ABSTRACT
Changes in cholesterol content of neuronal membranes occur during development and brain aging. Little is known about whether synaptic activity regulates cholesterol levels in neuronal membranes and whether these changes affect neuronal development and function. We generated transgenic flies that express the cholesterol-binding D4H domain of perfringolysin O toxin and found increased levels of cholesterol in presynaptic terminals of Drosophila larval neuromuscular junctions following increased synaptic activity. Reduced cholesterol impaired synaptic growth and largely prevented activity-dependent synaptic growth. Presynaptic knockdown of adenylyl cyclase phenocopied the impaired synaptic growth caused by reducing cholesterol. Furthermore, the effects of knocking down adenylyl cyclase and reducing cholesterol were not additive, suggesting that they function in the same pathway. Increasing cAMP levels using a dunce mutant with reduced phosphodiesterase activity failed to rescue this impaired synaptic growth, suggesting that cholesterol functions downstream of cAMP. We used a protein kinase A (PKA) sensor to show that reducing cholesterol levels reduced presynaptic PKA activity. Collectively, our results demonstrate that enhanced synaptic activity increased cholesterol levels in presynaptic terminals and that these changes likely activate the cAMP-PKA pathway during activity-dependent growth.
INTRODUCTION
Synaptic plasticity is a fundamental property of neurons that allows their ability to transmit information to change with experience. These changes can be due to functional or structural modifications of synapses. Numerous studies have examined how synaptic plasticity is regulated by protein–protein interactions, cytoskeletal dynamics, and changes in the expression and activation of various proteins (Costa-Mattioli et al., 2009; Gordon-Weeks and Fournier, 2014; Choi et al., 2014; Davis and Müller, 2015; Hafner et al., 2019). In contrast, the roles of membrane lipids in synaptic plasticity have been studied less, as the membrane lipid composition was thought to be relatively static. However, increasing evidence suggests that lipid content of neuronal membranes does not remain constant and is altered by synaptic activity (Sodero et al., 2012; Brachet et al., 2015; Vaughen et al., 2022).
Cholesterol is a prominent lipid component of neuronal membranes and affects numerous aspects of nervous system development and function (reviewed in Martín et al., 2014). The cholesterol content of neuronal membranes changes during development and aging (Martin et al., 2008; Nicholson and Ferreira, 2009; Sodero et al., 2011; Palomer et al., 2016). Defects in the synthesis and trafficking of cholesterol have been shown to affect the development of the nervous system (Mauch et al., 2001; Fan et al., 2002; Ko et al., 2005; Fünfschilling et al., 2012; Barszczyk et al., 2015). The cholesterol content of neuronal membranes also affects both synaptic vesicle exocytosis (Zamir and Charlton, 2006; Wasser et al., 2007; Smith et al., 2010; Dason et al., 2010; Linetti et al., 2010; Ormerod et al., 2012; Korinek et al., 2020) and endocytosis (Wasser et al., 2007; Dason et al., 2010; Petrov et al., 2010; Dason et al., 2014). Furthermore, cholesterol has been shown to regulate long-term potentiation (LTP; Frank et al., 2008; Brachet et al., 2015). Interestingly, NMDA receptor activation during LTP induction leads to a redistribution of cholesterol in postsynaptic hippocampal neurons (Brachet et al., 2015). It is unknown whether changes in the cholesterol content of neuronal membranes occur in response to other patterns of neuronal activity and whether these changes are a potential mechanism for other forms of synaptic plasticity such as activity-dependent synaptic growth.
The Drosophila larval neuromuscular junction (NMJ) is a well-established model system for studying synaptic growth that shares the basic molecular components found at most synapses across organisms (reviewed in Harris and Littleton, 2015). The Drosophila larval NMJ displays activity-dependent synaptic growth, which has previously been shown to be dependent on the presynaptic cAMP-protein kinase A (PKA) signaling pathway (Sigrist et al., 2003; Zhong and Wu, 2004; Vasin et al., 2014; Cho et al., 2015; Vasin et al., 2019). Here, we take advantage of this attractive model synapse to determine the role of cholesterol in activity-dependent synaptic growth. We generated transgenic flies to examine the distribution of cholesterol and found that cholesterol levels in presynaptic terminals increase in response to synaptic activity. We found that cholesterol is required for both synaptic growth and activity-dependent synaptic growth. Examination of several mutant and transgenic larvae reveal that cholesterol is likely regulating synaptic growth through the cAMP-PKA signaling pathway. Collectively, our data demonstrate that the cholesterol content of presynaptic terminal changes in response to synaptic activity and that these changes play a key role in development and activity-dependent synaptic plasticity.
RESULTS
Cholesterol distribution in the Drosophila larval nervous system
Drosophila, like other insects, cannot synthesize cholesterol directly, but instead acquires it from their diet and produces it from conversion of other sterols (Behmer and Nes, 2003). Despite being unable to synthesize cholesterol, Drosophila accumulate cholesterol in their membranes to a level comparable to that of mammalian membranes (Rietveld et al., 1999; Phillips et al., 2008). To investigate the distribution of cholesterol in the Drosophila larval nervous system, we created transgenic flies that express a cholesterol biosensor under the control of the GAL4/UAS system. Specifically, we used D4H, a mutated cholesterol-binding D4 domain of perfringolysin O toxin, which has a higher affinity for cholesterol than D4 and recognizes cholesterol in the cytosolic leaflet of the plasma membrane (Maekawa and Fairn, 2015). D4 and D4H are well-established cholesterol biosensors that have been used to examine cholesterol distribution in mammalian neurons (Smith et al., 2010; Brachet et al., 2015; Weber-Boyvat et al., 2022). Unlike cholesterol probes such as filipin, D4H is not easily photobleached, can be used for live imaging and can be transgenically expressed in specific neurons or cells (Maekawa and Fairn, 2015). Using neuronal (n-syb-GAL4) and glial (Repo-GAL4) GAL4s, we expressed D4H–GFP in the larval CNS and stained for D4H–GFP with an anti-GFP antibody. We found strong expression of D4H–GFP in the brain lobes and the ventral nerve cord of the larval CNS of n-syb-GAL4>UAS-D4H-GFP third-instar larvae (Fig. 1A). The ventral nerve cord consists of interneurons, projections from sensory neurons and cell bodies of motor neurons. D4H–GFP expression was also detected in the segmental nerves that contain motor axons that innervate muscle fibers (Fig. 1A). We also found strong expression of D4H–GFP in glia of the larval CNS and in glia that ensheath the segmental nerves of Repo-GAL4>UAS-D4H-GFP third-instar larvae (Fig. 1B). Next, we selectively drove D4H–GFP expression presynaptically (n-syb-GAL4), postsynaptically (24B-GAL4) or in glia (Repo-GAL4) at the larval NMJ. We examined third-instar larval NMJs of muscles 6 and 7 of abdominal segment 3, which are innervated by two types of axons that result in type 1b and type 1s boutons, which differ in both their morphological and physiological properties (Kurdyak et al., 1994). We found strong expression of D4H–GFP in both 1b and 1s boutons of n-syb-GAL4>UAS-D4H-GFP larvae (Fig. 1C). Expression was also detected in glia at the NMJ of Repo-GAL4>UAS-D4H-GFP larvae (Fig. 1D). These glia extend over non-synaptic areas of the muscle surface (Brink et al., 2012; Dason et al., 2019). Diffuse expression of D4H–GFP was detected in muscles of 24B-GAL4>UAS-D4H-GFP larvae (Fig. 1E). To verify that D4H was binding to cholesterol, we fed larvae 10 mM methyl-β-cyclodextrin (MβCD), a cholesterol-chelating agent, and examined D4H expression in 1b and 1s boutons. We found that D4H–GFP expression was significantly reduced in both 1b and 1s boutons of larvae fed MβCD in comparison to controls (Fig. 1F–H). These data confirm that D4H is a cholesterol biosensor that can be used to examine cholesterol levels in presynaptic boutons at the Drosophila larval NMJ. Overall, cholesterol is present in both neurons and glia in the Drosophila larval CNS and NMJ.
Cholesterol distribution in the Drosophila larval nervous system. (A) Representative image of fixed n-syb-GAL4>UAS-D4H-GFP larval CNS stained with an anti-GFP antibody. Strong expression of D4H–GFP was detected in the brain lobes and the ventral nerve cord of the larval CNS. (B) Representative image of fixed Repo-GAL4>UAS-D4H-GFP larval CNS stained with an anti-GFP antibody. Strong expression was detected in glia. (C–E) Representative images of larval NMJ stained with an anti-GFP antibody. (C) D4H–GFP expression was detected in both 1b and 1 s boutons of n-syb-GAL4>UAS-D4H-GFP larvae. (D) D4H–GFP expression was detected in glia of Repo-GAL4>UAS-D4H-GFP larvae. (E) D4H–GFP expression was diffuse and punctate in muscles of 24B-GAL4>UAS-D4H-GFP larvae. Images in A–E are representative of six repeats. (F) Representative images of n-syb-GAL4>UAS-D4H-GFP larval NMJ of larvae grown on control food or food supplemented with 10 mM MβCD. (G) The fluorescence intensity of D4H–GFP in 1b boutons of larvae grown on control food (n=6) was significantly higher (P=0.006) than the fluorescence of 1b boutons from larvae grown on food 10 mM MβCD (n=8). (H) The fluorescence intensity of D4H–GFP in 1 s boutons of larvae grown on control food (n=6) was significantly higher (P=0.015) than the fluorescence of 1 s boutons from larvae grown on food 10 mM MβCD (n=8). Values represent the mean±s.e.m. *P<0.05; **P<0.01 (two-tailed unpaired t-test). a.u., arbitrary units.
Cholesterol distribution in the Drosophila larval nervous system. (A) Representative image of fixed n-syb-GAL4>UAS-D4H-GFP larval CNS stained with an anti-GFP antibody. Strong expression of D4H–GFP was detected in the brain lobes and the ventral nerve cord of the larval CNS. (B) Representative image of fixed Repo-GAL4>UAS-D4H-GFP larval CNS stained with an anti-GFP antibody. Strong expression was detected in glia. (C–E) Representative images of larval NMJ stained with an anti-GFP antibody. (C) D4H–GFP expression was detected in both 1b and 1 s boutons of n-syb-GAL4>UAS-D4H-GFP larvae. (D) D4H–GFP expression was detected in glia of Repo-GAL4>UAS-D4H-GFP larvae. (E) D4H–GFP expression was diffuse and punctate in muscles of 24B-GAL4>UAS-D4H-GFP larvae. Images in A–E are representative of six repeats. (F) Representative images of n-syb-GAL4>UAS-D4H-GFP larval NMJ of larvae grown on control food or food supplemented with 10 mM MβCD. (G) The fluorescence intensity of D4H–GFP in 1b boutons of larvae grown on control food (n=6) was significantly higher (P=0.006) than the fluorescence of 1b boutons from larvae grown on food 10 mM MβCD (n=8). (H) The fluorescence intensity of D4H–GFP in 1 s boutons of larvae grown on control food (n=6) was significantly higher (P=0.015) than the fluorescence of 1 s boutons from larvae grown on food 10 mM MβCD (n=8). Values represent the mean±s.e.m. *P<0.05; **P<0.01 (two-tailed unpaired t-test). a.u., arbitrary units.
Drosophila larvae raised at high temperatures (30°C) have increased locomotion, which results in activity-dependent synaptic growth (Sigrist et al., 2003; Zhong and Wu, 2004). However, GAL4 activity is known to increase at higher temperatures (Duffy, 2002). Thus, we could not use our D4H–GFP probe to examine whether cholesterol levels change in response to increased synaptic activity at higher temperatures. Instead, we examined D4H expression in presynaptic boutons following a 5× spaced high KCl stimulation paradigm (five 90 mM K+ pulses for 5 min, spaced by 15 min of rest; Fig. 2A) that was previously shown to alter synaptic structure (Ataman et al., 2008). We found a significant increase in D4H expression in 1b and 1s boutons following the high KCl stimulation paradigm (Fig. 2B–E). Collectively, our cholesterol biosensor data revealed that cholesterol was present in presynaptic terminals at the larval NMJ and that presynaptic cholesterol levels increase following periods of increased synaptic activity.
Synaptic activity regulates the cholesterol content of presynaptic terminals at the Drosophila larval NMJ. (A) 5× spaced high KCl stimulation paradigm. (B,C) Representative images of D4H–GFP fluorescence of 1b and 1 s boutons prior to and after a 5× spaced high KCl stimulation paradigm. (D) The fluorescence intensity of D4H-GFP in 1b boutons was significantly higher following a 5× spaced high KCl stimulation paradigm (P<0.01) in comparison to 1b boutons that were not stimulated. n=7. (E) The fluorescence intensity of D4H–GFP in 1s boutons was significantly higher following a 5× spaced high KCl stimulation paradigm (P<0.01) in comparison to 1s boutons that were not stimulated. n=6. Fluorescence (F) was reported with background F subtracted. ΔF/Frest was calculated by dividing the change in fluorescence in response to stimulation (ΔF) by the value of fluorescence at rest (Frest) before the 5× spaced high KCl stimulation paradigm. Values represent the mean±s.e.m. **P<0.01 (two-tailed unpaired t-test).
Synaptic activity regulates the cholesterol content of presynaptic terminals at the Drosophila larval NMJ. (A) 5× spaced high KCl stimulation paradigm. (B,C) Representative images of D4H–GFP fluorescence of 1b and 1 s boutons prior to and after a 5× spaced high KCl stimulation paradigm. (D) The fluorescence intensity of D4H-GFP in 1b boutons was significantly higher following a 5× spaced high KCl stimulation paradigm (P<0.01) in comparison to 1b boutons that were not stimulated. n=7. (E) The fluorescence intensity of D4H–GFP in 1s boutons was significantly higher following a 5× spaced high KCl stimulation paradigm (P<0.01) in comparison to 1s boutons that were not stimulated. n=6. Fluorescence (F) was reported with background F subtracted. ΔF/Frest was calculated by dividing the change in fluorescence in response to stimulation (ΔF) by the value of fluorescence at rest (Frest) before the 5× spaced high KCl stimulation paradigm. Values represent the mean±s.e.m. **P<0.01 (two-tailed unpaired t-test).
Cholesterol regulates synaptic growth
A previous study found that supplementing 20 mM MβCD in food caused a reduction in the number of presynaptic boutons at the Drosophila larval NMJ without affecting larval and muscle size (Huang et al., 2018). However, that study did not examine whether cholesterol extraction affects the number of active zones or whether it differentially affects 1b and 1s presynaptic boutons. To reduce cholesterol levels, we raised larvae on fly food supplemented with either 5 or 10 mM MβCD, given that 20 mM MβCD was reported to result in mild developmental delays (Huang et al., 2018). We fixed and then stained NMJs from muscle fibers 6 and 7 of abdominal segment 3 from wandering third-instar larvae with anti-HRP (a neuronal membrane marker) and anti-Bruchpilot (Brp; an active zone marker; Wagh et al., 2006) antibodies. We found that 10 mM MβCD significantly reduced the number of both 1b (control, 28.6±0.70; 10 mM MβCD, 22.1±1.09; mean±s.e.m., P<0.001) and 1s boutons (control, 27.8±0.71; 10 mM MβCD, 19.8±0.99; P<0.001) compared to controls (Fig. 3A,B). Furthermore, 10 mM MβCD reduced the number of active zones per NMJ (Fig. 3A,C). No significant differences in the number of boutons or active zones were found between larvae raised on 5 mM MβCD and controls (Fig. 3A–C). Overall, a reduction in cholesterol results in fewer presynaptic boutons and fewer active zones.
Cholesterol loss reduces synaptic growth, but has mild effects on neurotransmission. (A) Representative images of fixed w1118 larval NMJs stained with FITC-conjugated anti-HRP and anti-Brp antibodies. (B) The number of presynaptic boutons in larvae fed 10 mM MβCD was significantly reduced in comparison to controls and larvae fed 10 mM MβCD (P<0.001; n=10). (C) The number of active zones in larvae fed 10 mM MβCD was significantly reduced in comparison to controls and larvae fed 5 mM MβCD (P<0.01; n=10). (D) Representative traces of EJPs and mEJPs. Preparations were maintained in HL6 (1 mM Ca2+) saline. (E) The amplitude of evoked EJPs from controls (n=8), larvae fed 5 mM MβCD (n=9), and larvae fed 10 mM MβCD (n=8) were not significantly different (P>0.05). (F) The amplitude of mEJPs from controls (n=6), larvae fed 5 mM MβCD (n=7), and larvae fed 10 mM MβCD (n=7) were not significantly different (P>0.05). (G) The frequency of mEJPs from larvae fed 10 mM MβCD (n=7) was significantly decreased (P=0.024) in comparison to controls (n=6), but no significant differences were found between larvae fed 5 mM MβCD (n=7) and controls (P>0.05). Values represent the mean±s.e.m. *P<0.05 (one-way ANOVA with Tukey post test).
Cholesterol loss reduces synaptic growth, but has mild effects on neurotransmission. (A) Representative images of fixed w1118 larval NMJs stained with FITC-conjugated anti-HRP and anti-Brp antibodies. (B) The number of presynaptic boutons in larvae fed 10 mM MβCD was significantly reduced in comparison to controls and larvae fed 10 mM MβCD (P<0.001; n=10). (C) The number of active zones in larvae fed 10 mM MβCD was significantly reduced in comparison to controls and larvae fed 5 mM MβCD (P<0.01; n=10). (D) Representative traces of EJPs and mEJPs. Preparations were maintained in HL6 (1 mM Ca2+) saline. (E) The amplitude of evoked EJPs from controls (n=8), larvae fed 5 mM MβCD (n=9), and larvae fed 10 mM MβCD (n=8) were not significantly different (P>0.05). (F) The amplitude of mEJPs from controls (n=6), larvae fed 5 mM MβCD (n=7), and larvae fed 10 mM MβCD (n=7) were not significantly different (P>0.05). (G) The frequency of mEJPs from larvae fed 10 mM MβCD (n=7) was significantly decreased (P=0.024) in comparison to controls (n=6), but no significant differences were found between larvae fed 5 mM MβCD (n=7) and controls (P>0.05). Values represent the mean±s.e.m. *P<0.05 (one-way ANOVA with Tukey post test).
Effects of loss of cholesterol on neurotransmission
Acute extraction of cholesterol from presynaptic plasma membranes with 10 mM MβCD was previously found to increase the amplitude of evoked neurotransmission and the frequency of spontaneous neurotransmission at the Drosophila larval NMJ (Dason et al., 2010). We next determined the effects of chronic reduction in cholesterol levels on neurotransmission by recording the compound excitatory junction potential (EJP) generated by tonic-like type 1b and phasic-like type 1s boutons from segment 3 of muscle fiber 6 or 7 in third-instar larvae in response to low frequency stimulation (0.05 Hz). Despite seeing a reduction in the total number of boutons and active zones (Fig. 3), evoked neurotransmission was not significantly different between control larvae and larvae grown on food supplemented with 5 or 10 mM MβCD (Fig. 3D,E). The frequency of spontaneously occurring miniature EJPs (mEJPs) was significantly decreased in larvae grown on food supplemented with 10 mM MβCD compared to in controls (Fig. 3D,G). No differences were found in the amplitude of mEJPs (Fig. 3D,F), suggesting that there were no postsynaptic effects. Overall, chronic reduction of cholesterol has mild effects on neurotransmission.
Cholesterol regulates activity-dependent synaptic growth
We next hypothesized that the activity-dependent changes in presynaptic cholesterol that we observed (Fig. 2B–E) would be required for activity-dependent synaptic growth. Consistent with previous findings (Sigrist et al., 2003; Zhong and Wu, 2004), we found that control larvae raised at 30°C had significantly more boutons and active zones than control larvae raised at 23°C (Fig. 4A,B,E). To determine whether cholesterol was required for this activity-dependent synaptic growth, we raised flies on food supplemented with 5 or 10 mM MβCD at 23°C and 30°C. The activity-dependent increase in the number of boutons and active zones was largely not seen in larvae raised on food supplemented with 5 or 10 mM MβCD (Fig. 4A,C,D,F,G). Collectively, our data demonstrates that larvae with reduced cholesterol levels have a reduction in the number of presynaptic boutons and active zones compared to in control larvae (Fig. 3A–C) and do no display activity-dependent synaptic growth (Fig. 4).
Reduction in cholesterol prevents activity-dependent synaptic growth. (A) Representative images of fixed w1118 larval NMJs stained with FITC-conjugated anti-HRP and anti-Brp antibodies. (B) The number of presynaptic boutons was significantly higher in control larvae grown at 30°C compared to control larvae grown at 23°C (P<0.01; n=16). (C) The number of presynaptic boutons was significantly higher in larvae fed 5 mM MβCD grown at 30°C compared to larvae fed 5 mM MβCD grown at 23°C (P<0.05; n=10). (D) The number of presynaptic boutons was not significantly different in larvae fed 10 mM MβCD grown at 30°C compared to larvae fed 10 mM MβCD grown at 23°C (P>0.05; n=10). (E) The number of active zones was significantly higher in control larvae grown at 30°C compared to control larvae grown at 23°C (P<0.05; n=16). (F) The number of active zones was not significantly different in larvae fed 5 mM MβCD grown at 30°C compared to larvae fed 5 mM MβCD grown at 23°C (P>0.05; n=10). (G) The number of active zones was not significantly different in larvae fed 10 mM MβCD grown at 30°C compared to larvae fed 10 mM MβCD grown at 23°C (P>0.05; n=10). Values represent the mean±s.e.m. **P<0.01 (two-tailed unpaired t-test).
Reduction in cholesterol prevents activity-dependent synaptic growth. (A) Representative images of fixed w1118 larval NMJs stained with FITC-conjugated anti-HRP and anti-Brp antibodies. (B) The number of presynaptic boutons was significantly higher in control larvae grown at 30°C compared to control larvae grown at 23°C (P<0.01; n=16). (C) The number of presynaptic boutons was significantly higher in larvae fed 5 mM MβCD grown at 30°C compared to larvae fed 5 mM MβCD grown at 23°C (P<0.05; n=10). (D) The number of presynaptic boutons was not significantly different in larvae fed 10 mM MβCD grown at 30°C compared to larvae fed 10 mM MβCD grown at 23°C (P>0.05; n=10). (E) The number of active zones was significantly higher in control larvae grown at 30°C compared to control larvae grown at 23°C (P<0.05; n=16). (F) The number of active zones was not significantly different in larvae fed 5 mM MβCD grown at 30°C compared to larvae fed 5 mM MβCD grown at 23°C (P>0.05; n=10). (G) The number of active zones was not significantly different in larvae fed 10 mM MβCD grown at 30°C compared to larvae fed 10 mM MβCD grown at 23°C (P>0.05; n=10). Values represent the mean±s.e.m. **P<0.01 (two-tailed unpaired t-test).
To determine whether the effects of MβCD on synaptic growth and activity-dependent synaptic growth were dependent on cholesterol-extracting activity of MβCD, we fed larvae a 1:10 cholesterol–MβCD complex to add cholesterol (Churchward et al., 2005). This 1:10 cholesterol–MβCD complex has previously been used as a control to show cholesterol-specific effects of MβCD (Suzuki et al., 2004; Dason et al., 2010; Smith et al., 2010). We found no significant differences in the number of presynaptic boutons or active zones of w1118 larvae grown at 23°C on control food in comparison to w1118 larvae grown at 23°C food supplemented with a 1:10 cholesterol–MβCD complex (Fig. 5A–C). Thus, the 1:10 cholesterol–MβCD complex did not inhibit synaptic growth. Similarly, we found no significant differences in the number of presynaptic boutons or active zones of w1118 larvae grown at 30°C on control food in comparison to w1118 larvae grown at 30°C food supplemented with a 1:10 cholesterol–MβCD complex (Fig. 5D–F). Thus, the 1:10 cholesterol–MβCD complex did not inhibit activity-dependent synaptic growth. The inability of 1:10 cholesterol–MβCD to reduce synaptic growth or block activity-dependent synaptic growth was not due to the failure of the 1:10 cholesterol–MβCD complex to extract some other lipophilic membrane constituent that was extracted by MβCD. We previously found that both 10 mM MβCD and 1:10 cholesterol–MβCD took up similar amounts of the lipophilic compound FM1-43 (Dason et al., 2010). Thus, even though MβCD might extract some other lipophilic molecule, the reduced synaptic growth and block of activity-dependent synaptic growth that we observed in larvae fed 10 mM MβCD is attributable to the extraction of cholesterol, given that both 10 mM MβCD and 1:10 cholesterol–MβCD can extract other lipophilic molecules.
Reduced synaptic growth and activity-dependent synaptic growth are cholesterol dependent. (A,D) Representative images of fixed w1118 larval NMJs stained with FITC-conjugated anti-HRP and anti-Brp antibodies. (B) The number of presynaptic boutons in larvae grown at 23°C and fed 1:10 cholesterol-MβCD was not significantly different in comparison to controls (P=0.502; n=12). (C) The number of active zones in larvae grown at 23°C and fed 1:10 cholesterol-MβCD was not significantly different in comparison to controls (P=0.341; n=12). (E) The number of presynaptic boutons in larvae grown at 30°C and fed 1:10 cholesterol-MβCD was not significantly different in comparison to controls (P=0.618; n=10-12). (F) The number of active zones in larvae grown at 30°C and fed 1:10 cholesterol-MβCD was not significantly different in comparison to controls (P=0.335; n=10–12). Values represent the mean±s.e.m. Significance was determined using two-tailed unpaired t-test.
Reduced synaptic growth and activity-dependent synaptic growth are cholesterol dependent. (A,D) Representative images of fixed w1118 larval NMJs stained with FITC-conjugated anti-HRP and anti-Brp antibodies. (B) The number of presynaptic boutons in larvae grown at 23°C and fed 1:10 cholesterol-MβCD was not significantly different in comparison to controls (P=0.502; n=12). (C) The number of active zones in larvae grown at 23°C and fed 1:10 cholesterol-MβCD was not significantly different in comparison to controls (P=0.341; n=12). (E) The number of presynaptic boutons in larvae grown at 30°C and fed 1:10 cholesterol-MβCD was not significantly different in comparison to controls (P=0.618; n=10-12). (F) The number of active zones in larvae grown at 30°C and fed 1:10 cholesterol-MβCD was not significantly different in comparison to controls (P=0.335; n=10–12). Values represent the mean±s.e.m. Significance was determined using two-tailed unpaired t-test.
Cholesterol regulates synaptic growth through the presynaptic cAMP-PKA pathway
Synaptic growth and activity-dependent synaptic growth have previously been shown to be dependent on the presynaptic cAMP-PKA kinase signaling pathway (Zhong and Wu, 2004; Yoshihara et al., 2005; Cho et al., 2015). Cholesterol enrichment has been shown to upregulate the cAMP-PKA pathway in cultured rat cerebellar granule neurons (Zhou et al., 2012). Furthermore, loss of cholesterol has been shown to inhibit adenylyl cyclase function (Whetton et al., 1983; Fagan et al., 2000). Consistent with previous studies (Zhong et al., 1992), we found that presynaptic knockdown of rutabaga (rut), which encodes an adenylyl cyclase, reduced synaptic growth (Fig. 6A) compared to controls (Fig. 3A). We hypothesized that if cholesterol was acting on adenylyl cyclase or in the cAMP-PKA kinase signaling pathway, then the effects of knocking down rut and reducing cholesterol would not be additive. Consistent with this hypothesis, we found the effects of knocking down rut and reducing cholesterol on the number of presynaptic boutons were not additive, as there were no significant differences in the number of presynaptic boutons between larvae with rut knocked down grown on control food or food supplemented with 5 or 10 mM MβCD (Fig. 6A,C). However, there was a further mild decrease in the number of active zones in larvae with rut knocked down and grown on food supplemented with 10 mM MβCD (Fig. 6A,D). Overall, our data suggests that rut (adenylyl cyclase) and cholesterol function in the same pathway.
Cholesterol regulates activity-dependent synaptic growth through the cAMP pathway. (A,B) Representative images of fixed n-syb-GAL4>UAS-rut-RNAi or dnc1 larval NMJs stained with FITC-conjugated anti-HRP and anti-Brp antibodies. (C) The number of presynaptic boutons in n-syb-GAL4>UAS-rut-RNAi larvae grown on control food was not significantly different in comparison to n-syb-GAL4>UAS-rut-RNAi larvae fed 5 or 10 mM MβCD (P>0.05; n=9). (D) The number of active zones in n-syb-GAL4>UAS-rut-RNAi larvae grown on control food was not significantly different in comparison to n-syb-GAL4>UAS-rut-RNAi larvae fed 5 mM MβCD (P>0.05; n=9) but was significantly higher in comparison to n-syb-GAL4>UAS-rut-RNAi larvae fed 10 mM MβCD (P<0.05; n=9). (E) The number of presynaptic boutons in dnc1 larvae fed 5 or 10 mM MβCD was significantly reduced compared to dnc1 larvae grown on control food (P<0.001; n=9). (F) The number of active zones in dnc1 larvae fed 5 or 10 mM MβCD was significantly reduced compared to dnc1 larvae grown on control food (P<0.001; n=9). Values represent the mean±s.e.m. *P<0.05; ***P<0.001 (one-way ANOVA with Tukey post test).
Cholesterol regulates activity-dependent synaptic growth through the cAMP pathway. (A,B) Representative images of fixed n-syb-GAL4>UAS-rut-RNAi or dnc1 larval NMJs stained with FITC-conjugated anti-HRP and anti-Brp antibodies. (C) The number of presynaptic boutons in n-syb-GAL4>UAS-rut-RNAi larvae grown on control food was not significantly different in comparison to n-syb-GAL4>UAS-rut-RNAi larvae fed 5 or 10 mM MβCD (P>0.05; n=9). (D) The number of active zones in n-syb-GAL4>UAS-rut-RNAi larvae grown on control food was not significantly different in comparison to n-syb-GAL4>UAS-rut-RNAi larvae fed 5 mM MβCD (P>0.05; n=9) but was significantly higher in comparison to n-syb-GAL4>UAS-rut-RNAi larvae fed 10 mM MβCD (P<0.05; n=9). (E) The number of presynaptic boutons in dnc1 larvae fed 5 or 10 mM MβCD was significantly reduced compared to dnc1 larvae grown on control food (P<0.001; n=9). (F) The number of active zones in dnc1 larvae fed 5 or 10 mM MβCD was significantly reduced compared to dnc1 larvae grown on control food (P<0.001; n=9). Values represent the mean±s.e.m. *P<0.05; ***P<0.001 (one-way ANOVA with Tukey post test).
We hypothesized that increasing cAMP levels would rescue the reduced synaptic growth seen following cholesterol extraction if cholesterol was acting upstream of cAMP. The dunce (dnc) gene encodes a phosphodiesterase II enzyme that hydrolyzes cAMP and dnc1 mutants have elevated cAMP levels (Byers et al., 1981). Consistent with previous studies (Zhong et al., 1992), we found dnc1 mutants grown on control food had significantly more boutons and active zones (Fig. 6B) than did controls (Fig. 3A). Similar to w1118 larvae, dnc1 larvae grown on food supplemented with 5 or 10 mM MβCD had significantly fewer boutons and active zones than dnc1 mutants grown on control food (Fig. 6B,E,F). Thus, the dnc1 mutation was unable to rescue the reduced synaptic growth caused by cholesterol extraction, suggesting that cholesterol is acting downstream of cAMP.
We hypothesized that reduced PKA activity could be responsible for the reduced synaptic growth seen when cholesterol is reduced. To measure PKA activity in presynaptic boutons at the larval NMJ, we employed the GAL4/UAS system to drive expression of PKA-SPARK (Zhang et al., 2018) using a neuronal n-syb-GAL4. The PKA-SPARK reporter is phosphorylated specifically by PKA and generates reversible GFP puncta (Zhang et al., 2018). Previous Drosophila studies have shown that these puncta can be imaged in fixed preparations (Sears et al., 2019; Sears and Broadie, 2020; Sears and Broadie, 2022). We examined 1b boutons because they are larger in diameter than 1s boutons, which allowed us to quantify the number of PKA-SPARK puncta in individual 1b boutons. We found a significant reduction in the number of PKA-SPARK puncta in 1b boutons when cholesterol was reduced in comparison to controls (Fig. 7), demonstrating a reduction in presynaptic PKA activity. Collectively, these data demonstrate that cholesterol extraction reduces presynaptic PKA activity and that this reduction in PKA activity is likely the cause of the reduced synaptic growth following cholesterol loss.
Cholesterol loss reduces presynaptic PKA activity. (A) Representative images of fixed 1b boutons of n-syb-GAL4>UAS-PKA-SPARK stained with an anti-GFP antibody. (B) 1b boutons of n-syb-GAL4>UAS-PKA-SPARK larvae fed 10 mM MβCD had significantly fewer PKA-SPARK puncta compared to n-syb-GAL4>UAS-PKA-SPARK larvae grown on control food (P<0.01; n=9). Values represent the mean±s.e.m. **P<0.01 (two-tailed unpaired t-test).
Cholesterol loss reduces presynaptic PKA activity. (A) Representative images of fixed 1b boutons of n-syb-GAL4>UAS-PKA-SPARK stained with an anti-GFP antibody. (B) 1b boutons of n-syb-GAL4>UAS-PKA-SPARK larvae fed 10 mM MβCD had significantly fewer PKA-SPARK puncta compared to n-syb-GAL4>UAS-PKA-SPARK larvae grown on control food (P<0.01; n=9). Values represent the mean±s.e.m. **P<0.01 (two-tailed unpaired t-test).
DISCUSSION
Our study is the first to show that activity-dependent changes in presynaptic cholesterol levels regulate synaptic growth. Specifically, we found increased levels of cholesterol in presynaptic terminals following periods of increased synaptic activity. Reduced levels of cholesterol impaired synaptic growth and prevented activity-dependent synaptic growth. These effects might be due to impaired cAMP-PKA signaling, as presynaptic PKA activity was reduced following cholesterol loss.
Activity-dependent changes in presynaptic cholesterol
The membrane lipid composition is thought to be relatively static. However, increasing evidence suggests that the cholesterol content of neuronal membranes can change in response to altered synaptic activity (Sodero et al., 2012; Brachet et al., 2015). Strong neuronal activity, linked to excitotoxicity, has been shown to reduce cholesterol levels in cultured hippocampal neurons (Sodero et al., 2012). Furthermore, a rapid redistribution of postsynaptic cholesterol in hippocampal neurons is triggered by NMDA receptor activation during LTP induction (Brachet et al., 2015). Niemann-Pick disease type C1 (NPC1) protein enables this cholesterol redistribution to occur during LTP (Mitroi et al., 2019). We found that presynaptic cholesterol levels increased following a 5× spaced high KCl stimulation paradigm (Fig. 2B–E) that was previously shown to alter synaptic structure (Ataman et al., 2008). Interestingly, although our stimulation paradigm resulted in an increase in presynaptic cholesterol content, a previous study used chemical induction of LTP and found decreased cholesterol in cultured hippocampal postsynaptic neurons (Brachet et al., 2015). It is tempting to speculate that these differences could be due to different stimulation paradigms having differential effects on the cholesterol content of neurons. Alternatively, cholesterol levels might be differentially regulated in presynaptic and postsynaptic neurons. The mechanisms responsible for the changes in presynaptic cholesterol content that we observed remain to be determined. These changes in presynaptic cholesterol content could be due to altered cholesterol metabolism or trafficking. The transgenic D4H–GFP flies that we generated in this study will be useful in future studies to screen for genes involved in cholesterol trafficking and transport in presynaptic terminals. A role for cholesterol in developmental plasticity has previously been shown in chicken auditory hair cells. Specifically, depletion of cholesterol by MβCD potentiated K+ currents in developing hair cells but had no effect on K+ currents in mature hair cells (Levic and Yamoah, 2011). Here, we found that cholesterol is required for activity-dependent synaptic growth at the Drosophila larval NMJ, as a reduction in cholesterol largely blocked this plasticity from occurring (Fig. 4). Our data and these previous studies point to a role for changes in the cholesterol content of neuronal membranes as a mechanism for synaptic and developmental plasticity.
Growing evidence suggests that modulation of lipid content of membranes is an underlying mechanism of synaptic plasticity. Specific sphingolipids have been shown to be elevated in the Drosophila adult brain during neurite retraction at dusk (Vaughen et al., 2022). Thus, the sphingolipid content of neuronal membranes can change in response to specific external cues, and this too can be a mechanism to alter synaptic structure and function. Cholesterol and sphingolipids are thought to self-aggregate together into membrane domains (Rohrbough and Broadie, 2005). Drosophila serine palmitoyltransferase long chain base subunit 2 (symbol lace) mutants have depleted levels of sphingolipids and reduced number of presynaptic boutons at the Drosophila larval NMJ (West et al., 2018). Phosphatidylserine has also been shown to regulate synaptic growth at the Drosophila larval NMJ (Park et al., 2021) and phosphatidylserine is required to retain cholesterol in the cytosolic leaflet of the plasma membrane in non-neuronal cell lines (Maekawa and Fairn, 2015). Thus, reduced levels of sphingolipids (West et al., 2018), phosphatidylserine (Park et al., 2021) or cholesterol (Fig. 3A–C) result in reduced synaptic growth at the Drosophila larval NMJ. Collectively, these data demonstrate the importance of the lipid content of neuronal membranes in synaptic growth and proper nervous system development.
Cholesterol and neurotransmission
Several studies have shown that homeostatic processes stabilize global synaptic strength when the total number of boutons and active zones are drastically altered at the Drosophila larval NMJ (Stewart et al., 1996; Davis and Goodman, 1998; Mosca et al., 2005; Goel et al., 2019a,b; Goel and Dickman, 2021; Wang et al., 2021). We found that reduced cholesterol levels did not affect global synaptic strength (evoked neurotransmission) (Fig. 3D,E), despite reducing the number of presynaptic boutons and active zones (Fig. 3A–C). Thus, it is possible that homeostatic processes are occurring in response to reduced synaptic growth in larvae with reduced cholesterol levels. Alternatively, cholesterol might have two distinct effects on neurotransmission and synaptic growth. We favor the latter explanation for several reasons. First, acute extraction of cholesterol from presynaptic plasma membranes has been shown to increase release probability in cultured cerebellar neurons (Smith et al., 2010) and increase evoked neurotransmission at the Drosophila larval NMJ (Dason et al., 2010). Therefore, it seems likely that evoked neurotransmission is increased at individual presynaptic boutons owing to a direct effect of loss of cholesterol instead of homeostatic changes. Second, acute extraction of cholesterol from presynaptic plasma membranes in several preparations has been shown to increase spontaneous neurotransmission (Zamir et al., 2006; Wasser et al., 2007; Smith et al., 2010; Dason et al., 2010). However, we found that chronic reduction in cholesterol caused a mild decrease in spontaneous neurotransmission (Fig. 3G). This mild decrease was likely due to the reduction in the number of boutons and active zones seen following chronic cholesterol extraction (Fig. 3A–C). Overall, cholesterol appears to have distinct effects on synaptic growth and neurotransmission.
Cholesterol loss did not appear to have any obvious postsynaptic effects (Fig. 3D,F). These findings are consistent with our previous studies that found acute cholesterol extraction did not affect the amplitude of mEJPs (Dason et al., 2010) or the distribution of postsynaptic glutamate receptors (Dason and Charlton, 2014). Similarly, cholesterol extraction from Purkinje cells does not have any postsynaptic effects (Smith et al., 2010). Nevertheless, acute cholesterol extraction has been shown to reduce the probability of the opening of postsynaptic NMDA receptors in hippocampal autaptic neurons (Korinek et al., 2020).
Cholesterol and the cAMP-PKA pathway
Several in vitro studies have linked cholesterol to the cAMP-PKA pathway (Burgos et al., 2004; Zhou et al., 2012). Cholesterol enrichment of membranes has been associated with increased PKA activity (Zhou et al., 2012). We hypothesized that the activity-dependent increase in presynaptic cholesterol that we observed (Fig. 2B–E) was regulating synaptic growth through PKA. We used an in vivo PKA activity sensor (PKA-SPARK) (Zhang et al., 2018) to show cholesterol depletion reduces PKA activity in presynaptic terminals (Fig. 7). Changes in the cholesterol content of membranes might have direct or indirect effects on PKA. Loss of cholesterol has been shown to inhibit adenylyl cyclase function (Whetton et al., 1983; Fagan et al., 2000). Consistent with these findings, we found that the effects of knocking down presynaptic adenylyl cyclase and reducing cholesterol on synaptic growth were not additive (Fig. 6A,C,D), suggesting that loss of cholesterol might affect adenylyl cyclase activity. However, our data suggests a downstream effect, as increasing cAMP levels using a dnc1 mutant with reduced phosphodiesterase activity (Fig. 6B,E,F) was unable to rescue the impaired synaptic growth caused by reducing cholesterol. The increased synaptic growth seen in dnc1 mutants is restored to wild-type levels in dnc1, rut double mutants (Zhong et al., 1992). If loss of cholesterol was acting solely through adenylyl cyclase, one would expect synaptic growth to similarly be restored to wild-type levels in dnc1 mutants with reduced cholesterol levels. Overall, our data suggests that cholesterol extraction reduces synaptic growth and blocks activity-dependent synaptic growth by reducing presynaptic PKA activity.
PKA-dependent phosphorylation of complexin has previously been shown to regulate activity-dependent synaptic growth at the Drosophila larval NMJ (Cho et al., 2015). In addition, presynaptic synapsin has been shown to facilitate the budding of new boutons via a cAMP-PKA-dependent pathway at the Drosophila larval NMJ (Vasin et al., 2014, 2019). The precise targets of PKA phosphorylation in cholesterol-dependent synaptic growth are beyond the scope of the current study, but both complexin and synapsin are potential candidates.
Collectively, our data demonstrate that changes in synaptic activity alter the cholesterol content of presynaptic terminals, which in turn regulate presynaptic PKA signaling during development (Fig. 8). These findings point to a key role for membrane lipids, such as cholesterol, in activity-dependent synaptic plasticity and in development of the nervous system.
Cholesterol is required for activity-dependent synaptic growth. Increased synaptic activity causes an increase in presynaptic cholesterol that activates PKA leading to increased synaptic growth. Adenylyl cyclase can also produce cAMP, leading to increased PKA activity and increased synaptic growth.
Cholesterol is required for activity-dependent synaptic growth. Increased synaptic activity causes an increase in presynaptic cholesterol that activates PKA leading to increased synaptic growth. Adenylyl cyclase can also produce cAMP, leading to increased PKA activity and increased synaptic growth.
MATERIALS AND METHODS
Fly stocks
Fly stocks were grown at 23°C or 30°C on molasses-based fly food medium (Dason et al., 2020). To reduce cholesterol levels in larvae, food was supplemented with either 5 or 10 mM MβCD (Sigma-Aldrich). For cholesterol–MβCD experiments, 1:10 cholesterol–MβCD was added to the food. Cholesterol–MβCD was prepared by dissolving cholesterol in chloroform as previously described (Churchward et al., 2005; Dason et al., 2010). 10 mM MβCD was then added to the dried film at a molar ratio of 1:10 (cholesterol–MβCD). Male or female wandering third-instar larvae were used for all experiments, as sex differences were not previously seen in NMJs of muscle fibers 6 and 7 (Lnenicka et al., 2006). w1118 was used as the control strain (Dason et al., 2009). The dnc1 hypomorphic allele contains an ethyl methanesulfonate (EMS)-induced mutation in the dnc gene (Dudai et al., 1976), which is known to decrease the phosphodiesterase II activity, leading to a higher concentration of cAMP (Zhong et al., 1992). The GAL4-UAS system (Brand and Perrimon, 1993) was used for tissue-specific expression of transgenes. n-syb-GAL4 (Verstreken et al., 2009) was used to drive expression of transgenes in neurons, Repo-GAL4 was used to drive expression in glia (Sepp et al., 2001), and 24B-GAL4 was used to drive expression in muscles (Sweeney et al., 1995). UAS-D4H-GFP (generated in the present study) was used as a cholesterol biosensor. UAS-rut-RNAi (Bloomington Drosophila Stock Center #80468) was used to knockdown rut (Zhao et al., 2020). UAS-SPARK-PKA was used to visualize PKA activity (Zhang et al., 2018).
Cloning
mCherry–D4H was isolated by digesting a pmCherry-C1 vector that contained D4H (Maekawa and Fairn, 2015) with restriction enzymes EcoRI and XhoI (New England BioLabs). D4H was then amplified by PCR using the forward primer 5′-AAGGGAAAAATAAACTTAGATCATAGTGGA-3′ and reverse primer 5′-TTAATTGTAAGTAATACTAGATCCAGGGTAT-3′. D4H was then amplified with the following primers to create KpnI and NotI restriction sites (underlined in primer sequences): 5′-TATGCGGCCGCAAGGGAAAAATAAACTTAGATC-3′ (sense primer) and 5′-ATAGGTACCTTAATTGTAAGTAATACTAGATCC-3′ (antisense primer). The PCR product was subsequently subcloned into the pJFRC14-10XUAS-IVS-GFP-WPRE vector (Addgene plasmid #26223; Pfeiffer et al., 2010) at KpnI and NotI restriction sites.
Immunohistochemistry
Immunohistochemistry was performed as previously described by Dason et al. (2009). Briefly, the wandering third-instar larvae were dissected and fixed with Bouin's fixative for 5 min. Preparations were incubated overnight at 4°C with fluorescein isothiocyanate (FITC)-conjugated anti-horseradish peroxidase (HRP) antibody (1:800 dilution, cat. no. 123-095-021, Jackson ImmunoResearch) to visualize neurons, the mouse monoclonal Bruchpilot (Brp) antibody (1:100 dilution; cat. no. nc82, Developmental Studies Hybridoma Bank, Iowa City, IA, USA; Wagh et al., 2006) to visualize active zones or mouse monoclonal 3E6 anti-GFP antibody (1:500 dilution; cat. no. A-11120, Invitrogen Thermo Fisher Scientific) to visualize GFP. Samples were mounted in Permafluor (Immunon, Pittsburgh, PA, USA) on a glass slide with a coverslip and viewed under an Olympus FV1000 confocal laser-scanning microscope with a 60× oil-immersion objective (1.42 NA) or a Leica TCS SP5 confocal laser-scanning microscope with a 63× oil-immersion objective (1.4 NA) or a Zeiss LSM 900 confocal laser-scanning with a 63× oil-immersion objective (1.4 NA).
Electrophysiology
Intracellular recordings were performed as previously described (Dason et al., 2009). Briefly, the ventral longitudinal muscle fiber 6 or 7 (abdominal segment 3) of dissected larvae was impaled with a sharp glass electrode filled with 3 M KCl (∼40 MΩ) to measure spontaneously occurring miniature excitatory junction potentials (mEJPs) and stimulus-evoked excitatory junction potentials (EJPs) in HL6 saline supplemented with 1 mM CaCl2 (Macleod et al., 2002). Cut segmental nerves were stimulated at 0.05 Hz using a suction electrode. Electrical signals were recorded using the Powerlab/8SP data acquisition system (ADInstruments).
Live imaging
Live D4H–GFP imaging experiments were performed using a Leica TCS SL confocal laser-scanning microscope with a 63× water dipping objective (0.90 NA). An image was taken of each preparation prior to stimulation in calcium free HL6 saline. Preparations were then stimulated by repetitive, spaced depolarizations, consisting of five high (90 mM) K+-pulses for 5 min spaced by 15 min of rest (Ataman et al., 2008). High K+ depolarization was induced using the following high K+ saline (25 mM NaCl, 90 mM KCl, 10 mM NaHCO3, 5 mM HEPES, 30 mM sucrose, 5 mM trehalose, 10 mM MgCl2, 2 mM CaCl2, pH 7.2) (Verstreken et al., 2008). During rest, preparations were maintained in Ca2+-free HL6 saline. An image was taken of each preparation in Ca2+-free HL6 saline at the end of the stimulation protocol. Images of live preparations in figures are single scans.
Statistical analysis
SigmaPlot (version 11.0; Systat Software) was used for statistical analysis. Unpaired two-tailed t-tests were used for comparing datasets of two groups, and one-way ANOVA tests (with a Kruskal–Wallis or Holm–Sidak post hoc test) were used for comparing datasets of more than two groups. n represents the number of preparations analyzed. Error bars in all figures represent s.e.m.
Acknowledgements
We thank Dr Xiaokun Shu and the Bloomington Drosophila Stock Center for fly stocks. We thank Dr Gregory Fairn and Dr Julie Brill for sharing the pmCherry-C1 vector with D4H.
Footnotes
Author contributions
Conceptualization: A. Shaheen, J.S.D.; Methodology: A. Shaheen, C.L.R.G., A. Sghaier, J.S.D.; Formal analysis: A. Shaheen, C.L.R.G., J.S.D.; Investigation: A. Shaheen, C.L.R.G., A. Sghaier, J.S.D.; Resources: J.S.D.; Writing - original draft: J.S.D.; Writing - review & editing: A. Shaheen, C.L.R.G., A. Sghaier, J.S.D.; Supervision: J.S.D.; Project administration: J.S.D.; Funding acquisition: J.S.D.
Funding
This work was supported by a grant from Natural Sciences and Engineering Research Council of Canada (NSERC) (RGPIN #06582 to J.S.D.) and Ontario Graduate Scholarships (A.S. and C.L.R.G.).
Data availability
All relevant data can be found within the article.
References
Competing interests
The authors declare no competing or financial interests.