The E3 ligase Highwire promotes synaptic transmission by targeting the NAD‐synthesizing enzyme dNmnat

Abstract

The ubiquitin ligase Highwire restrains synaptic growth and promotes evoked neurotransmission at NMJ synapses in Drosophila. Highwire regulates synaptic morphology by downregulating the MAP3K Wallenda, but excess Wallenda signaling does not account for the decreased presynaptic release observed in highwire mutants. Hence, Highwire likely has a second substrate that inhibits neurotransmission. Highwire targets the NAD⁺ biosynthetic and axoprotective enzyme dNmnat to regulate axonal injury responses. dNmnat localizes to synapses and interacts with the active zone protein Bruchpilot, leading us to hypothesize that Highwire promotes evoked release by downregulating dNmnat. Here, we show that excess dNmnat is necessary in highwire mutants and sufficient in wild‐type larvae to reduce quantal content, likely via disruption of active zone ultrastructure. Catalytically active dNmnat is required to drive defects in evoked release, and depletion of a second NAD⁺ synthesizing enzyme is sufficient to suppress these defects in highwire mutants, suggesting that excess NAD⁺ biosynthesis is the mechanism inhibiting neurotransmission. Thus, Highwire downregulates dNmnat to promote evoked synaptic release, suggesting that Highwire balances the axoprotective and synapse‐inhibitory functions of dNmnat.

Introduction

The E3 ubiquitin ligase PHR is a central regulator of neural circuit development, function, and maintenance with conserved activities in worms, flies, and mammals 1, 2. PHR is a huge, multidomain protein, but its best understood functions are to promote the turnover of target proteins via proteasomal degradation 2, 3, 4. In Drosophila, the PHR ortholog Highwire (Hiw) is a key regulator of neuromuscular junction (NMJ) morphology and physiology 5, 6, 7, 8. The NMJs of hiw mutant larvae are massively overgrown relative to wild type (WT) with a dramatic increase in synaptic terminal branching and bouton number 5, 8. In addition, hiw mutant NMJs exhibit reduced synaptic strength due to a decrease in quantal content, the number of vesicles released following an action potential 7, 8. We wished to identify the functionally relevant protein targets of Highwire to gain insight into the molecular mechanisms controlling synaptic morphology and function.

We identified the mitogen‐activated protein kinase kinase kinase (MAP3K) Wallenda (Wnd), the Drosophila ortholog of Dual leucine zipper kinase (DLK), as the Highwire target responsible for synaptic terminal overgrowth in highwire mutants 8. Wallenda protein levels are increased in hiw mutants, wnd is necessary for the synaptic terminal overgrowth in hiw mutants, and overexpression of wnd is sufficient to phenocopy this overgrowth in otherwise wild‐type larvae. To our surprise, however, Wallenda is not the Highwire target responsible for the defect in evoked transmitter release in hiw mutants 8. Even though hiw;wnd double mutants have NMJs that are morphologically indistinguishable from wild type, these double mutants are still defective in evoked synaptic transmission 8, 9, implying that Hiw must regulate a second substrate to promote synaptic release.

While Highwire and its orthologs were first studied due to their effects on neural circuit development and function 4, more recently it was discovered that Highwire is also a key determinant of axonal survival following injury 10. In the absence of Highwire or its vertebrate ortholog Phr1, Wallerian degeneration of injured axons is dramatically delayed 10, 11. For this function, Highwire targets nicotinamide mononucleotide adenyltransferase (dNmnat), an NAD⁺ biosynthetic enzyme which, along with its mammalian orthologs, is a potent axonal maintenance factor 10, 11, 12, 13. In addition, dNmnat also promotes synaptic maintenance, acting as a chaperone for the active zone scaffolding protein Bruchpilot (Brp) 14. While it is clear that dNmnat is necessary to maintain axons and synapses, we wondered whether the elevation in dNmnat levels in hiw mutants, which is apparent in both the synaptic terminal and axon 10, could also impact the synapse, inducing the defects in synaptic release in highwire mutants.

To investigate whether dNmnat could be the Highwire target inhibiting evoked release, we first demonstrated that excess dNmnat is sufficient to impair evoked synaptic transmission. Moreover, dNmnat is necessary for the defective evoked release in hiw mutants. Downregulation of dNmnat in the hiw mutant fully suppresses defects in evoked release, but has no impact on NMJ terminal morphology. This excess dNmnat leads to a decrease in release probability and disrupts the architecture of T‐bars at active zones. In addition, depletion of NAD⁺ Synthetase, the subsequent enzyme in the NAD⁺ biosynthesis pathway, also suppresses the defect in evoked release in hiw mutants. These findings support the model that excess NAD⁺ biosynthetic activity impairs evoked synaptic transmission, and suggests that Highwire locally tunes levels of dNmnat protein, and therefore local NAD⁺ levels, to promote efficient synaptic transmission. These findings identify an unexpected activity of dNmnat in the inhibition of evoked synaptic release and suggest that Highwire controls dNmnat levels to balance its promotion of axonal maintenance and inhibition of synaptic transmission.

Results

Excess presynaptic dNmnat is sufficient to reduce evoked release

Hiw mutants exhibit decreased synaptic strength at the larval NMJ; however, the substrate regulated by Hiw to promote synaptic transmission remains unknown. Given that the NAD⁺ synthesizing enzyme dNmnat is a known target for degradation by the E3 ligase Hiw, and that in hiw mutants there is an excess of dNmnat protein throughout the nervous system including at the synaptic terminal 10, we hypothesized that increased dNmnat protein levels decrease synaptic transmission. To begin probing this hypothesis, we first asked whether an excess of neuronal dNmnat is sufficient to induce a decrease in synaptic strength. We reasoned that if excess dNmnat protein in the hiw mutant synaptic terminal is the primary driver of diminished synaptic transmission, then overexpression of dNmnat protein in an otherwise wild‐type background could phenocopy mutations in hiw and cause a decrease in evoked release. To test this, we overexpressed a dNmnat transgene (UAS‐dNmnat) via the glutamatergic VGlut‐GAL4 driver 15 to express dNmnat in motor neurons and measured spontaneous and evoked synaptic transmission at the larval NMJ. Overexpression of dNmnat induces a decrease in EPSP amplitude, the postsynaptic response to the evoked release of many synaptic vesicles, that is not seen upon overexpression of a control transgene (UAS‐RFP; Fig 1A and B). To determine whether there was a postsynaptic contribution to the decrease in EPSP amplitude, we measured mEPSP amplitudes, the postsynaptic response to the transmitter content of single vesicles and therefore a measure of postsynaptic receptor responsivity to neurotransmitter. Expression of dNmnat did not lead to a significant difference in mEPSP amplitude (Fig 1A for representative traces, 1C). The quantal content, or average number of synaptic vesicles released upon an action potential, is calculated by dividing mean EPSP amplitude by mEPSP amplitude and demonstrates a significant decrease (Fig 1B). Hence, neuronal overexpression of dNmnat can impair presynaptic transmitter release. Overexpression of the MAP3K Wnd, known to be the Hiw substrate that promotes presynaptic terminal growth 8, did not elicit defects in quantal content. While EPSP amplitudes were decreased relative to controls (Fig 1A and B), this decrease was driven by a substantial decrease in mEPSP amplitudes (Fig 1A and C), leaving quantal content unchanged (Fig 1D). These results are consistent with recent work demonstrating that excess presynaptic Wnd drives reduced postsynaptic responsivity to transmitter 9, but does not decrease neurotransmitter release from the presynaptic motor neuron.

Figure 1. Excess presynaptic dNmnat is sufficient to depress evoked release without altering NMJ growth.

1. Representative EPSP and mEPSP electrophysiological traces from larval NMJs at muscle 6 in which a motoneuron‐specific driver was used to overexpress either a control transgene (DVGlut‐GAL4 > UAS‐RFP), dNmnat (DVGlut‐Gal4 > UAS‐dNmnat), or Wnd (DVGlut‐Gal4 > UAS‐Wnd).
2. Quantification of mean (± SEM) EPSP amplitudes, in which 75 consecutive evoked events were averaged per cell, and then, cell amplitudes were averaged per genotype [UAS‐RFP n = 13 cells, UAS‐dNmnat n = 14 cells, UAS‐Wnd n = 10 cells. One‐way ANOVA with Tukey's multiple comparisons, DF = 36, F = 12.66, P < 0.0001. UAS‐RFP vs. UAS‐dNmnat P < 0.0001 (****), UAS‐RFP vs. UAS‐Wnd P = 0.0093 (**)].
3. Quantification of mean (± SEM) mEPSP amplitudes, in which 75 consecutive spontaneous events were averaged per cell, and then, cell amplitudes were averaged per genotype [UAS‐RFP n = 15 cells, UAS‐dNmnat n = 14 cells, UAS‐Wnd n = 10 cells. One‐way ANOVA with Tukey's multiple comparisons, DF = 38, F = 11.66, P = 0.0001. UAS‐RFP vs. UAS‐dNmnat P = 0.994 (NS), UAS‐RFP vs. UAS‐Wnd P = 0.0003 (***)].
4. Quantification (mean ± SEM) of quantal content, which was calculated individually per cell by dividing the mean EPSP amplitude by the mean mEPSP amplitude and then averaged per genotype [UAS‐RFP n = 15 cells, UAS‐dNmnat n = 14 cells, UAS‐Wnd n = 10 cells. One‐way ANOVA with Tukey's multiple comparisons, DF = 38, F = 9.286, P = 0.0006. UAS‐RFP vs. UAS‐dNmnat P = 0.0015 (**), UAS‐RFP vs. UAS‐Wnd P = 0.9892 (NS).
5. Representative images of the NMJ synaptic terminal at muscle 4 in third‐instar larvae. The genotypes in this experiment were identical to those tested in the electrophysiology experiments in (A–D). NMJs were stained for the presynaptic bouton marker DVGLUT (green) and nerve membrane marker HRP (red).
6. Quantification of the mean (± SEM) number of DVGlut⁺ boutons per muscle 4 NMJ in each genotype. [UAS‐RFP n = 23, UAS‐dNmnat n = 21, UAS‐Wnd n = 20 NMJs. One‐way ANOVA with Tukey's multiple comparisons, DF = 63, F = 68.31, P < 0.0001. UAS‐RFP vs. UAS‐dNmnat P = 0.9818 (NS), UAS‐RFP vs. UAS‐Wnd P < 0.0001 (****)].

Given the effect of excess dNmnat on the function of NMJ synapses, we next investigated whether NMJ terminal morphology was impacted by overexpressing dNmnat. Synaptic terminal growth, as measured by counting the number Drosophila vesicular glutamate transporter (DVGlut)‐positive synaptic boutons localized within HRP‐positive terminal membrane 7, 8, was unchanged upon dNmnat overexpression (Fig 1E and F). In contrast, overexpressing Wnd in glutamatergic neurons was sufficient to drive synaptic overgrowth at the NMJ, consistent with previous findings (Fig 1E and F) 8, 9. Hence, overexpression of two distinct Hiw substrates in a wild‐type background can each phenocopy one aspect of the hiw mutant phenotype—excess neuronal dNmnat impairs synaptic release while excess neuronal Wnd promotes synaptic terminal overgrowth.

Evoked release defects in Hiw mutants are dependent on presynaptic dNmnat

Having demonstrated that excess dNmnat is sufficient to reduce presynaptic release, we next investigated whether dNmnat is necessary for the observed decrease in evoked release in a hiw mutant background. We genetically depleted dNmnat in a hiw mutant, using the strong hypomorphic allele hiw ^ND8 5, and measured synaptic transmission at the NMJ. dNmnat is an essential gene and mutants are embryonic lethal 16, so to decrease levels of dNmnat we used the pan‐neuronal driver Elav to express transgenic RNAis against dNmnat. This approach limits knockdown to the neuron while leaving dNmnat function intact in the postsynaptic muscle and other non‐neuronal cells. As previously published, hiw mutant NMJs have decreased synaptic strength relative to wild‐type NMJs, as measured by decreased EPSP amplitudes and a decrease in quantal content (Fig 2A–C, 7, 8). We confirmed previously published data that presynaptic depletion of Wnd via a transgenic RNAi is unable to suppress defective evoked release in a hiw mutant (Fig 2A–C), despite suppressing synaptic terminal morphology and bouton number (Fig 2D and E). Depletion of dNmnat in a hiw mutant background, however, fully suppresses the decrease in both EPSP amplitude and quantal content (Fig 2A–C). mEPSP amplitudes were not significantly different among all genotypes [Fig 2A for representative traces, mean (± SEM) mEPSP amplitudes for 75 consecutive spontaneous events per cell, averaged across genotype—WT n = 10 cells, mean amplitude = 0.7424 mV; hiw n = 10 cells, mean amplitude = 0.6329 mV; hiw > Wnd‐RNAi n = 10 cells, mean amplitude = 0.7685 mV; and hiw > dNmnat‐RNAi n = 10 cells, mean amplitude = 0.6709 mV. One‐way ANOVA with Tukey's multiple comparisons, DF = 39, F = 2.454, P = 0.0789. WT vs. hiw P = 0.2397 (NS), WT vs. hiw > Wnd‐RNAi P = 0.9670 (NS), WT vs. hiw > dNmnat‐RNAi P = 0.5914 (NS)]. We repeated the experiment in a second hiw mutant background (hiw ^∆N , which is a null allele, 7) and found that a second non‐overlapping RNAi line targeting dNmnat partially suppressed decreased EPSP amplitudes and completely suppressed the decrease in quantal content in the hiw null background (Fig EV1A–D).

Figure 2. Excess presynaptic dNmnat drives defects in evoked neurotransmission at the NMJ in Hiw mutants, but has no role in synaptic overgrowth.

1. Representative EPSP and mEPSP physiological traces from larval NMJs at muscle 6. Recordings were taken from WT control larvae (Elav‐Gal4 > UAS‐RFP), hiw mutants driving a control transgene (hiw ^ND8;Elav‐Gal4 > UAS‐RFP), and hiw mutants expressing RNAis to knockdown either dNmnat (hiw ^ND8;Elav‐Gal4 > UAS‐dNmnat‐RNAi [#29402]) or Wnd (hiw ^ND8;Elav‐Gal4 > UAS‐Wnd‐RNAi [#25396]).
2. Quantification of mean (± SEM) EPSP amplitudes, in which 75 consecutive evoked events were averaged per cell, and then, cell amplitudes were averaged per genotype [WT n = 10 cells, hiw n = 10 cells, hiw > Wnd‐RNAi n = 10 cells, hiw > dNmnat‐RNAi n = 10 cells. One‐way ANOVA with Tukey's multiple comparisons, DF = 39, F = 14.02, P < 0.0001. WT vs. hiw P = 0.0012 (**), WT vs. hiw > Wnd‐RNAi P < 0.0001 (****), WT vs. hiw > dNmnat‐RNAi P = 0.9880 (NS)].
3. Quantification of quantal content (± SEM) [WT n = 9 cells, hiw n = 9 cells, hiw > Wnd‐RNAi n = 10 cells, hiw > dNmnat‐RNAi n = 10 cells. One‐way ANOVA with Tukey's multiple comparisons, DF = 38, F = 2.508, P < 0.0001. WT vs. hiw P = 0.0361 (*), WT vs. hiw > Wnd‐RNAi P = 0.0001 (***), WT vs. hiw > dNmnat‐RNAi P = 0.9889 (NS)].
4. Representative images of the NMJ synaptic terminal at muscle 4 in third‐instar larvae. The genotypes in this experiment were identical to those tested in panels (A–C). NMJs were stained for the presynaptic bouton marker DVGLUT (green) and nerve membrane marker HRP (red).
5. Quantification of the mean (± SEM) number of DVGlut⁺ boutons per muscle 4 NMJ in each genotype [WT n = 23, hiw n = 21, hiw > Wnd‐RNAi n = 16, hiw > dNmnat‐RNAi n = 19. One‐way ANOVA with Tukey's multiple comparisons, DF = 78, F = 169.1, P < 0.0001. WT vs. hiw P < 0.0001 (****), WT vs. hiw > Wnd‐RNAi P = 0.9988 (NS), WT vs. hiw > dNmnat‐RNAi P < 0.0001 (****), hiw vs. hiw > dNmnat‐RNAi P = 0.9695 (NS)].

Figure EV1. A second RNAi targeting dNmnat suppresses defects in evoked release in the hiw null background.

1. Representative EPSP and mEPSP physiological traces from larval NMJs at muscle 6. Recordings were taken from WT control larvae (Elav‐Gal4 > UAS‐RFP), hiw null mutants driving a control transgene (hiw ^∆N;Elav‐Gal4 > UAS‐RFP), and hiw mutants expressing an RNAi to knockdown dNmnat (hiw ^ND8;Elav‐Gal4 > UAS‐dNmnat‐RNAi [KK#107262]). This RNAi is unique from the one presented in Fig 2 and targets a non‐overlapping portion of the dNmnat gene product (VDRC KK#107262).
2. Quantification of mean (± SEM) EPSP amplitudes, in which 75 consecutive evoked events were averaged per cell, and then, cell amplitudes were averaged per genotype [control n = 7 cells, hiw ^∆N n = 6 cells, hiw ^∆N > dNmnat‐RNAi n = 8 cells. One‐way ANOVA with Tukey's multiple comparisons, DF = 20, F = 46.58, P < 0.0001. Control vs. hiw P < 0.0001 (****), control vs. hiw > dNmnat‐RNAi P = 0.0013 (**), hiw vs. hiw > dNmnat‐RNAi P < 0.0001 (****)].
3. Quantification of mean (± SEM) mEPSP amplitudes, in which 75 consecutive spontaneous events were averaged per cell, and then, cell amplitudes were averaged per genotype [control n = 8 cells, hiw ^∆N n = 6 cells, hiw ^∆N > dNmnat‐RNAi n = 9 cells. One‐way ANOVA with Tukey's multiple comparisons, DF = 22, F = 6.729, P = 0.0058. Control vs. hiw ^∆N P = 0.0042 (**), control vs. hiw ^∆N > dNmnat‐RNAi P = 0.1901 (NS), hiw ^∆N vs. hiw ^∆N > dNmnat‐RNAi P = 0.1185 (NS)].
4. Quantification (mean ± SEM) of quantal content [control n = 7, hiw ^∆N n = 6, hiw ^∆N > dNmnat‐RNAi n = 8. One‐way ANOVA with Tukey's multiple comparisons, DF = 20, F = 11.58, P = 0.0006. Control vs. hiw ^∆N P = 0.0005 (***), control vs. hiw ^∆N > dNmnat‐RNAi P = 0.4282 (NS), hiw ^∆N vs. hiw ^∆N > dNmnat‐RNAi P = 0.0057 (**).]

We next examined whether this degree of dNmnat depletion influenced basal neurotransmission in an otherwise WT background (Elav‐GAL4 > UAS‐RFP vs. Elav‐GAL4 > UAS‐dNmnat‐RNAi). We observed no significant difference in EPSP amplitudes or quantal content upon dNmnat knock down in wild‐type motor neurons compared to a control transgene (Fig EV2A–C). These findings show that in the hiw mutant, excess dNmnat impairs evoked synaptic transmission, but in wild‐type larvae, the normal level of dNmnat does not disrupt transmission. Finally, we tested whether the synaptic overgrowth observed at the hiw mutant NMJ was altered by depletion of dNmnat. As expected from the overexpression studies (Fig 1), depleting dNmnat from a hiw mutant background did not suppress the dramatic morphological overgrowth at the larval NMJ; this phenotype is entirely Wnd‐dependent and can be suppressed through presynaptic expression of an RNAi against Wnd (Fig 2D and E). Thus, we conclude that Hiw targets two unique substrates to control different aspects of synaptic development; Hiw limits levels of the MAP3K Wnd to restrain synaptic growth, while limiting levels of dNmnat to promote evoked synaptic transmission.

Figure EV2. Neither Wnd nor dNmnat depletion impact basal evoked neurotransmission in an otherwise wild‐type background.

1. Representative EPSP traces from larval NMJs at muscle 6. Recordings were taken from a wild‐type control expressing a control transgene (elav > UAS‐RFP), wild type expressing an RNAi against Wnd (elav > Wnd‐RNAi [#25396]), and wild type expressing an RNAi against dNmnat (elav > dNmnat‐RNAi [#29402]).
2. Quantification of mean (± SEM) EPSP amplitudes, in which 75 consecutive evoked events were averaged per cell, and then, cell amplitudes were averaged per genotype [elav > RFP n = 10, elav > wnd‐RNAi n = 9, elav > dNmnat‐RNAi n = 11. One‐way ANOVA with Tukey's multiple comparisons, DF = 29, F = 0.4903, P = 0.6178. RFP vs. Wnd‐RNAi P = 0.7995 (NS), RFP vs. dNmnat‐RNAi P = 0.5992 (NS).
3. Quantification (mean ± SEM) of quantal content [elav > RFP n = 10, elav > wnd‐RNAi n = 9, elav > dNmnat‐RNAi n = 11. One‐way ANOVA with Tukey's multiple comparisons, DF = 29, F = 0.715, P = 0.4982. RFP vs. Wnd‐RNAi P = 0.4735 (NS), RFP vs. dNmnat‐RNAi P = 0.9016 (NS)].

Excess dNmnat decreases release probability in hiw mutants

We next probed the physiological mechanisms by which excess dNmnat impairs synaptic release in hiw mutants. Presynaptic strength is largely defined by two key parameters: N, which is the number of synaptic release sites, and P _r, the probability of release from each site. We first explored the impact of excess dNmnat on N at the hiw mutant NMJ. At the Drosophila NMJ, active zones are marked by the presence of the scaffolding protein Bruchpilot and are apposed to clusters of postsynaptic glutamate receptors17, 18. To estimate N, we defined a “release site” as a Brp‐positive puncta in close apposition to DGluRIII, an essential subunit of the glutamate receptor 19. We assessed Brp and DGluRIII at NMJs of wild‐type controls expressing RFP in neurons, hiw ^ND8 mutants expressing RFP in neurons, and hiw ^ND8 mutants expressing dNmnat‐RNAi in neurons. As expected, hiw mutant NMJs have an expanded synaptic surface area relative to control, and dNmnat knockdown does not suppress this morphological phenotype (Fig 3A and B). Interestingly, hiw mutant synaptic terminals exhibited no significant differences in the total number of Brp puncta per NMJ compared to controls (Fig 3A and C) and dNmnat knockdown did not increase the number of release sites in the hiw mutant background (Fig 3A and C). Further, neither hiw mutants nor hiw mutants in which dNmnat is knocked down exhibit a substantial misapposition defect; nearly all Brp puncta in the presynaptic terminal are apposed to DGluRIII clusters, with no significant difference from WT terminals (Fig 3A and D). Hence, we do not observe a significant change in N in hiw with or without dNmnat knockdown, suggesting that excess dNmnat is not impairing release by decreasing the total number of functional release sites.

Figure 3. hiw mutants exhibit dNmnat‐dependent defects in release probability (P _r) at the larval NMJ.

1. Representative images of immunostained larval NMJs and active zones from WT, hiw mutant, and hiw mutant driving an RNAi against dNmnat [#29402]. In the top three panels, grayscale images of entire terminal were acquired by staining for nerve marker HRP. Presynaptic Brp puncta (green) and postsynaptic dGluRIII clusters (magenta) were imaged to quantify N in each genotype.
2. Quantification (mean ± SEM) of NMJ synaptic terminal surface area in each genotype [WT n = 14, hiw n = 12, hiw > dNmnat‐RNAi n = 12. One‐way ANOVA with Tukey's multiple comparisons, DF = 37, F = 22.427, P < 0.0001. WT vs. hiw P < 0.0001 (****), WT vs. hiw > dNmnat‐RNAi P < 0.0001 (****), hiw vs. hiw > dNmnat‐RNAi P = 0.2907 (NS)].
3. Quantification (mean ± SEM) of BRP puncta per NMJ [WT n = 14, hiw n = 12, hiw > dNmnat‐RNAi n = 12. One‐way ANOVA with Tukey's multiple comparisons, DF = 37, F = 2.854, P = 0.0711. WT vs. hiw P = 0.0573 (NS), WT vs. hiw > dNmnat‐RNAi P = 0.4529 (NS), hiw vs. hiw > dNmnat‐RNAi P = 0.5030 (NS)].
4. Quantification (mean ± SEM) of percentage of Brp puncta that were apposed to a dGluRIII clusters at each NMJ [WT n = 14, hiw n = 12, hiw > dNmnat‐RNAi n = 12. One‐way ANOVA with Tukey's multiple comparisons, DF = 37, F = 2.093, P = 0.4716. WT vs. hiw P = 0.2168 (NS), WT vs. hiw > dNmnat‐RNAi P = 0.1882 (NS), hiw vs. hiw > dNmnat‐RNAi P = 0.9966 (NS)].
5. Representative electrophysiological traces from a synaptic facilitation experiment in which trains of five pulses were used to evoke consecutive action potentials at 100 ms ISI. Genotypes for these experiments were as follows: WT (elav>UAS‐RFP), hiw mutant (hiw; elav>UAS‐RFP), or hiw depleted of dNmnat (hiw; elav>dNmnat‐RNAi).
6. Quantification (mean ± SEM) of the facilitation index, in which the amplitude of the 5^th pulse is divided by the amplitude of the 1^st pulse [100 ms ISI: WT n = 11, hiw n = 9, hiw>dNmnat‐RNAi n = 12. One‐way ANOVA with Tukey's multiple comparisons, DF = 31, F = 7.546, P = 0.0023. WT vs. hiw P = 0.0020 (**), hiw vs. hiw> dNmnat‐RNAi P = 0.0229 (*), WT vs. hiw > dNmnat‐RNAi P = 0.5141 (NS). 50 ms ISI: WT n = 11, hiw n = 9, hiw>dNmnat‐RNAi n = 12. One‐way ANOVA with Tukey's multiple comparisons, DF = 31, F = 5.587, P = 0.0089. WT vs. hiw P = 0.0129 (*), hiw vs. hiw> dNmnat‐RNAi P = 0.0217 (*), WT vs. hiw> dNmnat‐RNAi P = 0.9546 (NS)].

Since hiw mutants do not have a defect in N, we next assessed release probability. Relative changes in release probability can be estimated by examining short‐term plasticity in a synaptic facilitation paradigm. There is an inverse correlation between synaptic facilitation and release probability, such that facilitation is more pronounced when probability of release is lower 20, 21. We calculated the facilitation index as the amplitude of the fifth pulse compared to the first in a train of stimuli at two different interstimulus intervals (50 and 100 ms) in wild type, hiw mutants, and hiw mutants expressing the dNmnat‐RNAi. Hiw mutants exhibit enhanced short‐term facilitation consistent with a decreased release probability (Fig 3E and F). Depletion of dNmnat rescues this enhanced facilitation in hiw mutants (Fig 3E and F). Hence, these data indicate that excess dNmnat in the hiw mutant impairs quantal content by decreasing release probability.

Hiw mutant active zones are enriched in dNmnat

Thus far, we have demonstrated that Hiw functions to keep dNmnat levels low to promote evoked synaptic release via maintenance of a normal synaptic release probability. It was previously demonstrated that dNmnat levels are elevated in axons and neuromuscular junctions of Hiw mutants 10. Given the effect of dNmnat on release probability, we investigated whether this excess dNmnat protein in hiw mutants accumulates near or at active zones, the sites controlling release probability. The UAS‐dNmnat protein is HA‐tagged, so we stained for HA‐dNmnat at the NMJ of wild‐type and hiw mutant larvae. As previously reported, dNmnat levels are higher at hiw NMJs than wild‐type NMJs (10, Fig 4A and B). We also stained for the active zone protein Bruchpilot as well as the entire NMJ, so that we could assess dNmnat abundance separately in two compartments: the entire terminal minus Brp‐positive puncta (Fig 4C, “Terminal regions excluding Brp‐positive active zones”) and the Brp‐positive active zones (Fig 4C, “Brp‐positive active zones only”). By using masks to select and separate Brp‐positive regions from the rest of the terminal compartment, we were able to assess whether Brp‐positive regions are differentially enriched for HA‐dNmnat compared to Brp‐negative, “non‐active zone” regions. We compared HA‐dNmnat levels between these two compartments in wild type and hiw NMJs. In hiw mutant terminals, there is a significant increase in dNmnat levels exclusively in the active zone compartment (Fig 4C). These data demonstrate that in hiw mutants, excess dNmnat is enriched near active zones, positioning dNmnat to potentially affect local synaptic substrates that control release probability.

Figure 4. dNmnat levels are increased near active zones in hiw mutant NMJ synaptic terminals.

1. Representative images of NMJ synaptic terminals in control (DVGlut‐Gal4 > UAS‐HA‐dNmnat) and hiw mutant (hiw^ND9, DVGlut‐Gal4 > UAS‐HA‐dNmnat) larvae. NMJs at larval muscle 4 were identified using HRP (blue in top panel) immunostaining. Insets highlight active zones within single boutons, which were identified by staining for the presynaptic active zone protein Brp (red). An anti‐HA antibody was used to stain for an HA‐tagged dNmnat (green) transgene to measure protein levels in control and hiw mutant active zones.
2. Quantification (mean ± SEM) of HA‐dNmnat levels across the entire NMJ synaptic terminal, normalized to total terminal area and presented as fold change over WT [WT > HA‐dNmnat n = 10 terminals, hiw > HA‐dNmnat n = 10 terminals. Unpaired t‐test, DF = 18, P = 0.0028 (**)].
3. Quantification (mean ± SEM) of HA‐dNmnat levels per unit area in WT and hiw for synaptic terminals excluding active zones (herein referred to as “terminals”) vs. active zones (AZs) only [WT n = 10 terminals, hiw n = 9 terminals, each parsed into terminal minus AZs and AZs only. Two‐way ANOVA with Tukey's multiple comparisons, DF = 34, P = 0.0005 for row factor (terminal vs. AZs) and P < 0.0001 for column factor (WT vs. hiw). WT terminal vs. hiw terminal P = 0.9830 (NS). Hiw terminal vs. hiw AZs P < 0.0001 (****). WT AZs vs. hiw AZs P < 0.0001 (****)].

Highwire regulation of dNmnat is required to promote proper T‐bar structure

The dNmnat‐dependent decrease in release probability in hiw mutants suggests that excess dNmnat is perturbing the ability of individual active zones to properly release vesicles. This hypothesis is supported by data showing that hiw mutant axons do not exhibit abnormalities in cytoskeletal integrity 10, nor are there gross defects in organelle distribution (Fig EV3A and B), suggesting that decreased release in hiw mutants is unlikely to be driven by widespread axonal abnormalities. Further, our expression studies demonstrate that excess dNmnat in hiw mutants is enriched near BRP‐positive active zones. Thus, we examined active zone ultrastructure in WT, hiw;wnd double mutants, and hiw;wnd mutants depleted of dNmnat (hiw;wnd > dNmnat‐RNAi). We conducted these studies in hiw;wnd double mutants because hiw mutants have gross morphological defects which are suppressed by elimination of excess Wnd, yet the defect in evoked release persists. Therefore, utilizing hiw;wnd double mutants allowed us to examine ultrastructural defects that should be retained if such defects are driving decreases in evoked release, independent of gross morphological defects at the NMJ. We employed electron microscopy at the NMJs of wild type, hiw;wnd double mutant larvae, and hiw;wnd > dNmnat‐RNAi larvae and assessed active zone ultrastructure. We examined T‐bars, which are electron‐dense structures at the active zone that are critical for localizing vesicles near the site of release and for promoting Ca⁺⁺ channel clustering at the synapse 22, 23, 24. Mutants with aberrant T‐bar structure have defects in evoked release and decreased release probability 18, 22, 25. Previous work demonstrated that hiw mutants can form T‐bars at active zones 5, but a quantitative assessment has not been performed. While we confirmed that T‐bars are present in hiw;wnd double mutants, we observed that hiw;wnd mutant synapses have numerous defects in T‐bar structure and organization at the NMJ (Fig 5A and B). Both T‐bar width and height are substantially reduced in hiw;wnd mutants (Fig 5B for representative images, 5C and E). Notably, there are many instances in which multiple abnormal T‐bar structures are present at a single active zone, and these T‐bars often lack the expected flat T‐bar top shape (Fig 5B, middle panels). Analysis of the frequency distribution of T‐bar width and height demonstrates that the entire distribution is shifted toward smaller sizes in hiw;wnd mutants compared to wild type, suggesting that defects are not restricted to a subset of T‐bars but instead affect the entire population. Depleting dNmnat in hiw;wnd mutants fully suppresses defects in T‐bar structure (Fig 5C–F). Both T‐bar height and width are restored to sizes observed in wild‐type T‐bars (Fig 5B for representative images, Fig 5C and E), and the frequency distributions for T‐bar width and height are restored to wild‐type values (Fig 5D and F). We did not observe any significant differences among genotypes in vesicle diameter [Fig 5A and B for representative vesicle images. WT: n = 22, mean diameter = 30.5 nm; hiw;wnd: n = 25, mean diameter = 32.3 nm; hiw;wnd elav > dNmnat‐RNAi: n = 23, mean diameter = 31.5 nm. One‐way ANOVA with Tukey's multiple comparisons, DF = 69, F = 0.352, P = 0.7043. WT vs. hiw;wnd— P = 0.6805 (NS), WT vs. hiw;wnd elav > dNmnat‐RNAi—P = 0.8854 (NS), hiw;wnd vs. hiw;wnd elav > dNmnat‐RNAi—P = 0.9300 (NS)] or vesicle density [WT: n = 22, mean # vesicles/bouton area = 92.2; hiw;wnd: n = 25, mean # vesicles/bouton area = 81.9; hiw;wnd elav > dNmnat‐RNAi: n = 23, mean # vesicles/bouton area = 81.9. One‐way ANOVA with Tukey's multiple comparisons, DF = 69, F = 0.665, P = 0.5175. WT vs. hiw;wnd—P = 0.5668 (NS), WT vs. hiw;wnd elav > dNmnat‐RNAi—P = 0.5850 (NS), hiw;wnd vs. hiw;wnd elav > dNmnat‐RNAi—P > 0.9999 (NS)], suggesting that defects in evoked release are unlikely to be due to aberrant vesicular formation or distribution. Hence, excess dNmnat in hiw mutants causes defects in T‐bar width and height, and, given the critical importance of T‐bars to localize vesicles and other synaptic proteins, it is likely that these abnormal T‐bars are responsible for the impaired synaptic release in hiw mutants.

Figure EV3. Hiw mutants exhibit normal mitochondrial distribution.

1. Representative images of larval nerves immunostained for nerve marker HRP (red) and MitoGFP (green). A UAS‐mitoGFP transgene was expressed in both WT and hiw larvae using the Elav‐GAL4 pan‐neuronal driver. Nineteen individual nerves from four different larvae were analyzed per genotype. We observed no significant differences in levels or distribution of MitoGFP between WT and hiw nerves.
2. Quantification of mean (± SEM) mitoGFP levels in nerves per nerve area and normalized to WT levels [WT n = 19 nerves, hiw n = 19 nerves. Unpaired t‐test, t = 1.752, df = 38, WT vs. hiw P = 0.0879 (NS)].

Figure 5. Excess dNmnat in a hiw mutant disrupts the ultrastructure of active zone T‐bars.

1. Representative electron micrographs of entire boutons in wild type, hiw;wnd, and hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi NMJs.
2. High magnification electron micrographs showing representative “T‐bar” structures at active zones in NMJs of the indicated genotypes. Red arrows denote the beginning and end of the T‐bar top, demonstrating changes in T‐bar width in the genotypes tested. As illustrated, a significant reduction in T‐bar size is observed in hiw;wnd, and this defect is suppressed in hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi.
3. Quantification of mean (± SEM) T‐bar width [WT n = 90, hiw;wnd n = 55, hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi n = 81. One‐way ANOVA with Tukey's multiple comparisons, DF = 225, F = 13.829, P < 0.0001. WT vs. hiw;wnd P < 0.0001 (****), WT vs. hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi P = 0.9318 (NS), hiw;wnd vs. hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi P < 0.0001 (****)].
4. Analysis of cumulative frequency distribution of T‐bar width shows a left shift in distribution in hiw;wnd but no change in hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi compared with WT NMJs [WT vs. hiw;wnd K‐S test, D = 0.5636, P < 0.0001 (****), WT vs. hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi K‐S test, D = 0.1815, P = 0.1240 (NS), hiw;wnd vs. hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi K‐S test, D = 0.4121, P < 0.0001 (****)].
5. Quantification of mean (± SEM) T‐bar height [WT n = 90, hiw;wnd n = 55, hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi n = 81. One‐way ANOVA with Tukey's multiple comparisons, DF = 225, F = 10.472, P < 0.0001. WT vs. hiw;wnd P = 0.0008 (***), WT vs. hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi P = 0.6694 (NS), hiw;wnd vs. hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi P < 0.0001 (****)].
6. Analysis of cumulative frequency distribution of T‐bar height shows a left shift in distribution in hiw;wnd but no change in hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi compared with WT NMJs [WT vs. hiw;wnd K‐S test, D = 0.3343, P = 0.001 (***), WT vs. hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi K‐S test, D = 0.1012, P = 0.7748 (NS), hiw;wnd vs. hiw;wnd, Elav‐GAL4 > dNmnat‐RNAi K‐S test, D = 0.3484, P = 0.0008 (***)].

Excess NAD⁺ biosynthesis drives defects in evoked release in hiw mutants

dNmnat and its mammalian NMNAT orthologs are critical enzymes in the NAD⁺ biosynthetic pathway. dNmnat catalyzes the conversion of nicotinic acid mononucleotide (NaMN) into nicotinic acid adenine dinucleotide (NaAD), which is then converted to nicotinamide adenine dinucleotide (NAD⁺), an essential coenzyme required in all cell types for proper metabolic redox reactions (16, 26, see Fig 6A for a schematic of the Drosophila NAD⁺ synthesis pathway). In Drosophila, dNmnat also can function as a chaperone to promote proper protein folding 27, 28, 29. Given this dual functionality of dNmnat, and given our data that overexpressing wild‐type dNmnat decreases evoked release (Fig 1), we wanted to test whether the canonical function of dNmnat, NAD⁺ biosynthesis, is required for the decrease in evoked synaptic transmission. To investigate this question, we overexpressed two different catalytic‐dead mutants of dNmnat (UAS‐dNmnat‐WR in which there are two mutations at residues W98G and R224A, and UAS‐dNmnat‐H30A, 16) in third‐instar larvae using the VGlut‐GAL4 driver. These transgenes encode proteins with ~ 1% of normal NAD⁺ synthesizing activity yet maintain their capacity to act as molecular chaperones 16, 27, 28, 29, 30. In contrast to overexpression of wild‐type dNmnat, which decreases EPSP amplitude and quantal content (Fig 1), these enzyme‐dead mutants did not impact EPSP amplitude or quantal content (Fig 6A–D). Likewise, we did not observe any differences in mEPSP amplitude when overexpressing catalytically inactive dNmnat proteins (Fig 6B for representative traces). We confirmed that catalytically inactive dNmnat transgenes express in the nervous system (Fig EV4). dNmnat protein is barely detectable in a WT larvae, but an increase in endogenous dNmnat can be observed in hiw mutants (Fig EV4A). Transgenic dNmnat is detectable at the same molecular weight as endogenous dNmnat in hiw mutants (Fig EV4B). Thus, the failure of catalytic‐dead mutants to impair transmitter release is not due to low levels of expression. We conclude that dNmnat enzyme function is required to inhibit synaptic release.

Figure 6. The NAD⁺ synthetic function of dNmnat is necessary to decrease evoked release.

1. Schematic of the NAD⁺ synthesis pathway in Drosophila. Red text indicates enzymes tested within the context of synaptic release at the larval NMJ.
2. Representative EPSP and mEPSP electrophysiological traces from larval NMJs at muscle 6 in which a motoneuron‐specific driver was used to overexpress either a control transgene (DVGlut‐GAL4 > UAS‐RFP), or one of the catalytic dNmnat mutants (DVGlut‐Gal4 > UAS‐WR‐dNmnat, DVGlut‐Gal4 > UAS‐H30A‐dNmnat).
3. Quantification of mean (± SEM) EPSP amplitudes, in which 75 consecutive evoked events were averaged per cell, and then, cell amplitudes were averaged per genotype [UAS‐RFP n = 8 cells, UAS‐WR‐dNmnat n = 7 cells, UAS‐H30A‐dNmnat n = 7 cells. One‐way ANOVA with Tukey's multiple comparisons, DF = 21, F = 0.8596, P = 0.4383. UAS‐RFP vs. UAS‐WR‐dNmnat P = 0.8810 (NS), UAS‐RFP vs. UAS‐H30A‐dNmnat P = 0.9997 (NS)].
4. Quantification of quantal content (± SEM) [UAS‐RFP n = 8 cells, UAS‐WR‐dNmnat n = 7 cells, UAS‐H30A‐dNmnat n = 7 cells. One‐way ANOVA with Tukey's multiple comparisons, DF = 21, F = 1.154, P = 0.3365. UAS‐RFP vs. UAS‐WR‐dNmnat P = 0.8696 (NS), UAS‐RFP vs. UAS‐H30A‐dNmnat P = 0.9697 (NS)].
5. Representative EPSP and mEPSP electrophysiological traces from larval NMJs at muscle 6. Recordings were taken from WT control larvae (Elav‐Gal4 > UAS‐RFP), hiw mutants driving a control transgene (hiw ^ND8;Elav‐Gal4 > UAS‐RFP), and hiw mutants expressing an RNAi to knockdown NADsyn (hiw ^ND8;Elav‐Gal4 > UAS‐NADsyn‐RNAi).
6. Quantification of mean (± SEM) EPSP amplitudes, in which 75 consecutive evoked events were averaged per cell, and then, cell amplitudes were averaged per genotype [WT n = 10 cells, hiw n = 10 cells, hiw > NADsyn‐RNAi n = 10 cells. One‐way ANOVA with Tukey's multiple comparisons, DF = 29, F = 14.02, P < 0.0001. WT vs. hiw P < 0.0001 (****), WT vs. hiw > NADsyn‐RNAi P = 0.5147 (NS)].
7. Quantification (mean ± SEM) of quantal content [WT n = 10 cells, hiw n = 10 cells, hiw > NADsyn‐RNAi n = 10 cells. One‐way ANOVA with Tukey's multiple comparisons, DF = 29, F = 6.635, P = 0.0047. WT vs. hiw P = 0.0063 (**), WT vs. hiw > NADsyn‐RNAi P = 0.8942 (NS)].
8. Representative images of the NMJ synaptic terminal at muscle 4 in third‐instar larvae. The genotypes in this experiment were identical to those tested in panels (E–G). NMJs were stained for the presynaptic bouton marker DVGLUT (green) and nerve membrane marker HRP (red).
9. Quantification of the mean (± SEM) number of DVGlut⁺ boutons per muscle 4 NMJ in each genotype [WT n = 25, hiw n = 21, hiw > NADsyn‐RNAi n = 24. One‐way ANOVA with Tukey's multiple comparisons, DF = 67, F = 141.3, P < 0.0001. WT vs. hiw P < 0.0001 (****), WT vs. hiw > NADsyn‐RNAi P > 0.0001 (****)].

Figure EV4. Catalytically inactive dNmnat transgenes are overexpressed relative to endogenous dNmnat using the DVGlut‐GAL4 driver in Drosophila ventral nerve cords.

1. Endogenous dNmnat is virtually undetectable in the VNC of WT larvae, however endogenous dNmnat can be detected in hiw mutant VNCs at the expected molecular weight of 30 kD 16, consistent with prior studies demonstrating that hiw mutants have increased levels of dNmnat in the nervous system 10.
2. While endogenous dNmnat remains undetectable in WT animals, overexpressed WT‐dNmnat can be detected as a 30 kDa band, consistent with the molecular weight of endogenous dNmnat in hiw mutants and consistent with initial characterization of the transgene 16. The WR‐dNmnat transgene is expressed at least as strongly as the WT transgene and can also be detected most strongly at 30 kDa; however, a faint band for both WT and WR can be observed at around 35 kDa. The H30A‐dNmnat transgene can be detected as 30‐ and 35‐kDa bands with overall levels being lower than that of WT or WR‐dNmnat, and the presence of two bands for all transgenes is consistent with previous expression studies of these transgenes 16. Blotting for β‐tubulin was used for loading control.

If excessive NAD⁺ synthesis is responsible for decreased evoked synaptic transmission in hiw mutants, then we postulate that depletion of another NAD⁺ biosynthetic enzyme could also suppress defective synaptic release in the hiw mutant. NAD⁺ synthetase (NADsyn) is the enzyme that follows dNmnat in the NAD⁺ biosynthetic pathway, converting NaAD to NAD⁺ (Fig 6A, 16). In Drosophila, this is likely an essential enzyme for NAD⁺ biosynthesis because flies do not appear to encode the enzymes required for the NAD⁺ salvage pathway 31. We employed a transgenic RNAi targeting NADsyn to deplete the synthetase in neurons and measured evoked and spontaneous release at the NMJ of wild type and hiw larvae. NADsyn depletion suppressed defects in evoked release in the hiw mutant background, restoring both EPSP amplitudes and quantal content to wild‐type levels (Fig 6E–G) while leaving mEPSP amplitudes unchanged (Fig 6E for representative traces). Mirroring the results with dNmnat, NADsyn depletion did not suppress synaptic overgrowth at the hiw mutant NMJ (Fig 6H and I). Taken together, these studies reveal an unexpected role for excess NAD⁺ biosynthesis as an inhibitor of evoked synaptic release. Moreover, we identify a mechanistic model by which Highwire maintains normal levels of synaptic transmission: The Highwire ligase promotes degradation of dNmnat, thereby restraining synthesis of NAD⁺ to levels that do not induce ultrastructural defects at active zones and impair synaptic release.

Discussion

Here, we demonstrate that the E3 ubiquitin ligase Highwire targets distinct substrates to regulate synaptic terminal morphology and synaptic function. We previously demonstrated that Highwire targets the MAP3K Wnd to restrain synaptic terminal growth; however, the substrate that Hiw targets to promote evoked release remained unknown. We show here that the NAD⁺ biosynthetic enzyme dNmnat is that substrate. Excess dNmnat in the hiw mutant disrupts the structure of the dense body T‐bars at active zones and decreases release probability. Our findings suggest that excess NAD⁺ biosynthesis is responsible for these synaptic defects.

A role for dNmnat and NAD⁺ synthesis in synaptic function

The Nmnat family of proteins are required for the synthesis of NAD⁺, an essential coenzyme required for cellular energy homeostasis as well as a substrate for NADase enzymes such as sirtuins, parps, and SARM1 32, 33. In Drosophila, dNmnat can also act as a chaperone 27. Loss of dNmnat and its biosynthetic and chaperone activities are devastating to the nervous system, leading to profound synaptic and neuronal degeneration 16, 28, 34. Here, we make the unexpected discovery that not only does loss of dNmnat disrupt synapses, but also excess dNmnat diminishes synaptic function and this effect is mediated by increased NAD⁺ biosynthesis. Three lines of evidence point to NAD⁺ synthesis as the culprit in the impaired transmitter release. First, at otherwise wild‐type synapses, overexpression of catalytically active dNmnat, but not inactive dNmnat, is sufficient to decrease transmitter release. Second, depletion of dNmnat in hiw mutants, in which dNmnat levels are elevated, suppresses the defects in synaptic release and active zone ultrastructure. Third, depletion of NADsyn, the final enzyme in the NAD⁺ biosynthesis pathway, also suppresses the defect in transmitter release in the hiw mutant. Moreover, the finding that dNmnat is enriched near active zones suggests that local regulation of NAD⁺ may impact the function of individual release sites.

How might excess NAD⁺ inhibit synaptic release? There is precedence in the literature for NAD⁺ and its metabolic byproducts, such as adenosine, to inhibit synaptic transmission 35, 36. However, these studies added extracellular NAD⁺ to cultured neurons, and NAD⁺ is not expected to cross cell membranes. Instead, these effects may be due to the breakdown of NAD⁺ to adenosine, which may then inhibit release via presynaptic adenosine receptors 37. In contrast, our results demonstrate that excess NAD⁺ synthesis capacity in the neuron alters the structure of the dense body T‐bars at the active zone, which likely accounts for the decreased release probability and impaired synaptic release. NAD⁺ both participates as a cofactor in a slew of metabolic reactions and also can serve as a substrate for enzymes such as sirtuins, PARPs, and SARM1‐family NADases 33, 38, 39. Among these, sirtuins are particularly interesting candidates, as they regulate many cellular processes. Sirtuins are NAD‐dependent deacetylases 38, and so elevated levels of NAD⁺ at synaptic terminals could activate a sirtuin and induce deacetylation of a target protein. Interestingly, Bruchpilot is a key constituent of T‐bars and is regulated by acetylation 40, 41. It is plausible that when dNmnat levels are elevated, excess NAD⁺ synthesis perturbs normal T‐bar assembly by altering the normal regulatory mechanisms of Bruchpilot. While this model is attractive, current studies indicate that deacetylated Bruchpilot is better able to mediate synaptic release 40, which is not consistent with excess Sirtuin activity disrupting Bruchpilot function. However, these findings come from studies of HDAC6, a non‐NAD‐dependent deactylase, leaving open the possibility of differential regulation of acetylation sites by HDAC6 and Sirtuins that could have different effects on Bruchpilot function. Of course, there are many other synaptic proteins that could be influenced by excess NAD⁺ or the mechanism could be less direct and reflect metabolic changes in the synaptic terminal. Further biochemical and mechanistic studies are required to resolve these questions.

Highwire: balancing axonal maintenance and synaptic strength

dNmnat and its vertebrate ortholog NMNAT2 play critical roles in axonal maintenance. Current models for the mechanism of injury‐induced axon degeneration posit that dNmnat/NMNAT2 protects axons by inhibiting activation of the central executioner of axonal degeneration, SARM1 42, 43, 44. dNmnat and NMNAT2 are labile proteins that are constantly delivered to the axon via fast axonal transport. Upon injury or disease that disrupts this transport, levels of dNmnat/NMNAT2 fall, allowing for activation of SARM1, thereby triggering axon destruction 42, 45. Highwire and its vertebrate ortholog Phr1 are ubiquitin ligases targeting dNmnat/NMNAT2 for degradation 11, 46, 47, hence keeping the levels of this potent axoprotective molecule low. In hiw and Phr1 mutants, the degeneration of injured axons is dramatically delayed due to this increase in dNmnat/NMNAT2 10, 11. While very strong data support this model, the purpose of the regulatory relationship between Highwire/Phr1 and dNmnat/NMNAT2 has been difficult to rationalize. Why would the neuron have retained this mechanism to constantly degrade a potent axonal maintenance factor across 400 million years of evolution? Our finding that excess dNmnat depresses synaptic transmission provides a plausible explanation. Hiw must restrict the levels of dNmnat near active zones to maintain efficient synaptic release. This suggests that in an uninjured neuron, Hiw maintains an appropriate level of dNmnat protein to balance its axoprotective and synapse‐inhibitory functions.

Highwire is an important regulator of axon injury signaling 1, 2. Upon injury, Hiw levels fall 48, which would relieve the negative regulation of both Wnd and dNmnat. Activated Wnd signaling reduces postsynaptic sensitivity to neurotransmitter 9, while here we show that excess dNmnat can depress presynaptic release, suggesting that Hiw coordinately downregulates both presynaptic release efficacy and postsynaptic neurotransmitter sensitivity as a response to injury. Moreover, increased Wnd signaling or that of its vertebrate ortholog DLK activates a pro‐regenerative gene expression program 48, 49. Taken together, these findings suggest that Highwire regulates a coordinated response to axon injury, promoting the transition from a stable, synaptically connected neuron to an exploratory, pro‐regenerative neuron.

We do not know whether excess NAD⁺ synthesis will impair synaptic release at vertebrate synapses. However, this will be an important topic to address because there is currently great interest in boosting NMNAT2 levels as a potential neuroprotective strategy 50. In addition, nutritional supplements with NAD⁺ precursors are currently marketed to enhance NAD⁺ synthesis in hopes of improving cellular metabolism and potentially delaying aging 30, 42, 50, 51, 52. While there is great promise in enhancing NMNAT2 and NAD⁺ biosynthesis as a neuroprotective strategy, our findings in Drosophila serve as a reminder that caution is in order.

Materials and Methods

Drosophila stocks and genetics

Drosophila were cultured employing standard techniques, and all crosses were maintained at 25°C in 60% relative humidity. The following stocks were used in this study: hiw ^ΔN (null allele of hiw, 7), hiw ^ND8 (strong hypomorphic allele of hiw, 5), wnd(dfED228) and wnd‐3 8, DVGlut‐Gal4 15, elav3E‐Gal4 53, and UAS‐WndE 8. Flies obtained from the Bloomington Stock Center included the following: UAS‐RFP (#32218), UAS‐Wallenda‐RNAi (TRiP Collection #25396), UAS‐dNmnat‐RNAi (TRiP Collection #29402), UAS‐NADsyn‐RNAi (TRiP Collection #62265), UAS‐HA‐dNmnat (#39702), UAS‐WR‐dNmnat (#39700), UAS‐H30A‐dNmnat (#39701), and UAS‐mitoGFP (#8442). Flies obtained from the Vienna Drosophila RNAi Center included UAS‐dNmnat‐RNAi #2 (KK #107262). In all experiments, control larvae were generated by crossing virgins of the respective driver used to UAS‐RFP males.

Electrophysiology

Intracellular recordings were performed as described previously 54. In brief, wandering 3^rd instar larvae were dissected in Ca⁺⁺ free HL3.1 buffer 55 (70 mm NaCl, 5 mm KCl, 8 mm MgCl₂, 10 mm NaHCO₃, 5 mm trehalose, 5 mm HEPES, and 0 mm Ca⁺⁺, pH 7.2) and then washed with and recorded in HL3.1 buffer containing the concentration of Ca⁺⁺ indicated for each experiment. Spontaneous mEPSPs and evoked EPSPs were recorded from muscle 6 in segments A2, A3, and A4 using borosilicate sharp electrodes (15–20 MΩ) in 0.35 mM Ca⁺⁺ HL3.1. Intracellular recordings were only used if a resting membrane potential between −60 and −80 mV could be maintained through the duration of the recording and if the muscle input resistance was > 5 MΩ. We did not observe significant differences in resting membrane potential or input resistance between genotypes in any experiment. Seventy‐five consecutive spontaneous events were measured per cell using MiniAnalysis Software (Synaptosoft, Decatur, GA) to determine mEPSP mean amplitude and frequency. Evoked EPSPs were recorded by sucking the end of a cut segmental nerve into a stimulating electrode and stimulating with a 1‐ms pulse of sufficient amperage to recruit both axons innervating muscle 6. EPSP amplitude was calculated by averaging 75 consecutive evoked events at 2 Hz using pClamp9 software (Molecular Devices). Quantal content (QC) for each individual cell was estimated by dividing the average EPSP amplitude by the average mEPSP amplitude. Short‐term facilitation experiments were conducted in HL3.1 containing 0.15 mM Ca⁺⁺ and 4 mM Mg⁺⁺, and the nerve was stimulated with 10 trains of five pulses at either 10 Hz (100 ms interstimulus interval) or 20 Hz (50 ms ISI), and 5 s of rest was given between trains. The facilitation index was calculated by dividing the amplitude of the response to the 5^th pulse in the train by the amplitude of the response to the 1^st pulse. The amplitude of the response to the 5^th pulse was calculated by using an exponential function to calculate, and then subtract, the decay phase contribution of the response to the preceding (4^th) pulse 56.

Immunocytochemistry

Third‐instar larvae were dissected in either cold PBS or Ca⁺⁺‐free HL3 and fixed in either Bouin's fixative for 10 min at RT or for 20 min in ice‐cold 4% paraformaldehyde. After fixing, larval preps were washed in PBS + 0.1% Triton X‐100 (PBST), and then blocked and stained in PBST + 5% goat serum. Larvae were incubated overnight in primary antibodies, including rabbit α‐DVGLUT antibody 54 at 1:5,000 dilution, rabbit α‐HA (C29F4, Cell Signaling Technology) at 1:500 dilution, rabbit α‐dGluRIII (generated and affinity purified by Cocalico Biologicals, 9) at 1:2,000 dilution, and mouse α‐Brp (also called nc82, Developmental Hybridoma Bank) at 1:100 dilution. After incubation with primary, larvae were washed in PBST and incubated for 1 h in the following secondary antibodies: Cy3 and Cy5‐conjugated goat α‐HRP (1:1,000, Jackson ImmunoResearch), Alexa Fluor 488‐conjugated α‐rabbit (1:1,000, Invitrogen), Alexa Fluor 488‐conjugated α‐GFP (1:1,000, Invitrogen), and Cy3 α‐mouse (1:1,000, Jackson ImmunoResearch). After secondary, larval preps were equilibrated in 70% glycerol in PBS, and mounted and imaged in Vectashield (Vector Laboratories).

Imaging and analysis

To assess synaptic growth, larvae stained for DVGlut and HRP were imaged using a 40×‐oil immersion objective on a Leica TCS SPE confocal microscope. Images shown are maximal Z‐projections of confocal stacks. Images for a given experiment were taken simultaneously using identical laser power, gain, and offset settings using the same step size. Synaptic growth was quantified as previously described 8, 57; briefly, DVGlut‐positive boutons were manually counted at larval muscle 4, segments A2 through A4, from at least four larvae per genotype.

To assess dNmnat levels at active zones, larvae stained for HA‐tagged dNmnat, Brp, and HRP were imaged using a 63×‐oil immersion objective on a Leica TCS SPE confocal microscope. Maximal Z‐projections of confocal stacks were taken using identical settings within an experiment, and the images were analyzed for fluorescence intensity using ImageJ software (NIH). To measure HA‐dNmnat fluorescence intensity at active zones, a mask was created using the Brp channel to select only the areas comprising active zones within the presynaptic membrane. Thus, only HA‐positive puncta within this mask were measured as contributing HA‐dNmnat levels at active zones. These masked areas were further subtracted from the total synaptic terminal, and “non‐active zone” areas were also analyzed for HA‐dNmnat levels. The Brp mask was also used to quantitate active zone area for each synaptic terminal analyzed; absolute HA intensity per synaptic terminal was divided by its individual active zone area to calculate HA‐dNmnat intensity per active zone and then presented as the fold change over the mean HA‐dNmnat intensity in controls.

To measure active zone density, synapse number, and apposition of active zones to receptor clusters, larvae were stained for Brp, HRP, and dGluRIII and NMJs at segments A2–A3 were imaged using a Nikon A1R Resonant Scanning Confocal Microscope equipped with NIS Elements software and a 100× APO 1.4NA oil immersion objective with separate channels for three laser lines (488, 561, and 637 nm). All genotypes were immunostained in the same tube with identical reagents, and then mounted and imaged in the same session. z‐stacks were obtained using identical settings for all genotypes with z‐axis spacing between 0.15 and 0.2 μm within an experiment and optimized for detection without saturation of the signal. Maximum intensity projections were used for quantitative image analysis with the NIS Elements software General Analysis toolkit. Neuronal membrane area was calculated by creating a mask based on intensity threshold of the 637 nm HRP channel that labels the neuronal membrane. Intensity thresholds and filters (Rolling Ball Correction) were applied on the 488 channel for BRP, to accurately detect and quantify BRP puncta number per NMJ. To quantify BRP‐dGluRIII apposition, BRP puncta co‐localizing with dGluRIII puncta were taken as a percentage of total BRP puncta per NMJ. Measurements based on confocal images were taken from at least twelve synapses acquired from at least six different animals.

Electron microscopy

EM analysis was performed as described 58. Wandering third‐instar larvae were dissected in Ca⁺⁺‐free HL‐3 and then fixed in 2.5% glutaraldehyde/0.1 M cacodylate buffer at 4°C. Larvae were then washed three times for 20 min in 0.1 M cacodylate buffer. The larval pelts were then placed in 1% osmium tetroxide/0.1 M cacodylate buffer for 1 h at room temperature. After washing the larvae twice with cacodylate and twice with water, larvae were then dehydrated in Ethanol. Samples were cleared in propylene oxide and infiltrated with 50% Eponate 12 in propylene oxide overnight. The following day, samples were embedded in fresh Eponate 12. EM sections were obtained on a Morgagni 268 transmission electron microscope (FEI, Hillsboro, OR). The NMJs were serially sectioned at a 60–70 nm thickness. The sections were mounted on Formvar‐coated single‐slot grids and viewed at a 23,000× magnification and were recorded with a Megaview II CCD camera. Images were analyzed blind to genotype using the general analysis toolkit in the NIS Elements software and Image J software.

Western blots

To analyze protein levels in the third‐instar larval nervous system, ventral nerve cords (VNCs) were dissected and homogenized on ice in lysis buffer (100 mM NaCl, 100 mM Tris ph 8.0, 1 mM EDTA pH 8.0, 2 M urea, 1% v/v SDS, 20 mM Na₃VO₄, 1× Roche protease inhibitor) with a pestle. VNCs from 10 genetically identical flies were pooled into one lysate to achieve sufficient protein concentration, and the lysate was boiled for 10 min and centrifuged at RT for 10 min at 15,700 g. Gel electrophoresis and transfer were performed by standard procedures 7. Membranes were blocked for 1 h and incubated overnight at 4C in primary antibody solution consisting of 5% milk solution in Tris‐buffered saline with 0.2% Tween‐20 (TBST) and one of the following antibodies: guinea pig anti‐dNmnat (1:1,000, a gift from Hugo Bellen) and mouse anti‐ β‐tubulin E7 (1:100, Developmental Studies Hybridoma). Primary antibody solution was removed, membranes were washed 3× with TBST, and membranes were then incubated before incubation with HRP‐conjugated secondary antibodies to either anti‐mouse or anti‐guinea pig antibody (Jackson, 1:10,000). After 1 h of incubation, membranes were washed 3× in TBST, 1× in TBS, and developed using Immobilon Western Chemiluminescent Substrate (EMD Millipore). Membranes were imaged on a G:Box Chemi‐XX6 (Syngene).

Experimental design and statistical analysis

For electrophysiological and immunostaining experiments, each NMJ terminal (muscle 6 for physiology and muscle 4 for immunostaining analyses of synaptic terminals and active zones) is considered an N of 1 since each presynaptic motor neuron terminal is confined to its own muscular hemisegment. For these experiments, muscle 4 or 6 was analyzed from hemisegments A2–A4 for each larva, and thus, each larva contributes 4–6 NMJs per experiment. To control for variability between larvae within a genotype, for all electrophysiological and immunostaining experiments presented, NMJs were analyzed from no less than four individual larvae. For these experiments, muscle 4 from hemisegment A3 was analyzed for each larva, and thus, each larva contributes an N of 2 per experiment. To control for variability between larvae within a genotype, for immunostaining experiments involving BRP and dGluRIII, NMJs were analyzed from no less than six individual larvae.

For electron microscopy, T‐bars at active zones of boutons at muscle 6/7 were analyzed from hemisegment A3. Each T‐bar is considered an N of 1 resulting in at least 50 T‐bars for analysis per genotype taken from five to six boutons per NMJ. To control for variability between larvae within a genotype, two to three NMJ terminals were assessed, each from a different larvae per genotype.

All data are presented as mean ± SEM. The experiment in Fig 4A–C, in which there are two genotypic conditions, was analyzed for statistical significance with a two‐tailed, unpaired t‐test. The data presented in Fig 4D were analyzed for statistical significance with a two‐way ANOVA using Tukey's post hoc test to correct for multiple comparisons between genotypes and condition tested. The experiment in Fig EV3 was analyzed for statistical significance using a two‐tailed, unpaired t‐test. For all other experiments presented, in which at least three conditions were tested, a one‐way ANOVA with a Tukey post hoc test was used to determine significance and correct for multiple comparisons of experiments with three or more groups (Figs 1, 2, 3, 5 and 6, and EV1 and EV2). *, **, *** and **** indicate P‐values < 0.05, 0.01, 0.001, and 0.0001, respectively, and exact P‐values are reported in the Results and figure legends.

Author contributions

AR, EJB, PG, DD, and AD designed the research; AR, CB, and PG performed the research; AR and PG analyzed data; AR, PG, and AD wrote the paper.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Review Process File

EMBO Reports (2019) 20: e46975