAT2 acetyltransferase activity regulates synapse maintenance (Figure 6B). We added many other new experiments and made several other changes to further increase the impact and novelty of our manuscript. These include:

1) A more detailed, thorough analysis of synapse maintenance defects allowed us to discover that subtle changes in the active zone marker UNC-10/RIM and presynaptic terminal size precede synapse destabilization in rpm-1 (Figure 3G, H) and atat-2 mutants (Figure 6—figure supplement 1B). These new findings add mechanistic insight into why synapse maintenance fails in rpm-1 and atat-2 mutants.

2) We add a new figure (Figure 3—figure supplement 1) showing that another active zone marker, SYD-2/Liprin, is present in rpm-1 mutants during initial synapse assembly. This further strengthens the concept that synapse assembly is occurring in rpm-1 mutants, and that synapse destabilization is the primary phenotype in the mechanosensory neurons of these animals.

3) We add numerous genetic experiments and show rpm-1 is a hub in a genetic network that regulates synapse maintenance. Results indicate that ATAT-2 functions in the same pathway as RPM-1, but in parallel pathways to both PTRN-1 and DLK-1 (Figure 5C, D, E). Unexpectedly, our analysis also revealed that PTRN-1 functions in the same pathway as RPM-1 (Figure 5C).

4) Importantly, genetic results indicate that while dlk-1 (lf) suppresses rpm-1 (consistent with prior work), dlk-1 fails to suppress atat-2 (Figure 5E). This provides experimental support for the model that ATAT-2 functions downstream of RPM-1, and importantly indicates that ATAT-2 is a novel, DLK-1 independent mechanism RPM-1 utilizes to regulate synapse maintenance.

5) Our discovery that the acetyltransferase activity of ATAT-2 is required for synapse maintenance is a novel and particularly interesting observation for two reasons. First, synapse instability is a hallmark of many neurodegenerative diseases. Second, altered microtubule acetylation is associated with neurodegenerative diseases, such as Alzheimer’s and Parkinson’s.

We hope the reviewers agree that our extensive experimental and textual revisions now make our paper suitable for eLife. Below we detail specific responses to each reviewer’s individual comments.

[…] there was also general agreement that the study relies heavily on a limited number of assays and would benefit greatly from a more in depth analysis of the microtubule cytoskeleton and the synapse. There are a number of studies that have pursued direct visualization of the microtubule cytoskeleton in the worm, either live or by EM analyses, and these have in some cases been performed in the context of synapse remodeling, which closely resembles the phenomenon being studied here. By directly addressing the phenotypes associated with the central argument of your work, it is expected that the interpretation of the genetics will be simplified and more powerful. There was general agreement that the behavioral analysis was difficult to interpret, both as a phenomenon and in terms of 'regulatory control' by the genes you are studying. Finally, there were concerns raised about prior work on the topic that seem directly relevant but not acknowledged.

Reviewer #1:

[…] Overall the data are clean and convincing. That said, this work is very descriptive – it defines new pieces involved in synapse maintenance but it does not provide new mechanistic insights into how these new players regulate synapse maintenance. As such, it would be much more appropriate for solid journal focusing on genetic mechanisms (e.g. Genetics).

We appreciate the reviewer’s support for the significance and quality of our results. We agree that further experiments providing more mechanistic insight would be valuable. We detail changes to our manuscript and new experiments we have added to address these issues.

1) The work describes new players but does not establish how these new players impact synapse stability. The lack of such a mechanistic understanding limits the impact of the discovery. One possibly productive area would be to measure microtubule dynamics in the various genetic backgrounds to test whether and how these treatments impact these. Alternatively, identifying key players or cargoes whose trafficking to synapses requires RPM-1 and/or ATAT-2 could provide mechanistic insights.

We thank the reviewer for making an important point.

We attempted to analyze three different microtubule markers: EBP-2, PTRN-1, and UNC-104 in developing PLM neurons. We also attempted to analyze transport of RAB-3 in PLM neurons using time-lapse imaging. Unfortunately, our experiments were unsuccessful for a variety of reasons.

With regard to RAB-3, we actually found very little transport into and out of the presynaptic terminals in time frames examined. While somewhat unexpected, this suggests that the presynaptic terminals of PLM neurons might have limited or very slow trafficking that is difficult to detect with our existing reagents. In the future, we will consider trying FRAP to detect more subtle trafficking events into and out of presynaptic terminals.

With regard to EBP-2, PTRN-1 and UNC-104 markers, we were unable to get suitable expression with transgenic constructs for analysis. We incurred problems with detecting markers when single copy Mos insertion was used, and encountered many issues with the promoter we chose for traditional transgenic expression. Some of this could be due to our use of laser scanning confocal microscopy and might be improved with spinning disc confocal. However, our tri-institutional campus (Scripps Florida, Max Planck Florida and Florida Atlantic University) does not have a spinning disc microscope, which prevented us from testing this possibility in a reasonable timeframe. Thus, while the reviewer’s point is well taken, and a direction we hope to pursue further in the future, it could not be addressed as part of this study. Consistent with the difficulties we have encountered, we note that Nonet and colleagues made the following statement in their previous paper that examined the same mechanosensory neurons (Marcette, Chen and Nonet, 2014): “investigating neuronal microtubule dynamics in C. elegans neurons in vivo is technically difficult”.

As a result of these issues, we took a different direction in addressing the reviewer’s concern about mechanism. We engaged in a much more in depth developmental analysis of RAB-3 and UNC-10/RIM at the presynaptic terminals of wt, rpm-1 and atat-2 mutants. Our new results (Figure 3G, H) revealed subtle defects in the active zone and presynaptic bouton size of rpm-1 mutants. These small, but significant defects precede synapse deterioration and provide a further explanation as to why synapse deterioration occurs in rpm-1 mutants. Similar subtle defects in UNC-10 accumulation at presynaptic terminals of atat-2 mutants were also observed at a critical time point just prior to synapse deterioration (Figure 6—figure supplement 1B).

Further, we add extensive new genetic analysis with dlk-1 and ptrn-1. These experiments revealed that ATAT-2 functions in a linear pathway with RPM-1, but does so independently of DLK-1 and in parallel to PTRN-1. This indicates that the RPM1/ATAT-2 pathway regulates synapse maintenance in a DLK-1 independent manner, and provides experimental support for the concept that ATAT-2 is likely to function downstream of RPM-1. We comment on this point in both the Results and the Discussion in detail.

We now add further experiments and better present our novel discovery that ATAT-2 regulates synapse maintenance. Mechanistically, we show ATAT-2 functions via enzymatic tubulin acetyltransferase activity in the presynaptic mechanosensory neurons to regulate synapse maintenance (Figure 6, Figure 6—figure supplement 1).

Finally, more in depth analysis revealed that atat-2 mutants, like rpm-1 mutants, show subtle loss of the active zone marker UNC-10/RIM that precedes synapse destabilization (Figure 6—figure supplement 1).

2) rpm-1 and atat-2 mutants have similar defects in PML synapse destabilization. The finding that habituation is compromised in rpm-1 much more than in atat-2 mutants would suggest that this phenotype is not simply explained by loss of these synapses. How do the authors reconcile this.

The reviewer makes a fair point. We have updated the text to address this concern: “Prior studies (Crawley et al., 2017; Giles et al., 2015; Rankin and Wicks, 2000) and our results here are all consistent with impaired chemical synapses in mechanosensory neurons affecting habituation to repeated tap, a simple form of short-term learning. […] This idea is supported by previous studies in Drosophila which have shown that loss of function in the RPM-1 ortholog Highwire results in synapse formation defects and synaptic transmission defects that are mediated by distinct molecular mechanisms (Borgen et al., 2017a; Collins et al., 2006).”

3) The finding that axon elongation and synapse formation are inversely coupled is very interesting observation but would be even more interesting if the authors could identify the underlying mechanisms.

We agree with the reviewer. However, this data was presented a bit prematurely, and does not fit well with the focus of our new revised manuscript. This data also received conflicting reactions from our two reviewers. Therefore, we have removed this data from our revised manuscript. We hope to present this data with more mechanistic insight in the future.

We also opted to remove this data and not focus further on this part of our original paper because time was needed to address the extensive concerns raised by reviewer 2.

Reviewer #2:

[…] While this topic is in principle of great interest for the neuroscience community the current manuscript does not extend significantly enough beyond prior published work. It fails to sufficiently acknowledge prior work and does not incorporate or extend on these published results that addressed at least parts of this important question.

Three papers previously identified microtubule stability as a key factor for synapse maintenance using the same PLM synapse as a model system (Richardson et al., 2014; Marcette, Chen and Nonet, 2014; Chen et al., 2014). All three studies demonstrated that impairing MT stability by manipulating either Tubulin, Tubulin-binding proteins or Tubulin-modifying proteins results in defects in synapse maintenance. Importantly, these studies also demonstrated that pharmacological alterations mimic and/or modulate these effects, that synapse formation occurs normally and is followed by synapse retractions (e.g. Marcette Figure 5), that these effects are coupled to defects in the neurite extension and that rpm-1, ptrn-1, mec-17 and atat-2 are part of the signaling machinery controlling synapse maintenance. Furthermore, these studies demonstrated that defects in microtubule stability result in dlk-1 activation that then in turn induces neuronal remodeling (neurite extension and synapse retraction). Mutations in dlk-1 effectively suppress synaptic retraction phenotypes caused by impairments of microtubule stability. As Dlk-1 is the main effector of Rpm-1 (Nataka et al., 2005; rpm-1 inhibits dlk-1 activation) these studies already place rpm-1 upstream of dlk-1 in synapse maintenance (see Marcette, Chen and Nonet, 2014).

We thank the reviewer for their feedback. We appreciate the reviewer noting that the topic of our study is of “great interest to the neuroscience community”.

We have addressed the reviewer’s specific comments below extensively with both revisions to our text and new experiments. The result is a study that more comprehensively evaluates the relationship between rpm-1 and a genetic network that affects synapse maintenance. Importantly, we now focus more on our novel discovery that ATAT-2 regulates synapse maintenance via its tubulin acetyltransferase activity. Unfortunately, our prior manuscript was not focused properly around this novel discovery. To further address this, we added several pieces of new data on ATAT-2 (Figures 5D, E, Figure 6, and Figure 6—figure supplement 1).

Our novel finding that ATAT-2 functions in the RPM-1 pathway (Figure 5) is now bolstered by new data showing that both atat-2 and rpm-1 mutants have subtle defects in accumulation of the active zone marker UNC-10/RIM that occurs just prior to synapse destabilization (Figure 3G and Figure 6—figure supplement 1).

Moreover, we now include new data on both PTRN-1 and DLK-1, which the reviewer notes were studied previously. The reviewer’s point, that our prior manuscript did not address the relationship between RPM-1 and ATAT-2 with regard to these important players, is well taken. Several interesting findings have resulted from our new experiments (Figure 5D, E). This has allowed us to paint a clearer, more comprehensive picture of how RPM-1 and ATAT-2 relate to PTRN-1 and DLK-1.

Our results indicate that RPM-1 and ATAT-2 tubulin acetyltransferase activity function in a linear genetic pathway presynaptically to regulate synapse maintenance (novel discovery). Genetic results indicate that PTRN-1 functions in the RPM-1 pathway (new finding), but acts in parallel to ATAT-2 (novel finding). Perhaps of particular interest to the reviewer is our finding that synapse maintenance defects in atat-2 mutants are not suppressed by dlk-1 (novel finding), although defects caused by rpm-1 (lf) are suppressed by dlk-1 (consistent with prior results). This indicates that ATAT-2 functions in the RPM-1 pathway, but independently of DLK-1. As the reviewer notes DLK-1 is a prominent mechanism of RPM-1 function, our results show that ATAT-2 is an entirely new mechanism of RPM-1 function that does not rely upon DLK-1. Our results also provide evidence suggesting that ATAT-2 functions downstream of RPM-1 since we would expect atat-2 to be suppressed by dlk-1 if it functioned upstream of RPM-1.

Importantly, it is not unreasonable that RPM-1 interacts genetically with several molecules other than DLK-1, as numerous studies in C. elegans have now established that RPM-1 utilizes 6 different downstream mechanisms (GLO-4, PPM-2, ANC-1, RAE1, MLK-1 and DLK-1; for review see Grill et al., 2016). While DLK-1 is a prominent player, it is not the only player in the RPM-1 signaling network.

The reviewer’s comments prompted us to re-evaluate and substantially overhaul our entire manuscript, as well as add an extensive body of new experiments. We think this has dramatically improved the novelty and impact of our paper, which we hope the reviewer agrees is suitable for publication in eLife.

Below we detail our specific experimental and textual updates to this manuscript.

Thus the majority of phenotypes characterized in the Borgen et al. study have been previously described:

1) Rpm-1 required for synapse maintenance;

The reviewer makes a fair point that a previous study (which we cite) described a role for rpm-1 in synapse formation (Schaefer, et al., 2000). However, the data from this study are more consistent with impaired synapse assembly in rpm-1 mutants and mild effects on synapse maintenance (See Schaefer et al., 2000, Figure 3).

Importantly, this prior study by the Nonet group only examined the presence or absence of presynaptic boutons.

Our study expands extensively from this prior work in several ways. 1) Our results differ with the findings of Schaefer et al., as we now show that the principle defect in rpm-1 mutants is failed synapse maintenance rather than impaired synapse assembly. 2) We do this using developmental time course analysis for not only bouton morphology (two transgenes tested Figure 1 and Figure 3), but also three different presynaptic proteins marking synaptic vesicles (RAB-3, Figure 3), and the active zone (UNC-10/RIM, Figure 3; and SYD-3/Liprin, Figure 3—figure supplement 1). None of these markers were analyzed in rpm-1 mutants in the original Nonet study, and none were tested over developmental time course previously. 3) We perform new quantitative analysis on bouton size and the presence of UNC-10 at presynaptic boutons in rpm-1 mutants during development, and show that at the critical 16 hour PH time point rpm-1 mutants have small defects in bouton size and active zone accumulation (Figure 3G, H). Thus, subtle presynaptic defects occur at a key time point just before major loss of synapse stability in rpm-1 mutants. This is also a novel dataset that significantly informs our mechanistic understanding of why synapse maintenance fails in rpm-1 mutants.

2) Coupling of neurite extension and retraction;

We think there is some confusion about our presentation of growth cone frequency with synaptic bouton frequency. Both reviewers have commented on this data with differing positive and negative reactions. Since this is not at the core of discoveries made in this paper, we have removed this data from our revised paper.

We hope to present this data in the future more clearly and with better mechanistic insight.

3) Peri-active zone localization of the rpm-1/Dlk-1 complex;

We have adjusted the text to more accurately reflect our own analysis (done in mechanosensory neurons for the first time) with previous studies that focused on motor neurons (Abrams et al., 2008).

“In adult animals, RPM-1 localized directly adjacent to UNC-10 at presynaptic terminals (Figure 2C). This indicates RPM-1 localizes to the periactive zone of presynaptic terminals in mechanosensory neurons. Our observation is consistent with prior studies that examined RPM-1 localization in motor neurons (Abrams et al., 2008).”

4) Time course of retraction phenotype for MT regulators (but not for rpm-1);

We agree with the reviewer that we did a quantitative time course of presynaptic bouton morphology and RAB-3 presence in rpm-1 mutants, but more quantitation for rpm-1 mutants would be valuable. We now include new quantitative developmental time course data on bouton size, and UNC-10 accumulation in rpm-1 mutants (Figure 3G, H). We also include new data on quantitation of SYD-2 active marker in rpm-1 mutants (Figure 3—figure supplement 1).

Given the focus of our revised manuscript on our novel discovery that ATAT-2

acetyltransferase activity regulates synapse maintenance, we opted to also perform more in depth analysis of synapse maintenance defects in atat-2 mutants (Figure 6, Figure 6—figure supplement 1). This showed that UNC-10 and RAB-3 accumulate at presynaptic terminals of atat-2 mutants, but by adulthood synapses are lost. Moreover, more in depth analysis showed that subtle UNC-10 defects occur at a small number of synapses in atat-2 mutants (Figure 6—figure supplement 1B). This is strikingly similar to what we observe in rpm-1 mutants (Figure 3G).

Thus, we now add an extensive amount of new data to address this concern.

5) Pharmacological manipulation of MTs causing synaptic retractions;

We agree with the reviewer that this phenomenon has been shown previously, which we cite in the Results section.

“Consistent with prior work (Chen et al., 2014; Richardson et al., 2014), treating wt animals with colchicine, a microtubule-destabilizing drug, resulted in loss of PLM synapses (Figure 4A).”

However, the relationship between colchicine, taxol and rpm-1 loss of function has not been evaluated previously for synaptic defects in any system. Notably, we did do this type of experiment with regard to axon termination (Borgen, Wang and Grill, 2017). We now show our data for the effects of these drugs on rpm-1 mutants in the context of synapse maintenance (Figure 4). However, it is notable that our findings on the relationship between rpm-1 (lf) and these drugs is the opposite of what occurs in the context of axon termination (Borgen, Wang and Grill, 2017). Thus, between our prior study (Borgen, Wang and Grill, 2017) and this new study we unveil a surprising finding: The relationship between RPM-1 and microtubules varies with the phenotype and where it occurs in the axon (axon termination versus synapse maintenance) even though both phenotypes were evaluated in the same neurons.

To add further novelty, we add new experiments showing colchicine treatment of atat-2 mutants (Figure 6C). Our results on RPM-1 (Figure 4) and ATAT-2 (Figure 6C) indicate RPM-1 and ATAT-2 have similar relationships to pharmacological manipulation of microtubules in the context of synapse maintenance. This is consistent with another novel discovery we present, that RPM-1 and ATAT-2 function in the same pathway to regulate synapse maintenance.

Importantly, the reviewer’s comments helped us realize a major problem with our prior paper is that it was written to focus much too heavily on microtubule effects, and neglected to properly emphasize our novel discoveries regarding synapse maintenance. To address this, we have removed conclusive strong statements regarding RPM-1, ATAT-2 and microtubules from our paper and instead focus on several novel discoveries including that ATAT-2 regulates synapse maintenance via its acetyltransferase activity, that ATAT-2 functions in the RPM-1 pathway to regulate presynaptic maintenance, and that ATAT-2 functions independently of DLK-1 to regulate synapse maintenance.

6) Identification of mec-17, ptrn-1 and atat-2 as synapse stability genes).

We agree that ptrn-1, which is a microtubule minus end binding protein, has been shown to regulate synapse maintenance (Marcette, Chen and Nonet, 2014). Indeed, we cite this paper numerous times in our revised manuscript.

Marcette and colleagues do look at mec-17 and atat-2 mutants in their study. However, it is important to emphasize they only examined axon termination defects (neurite overgrowth defects) and did not analyze synaptic defects. It is important to point this out as Marcette and colleagues use the term “neurite remodeling”. This can lead to confusion, since it could imply that both axon termination and synapse formation were evaluated, when this is not the case.

Moreover, Marcette et al. do not explore whether ATAT-2 or MEC-17 regulate synapse assembly or maintenance in their paper. Nor do they explore the genetic relationship between rpm-1 and atat-2 or mec-17. Notably, they did examine rpm-1; ptrn-1 double mutants, but conclude these genes are enhancers. We now add new data (Figure 5C), which shows that rpm-1 and ptrn-1 function in the same pathway (i.e.synapse maintenance defects are not significantly increased in rpm-1; ptrn-1 double mutants compared to rpm-1 single mutants).

As noted above, we also add numerous new experiments on ATAT-2 showing it regulates synapse maintenance using multiple synaptic markers, and quantitative analysis (Figure 6, Figure 6—figure supplement 1). Most importantly, we show that ATAT-2 functions via its tubulin acetyltransferase activity to regulate synapse maintenance (Figure 6A, B).

To further increase novelty, we add new experiments on ptrn-1; atat-2 double mutants. Our results indicate the frequency of synapse maintenance defects is enhanced in these double mutants. The genetic relationship between ptrn-1 and atat-2 has not been assessed previously.

In summary, we have added numerous experiments that increase novelty and substantially expand on prior work. We have also paid careful attention to citing prior work and ensuring we are fair and accurate in the presentation of our findings that are novel, and well as differences with prior work.

In addition, there is convincing evidence that dlk-1 upregulation is the main mediator of neuronal remodeling as a consequence of the MT perturbations. Rpm-1 as the main dlk-1 inhibitor can thus clearly be integrated into this model (see Marcette, Chen and Nonet, 2014).

We thank the reviewer for highlighting the importance of addressing DLK-1 in our study. We agree that DLK-1 is an important mediator, although not the only mediator, of RPM-1 function (Grill et al., 2016).

We now test the genetic relationship between atat-2 and dlk-1. Interestingly, while atat-2 is in the rpm-1 pathway (Figure 5C), atat-2 is not dependent upon dlk-1 (i.e. atat-2; dlk-1 double mutants are not suppressed while rpm-1; dlk-1 double mutants are suppressed, see Figure 5E).

This is genetic evidence that ATAT-2 acts independently of DLK-1. It is also consistent with ATAT-2 functioning downstream of RPM-1. This new experiment was extremely valuable and provided further novelty by showing the RPM-1/ATAT-2 pathway is a new, DLK-1 independent mechanism for regulating synapse maintenance.

Unfortunately, Borgen et al., do not test and extend the current model but use (weak) genetic interaction data to suggest a reverse order of the signaling complex with rpm-1 being upstream or parallel to the control of MT stability. In addition, the current manuscript does not provide sufficient evidence to support their proposed conclusions.

We aren’t certain what the reviewer is referring to, but we suspect they disagree with our model that RPM-1 most likely functions upstream of ATAT-2. While we thought this was most likely, it is a fair point that our original paper lacked data to directly address this model.

We now show new data indicating that synapse maintenance defects in atat-2 mutants are not suppressed by dlk-1 (Figure 5E). In contrast, defects in rpm-1 mutants are suppressed by dlk-1 (Figure 5E). Given these findings and evidence that rpm-1 and atat2 function in the same pathway (Figure 5C), the simplest model that explains our results is that ATAT-2 and DLK-1 are part of a signaling network that is differentially regulated downstream of RPM-1 (summarized in revised Figure 8). Because dlk-1 suppresses rpm-1 but not atat-2, it is particularly likely that ATAT-2 functions downstream of RPM-1. If this were not the case, one would expect suppression of both rpm-1 and atat-2 by dlk1, which did not occur.

To try and provide further evidence that RPM-1 functions upstream of ATAT-2, we attempted transgenic bypass experiments in which ATAT-2 was overexpressed in rpm-1 mutants. Unfortunately, despite extensive efforts on this front our experiments were not conclusive. However, these experiments could have failed for many reasons and negative outcomes here do not invalidate our model. Although, positive outcomes would have made us more confident.

To further address the reviewer’s concern, we have substantially revised our summary figure (Figure 8).

We have also updated our Discussion to handle this issue carefully and note the caveat to our model:

“Our results do not provide definitive evidence for the order of ATAT-2 and RPM-1 within this novel pathway. […] Despite cumulative reasons for favoring the model that RPM-1 functions upstream of ATAT-2, we cannot entirely rule out the alternative possibility.”

We ask the reviewer to note that it is not unreasonable some regulators of microtubules could lie upstream of RPM-1 while others, such as ATAT-2, function downstream. Indeed, our prior work (Borgen et al., 2017) suggests that PTL-1/Tau is a potential upstream inhibitor of RPM-1 in the context of axon termination, but data here indicate a different relationship between RPM-1 and PTL-1/Tau during synapse maintenance. Given that our Discussion is already quite lengthy, more in depth commentary on these ideas is beyond the scope of this paper. However, we hope to incorporate a discussion on these ideas in a review article in the future.

For these reasons I cannot recommend publication of this manuscript in eLife.

Additional comments:

1) Figure 1 there is no clear distinction between synapse retraction and axon branch degeneration. Throughout the paper it is not clear which comes first and all phenotype could be explained by axon degeneration (at 60h the axon is gone but remnants of the synapse remain).

The reviewer notes that a bouton is present at 60hr (Figure 1), but this is the bouton from the PLM on the back side of the animal. A synaptic branch is present, but is outside our confocal imaging depth. We have updated our schematic in the figure and the legend to make this clear. =

We note that during the developmental time when maintenance fails (between 16 and 48 hours PH), we commonly observe synaptic branches that lack boutons. In contrast, we never observe presynaptic boutons that lack synaptic branches. Thus, it is unlikely synaptic branch degeneration is triggering synapse loss. On this particular point we are in agreement with the Nonet group: when synapse maintenance is impaired the synaptic branch retracts (Marcette, Chen and Nonet, 2014, Luo et al., 2014).

In addition, there is a clear synapse formation phenotype. This should be quantified. Does the synaptic bouton region ever reach wild type dimensions? Impaired synapse formation might thus contribute to synaptic retraction.

The reviewer makes a good point. We have significantly expanded our analysis by quantitatively assessing both presynaptic bouton size and the presence of UNC-10 at 12 and 16 hours PH. We now show that rpm-1 synapses start out normal in size and are similar to wt initially (5 and 7hr PH, Figure 3H). Synapses in rpm-1 mutants accumulate UNC-10 at normal frequency through 12 hours PH (Figure 3G). However, UNC-10 levels begin to drop a small but significant amount just prior to synapse destabilization at 16 hours PH (Figure 3G). Thus, just prior to branch loss, we observe small reductions in bouton size (Figure 3H) and the UNC-10 active zone marker (Figure 3G). These results indicate that while synapses are assembling, they begin to show subtle abnormalities corresponding with onset of synapse destabilization.

To further address this concern, we now add a third presynaptic marker, SYD-2/liprin, to our analysis (Figure 3—figure supplement 1). In rpm-1 mutants, SYD-2 accumulates at presynaptic terminals normally prior synapse destabilization.

2) Synapse retraction should be demonstrated using unc-10 as a marker. Currently the authors only show that unc-10 assembles at the synapse by 16 hr but do not demonstrate that unc-10 (and thus the presynaptic active zone) is indeed disassembled in rpm-1 mutants. This would also enable to differentiate between induced axonal degeneration and/or synaptic disassembly.

We have addressed this with new data showing UNC-10 quantitation at 12 and 16hr PH (Figure 3G). Our data suggest that UNC-10 is present at all presynaptic terminals in rpm-1 mutants at 12 hours PH, and then subtle defects in UNC-10 accumulation occur at 16 hours PH, the key time point just prior to loss of synapses. As noted above, this data is consistent with synapse assembly being largely normal in rpm-1 mutants, and subtle abnormalities in presynaptic terminals occurring just before the onset of synapse destabilization which culminates in total loss of the presynaptic terminal and the synaptic branch. Above we have also discussed why it is unlikely that axon degeneration is leading to synapse loss.

3) A large number of results have been previously published by Marcette, Chen and Nonet, 2014/Chen et al., 2014 – these findings are not acknowledged in this study (see general comment above).

We appreciate the reviewer’s concern. We cited both of these studies and their findings in our initial manuscript. We reread both these papers front to back, and pay very careful attention to citing these prior studies in our revised paper.

Further, we have removed the prior emphasis on microtubule stability/dynamics, and instead focus on synapse maintenance in our revised manuscript. Importantly, both reviewers agreed that synapse maintenance is a very important topic in neuroscience.

Below we detail where we now cite these papers. We must note, that some of our results do not agree with the Marcette paper. We have tried to note this with careful, reasonable language. Indeed, this is a delicate issue, as we greatly respect the tremendous contributions Mike Nonet and his lab have made to understanding synapse formation.

Chen at al. and Marcette et al. are cited in the Introduction:

“For instance, pharmacological and genetic perturbation of microtubules impairs presynaptic bouton maintenance in these cells (Chen et al., 2014). Genetic screens using mechanosensory neurons revealed that the microtubule minus end binding protein PTRN-1/CAMSAP and the actin binding protein ZYX-1 are required for synapse maintenance (Luo et al., 2014; Marcette et al., 2014; Richardson et al., 2014).”

Chen at al. is cited in the Results:

“Consistent with prior work (Chen et al., 2014; Richardson et al., 2014), treating wt animals with colchicine, a microtubule-destabilizing drug, resulted in loss of PLM synapses (Figure 4A).”

Marcette et al., is cited in the Results:

“Our observation that ptrn-1 affects PLM synapse maintenance is consistent with a prior study (Marcette et al., 2014).”

Due to differing results with rpm-1; ptrn-1 double mutants compared to Marcette et al., we also note this discrepancy in the Results:

“the frequency of synapse maintenance defects was not increased in rpm-1; ptrn-1 double mutants compared to single mutants, which suggests PTRN-1 and RPM-1 function in the same pathway (Figure 5C). We note that this result differs with a prior study that suggested RPM-1 and PTRN-1 function in parallel pathways to regulate synapse formation (Marcette, Chen and Nonet, 2014).”

4) The behavioral data is relatively meaningless – if these synapses are no longer present it is not surprising that the stimulus cannot evoke behavioral responses. It remains unclear why there are differences between rpm-1 and atat-2 mutants that display the same frequency of synaptic retractions.

Reviewer 1 also asked for clarification on the behavioral habituation experiments. We addressed this by rewriting our Results on habituation much more carefully and clearly.

We also specifically note the following statement and explanation:

“Prior studies (Crawley et al., 2017; Giles et al., 2015; Rankin and Wicks, 2000) and our results here are all consistent with impaired chemical synapses in mechanosensory neurons affecting habituation to repeated tap, a simple form of short-term learning. […] This idea is supported by previous studies in Drosophila which have shown that loss of function in the RPM-1 ortholog Highwire results in synapse formation defects and synaptic transmission defects that are mediated by distinct molecular mechanisms (Borgen et al., 2017a; Collins et al., 2006).”

We think this data remains valuable because it provides independent behavioral evidence that RPM-1 and ATAT-2 function in the same pathway. We now emphasize this point, rather than focusing on the link between synaptic effects in mechanosensory neurons and behavioral habituation to repeated tap.

5) The model is only based on genetic interactions and at least partly contradict findings from Marcette, Chen and Nonet, 2014, Borgen et al., 2017 and Chen et al., 2014. This should have been addressed experimentally.

We agree that our genetic results differ with some findings from Marcette, Chen and Nonet, 2014. We note this in our revised paper.

As noted above, we now provide new data on ptrn-1, atat-2, dlk-1 and their genetic relationships with rpm-1 and one another. This has resulted in a much more novel and comprehensive study focused on molecular genetic mechanisms regulating synapse maintenance.

Our findings agree with Marcette and colleagues that ptrn-1 mutants have defects in synapse maintenance. However, our results disagree about ptrn-1 functioning in parallel to rpm-1, as mentioned above. Importantly, we note that all our experiments were done after 4X outcrossing to wt (N2) and show that ptrn-1 is epistatic to rpm-1 instead of functioning in parallel. We feel it’s important to share these results with the community.

With respect to our prior study, we were equally surprised that the genetic relationship between rpm-1 and mutants that affect microtubule stability differ based on the phenotype and subcellular location examined. We think it is critical to report our unexpected, but important finding to the community. We think it is plausible that a molecule such as RPM-1 that functions as a signaling hub to regulate axon termination and synapse formation (Grill, et al., 2016) could be a sophisticated regulator of signaling that influences microtubule stability with differing outcomes in different axonal compartments. Indeed, our own recent paper argues that effects of RPM-1 on axon termination can be mechanistically distinguished from synapse formation/maintenance via the MIG-15/JNK-1 pathway (Crawley et al., 2017).

In the Discussion we attempt to address this more clearly and note its importance for the field with the following paragraph:

“It is intriguing that the functional genetic relationship between RPM-1 and molecules that affect microtubule stability differs between synapse maintenance in mechanosensory neurons described here (Figure 8), and axon termination of the same neurons that occurs in a different anatomical location (Borgen et al., 2017b). […] These findings in C. elegans might explain why studies in fish and mice that analyzed different types of neurons arrived at opposing conclusions about how PHR protein signaling influences microtubule stability (Hendricks and Jesuthasan, 2009; Lewcock et al., 2007).”

https://doi.org/10.7554/eLife.44040.017

Article and author information

Author details

  1. Melissa A Borgen

    Department of Neuroscience, The Scripps Research Institute, Jupiter, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Methodology, Writing—original draft
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2600-9946
  2. Andrew C Giles

    Department of Neuroscience, The Scripps Research Institute, Jupiter, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing—original draft
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0875-0696
  3. Dandan Wang

    Department of Neuroscience, The Scripps Research Institute, Jupiter, United States
    Contribution
    Formal analysis, Validation, Investigation, Methodology, Writing—original draft
    Competing interests
    No competing interests declared
  4. Brock Grill

    Department of Neuroscience, The Scripps Research Institute, Jupiter, United States
    Contribution
    Conceptualization, Supervision, Funding acquisition, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    bgrill@scripps.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0379-3267

Funding

National Institute of Neurological Disorders and Stroke (R01 NS072129)

  • Brock Grill

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank the C elegans knockout consortium for several alleles, and the C elegans Genetics Center (NIH Office of Research Infrastructure Programs, P40 OD010440) for providing strains. BG was supported by a grant from the NIH (R01 NS072129). MB is a Neuroscience Scholar of the Esther B O’Keeffe Charitable Foundation.

Senior Editor

  1. K VijayRaghavan, National Centre for Biological Sciences, Tata Institute of Fundamental Research, India

Reviewing Editor

  1. Graeme W Davis, University of California, San Francisco, United States

Publication history

  1. Received: November 30, 2018
  2. Accepted: January 15, 2019
  3. Accepted Manuscript published: January 17, 2019 (version 1)
  4. Accepted Manuscript updated: January 18, 2019 (version 2)
  5. Version of Record published: January 31, 2019 (version 3)

Copyright

© 2019, Borgen et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 610
    Page views
  • 120
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Developmental Biology
    2. Evolutionary Biology
    Suparna Ray et al.
    Research Communication
    1. Cell Biology
    2. Developmental Biology
    Torey R Arnold et al.
    Research Article Updated