Characterization of developmental and molecular factors underlying release heterogeneity at Drosophila synapses

We appreciate the reviewers’ comments on our manuscript and suggestions for improvement. All three reviewers were positive about the study and suggested several avenues for improving the presentation and extending the science, as well as refocusing how we presented the data. We have addressed their concerns in a point-by-point response below. In addition, we have added a host of new data and figures to extend our initial findings as suggested by the reviewers.

Reviewer 1 described the study as “a technically impressive study that reveals correlations between functional and structural features of AZs and PSDs at unprecedented resolution”. The reviewer requested three major clarifications and some additional minor comments we have now addressed as described below.

Framing of the manuscript: “The Introduction needs to be significantly revised to lay out from the outset what is known and suspected and give proper credit in setting up the motivation for this study.”

In our revised manuscript, we have focused our findings in the context of what is known in the field as suggested by the reviewer. We indicate the well-described previous links in the field to Ca²⁺ channels and Ca²⁺ influx in relation to synapse strength. We have reframed our data presentation as the reviewer suggests to lay out the relative importance of the AZ components versus other factors. In addition, we have shifted the last half of the study to investigate how activity alters PSD maturation during development to provide new insights that further separate our study from prior work.

Investigate P_r heterogeneity when AZ size/number is altered: "It would be a great addition to show to what extent P_r heterogeneity persists in a converse experiment – when AZ size is greatly enlarged (including BRP/Cac abundance), while AZ number/density is reduced in rab3 mutants (Graf et al., 2009). This would be of major interest and would nicely complement their study on syt mutants while serving to address the question on the relative importance of AZ size/abundance vs. synaptic vesicle/functional modifications in determining P_r heterogeneity. "

As requested by the reviewer, we completed a new set of experiments in the rab3 mutant and added new data and figures on this topic. Prior work with rab3 (Peled and Isacoff, 2011) had been done with optical quantal imaging to examine Pr in the mutant, revealing that the majority of sites have low Pr, with a smaller population of large release sites with much higher Pr. We observed a similar phenotype in our study and were able to extend the analysis by using the redistribution of Pr seen in the rab3 NMJ as atool to examine the activity-dependence of postsynaptic glutamate receptor field maturation on presynaptic activity using intravital imaging across development (new Figure 10 and Figure 10—figure supplement 1). While prior studies of rab3 mutants examined the third instar stage, we used sequential intravital imaging throughout the whole developmental window beginning in first instar larvae to determine how the NMJ develops with only a few large AZs containing Cac and BRP, while the rest of the AZs lack these components. We observed that glutamate receptor fields apposing BRP-enriched sites developed into a more mature state than PSDs apposing sites lacking BRP and Cac. This observation suggests that PSD maturation depends on either presynaptic components or presynaptic release at the level of single release sites. Coupled with new data we provide on other mutants that enhance or decrease neuronal activity, our findings indicate that the level of synaptic activity at single AZs is a key factor in controlling the rate of postsynaptic maturation of the glutamate receptor field.

Measuring calcium influx at AZs: “A very likely alternative explanation for the results shown in Figures 5 and 6 is that a similar concentration of calcium is experienced at each GCaMP-BRP puncta following the 10Hz stimulation, and that the abundance of the GCaMP reporter itself varies according to BRP/AZ size. I don't think this is major point that detracts from the overall study, as higher BRP/Cac size/abundance almost certainly correlates with elevated calcium influx at individual active zones. But this section needs to be thoroughly revised to take into account this major possibility and the many caveats that the authors need to consider in their attempt to measure and interpret calcium influx at single active zones.”

As the reviewer correctly points out, differences in BRP levels at individual AZs may affect our measurement. Although the use of ratiometric GCaMP-mcherry sensors for this work would be great, it is not feasible for our Pr mapping as we need the red channel for postsynaptic Ca²⁺ measurements to assign synaptic vesicle fusion events to individual AZs. As such, we have reworked this section to highlight this important caveat noted by the reviewer. We note that the GCaMP-BRP sensor data is used to support our conclusions drawn from the Cac localization studies and Pr mapping.

In addition, we experimentally addressed the reviewers comment as well. The nc82 antibody recognizes an epitope at the C-terminus of BRP and thus also recognizes BRP^short. Therefore, we were unable to stain for endogenous BRP in this line without also labeling GCaMP-BRP^short. We therefore approached this question using a different method; we applied the Ca²⁺ ionophore ionomycin to uniformly elevate intracellular Ca²⁺ throughout the presynaptic terminal. We then measured the heterogeneity in fluorescence from GCaMP-BRP^short before applying ionomycin (baseline fluorescence), during 10 Hz stimulation, and after ionomycin application. In the presence of ionomycin, differences in fluorescence signals between AZs should be entirely due to heterogeneity in sensor abundance. We observed a rightward shift in the GCaMP-BRP^short intensity distribution among AZs upon ionomycin application compared to 10 Hz stimulation (Figure 5—figure supplement 1A, B), indicating that detection of Ca²⁺ by GCaMP-BRP^short fluorescence during 10 Hz stimulation is not limited by sensor abundance. Furthermore, we observed a significant difference in the shape of the distribution during 10 Hz stimulation compared to both before stimulation and after ionomycin. The distribution of fluorescence intensities is narrower both at rest and upon ionomycin application; these two distributions should primarily reflect sensor distribution. In contrast, the distribution of GCaMPBRP^short fluorescence upon 10 Hz stimulation is wider, indicating that the sensor is reporting local changes in Ca²⁺ influx and not just sensor distribution (Figure 5—figure supplement 1C, D). Thus, although GCaMP-BRP^short abundance is likely to contribute to the levels of Ca²⁺ influx detected, these results are consistent with heterogeneity in Ca²⁺ influx across individual AZs. We have modified the text to clearly explain this caveat and have added a new supplemental figure for the ionomycin experiment (Figure 5—figure supplement 1).

We also added another experimental approach to further strengthen the connection between Pr and Ca²⁺ influx. We reasoned that if Ca²⁺ influx at AZs, and not just Cac abundance, is a driving factor in release heterogeneity, then specifically reducing the conductance of Cac should reduce Pr across the entire range of release, effectively shifting the distribution towards lower Pr but maintaining the overall shape of the distribution. As such, we performed Pr mapping in the cac^NT27 mutant, which has a point mutation in the S4 voltage sensor that reduces Ca²⁺ influx and observed the expected change. These data provide functional validation that Ca²⁺ levels are important for overall Pr and are not simply correlated (not that this is a controversial assertion given the large number of studies demonstrating this point).

Reviewer 2 found the developmental analysis fascinating and suggested some valuable additional experiments and some clarifying modifications to the text.

1a) Correlation between pre- and post-synaptic calcium events. In evoked events, presynaptic calcium influx at the AZ is required for synaptic vesicle fusion, and AZ calcium transients are necessary (but not sufficient) for EPSP, therefore, one would expect a stronger correlation between pre- and postsynaptic calcium events (determined in Figure 6) than that between calcium channel localization/abundance and postsynaptic calcium events (Figure 4). However, the authors observed almost opposite results where postsynaptic calcium events seem to correlate better with channel localization/abundance (r=0.62 Figure 4B) than with calcium influx (r=0.56, Figure 6B). The authors didn't offer any interpretation.

The results represent correlations over a very large number of AZs across multiple animals. Although it is unclear why the correlation with Ca²⁺ influx is slightly lower than with Ca²⁺ channel density, it may be reflective of the fact that we have to measure Ca²⁺ influx during a short train of action potentials rather than with the single stimuli we use to measure Pr. As such, the Ca²⁺ influx measurements are not direct assays for single action potential Ca²⁺ transients, and certainly reflect summation of signals over the short train. However, they do support the model that AZs with higher Ca²⁺ channel density and overall higher Ca²⁺ influx are likely to be higher Pr synapses. As described in reviewer 1 comment 3, we also added new functional data on the effect of a Cac mutant and its ability to support release as a mechanism to strengthen the model.

1b) Clarifying correlation vs. causation in release. The conclusion that "Ca²⁺ influx is the key factor that regulates evoked release at individual AZs" is not fully supported, as the results support a "correlation" between Ca²⁺ channel abundance/Ca²⁺ influx and P_r.

We agree with the reviewer that causation cannot be inferred directly from the correlation between Cac abundance/Ca²⁺ influx and Pr. For example, Ca²⁺ influx could vary between AZs without serving as the limiting reagent for vesicle fusion. In response to this concern, we performed an additional genetic manipulation to test whether Ca²⁺ influx through presynaptic Ca²⁺ channels, independent of the structural presence of the channel, is truly a causative factor in evoked release at Drosophila NMJs. When we mapped release probability in cac^NT27 mutants, which have lower conductance due to a point mutation in the S4 voltage sensor, we observed a significant downward shift in the distribution of Pr at the NMJ (Figure 6—figure supplement 1). This suggests that amount of Ca²⁺ influx through Ca²⁺ channels, rather than the mere presence of the channels, directly effects release at AZs across the release spectrum.

1c) The regulator of presynaptic calcium channel abundance. The actual "regulator", which determines the presynaptic Ca²⁺ channel abundance and local Ca²⁺ influx, still remains unknown.

Through our developmental analysis of AZ maturation using intravital imaging, we have shown that one key primary regulator of presynaptic Ca²⁺ channel abundance is age. Ca²⁺ channels accumulate at AZs throughout their maturation process. We demonstrated that AZs add Cac throughout their development by following Cac-GFP and GluRIIA-RFP over time. Unlike the Cac-GFP area, the GluRIIA-RFP field is large enough to be reliably resolved by light microscopy, and we observed that the size and intensity of the GluRIIA field increases over time as AZs mature. Moreover, we saw a strong correlation between Cac-GFP intensity and GluRIIA intensity, suggesting these two components are tightly coupled during synapse maturation. Based on this information, we suggest that Cac abundance at AZs increases during maturation and that AZ age is a major factor in determining AZ Ca²⁺ channel abundance. There are likely to be many interesting mechanisms further regulating the rate of Ca²⁺ channel accumulation and persistence at AZs, as well as the regulation of Ca²⁺ channel gating at individual AZs, and these are topics of interest moving forward.

1d) Clarifying expression of transgenes. “It is unclear how the transgenes are expressed separately in the neuron and the muscle. The methods listed both transgenes, UAS-CactdTomato, UAS-GCaMP, but no indication of GAL4 drivers used to express both either together or separately.”

We thank this reviewer for pointing out the lack of clarity in our methods of transgene expression. In this study, all presynaptically driven transgenes (UAS-Cac-tdTomato, UAS-CacGFP, and UAS-GCaMP-BRP^short) were expressed using the pan-neuronal driver elav-GAL4. Postsynaptically, UAS-myr-GCaMP6s was driven with the muscle driver mef2-GAL4 when coexpressed with GluRIIA-RFP, which was inserted on chromosome III and driven using its endogenous promoter. Myr-GCamP6s was driven using the LexA/LexOP system in muscle when combined with GluRIIA-RFP and GluRIIB-GFP which were both inserted on chromosome III and driven using their endogenous promoters. This has been clarified throughout the text.

Developmental acquisition of P_r property. The finding of developmental clustering of GluRII is fascinating. If the birth order of AZs determines the level of maturation and the accumulation of GluRII, then the birth order also determines P_r and its heterogeneity… It needs to be determined whether the presynaptic terminal influences/contributes to the postsynaptic GluRII clustering. The authors indicated feasibility issues with P_r map, however, measuring developmental/birth order clustering of Cac should be feasible, since Cac abundance is correlated with P_r. In addition, the developmental clustering of Calcium channels and GluRII should be done in syt mutant background to determine the effect of transmission (evoked or spontaneous)."

This was an excellent set of suggestions, and we performed an extensive number of experiments to address these questions; this new addition enhanced the novelty of this work and we are grateful to the reviewer for contributing these ideas. We addressed the question of whether differences in presynaptic activity levels can alter the maturation rate of synapses at the level of individual AZs. Instead of monitoring Cac-GFP and GluRIIA-RFP over time like the reviewer suggested, we chose to perform these experiments using GluRIIA-RFP and GluRIIB-GFP for two reasons. First, GluRIIA and GluRIIB co-labeling allows us to identify PSDs that display receptor subtype segregation, which is a useful marker for the progression of PSDs through maturation. Second, Cac-GFP puncta are smaller than the resolution limit of conventional light microscopy, and so all quantification through the cuticle would rely on comparisons in Cac puncta intensity that could be affected by bleaching and by varying cuticle thickness.

We began by measuring the rate of PSD growth (24-hour fold-increase in GluRIIB area) and maturation (percent of PSDs displaying GluRIIA/GluRIIB segregation) in BRP^69/def mutants that (1) alter presynaptic structure and (2) decrease release by reducing Cac abundance at AZs. In this mutant we observed a significant reduction in both PSD growth and ring formation. In syt1^null, and nap^TS mutants that reduce presynaptic release without structural changes to the AZ, we also observed a significant reduction in postsynaptic maturation rate (Figure 10). In the shaker, eag double mutant with increased presynaptic excitability, we observed a significantly increased rate of GluRIIB field size increase when compared to control, and a significant increase in GluRIIB rings at the early second instar stage (Figure 10). Finally, we used the rab3 null mutant to investigate whether differences in PSD growth and maturation rate could occur between AZs containing BRP and Cac versus neighboring AZs within the same NMJ that are deficient in these components. In the first instar stage when AZs are roughly age matched, we saw that release sites enriched in presynaptic components had developed large and mature PSDs in stark comparison with un-enriched AZs, whose PSDs were underdeveloped (Figure 10, supplementary figure 1). Overall, these results strongly suggest that presynaptic activity drives postsynaptic maturation rate at an AZ-specific level and opens many exciting questions for future work.

NMJ specificity. Majority of the analysis focuses on the NMJ formed only by MN4-Ib onto muscle 4 at one time point (early L3 larvae). It will be valuable to at least include another NMJ, and include a different development stage (eg, late L3 larve) for the P_r map (like Figure 1A-C).

We agree that Prmapping at a different developmental stage and another NMJ are valuable additions to this study, even beyond their ability to determine how generalizable our observations of release heterogeneity are across synapses and across developmental stages. Mapping release at muscle 26 serves to connect the work we have done building correlations between release and synapse component density in muscle 4 with the work performed following AZ maturation over development at muscle 26. We were pleased to see that when we mapped release at muscle 26 in the third instar stage, the distribution in Prwas indistinguishable from the distribution seen at the NMJ on muscle 4 (Figure 9—figure supplement 1). This suggests that our findings in these two NMJs are indeed comparable, and that Pr heterogeneity is not specific to muscle 4. Mapping release in the second instar provides another way to test our model that AZ age and developmental maturation are key determinants of Pr heterogeneity during NMJ development. Based on our observation that the number of AZs roughly doubles every 24 hours of development, and that AZs are born weak and undergo a multi-day maturation process where they develop high Pr, we would expect the second instar NMJ to show a shift in Pr distribution when compared to the third instar NMJ. Specifically, we would expect a lower percentage of low Pr sites in second instar stage (as the number of AZs will double over the next 24 hours and the newly added AZs will be low Pr). Indeed, when we mapped Pr in the second instar stage, we observed the expected shift in the distribution ((Figure 9—figure supplement 1)), providing support for the model that developmental maturation and new AZ addition over time generates heterogeneity at the NMJ.

Reviewer 3 indicated the manuscript was technically done very well but wanted clarification on the biological significance of the findings and the novelty of the current data.

1a) Describing biological significance.

We thank this reviewer for pointing out the lack of discussion on the novelty of this work, and we have altered the text significantly to further contextualize these results. Though it has already been shown that Pr heterogeneity exists across the NMJ (Peled 2011, Melom 2013) and that Pr correlates with levels of BRP (Reddy-Alla et al., 2017, Muhammad et al., 2015, Peled et al., 2014), the work presented here extends this body of knowledge in multiple important ways.

Combining our green Ca²⁺ sensor with GluRIIA-RFP to mark individual release sites has allowed us to provide a higher resolution view of the correlation between single AZ evoked release and BRP, Cac, and GluRIIA abundance. GluRIIA-RFP also allowed us to employ superresolution structured illumination microscopy after quantal imaging to confirm that high releasing sites correspond to single AZs. This increase in resolution is essential for understanding release heterogeneity at the NMJ, where single BRP T-bars are smaller than the resolution of conventional light microscopy and can be closely clustered.

We have assessed the role of local synaptic vesicle pools in determining local release heterogeneity. Though it has been established that certain synaptic vesicle components (like Syt1) are dominant factors in determining Pr at synapses, the role of local synaptic vesicle pools in determining release heterogeneity across AZs of the NMJ has not been explored. By showing that heterogeneity in release is retained in syt1 null mutants, and is stable during extensive synaptic vesicle cycling, the study indicates that AZ-specific, differentially fusogenic synaptic vesicle populations are unlikely to underlie the heterogeneity in release observed at the NMJ.

We have shown that AZ maturation throughout the three stages of larval development plays a dominant role in generating the extreme release heterogeneity that we and other labs have observed in the third instar larval stage. The Sigrist lab has previously demonstrated that AZs gain presynaptic and postsynaptic components over time, generating the hypothesis that AZ aging process may lead to increasing AZ Pr. However, to demonstrate that the heterogeneity observed at the third instar stage results largely from AZ birth date, we needed to extend this intravital imaging of individual release sites through a longer time period starting much earlier in development to show that the heterogeneity in developmental birth order is a major source of heterogeneity in AZ release probability at the third instar stage. This long-term developmental perspective, extending from early first to third instar stages, provides an important developmental picture of how release heterogeneity arises. Since the field typically uses the third instar NMJ as a model to study presynaptic release, understanding how functional release heterogeneity seen at the third instar NMJ is established from a molecular and developmental perspective provides valuable information about this common model synapse.

We have included additional experiments to show that postsynaptic maturation rate is activity-dependent and can occur at an active zone-specific level (this is discussed below). To our knowledge, this has not been previously demonstrated at the NMJ. Overall, these results provide a more developmentally complete and technically higher resolution understanding of how release heterogeneity at the NMJ arises during development.

1b) Furthering biological significance through more genetic analysis. “Further genetic analysis could be used to delineate exactly how the different components (BRP, Cacophony, GlurII etc.) may affect P_r and. whether these proteins play independent or dependent roles in controlling the heterogeneity of P_r across AZ… The authors could investigate how the various BRP alleles affect the P_r.The BRPnude allele does not seem to affect vesicle release but does affect synaptic depression. A conditional reversible knockdown of BRP has been documented (Nagarkar-Jaiswall et al., 2015). This allows one to control Brp levels."

This reviewer encouraged us to move beyond building correlations between AZ components and release probability in wildtype animals and to begin using genetic manipulations to further understand the generation of heterogeneous Pr at the NMJ. Since we have shown that AZ age is a strong contributor to the abundance of pre- and postsynaptic proteins that our lab and others have shown to correlate with Pr heterogeneity at the NMJ, we added an entire set of new experiments to investigate what parameters affect age-dependency of Pr. First, we established confidence in the age-dependency of Pr using two new experiments; we showed that all new active zones (under 24 hours old) are low Pr (Figure 9F, G). Second, we mapped Pr in the second instar stage and showed that, in line with what our model predicts, the heterogeneity in release in the second instar is shifted towards higher releasing sites, with a reduction in low-releasing sites (Figure 9—figure supplement 1).

After further documenting the age-dependence of Pr, we addressed the concern that further genetic manipulations could be used to determine how overall component accumulation during development could shape Pr distribution by assaying whether mutants that alter presynaptic activity can influence synapse maturation rate (Figure 10). We measured the rate of PSD growth (24-hour fold-increase in GluRIIB area) and GluRIIA/B segregation in mutants with altered presynaptic activity. In the BRP^69/def, syt1^null, and nap^TS mutants with decreased presynaptic release, we observed a significant reduction in postsynaptic maturation rate and a significant reduction in GluRIIA/IIB rings (Figure 10). In the shaker, eag double mutant with increased presynaptic excitability, we observed a significantly increased rate of GluRIIB field size increase compared to control, and a significant increase in GluRIIB rings at the early second instar stage (Figure 10). Finally, we used the rab3 null mutant to investigate whether differences in PSD growth and maturation rate could be seen between AZs containing BRP and Cac vs. neighboring AZs within the same NMJ that are deficient in these components. In the first instar stage when AZs are roughly age matched, we observed that release sites enriched in presynaptic components had developed large and mature PSDs in stark comparison with un-enriched AZs, whose PSDs were underdeveloped (Figure 10—figure supplement 1).

Overall, these results add valuable new information to the study; we have shown that not only is the heterogeneity in AZ Pr explained by differences in AZ component density that reflect differences in AZ age, but we have also shown that this dependence on age can be influenced by changes in presynaptic activity.

1c) Functional demonstration that calcium influx determines P_r. “Does the reduction in Cacophony levels by RNAi or mutants change the heterogeneity of P_r? would it be possible to use a calcium channel mutant with reduced activity to specifically differentiate whether the presence of the calcium channel is enough to define the high P_r or the influx of calcium is needed?"

This is an excellent suggestion. As another approach to test whether the level of Ca²⁺ influx rather than the structural presence of the Ca²⁺ channel is responsible for determining Pr, we generated Pr maps in the cac^NT27 mutant using dual color quantal imaging, with GluRIIA-RFP and myrGCamP6s expressed postsynaptically in the muscle. Cac^NT27 channels have reduced Ca²⁺ conductance due to a point mutation in the Cac S4 voltage sensor (Rieckhof et al. 2003). We observed that cac^NT27 results in a global decrease in Pr across AZs; WT evoked Pr ranged from 0 to 0.73 with an average of 0.073, while Cac^NT2 Pr ranged from 0 to 0.47 with a significantly lower average of 0.049 (Figure 6—figure supplement 1). In this mutant, the entire Pr distribution was shifted towards low Pr when compared to controls, and the resulting NMJ was almost entirely populated with low-releasing AZs. This result strongly suggests that the levels of Ca²⁺ influx through Cac channels, and not the physical presence of the channels, is a primary determinant of Pr at AZs.

1d) Differential short term plasticity across AZs “Furthermore, How does this developmentally controlled heterogeneity compare to synaptic plasticity? Are the identified high P_r sites the most prone to synaptic depression? Are the low P_r sites preferentially recruited to induce synaptic potentiation?”

This is a compelling question, and the Drosophila NMJ provides an excellent model system to dissect the short-term plasticity at low vs. high Pr sites. These questions have been previously addressed by Peled and Isacoff (2011); here, the authors demonstrated that high releasing AZs tend to depress during paired pulse stimulation, while lower releasing AZs facilitate. Thus, the developmentally generated heterogeneity in Pr across AZs results in a population of AZs with distinct short-term plasticity properties. We have included a description of this result in the text.

1e) In the last section of the Results section, the authors correlate the number of mature AZs at the larval NMJ seen by in-vivo imaging with the number of high P_r AZs observed in the previous P_r mapping. They must avoid concluding that because both numbers are similar (14.7% and 10% respectively), it supports their hypothesis about the link between AZ maturation and high P_r. These are very different experiments (in-vivo imaging vs. larval preparations) and are also conducted at different NMJs (Muscle 4 vs. Muscle 26). The authors should follow up their in-vivo imaging with calcium imaging on larval preps to show that the mature AZ do in fact have high P_r.

This is an excellent suggestion, and we have altered the text and performed two additional experiments to address this point. First, as the reviewer pointed out, it is not ideal to compare release distributions at muscle 4 versus muscle 26 without directly demonstrating that these two NMJs have a similar distribution of Pr. To address this issue, we mapped release at muscle 26 using the same methods that were used to map release in our initial experiments at muscle 4. When we compared the release distributions at the two NMJs, we saw that the distributions were not significantly different (Figure 9—figure supplement 1). Second, we followed the suggestion of this reviewer to directly assess whether AZ age corresponds to Pr by mapping Pr after determining the developmental age of AZs using intravital imaging. We followed PSDs in animals expressing GluRIIA-RFP, GluRIIB-GFP, and myr-GCaMP for 24 hours using in vivo imaging, and then dissected the animal and mapped Pr (Figure 9F). We observed that PSDs that appeared on the second day of imaging (less than 24 hours old) were consistently associated with very low-Pr AZs. The mean Prof newly-formed AZs was 0.035, and the Pr for AZs older than 24 hours was 0.15 (Figure 9G). This result supports a model in which a vast majority of AZs are born weak, and higher Pr takes more than 24 hours to develop.