Glycolytic enzymes localize to synapses under energy stress to support synaptic function

Glycolytic proteins are required to maintain synaptic vesicle protein clusters during energy stress

We conducted unbiased forward genetic screens to identify molecules required for the localization of synaptic vesicle proteins in the serotonergic NSM neurons in the nematode C. elegans (Figure 1A–1D, and 1H; genetic screen described in Figure S1 and Supplemental Experimental Procedures). From this screen we identified allele ola72, which displayed diffuse distribution of the synaptic vesicle proteins VMAT/CAT-1 and synaptobrevin under hypoxic conditions (Figure 1E–1G, 1I, and 1N). Positional cloning of the ola72 allele revealed a missense mutation (C562Y) in pfk-1.1 (Figure S2A) – one of two C. elegans genes that encode phosphofructokinase-1. In mutant animals carrying the ola72 allele, synaptic vesicle proteins cluster normally under normoxic conditions (Figures 1J, 1M, 1O, S1C, and S1L). However, under hypoxic conditions (induced by mounting animals under a glass coverslip (Pitts and Toombs, 2004), or by incubation in a hypoxia chamber) these synaptic vesicle proteins become diffusely distributed throughout the neurite in ola72 animals (Figures 1J, 1N, 1P, and S1F; Movie S1). This phenotype was not observed in wild-type animals, which maintained punctate localization of synaptic vesicle proteins under the same hypoxic conditions (Figures 1J–1L, and S1E). Three independent alleles, pfk-1.1(gk549413), pfk-1.1(gk758818), and pfk-1.1(gk922689), phenocopy and fail to complement the ola72 allele (Figures 2C, 2I, and S2C). Expression of a wild-type copy of the pfk-1.1 gene in ola72 mutant animals rescues punctate localization of synaptic vesicle proteins under hypoxic conditions (Figure S2B). These findings indicate that phosphofructokinase-1 is required to maintain the localization of synaptic vesicle proteins at synapses under hypoxic conditions.

The pfk-1.1 phenotype is not due to a disruption of the synapses themselves, since the localization of the presynaptic active zone protein ELKS is not altered in pfk-1.1 mutants (Figures S3A–S3E). In addition, the requirement for pfk-1.1 is not limited to the NSM neuron, as it is also required for the maintenance of vesicle protein clustering in all neurons examined (Figures 1K–1P, S3A–S3J). Consistent with the pan-neuronal phenotype, a GFP reporter under the pfk-1.1 promoter is expressed in all examined neurons, including the NSM neuron (Figures S2D–S2G). Moreover, NSM neuron-specific expression of the pfk-1.1 cDNA in pfk-1.1(ola72) mutants led to rescue of the synaptic vesicle phenotype in the NSM neuron (Figures 2B and 2I). Our findings indicate that the phosphofructokinase enzyme is required cell autonomously in neurons to cluster synaptic vesicle proteins under hypoxic conditions.

Phosphofructokinase-1 is a rate-limiting enzyme that catalyzes the first committed step during glycolysis. To determine if the glycolytic pathway is required for the maintenance of synaptic vesicle protein clusters in neurons under hypoxic conditions, we examined mutants of other glycolytic enzymes (Figure S2H). Mutants for phosphofructokinase-2/fructose-2,6-bisphosphatase/pfkb-1.1(ok2733), glyceraldehyde 3-phosphate dehydrogenase/gpd-3(ok2870), aldolase/aldo-1(tm5782), and phosphoglycerate kinase/pgk-1(tm5613) phenocopy the hypoxia-dependent synaptic vesicle protein phenotype of pfk-1.1 mutant animals (Figures 2C–2I).

Hypoxia inhibits oxidative phosphorylation(Cohen, 1972). To determine if disruption of oxidative phosphorylation phenocopied the effects of hypoxia, we pharmacologically blocked cytochrome oxidase with sodium azide (NaN₃) and the ATP synthase with oligomycin (Bogucka and Wojtczak, 1966; Chappell and Greville, 1961). We observed that pfk-1.1 mutants exhibit hypersensitivity to sodium azide and oligomycin, similar to their acute response to lowered oxygen, but do not exhibit hypersensitivity to glycolysis inhibitor 2-deoxy-D-glucose (2-DG) (Woodward and Hudson, 1954) (Figures 1J and S3K). Together our findings indicate that in vivo, under conditions of energy stress in which the activity of the oxidative phosphorylation pathway is decreased (either pharmacologically or by hypoxia), the glycolytic pathway is required in C. elegans neurons for the maintenance of synaptic vesicle protein clusters.

Endocytosis is disrupted in pfk-1.1 mutants during hypoxia

Synaptic vesicle endocytosis is vulnerable to ATP levels, and pharmacological or genetic inhibition of glycolysis dramatically and specifically reduces endocytosis at presynaptic sites (Rangaraju et al., 2014; Wang et al., 2004). To examine if pfk-1.1 mutants disrupt synaptic vesicle endocytosis, we first visualized vesicle protein clusters in endocytic mutant unc-57/endophilin and pfk-1.1;unc-57/endophilin double mutants. Mutants lacking unc-57/endophilin display a diffuse distribution of synaptic vesicle proteins throughout the neurite even under normoxic conditions (Schuske et al., 2003) (Figures 3A–3F). This phenotype resembles that seen for glycolytic mutants under conditions of energetic stress (Figures 2C–2I). Consistent with endophilin/unc-57 and pfk-1.1 acting in the same genetic pathway, we observed that mutations in endophilin did not enhance the hypoxia-induced pfk-1.1 phenotype (Figure 3J). These observations suggest that the synaptic vesicle defect observed in pfk-1.1 mutant animals could result from inhibition of endocytosis due to decreased rates of glycolysis.

If pfk-1.1 mutants disrupt synaptic vesicle endocytosis during energy stress, vesicle proteins would remain trapped on the plasma membrane, as has been observed for other endocytosis mutants (Bai et al., 2010; Kraszewski et al., 1996). To examine this hypothesis, we analyzed the mobility of vesicle protein synaptobrevin using Fluorescence Recovery After Photobleaching (FRAP). Proteins associated with the plasma membrane exhibit greater mobility than proteins associated with synaptic vesicles (Bai et al., 2010; Kraszewski et al., 1996). We observed that synaptobrevin exhibited greater mobility in endocytic mutants than in wild type animals, as expected (Figures 3K and 3L). Consistent with the hypothesis that pfk-1.1 mutants affect endocytosis during energy stress, we observed that synaptobrevin also exhibited greater mobility in pfk-1.1 mutants under hypoxic conditions, and that the mobility phenocopied that seen for endocytosis mutants (Figures 3K and 3L). If pfk-1.1 mutants affect endocytosis during energy stress, we would expect that blocking exocytosis would suppress the pfk-1.1 mutant phenotype. Indeed, we observed that mutants for the synaptic vesicle docking protein UNC-13 partially suppressed the hypoxia-induced pfk-1.1 phenotype (Figures 3G–3J). The observed partial suppression may be due to the hypomorphic nature of the unc-13 allele, or to potential contributions to the phenotype from other, unc-13 independent pathways. The phenotype in pfk-1.1; unc-13 double mutants, however, indicates that a defect in the synaptic vesicle cycle significantly contributes to the pfk-1.1 mutant phenotype. Our results are consistent with previous studies (Rangaraju et al., 2014; Wang et al., 2004) and suggest that under conditions of energy depletion, the glycolytic pathway becomes critical in maintaining the energy supplies necessary for sustaining endocytosis and the synaptic vesicle cycle.

Glycolytic enzymes localize to presynaptic sites during energy stress

Where is PFK-1.1 localized to meet local energy demands at synapses? We examined the subcellular localization of PFK-1.1 in neurons. We observed that under normoxic conditions, PFK-1.1 is localized in a punctate pattern at some of the cell somas (Figure S4A and S4A′) and largely diffuse throughout the neurites (Figures 4A and 4C). However, upon exposure to hypoxia, PFK-1.1 became clustered in neurites (Figures 4A–4E; Movie S2), preferentially localizing to synaptic rich regions (Figure 4F) and within 0.2 um from vesicle release sites (Figures 4G and 4H).

Glycolytic proteins have been hypothesized to interact and form functional super-complexes, termed glycolytic metabolons (or glycolons), which sustain the accelerated rates of glycolysis (Clarke and Masters, 1975; Kurganov et al., 1985; Ureta, 1985). The concept of the glycolytic metabolon derives from evidence generated primarily from work in fixed tissues or biochemical assays, and less is known about the existence in vivo of this complex or its physiological importance (Bronstein and Knull, 1981; Campanella et al., 2008; Knull et al., 1980; Kurganov et al., 1985; Masters, 1984; Sullivan et al., 2003; Zhou et al., 2005). To examine if other glycolytic enzymes colocalize with PFK-1.1 in vivo, we visualized the subcellular localization of ALDO-1 and GPD-3. ALDO-1 and GPD-3, like PFK-1.1, were diffusely distributed throughout the cytosol under normoxic conditions (data not shown). Upon exposure to hypoxia, both enzymes clustered near synapses (Figures 4J and 4N). The GPD-3 and ALDO-1 clusters colocalized with the PFK-1.1 clusters (Figures 4I–4P), suggesting that these enzymes dynamically form a complex near presynaptic sites in response to demands for ATP.

Aerobic glycolysis normally occurs in the brain, and increases locally upon conditions of increased neuronal activity (Vaishnavi et al., 2010). To test if the glycolytic metabolon is necessary at synapses in response to high levels of neuronal activity we stimulated neurons using a pharmacological approach. C. elegans GABA neurons can be stimulated through the bath application of levamisole, an agonist of acetylcholine receptors (such as lev-8) present in the GABA neurons (Towers et al., 2005). By applying levamisole under normoxic conditions, we examined whether pfk-1.1 was required to support vesicle cycling during periods of strong neuronal stimulation. We observed that pfk-1.1 mutant animals were sensitized to levamisole treatment compared to wild-type animals (Figures 5A–5C). This sensitization is specific to the GABA neurons that express levamisole-sensitive acetylcholine receptors (data not shown). These findings suggest that even in aerobic conditions, glycolytic proteins are required to maintain the synaptic vesicle cycle during periods of high neuronal activity, which are known to result in increased energy demands at synapses (Rangaraju et al., 2014). Together, our findings suggest that glycolysis is critical for the maintenance of the synaptic vesicle cycle under conditions that either reduce energy supplies (such as hypoxia), or increase energy demands (such as neuronal stimulation).

Does PFK-1.1 cluster at synapses upon increased neuronal stimulation? Optogenetic stimulation of GABA neurons expressing channelrhodopsin caused clustering of PFK-1.1 in normoxic conditions (Figures 5D–5F). Therefore PFK-1.1 can dynamically localize near presynaptic sites upon neuronal stimulation. To test whether this relocalization is dependent on metabolic needs at the synapse, we examined, in hypoxic conditions, PFK-1.1 localization in unc-13 mutants, in which exocytosis and synaptic activity are greatly reduced (Richmond et al., 1999). We observed that the subcellular localization of PFK-1.1 to presynaptic sites is reduced in the neurites (but not the somas) of unc-13(e51) mutants (Figures 5G, 5H, and S4A). In normoxic conditions, optogenetic stimulation of GABA neurons did not promote PFK-1.1 clustering in unc-13 mutants, indicating that synaptic PFK-1.1 clustering does not simply depend on depolarization (Figure S4B). These results suggest that PFK-1.1 clustering to presynaptic sites relies on energy demands at synaptic sites.

Presynaptic localization of PFK-1.1 depends on synaptic scaffolding proteins

How do glycolytic proteins localize to presynaptic sites? Glycolytic proteins co-purify with synaptic vesicles and are required for their transport (Burre and Volknandt, 2007; Ikemoto et al., 2003; Ishida et al., 2009; Zala et al., 2013). Glycolytic proteins are also known to associate with the mitochondria, which are actively transported to synaptic sites (Giege et al., 2003). Therefore we next examined if the localization of glycolytic proteins to presynaptic sites depended on synaptic vesicle transport or on mitochondrial transport to synapses. We observed that in unc-104/kinesin 3/kif1A mutant animals, synaptic vesicles fail to cluster at presynaptic sites, as previously reported (Hall and Hedgecock, 1991; Nelson and Colón-Ramos, 2013). However, PFK-1.1 clustering under hypoxic conditions was not affected in unc-104(e1265) mutant animals (Figures 6A–6D, and 6O). Our data indicate that the clustering of PFK-1.1 is not dependent on vesicle transport to synapses.

To determine if the localization of PFK-1.1 to presynaptic sites depends on the mitochondria, we simultaneously visualized the subcellular localization of PFK-1.1 and mitochondrial outer-membrane protein TOM-20 in mutant backgrounds that affect mitochondria localization in neurites. We observed that in the mutant backgrounds of kinesin klp-6(sy511), ric-7(nu447) and mitochondrial fission drp-1(tm408) there was a reduction (or altered distribution) of mitochondria in the neurites, as previously reported ((Labrousse et al., 1999; Rawson et al., 2014; Tanaka et al., 2011) and data not shown). While the mitochondria localization was affected in these animals, we did not detect a reduction in the number of animals displaying PFK-1.1 clusters in the neurites (Figures 6E–6J, and 6O). Instead, we observed that klp-6(sy511), ric-7(nu447), and drp-1(tm408) mutants displayed PFK-1.1 clustering in neurites even before animals were exposed to hypoxic conditions (Figure 6O). Our findings indicate that glycolytic protein localization to presynaptic sites is not dependent on the mitochondria or unc-104/kinesin 3/kif1A synaptic vesicle transport. Our findings also uncover a relationship between PFK-1.1 cluster formation and mitochondrial dysfunction.

SYD-2 is a synaptic scaffolding protein required for synaptic release sites formation and maintenance (Zhen and Jin, 1999). To examine if the integrity of the presynaptic sites was necessary for the localization of glycolytic clusters in the neurite, we visualized PFK-1.1 in syd-2 mutant animals. We observed that in the syd-2(ok217) null allele (Wagner et al., 2009) and the syd-2(ju37) loss of function allele (Zhen and Jin, 1999), PFK-1.1 clustering at the neurites was significantly suppressed (Figures 6K, 6L, 6O, and 6P). Cell-specific expression of the wild-type SYD-2 in GABA neurons rescued the PFK-1.1 clustering in GABA neurons of syd-2 (ju37) mutants (Figures 6M–6O), demonstrating the cell autonomous role of this synaptic scaffolding protein in PFK-1.1 localization. Similar results were obtained for presynaptic protein SYD-1 (data not shown) (Hallam et al., 2002). Our findings demonstrate that presynaptic scaffolding proteins are cell autonomously required in neurons for the clustering of PFK-1, and underscore a functional link between the presynaptic release sites and the clustering of glycolytic proteins.

Localization of PFK-1.1 to synaptic sites is important for maintaining the synaptic vesicle cycle during energy stress

To test the importance of the subcellular localization of PFK-1.1 in sustaining energy levels at synapses, we identified a dominant negative version of PFK-1.1 (PFK-1.1G532E). PFK- 1.1G532E contains a lesion in the regulatory domain of PFK-1.1, fails to form clusters (Figures S4C–S4E), and fails to rescue the pfk-1.1 mutant phenotype (data not shown). We observed that expression of PFK-1.1G532E in wild-type animals altered wild-type PFK-1.1 localization to presynaptic sites, suggesting that PFK-1.1G532E acts as a dominant negative, affecting the subcellular localization of wild-type PFK-1.1 (Figures 7A–7D). In addition, these wild-type animals expressing PFK-1.1G532E display a dominant negative synaptic vesicle protein clustering phenotype similar to that observed for pfk-1.1 loss-of-function mutant animals (Figures 7E–7I). Because expression of PFK-1.1G532E affects wild-type PFK-1.1 localization, we hypothesize that the resulting phenotype occurs because of a disruption in the localization of the wild-type PFK-1.1 to presynaptic sites and, therefore, its capacity to sustain energy levels at the synapse. Consistent with these findings, we also observed that when we alter the localization of a wild-type version of PFK-1.1 with a nuclear localization sequence, the wild-type PFK-1.1 is incapable of rescuing the synaptic vesicle phenotype of pfk-1.1 mutants (Figures S4F–S4H). Our findings are consistent with studies in Drosophila flight muscles, which demonstrated that even when the full complement of glycolytic enzymes is present in the muscle, disruption of glycolytic protein colocalization to sarcomeres results in inability to fly (Wojtas et al., 1997). Together our findings suggest that the presence of functional glycolytic enzymes in the neurons is not sufficient to power synaptic function. Instead, subcellular localization of these glycolytic enzymes near synapses is necessary to power synaptic function.

PFK-1.1 is required to sustain synaptic activity and locomotion during energy stress

Is PFK-1.1 required to sustain synaptic activity? We examined the requirement of pfk-1.1 in synaptic recovery following fatigue. Animals expressing the channelrhodopsin variant ChIEF were dissected to expose neuromuscular synapses of the body muscles (Richmond et al., 1999; Richmond and Jorgensen, 1999). Postsynaptic muscles were voltage-clamped and high-frequency pulses of blue light were delivered to stimulate synapses for 30 seconds at 10Hz. This stimulation caused rapid fatigue of acetylcholine neuromuscular junctions (Liu et al., 2009), presumably by the depletion of vesicle pools. We applied test-pulses at increasing time increments after fatigue and assayed the recovery of evoked currents (Figures 8A and 8B). We observed that the absolute value of the first evoked response in the pfk-1.1 mutant is similar to the wild-type, demonstrating that the synapses of both wild-type and pfk-1.1 mutants are healthy and normal in response to an evoked stimulus (Figure 8C). Moreover, under normoxic conditions, both wild-type and pfk-1.1 mutant synapses reached full recovery (plateau amplitude +/−SEM: WT=1.1 +/−0.037; pfk-1.1=1.0 +/−0.046) with time constants of 8.9 and 8.6 seconds respectively (Figure 8D). However, with two and three repetitions of the 30 second stimulation protocol, pfk-1.1 mutants exhibited reduced recovery (72% (plateau amplitude: WT=0.88 +/−0.071; pfk-1.1=0.63 +/−0.086) and 40% (WT=0.55 +/−0.071; pfk-1.1=0.22 +/−0.043) of wild-type levels, respectively) (Figures 8E and 8F). Next we included oligomycin to block oxidative phosphorylation. The rate of recovery at wild-type synapses was slightly decreased but not significantly different from control conditions (time constant: oligo=11.9 sec vs. control=8.9 sec). Furthermore, the drug did not affect the maximum recovery (plateau amplitude: oligo=1.0 +/−0.047 vs. control=1.1 +/−0.037) (Figure 8G). By contrast, pfk-1.1 mutants recovered more slowly (time constant >35 seconds vs. WT = 12 seconds) and failed to reach full recovery (0.60 +/−0.17 of wild-type plateau amplitude) (Figure 8G). Therefore oxidative phosphorylation or glycolysis alone support ATP demands under moderate stimulation. However, glycolysis is required for synaptic function under persistent stimulus conditions, or upon inhibition of oxidative phosphorylation.

To determine if glycolysis is required for normal behavior, we tested whether pfk-1.1 mutant animals exhibited fatigue during swimming. Worms thrash rapidly in liquid, so we scored the number of body bends in solution per minute (Figure S5A). Consistent with behavioral impairments in pfk-1.1 mutants, we observed a significant reduction in the number of body bends in pfk-1.1 mutants compared to wild-type animals when oxidative phosphorylation was inhibited (Figures S5B and S5C). Together, these data suggest that disruption of the glycolytic metabolon impairs synaptic recovery and affects locomotion.

Inducing hypoxia with glass coverslips and slides

Glass coverslips have been used to induce hypoxia in cell cultures (Pitts and Toombs, 2004). The reduced environment generated by mounting 10–15 live worms on glass slides was examined by using redox indicator resazurin (25ug/mL) dissolved in water. Resazurin when reduced to resorufin gives off fluorescence under yellow-green light (O’Brien et al., 2000). Using resazurin, we confirmed that worms between a glass coverslip and slide experience reduced, or hypoxic, conditions; Leica DM500B compound fluorescent microscope was used to acquire the images (Figures S1M and S1N). Gas permeable slides were made with Sylgard-184 (polydimethylsiloxane, or PDMS) (Dow Corning) according to manufacturer instructions.

Inhibiting oxidative phosphorylation or glycolysis using a hypoxia chamber or pharmacological treatments

Worms were mounted on a slide, and exposed to nitrogen gas, carbon dioxide, 10mM sodium azide (NaN₃), 10uM 2-deoxy-D-glucose (2-DG), or 1uM of oligomycin for 10 minutes. Phenotype was scored immediately after the treatment.

Neuronal stimulation with pharmacological treatments or optogenetics

To pharmacologically stimulate GABA neurons, worms were mounted on a slide in 1mM levamisole (Sigma) and imaged immediately in spinning-disc confocal microscope (PerkinElmer Life and Analytical Sciences). As a control, 50mM of muscimol (Abcam) was used. To optogenetically stimulate GABA neurons, a strain expressing channelrhodopsin 2 in GABA neurons (oxIs352) (Liu et al., 2009) was stimulated with blue light (~0.6mW/mm²) for 5 minutes.

Quantification of phenotypic expressivity

To quantify synaptic enrichment of the indicated proteins (synaptic vesicle proteins or PFK-1.1), fluorescence values for individual neurites (ventral neurite for the NSM neuron, Zone 3 for the AIY neuron, and dorsal and ventral neurite for GABA neurons) were obtained through segmented line scans using ImageJ. A sliding window of 2μm was used to identify all the local fluorescence peak values and trough values for an individual neuron (the maximum and the minimum fluorescence values in a 2μm interval, respectively). Synaptic enrichment was then calculated as ΔF/F as previously described (Bai et al., 2010; Dittman and Kaplan, 2006).

Electrophysiology

Worms expressing the channelrhodopsin I/II chimera ChIEF in acetylcholine neurons (oxSi91[Punc-17::ChIEF]) and lacking GABA inputs to the muscle (unc-49(e407)) were dissected and patch-clamp physiology was conducted as previously described (Richmond et al., 1999; Richmond and Jorgensen, 1999). For oligomycin experiments, the dissected preps were exposed to 1uM oligomycin for 5 minutes prior to patch clamp recordings.