Oxygen influences behaviour in many organisms, with low levels (hypoxia) having devastating consequences for neuron survival. How neurons respond physiologically to counter the effects of hypoxia is not fully understood. Here, we show that hypoxia regulates the trafficking of the glutamate receptor GLR‐1 in C. elegans neurons. Either hypoxia or mutations in egl‐9, a prolyl hydroxylase cellular oxygen sensor, result in the internalization of GLR‐1, the reduction of glutamate‐activated currents, and the depression of GLR‐1‐mediated behaviours. Surprisingly, hypoxia‐inducible factor (HIF)‐1, the canonical substrate of EGL‐9, is not required for this effect. Instead, EGL‐9 interacts with the Mint orthologue LIN‐10, a mediator of GLR‐1 membrane recycling, to promote LIN‐10 subcellular localization in an oxygen‐dependent manner. The observed effects of hypoxia and egl‐9 mutations require the activity of the proline‐directed CDK‐5 kinase and the CDK‐5 phosphorylation sites on LIN‐10, suggesting that EGL‐9 and CDK‐5 compete in an oxygen‐dependent manner to regulate LIN‐10 activity and thus GLR‐1 trafficking. Our findings demonstrate a novel mechanism by which neurons sense and respond to hypoxia.
Aerobic organisms are sometimes confronted with an oxygen supply that is insufficient to meet the physiological demands of their cells, a state termed hypoxia. Hypoxia has a particularly strong impact on neurons, which are highly dependent on oxygen, have little in the way of glycolytic reserves, and are sensitive to excitotoxic death via overactivated glutamate receptors, including AMPA‐type receptors (AMPARs) during ischaemic stroke (Buchan et al, 1991; Nellgard and Wieloch, 1992; Li and Buchan, 1993; Sheardown et al, 1993; Gill, 1994; Xue et al, 1994; Lees, 2000; Takahashi et al, 2002). In many species, neurons have mechanisms to combat oxygen deprivation and its consequences. For example, the cerebrocortical neurons of the freshwater painted turtle Chrysemys picta bellii can tolerate low oxygen, surviving months of oxygen deprivation during periods of winter dormancy by reducing GluR current amplitude, although the precise mechanism is unclear (Shin and Buck, 2003; Pamenter et al, 2008; Zivkovic and Buck, 2010). One possible hypoxia protection mechanism could involve regulated AMPAR trafficking, a process already known to mediate multiple forms of synaptic plasticity (Mammen et al, 1997; Lissin et al, 1998, 1999; ‘O'Brien et al, 1998; Rongo and Kaplan, 1999; Shi et al, 1999; Hayashi et al, 2000), although such a protective role for trafficking has not been fully explored in vivo.
Metazoan cells in general use a well‐studied and conserved hypoxia protection network. When oxygen levels are sufficiently high, the EGL‐9/PHD family of prolyl hydroxylases use molecular oxygen, 2‐oxoglutarate, and iron to hydroxylate key proline side chains on the hypoxia‐inducible factor alpha (HIFα) proteins (Bruick and McKnight, 2001; Epstein et al, 2001; Fong and Takeda, 2008; Aragones et al, 2009). Once hydroxylated, HIFα is a substrate for the Von Hippel‐Lindau (VHL) ubiquitin ligase, which ubiquitinates HIFα, targeting it for degradation. Under hypoxic conditions, HIFα protein remains stable, acting as a transcription factor to regulate gene expression (Fandrey and Gassmann, 2009; Semenza, 2009). Several studies have suggested that, as cellular oxygen sensors, EGL‐9/PHD proteins might have other, less well‐understood functions besides HIFα destabilization (Lee et al, 2005; Cummins et al, 2006; Fu et al, 2007; Koditz et al, 2007; Fu and Taubman, 2010).
Given that mammalian neurons in culture die so easily under hypoxia, it has been challenging to determine a cell biological link between hypoxia, the hypoxia response pathway, and AMPAR trafficking. Here, we have taken advantage of the genetically tractable and hypoxia tolerant model organism C. elegans to investigate how AMPAR trafficking is regulated in response to hypoxia. The C. elegans AMPAR subunits GLR‐1 and GLR‐2 act in the command interneurons to receive synaptic input and direct overall locomotory reversal behaviour (Hart et al, 1995; Maricq et al, 1995; Mellem et al, 2002; Chang and Rongo, 2005). Mutants that lack AMPAR function or fail to localize AMPARs at synapses have a depressed frequency of spontaneous reversals, and reversal frequency has been used to infer the abundance of AMPARs at synaptic membranes (Zheng et al, 1999; Burbea et al, 2002; Shim et al, 2004; Glodowski et al, 2005; Schaefer and Rongo, 2006). AMPAR synaptic trafficking has also been monitored directly in vivo using a GLR‐1∷GFP rescuing transgene (Rongo et al, 1998).
GLR‐1 receptor synaptic abundance is regulated at the level of endocytosis and recycling (Burbea et al, 2002; Glodowski et al, 2007), and such regulation is used to modulate behaviour (Grunwald et al, 2004; Emtage et al, 2009). LIN‐10, a PTB/PDZ domain protein homologous to mammalian Mints, mediates the recycling of endocytosed GLR‐1 back to the plasma membrane, and lin‐10 mutants have GLR‐1 receptors that remain trapped in endosomal compartments (Glodowski et al, 2007; Park et al, 2009). GLR‐1 recycling is also regulated by the cyclin‐dependent kinase CDK‐5 and its p35 activator CDKA‐1, which phosphorylate LIN‐10, disrupting its subcellular localization and function (Juo et al, 2007).
Nematodes naturally encounter environmental conditions with variable oxygen concentrations, including hypoxia (Anderson and Dusenbery, 1977; Van Voorhies and Ward, 2000). Here, we show that hypoxia results in the retention of GLR‐1 receptors in internal subcellular compartments. Loss of function mutations in egl‐9, which encodes a PHD protein that normally inhibits the cellular response to hypoxia, mimics the effects of hypoxia on GLR‐1 trafficking and depresses glutamate‐activated currents and behavioural output. Unexpectedly, HIF‐1, the canonical substrate of EGL‐9 activity, is not required for this effect. Instead, EGL‐9 interacts with LIN‐10, a regulator of GLR‐1 recycling. We show that oxygen levels control the subcellular localization of EGL‐9 and LIN‐10 in neurons, and the effect of hypoxia and egl‐9 mutations on LIN‐10 localization and GLR‐1 trafficking requires the activity of the CDK‐5 kinase. Our results support a model in which EGL‐9 and CDK‐5 compete in an oxygen‐dependent manner to regulate LIN‐10 subcellular localization, GLR‐1 membrane recycling, and GLR‐1‐mediated exploratory behaviour, demonstrating a novel mechanism by which neurons sense and respond to hypoxia.
Hypoxia and EGL‐9 regulate GLR‐1 trafficking
In normoxia (room air: 21% oxygen), full‐length, functional GLR‐1 receptors tagged with GFP (GLR‐1∷GFP) are localized to discrete puncta along the ventral cord dendrites of interneurons (Figure 1A), with 85% of such puncta colocalized with synaptic markers (Rongo et al, 1998; Burbea et al, 2002). To determine if hypoxia alters GLR‐1 trafficking, we incubated cultures of nematodes under hypoxic conditions using a published nitrogen displacement approach (Pocock and Hobert, 2008). We found that animals exposed to 0.5% oxygen (hypoxia) localized GLR‐1∷GFP to elongated accumulations along the ventral cord (Figure 1C), similarly to the GLR‐1∷GFP accumulation in enlarged endosomes observed in mutants for membrane recycling factors (Rongo et al, 1998; Glodowski et al, 2007; Park et al, 2009; Kramer et al, 2010; Shi et al, 2010). These GLR‐1‐containing compartments are distinct in shape and size (2.5±0.5 μm) from >95% of the GLR‐1 synaptic puncta (0.6±0.2 μm) observed in normoxic wild‐type animals, and thus can be easily distinguished and quantified (Figure 1E). The observed changes in GLR‐1 localization are unlikely to be due to gross defects in synapse formation or overall cell polarity, as the localization of a synaptobrevin–GFP reporter (SNB‐1∷GFP), which decorates synaptic vesicles at interneuron presynaptic elements, does not change upon shift from normoxia to hypoxia (Supplementary Figure S1A–C). In addition, we detected similar levels of glr‐1 mRNA in hypoxic and normoxic animals, indicating that the observed changes are post‐transcriptional (Supplementary Figure S1D).
We reasoned that hypoxia might be activating the PHD/HIF pathway to regulate GLR‐1. If this were the case, then mutations in the lone C. elegans PHD gene egl‐9 should cause a similar effect in normoxic animals to that seen in hypoxic wild‐type animals. We examined GLR‐1∷GFP in two strong loss of function egl‐9 mutants: egl‐9(sa307) and egl‐9(n571) (Darby et al, 1999). Both mutations caused GLR‐1 to be localized to elongated accumulations similar to those seen in hypoxic animals (Figure 1B and E; Supplementary Figure S1E and F). As in hypoxic animals, egl‐9 mutants had an SNB‐1∷GFP punctate pattern similar to that of wild‐type animals (Supplementary Figure S1A–C), and similar levels of glr‐1 mRNA (Supplementary Figure S1D). Introduction of a wild‐type egl‐9 transgene fully rescued the GLR‐1 localization defects of egl‐9 mutants (Figure 1E). Interestingly, there was no additive effect of combining egl‐9 mutations with hypoxia (Figure 1D and F), suggesting that the effect of hypoxia on GLR‐1 is through EGL‐9.
EGL‐9 regulates multiple developmental and behavioural processes, each of which depends on HIF‐1 function (Shen et al, 2006; Chang and Bargmann, 2008; Gort et al, 2008; Pocock and Hobert, 2008; Shao et al, 2010). If HIF‐1 were also involved in regulating GLR‐1 trafficking, then the effects of hypoxia or egl‐9 mutations on GLR‐1 trafficking should be suppressed by a deletion mutation in hif‐1. We examined GLR‐1∷GFP in a hif‐1 molecular null (Jiang et al, 2001). Surprisingly, this hif‐1 mutation did not alter the GLR‐1 trafficking defects of hypoxic animals or egl‐9 mutants, and did not affect GLR‐1 under normoxic conditions (Figure 1F). We also examined the GLR‐1 localization phenotype in mutants that are deficient for VHL‐1 and RHY‐1, as they inhibit HIF‐1 (Epstein et al, 2001; Shen et al, 2006). Loss of function mutations in these genes did not affect GLR‐1 localization (Figure 1F; Supplementary Figure S1E and F). Collectively, these data strongly suggest that EGL‐9 and hypoxia regulate GLR‐1 subcellular localization by a novel pathway.
Hypoxia and EGL‐9 regulate GLR‐1 function
The elongated accumulations of GLR‐1 in hypoxic animals and egl‐9 mutants are similar to those observed in mutants in which GLR‐1 recycling is impaired, suggesting that they might represent internalized receptors. If this were the case, then synapses should be partially depleted of GLR‐1 and thus depressed for GLR‐1 function. We tested this several different ways. First, we monitored functional synaptic GLR‐1 by measuring two GLR‐1‐mediated behaviours: the frequency of spontaneous reversals in locomotion during foraging and the frequency of reversals in response to nose touch stimuli (i.e., nose touch mechanosensation). Both hypoxic animals and normoxic egl‐9 mutants showed depressed reversal rates and depressed nose touch mechanosensation rates compared with those of wild‐type animals in normoxia, although not quite as severe as those observed in glr‐1 null mutants (Figure 2A and B). Thus, hypoxic conditions repress GLR‐1‐dependent behaviours. Double mutants between egl‐9 and glr‐1 resembled glr‐1 single mutants, consistent with EGL‐9 regulating these behaviours in the same genetic pathway as GLR‐1 (Figure 2A and B). Mutations in hif‐1 did not affect reversal behaviour, nor did they suppress the effect of egl‐9 mutations or hypoxia on reversal behaviour (Supplementary Figure S1G), indicating that this role of EGL‐9 is independent of HIF‐1.
Second, we directly tested whether EGL‐9 can influence GLR‐1 channel activity. Glutamate‐activated currents can be recorded from the interneuron AVA (Mellem et al, 2002). If endogenous GLR‐1 is retained internally (internalized) in egl‐9 mutants, then mutant interneurons should have reduced glutamate‐activated currents. We used a previously described whole‐cell recording protocol to measure the size of GLR‐1‐mediated currents in the AVA interneuron elicited by the application of 1 mM glutamate (Wang et al, 2010), and we found that there was a significant reduction in current amplitude in egl‐9 versus wild type when AVA was voltage clamped at −70 mV (Figure 2C–E). Current–voltage relationship in neurons from wild‐type animals and egl‐9 mutants did not differ, indicating that egl‐9 mutants do not have general defects in conductance (Supplementary Figure S1H).
Third, we directly examined the amount of GLR‐1 receptors exposed on the plasma membrane by performing cell surface antibody labelling (Gottschalk and Schafer, 2006). We examined transgenic animals that express GLR‐1 dual‐tagged with an extracellular HA epitope and an intracellular GFP (Wang et al, 2008). Using anti‐HA antibodies under non‐permeabilized conditions, we detected both total HA∷GLR‐1∷GFP (Figure 2F) and membrane surface receptor (Figure 2F′) in wild type. By contrast, we detected little surface HA∷GLR‐1∷GFP in egl‐9 mutants (Figure 2G′), even though significant levels of HA∷GLR‐1∷GFP were present (Figure 2G). Taken together, our results indicate that GLR‐1 receptors accumulate in elongated, internal compartments along dendrites of egl‐9 mutants.
A specific EGL‐9 isoform regulates GLR‐1 trafficking
EGL‐9 might have multiple molecular or biochemical functions, as the egl‐9 gene produces a number of different protein isoforms (Figure 3A). We tested whether a specific EGL‐9 isoform regulates GLR‐1 trafficking by generating multiple isoform‐specific transgenes, each under the control of the glr‐1 promoter, and testing them for their ability to rescue the GLR‐1 localization phenotype of egl‐9 mutants. We found that isoform EGL‐9E, which lacks the first two exons, completely rescued both the GLR‐1 localization phenotype and the behavioural defects of egl‐9 mutants (Figure 3B–G). By contrast, both isoforms EGL‐9A, which is the longest isoform, and EGL‐9C, which lacks exons 1–4, failed to rescue the GLR‐1 trafficking defects of egl‐9 mutants (Figure 3C).
We also tested the localization pattern of the different EGL‐9 isoforms by tagging them with GFP and examining their subcellular localization in interneurons. EGL‐9A and EGL‐9C, which failed to rescue GLR‐1 trafficking defects, were localized to nuclei, but were diffusely distributed throughout ventral cord dendrites (Figure 3I–M). By contrast, EGL‐9E, which rescues GLR‐1 trafficking defects, was localized to puncta along the ventral cord (Figure 3J and L) in addition to the nucleus. We noted that EGL‐9E contains exons 3 and 4, which are not found in EGL‐9C, suggesting that the sequences encoded by these exons might be critical for localizing EGL‐9E to puncta in neurons. We analysed the sequence of exons 3 and 4, and found that it encodes a potential C2 domain (Supplementary Figure S2), which could be mediating protein–protein or protein–phospholipid interactions (Rizo and Sudhof, 1998). Taken together, our results suggest that, in addition to the prolyl hydroxylase catalytic domain, EGL‐9 contains additional protein domains that can conduct specialized functions, including the direction of subcellular localization. We also speculate that the amino acids encoded by exons 1 and 2 antagonize EGL‐9 punctate localization in neurons (mediated by the C2 domain in exons 3 and 4).
To examine the role of oxygen sensation in the regulation of GLR‐1 trafficking, we tested the importance of EGL‐9E catalytic function. The prolyl hydroxylase catalytic domain is an oxygen sensor comprising histidine and arginine side chains that stably bind to 2‐oxoglutarate and iron, which in turn interact with molecular oxygen (Epstein et al, 2001; McDonough et al, 2006; Chowdhury et al, 2009). We therefore treated animals with the iron chelator 2,2′‐dipyridyl (DIP) and found that such treatment resulted in the accumulation of GLR‐1∷GFP in internal compartments similarly to those observed in egl‐9 mutants and hypoxic animals (Supplementary Figure S1E and F). Next, we tested the ability of EGL‐9E(H487A) to rescue the GLR‐1 localization phenotype of egl‐9 mutants. Histidine 487 (amino‐acid numbering based on the longest EGL‐9 isoform) is a highly conserved residue that coordinates iron in the catalytic core and is required for EGL‐9 to regulate HIF‐1; the equivalent mutation in mammalian PHD2 reduces iron binding, hydroxylase activity, and oxygen sensing (McDonough et al, 2006; Shao et al, 2009). We found that EGL‐9E(H487A) failed to rescue (Figure 3C and H). Finally, we examined the role of oxygen on EGL‐9E subcellular localization. We found that hypoxia resulted in a decrease in the number of EGL‐9E puncta in the ventral cord (Figure 3K). Our findings indicate that EGL‐9E catalytic activity is required to regulate GLR‐1 trafficking, and that oxygen can regulate the subcellular localization of EGL‐9E itself.
EGL‐9 interacts with LIN‐10
To identify the substrates of EGL‐9 that regulate GLR‐1, we generated a yeast two‐hybrid bait plasmid containing just the catalytic domain of EGL‐9 (amino acids 347–608) and used it to screen a C. elegans prey cDNA library, identifying LIN‐10, a PDZ/PTB domain protein of the Mint family and a known mediator of GLR‐1 membrane recycling (Rongo et al, 1998; Glodowski et al, 2007; Park et al, 2009). To characterize the interaction between LIN‐10 and EGL‐9, we generated a collection of yeast two‐hybrid prey vectors containing different regions of the LIN‐10 protein, and then examined the ability of these preys to support the expression of multiple markers when paired with the EGL‐9 catalytic domain as bait. We found that amino acids 1–138 of LIN‐10 showed a strong interaction with EGL‐9, and that amino acids 1–90 as well as amino acids 108–280 were capable of weak interactions with EGL‐9, suggesting that the amino‐terminus of LIN‐10 might contain multiple sites for EGL‐9 interaction (Figure 4A).
To explore whether EGL‐9 regulates GLR‐1 through its interaction with LIN‐10, we re‐examined GLR‐1 localization in lin‐10 mutants. Like in hypoxic animals (Figure 4D) and egl‐9 mutants, GLR‐1∷GFP accumulates in elongated, internal compartments in lin‐10 mutants (Figure 4C), although at higher numbers (Figure 4F). Similarly, mutants for lin‐10 demonstrate behavioural defects consistent with diminished GLR‐1 function (Figure 4M and N). We used cell surface antibody labelling to directly examine the amount of HA∷GLR‐1∷GFP on the plasma membrane of lin‐10 mutants, but detected little surface HA∷GLR‐1∷GFP (Figure 2H′). We also examined lin‐10 single mutants and egl‐9 lin‐10 double mutants under conditions of either normoxia or hypoxia. Neither hypoxia treatment nor egl‐9 mutations increased the severity of the GLR‐1 localization and behavioural defects in animals lacking LIN‐10, suggesting that LIN‐10 and EGL‐9 act in the same genetic pathway to regulate oxygen‐sensitive GLR‐1 trafficking (Figure 4F, M, and N; Supplementary Figure S1E and F). Finally, if EGL‐9 interacts with LIN‐10, then we might expect both proteins to be colocalized in neurons. We examined GFP∷EGL‐9E and LIN‐10∷mCherry in interneuron cell bodies (Figure 4G, I, and K) and dendrites (Figure 4H, J, and L), and found that the two proteins were colocalized to puncta.
EGL‐9 could require its interaction with LIN‐10 solely to regulate GLR‐1, or alternatively LIN‐10 could be essential for all EGL‐9‐dependent functions, including its regulation of HIF‐1 protein levels and the hypoxia response. We tested these possibilities by examining several phenotypes of egl‐9(sa307) that are rescued by EGL‐9A (Epstein et al, 2001; Shao et al, 2009). First, we measured HIF‐1 protein levels and found that they were not altered in lin‐10 mutants (Supplementary Figure S3A). Second, we found that lin‐10 mutations did not affect HIF‐1 levels in mutants for the VHL‐1 ubiquitin ligase, which targets HIF‐1 for degradation (Supplementary Figure S3A). Third, we found that expression from Pnhr‐57∷GFP, a reporter for HIF‐1‐dependent transcription during hypoxia that shows elevated expression in both egl‐9 and vhl‐1 mutants (Shen et al, 2006), did not significantly change in lin‐10 mutants relative to wild type under both normoxic and hypoxic conditions (Supplementary Figure S3B–D). Our results indicate that LIN‐10 is not required for the canonical HIF‐1 hypoxia response pathway.
EGL‐9 and CDK‐5 regulate LIN‐10 subcellular localization
The LIN‐10 amino‐terminus directs LIN‐10 subcellular localization to Golgi and endosomal puncta (Whitfield et al, 1999; Glodowski et al, 2005; Park et al, 2009). Since EGL‐9 binds to the LIN‐10 amino‐terminus, we reasoned that EGL‐9 and oxygen might regulate LIN‐10 subcellular localization along dendrites. We found that LIN‐10∷GFP resides at puncta under normoxic conditions (Figure 5A), whereas it is diffusely distributed throughout dendrites under hypoxic conditions (Figure 5B). We quantified this effect by measuring the number of LIN‐10∷GFP puncta (Figure 5G) and the size and fluorescence intensity of LIN‐10∷GFP puncta as a combined integrated optical density (IOD), summing the pixel values for each punctum, which were then totalled for each animal (Figure 5H). We also examined LIN‐10∷GFP in egl‐9 mutants and found a diffuse distribution of LIN‐10 similar to that of hypoxic animals (Figure 5E, G, and H). Our quantification of LIN‐10∷GFP fluorescence suggests that LIN‐10 shifts from endosomal localization to a cytosolic distribution in egl‐9 mutants. To confirm that this change in LIN‐10 localization indicates a redistribution of LIN‐10 rather than a loss of LIN‐10 protein levels in egl‐9 mutants, we examined LIN‐10∷GFP protein levels in nematodes by western blot, using transgenic animals that only express LIN‐10∷GFP in the GLR‐1‐expressing interneurons (Figure 5I). Relative to actin levels, we did not detect a significant change in LIN‐10 protein levels in egl‐9 mutants compared with wild type (Figure 5J), indicating that EGL‐9 does not regulate LIN‐10 protein levels.
The amino‐terminus of LIN‐10 also interacts with the proline‐directed serine‐threonine kinase CDK‐5, which negatively regulates LIN‐10 by phosphorylating multiple proline‐proximal sites on the amino‐terminus (Juo et al, 2007). As EGL‐9 can hydroxylate prolines, we reasoned that EGL‐9 could prevent the inhibitory phosphorylation of LIN‐10 by CDK‐5, thereby keeping LIN‐10 active. If correct, then loss of CDK‐5 should protect LIN‐10 from hypoxia‐mediated downregulation. We tested this model by first examining LIN‐10∷GFP localization in cdk‐5 mutants. We found that LIN‐10∷GFP was localized to puncta in cdk‐5 and cdka‐1 mutants under both normoxic and hypoxic conditions (Figure 5C, D, G, and H). We also examined LIN‐10∷GFP in egl‐9 cdk‐5 double mutants and found that LIN‐10 was localized to puncta in a manner similar to that of cdk‐5 single mutants (Figure 5F–H). LIN‐10∷GFP protein levels remained constant in these genotypes (Figure 5I and J). Our results indicate that the regulation of LIN‐10∷GFP subcellular localization by hypoxia and EGL‐9 requires the activity of CDK‐5.
If EGL‐9 promotes LIN‐10 function by antagonizing the ability of CDK‐5 to phosphorylate LIN‐10, then we would expect that the loss of CDK‐5 would block the effects of hypoxia and egl‐9 mutations on GLR‐1 trafficking. Whereas hypoxia caused wild‐type neurons to accumulate GLR‐1 in internal compartments (Figure 6A and B), mutations in cdk‐5 completely blocked this effect (Figure 6C, D, and K). In addition, mutations in cdk‐5 suppressed the GLR‐1 trafficking defects caused by egl‐9 mutations (Figure 6E, F, and K). Consistent with the GLR‐1 trafficking data, we observed that mutation in cdk‐5 suppressed the spontaneous reversal and nose touch mechanosensation behavioural defects in both hypoxic wild‐type animals and egl‐9 mutants (Figure 6L and M), placing CDK‐5 genetically downstream of EGL‐9 and oxygen. Importantly, mutations in cdk‐5 did not suppress the defects in GLR‐1 trafficking and GLR‐1‐mediated behaviours observed in lin‐10 mutants (Figure 6L and M; data not shown), indicating that LIN‐10 is acting genetically downstream of CDK‐5.
We introduced a transgene, Pglr‐1∷cdk‐5(+), into nematodes to overexpress wild‐type CDK‐5 (Juo et al, 2007). Overexpression of wild‐type CDK‐5 results in a phenotype similar to that of egl‐9 mutants and hypoxic animals (Figure 6G, H, and K). Interestingly, CDK‐5 was not able to increase the amount of internalized GLR‐1 when overexpressed in egl‐9 mutants, hypoxic animals, or a combination of both (Figure 6I–K), indicating that the effect of CDK‐5 overexpression is not additive with the effects of egl‐9 mutations or hypoxia, consistent with CDK‐5 and EGL‐9 regulating the same target. Our results show that in the absence of EGL‐9 function or oxygen, CDK‐5 overly depresses GLR‐1 recycling, suggesting that EGL‐9 and CDK‐5 normally compete against each other to regulate GLR‐1 trafficking.
EGL‐9 could regulate LIN‐10 localization by preventing LIN‐10 phosphorylation by CDK‐5, perhaps by hydroxylating prolines within each CDK‐5 phosphorylation consensus site. In this model, we would predict that removing the proline residues from these consensus sites would block the phosphorylation of LIN‐10 by CDK‐5. Ten putative CDK‐5 phosphorylation sites (serine or threonine with an adjacent proline) are located in the LIN‐10 amino‐terminus (Figure 4A). We tested the ability of CDK‐5 to phosphorylate the LIN‐10 amino‐terminus in vitro, using PhosTag electrophoresis to detect phosphorylated substrates based on a decrease in electrophoretic mobility (Kinoshita et al, 2006). Whereas CDK‐5 can phosphorylate the LIN‐10 amino‐terminus, it could not phosphorylate a mutated version of LIN‐10, LIN‐10(Pmut,1–574), in which the prolines were substituted for alanines (Figure 7A).
If CDK‐5 regulates LIN‐10 subcellular localization by direct phosphorylation, then we would expect that mutations in the CDK‐5 consensus site prolines should mimic the cdk‐5 mutant phenotype and block the effects of hypoxia and egl‐9 mutation on LIN‐10 in vivo. We therefore introduced the identical proline to alanine substitutions into our LIN‐10∷GFP transgene and observed a punctate localization of the resulting LIN‐10(Pmut)∷GFP protein similar to that of wild‐type LIN‐10∷GFP (Figure 7B, D, and H). However, unlike for wild‐type LIN‐10∷GFP, LIN‐10(Pmut)∷GFP remained localized in egl‐9 mutants and under conditions of hypoxia (Figure 7D, E, and H), indicating that the proline residues are required for EGL‐9 and oxygen to regulate LIN‐10. Importantly, neuronal expression of LIN‐10(Pmut) robustly rescued the GLR‐1 trafficking defects of lin‐10 mutants, indicating that the proline substitutions impaired the ability of LIN‐10 to be regulated by EGL‐9, CDK‐5, and oxygen levels, but did not impair LIN‐10 function (Figure 7I). We also examined a mutant version of LIN‐10∷GFP, LIN‐10(S/Tmut)∷GFP, in which the CDK‐5 sites contain either serine or threonine to alanine substitutions (Juo et al, 2007). Like LIN‐10(Pmut)∷GFP (or wild‐type LIN‐10∷GFP in egl‐9 mutants), LIN‐10(S/Tmut)∷GFP was localized even in egl‐9 mutants and hypoxic animals (Figure 7F and G). Our results indicate that phosphorylation of LIN‐10 by CDK‐5 disrupts LIN‐10 subcellular localization, and suggest that EGL‐9 regulates LIN‐10 by antagonizing its phosphorylation by CDK‐5 when oxygen is present, either by direct hydroxylation of key proline sites on LIN‐10, or perhaps by steric hindrance of CDK‐5 away from LIN‐10 due to oxygen‐dependent binding of EGL‐9 to LIN‐10. This represents a novel pathway for glutamate receptor trafficking in response to fluctuating oxygen levels.
Here, we have shown that oxygen levels regulate GLR‐1 AMPAR trafficking and GLR‐1‐mediated behaviour via a non‐canonical variant of the hypoxia response pathway (Figure 8). When oxygen levels are sufficiently high, the EGL‐9 enzyme uses dioxygen as a molecular substrate to hydroxylate a specific proline residue on HIF‐1 to 4‐hydroxyproline (Epstein et al, 2001). This modification enables VHL‐1 to bind and ubiquitinate HIF‐1, resulting in its proteolysis (Maxwell et al, 1999; Ohh et al, 2000; Ivan et al, 2001; Jaakkola et al, 2001; Min et al, 2002; Bishop et al, 2004; Shen et al, 2006). Here, we report that EGL‐9 has another critical function in neurons that is independent of HIF‐1: the regulation of AMPAR recycling via its interaction with LIN‐10. As illustrated in Figure 8A, we favour a model in which EGL‐9 uses oxygen to hydroxylate LIN‐10 at prolines that comprise CDK‐5 phosphorylation sites, precluding CDK‐5 phosphorylation at neighbouring serine or threonine residues. In the absence of CDK‐5 phosphorylation, the LIN‐10 amino‐terminus directs its localization to endosomes, where LIN‐10 promotes GLR‐1 recycling. When oxygen levels are low (Figure 8B), EGL‐9 cannot hydroxylate LIN‐10, thereby allowing CDK‐5 to phosphorylate the LIN‐10 amino‐terminus and disrupt LIN‐10 endosomal localization. In effect, EGL‐9 and CDK‐5 compete for LIN‐10, with oxygen tipping the balance towards EGL‐9. This mechanism keeps LIN‐10 at endosomes so as to constantly resupply synapses with GLR‐1 receptors when oxygen is plentiful, and it results in LIN‐10 inhibition and hence GLR‐1 internalization when oxygen is scarce.
The pathway by which EGL‐9 regulates LIN‐10 is distinct from the oxygen‐dependent degradation of HIF‐1 in several ways. First, the pathway is independent of HIF‐1, with EGL‐9 directly interacting with LIN‐10 protein instead. Mutations in hif‐1 do not appear to alter GLR‐1 trafficking, and mutations in lin‐10 do not appear to alter the stability of HIF‐1 protein or the expression of HIF‐1 target genes, indicating that these pathways function independently from one another. Second, the regulatory outcome of this pathway is at the level of a post‐translational response—GLR‐1 trafficking—rather than a transcriptional response. Third, in the canonical pathway, EGL‐9 is negatively regulating its target, HIF‐1, whereas during the regulation of GLR‐1, EGL‐9 is positively regulating its target, LIN‐10. Fourth, the cell biological outcome of EGL‐9 activity on its target involves protein turnover in the case of HIF‐1, whereas it involves altered subcellular localization in the case of LIN‐10. The discovery of this novel pathway suggests that the mechanisms by which cells and prolyl hydroxylase proteins like EGL‐9 respond to changes in oxygen are likely more diverse than originally appreciated.
What is the function of this novel regulation? C. elegans, as a soil nematode, does encounter environments of low oxygen, including the buried spaces and rotting material where it feeds on bacteria (Anderson and Dusenbery, 1977; Van Voorhies and Ward, 2000). Since C. elegans lack specialized respiratory structures, gaseous exchange is believed to occur through the body cuticle directly into tissues and the fluid of the pseudocoelomic body cavity. A broad range of oxygen concentration is found below the soil surface depending on variables such as aeration, temperature, moisture levels, and the bacterial population. Nematodes avoid both hyperoxic (>15% oxygen) and hypoxic (<4% oxygen) environments, showing a preference for more intermediate levels (Gray et al, 2004; Cheung et al, 2005). The neurons AQR, PQR, and URX are exposed in the body cavity and express soluble guanylate cyclases that bind and sense oxygen in the body cavity fluid; these neurons are thought to mediate a rapid aerotaxis response to changing oxygen levels (Cheung et al, 2004; Gray et al, 2004). By contrast, the EGL‐9/HIF‐1 canonical hypoxia response pathway is broadly expressed in C. elegans and is thought to elicit slower, cell‐autonomous stress responses to low oxygen, but not directly mediate rapid aerotaxis responses (Branicky and Schafer, 2008). However, these two mechanisms are partly linked, as the activity of HIF‐1 in C. elegans neurons can adjust the preferred oxygen range to which nematodes will aerotax by reorganizing the neural circuits required for oxygen sensation (Chang and Bargmann, 2008; Pocock and Hobert, 2008). Our findings here indicate that oxygen sensation and behaviour are linked by a third mechanism: the EGL‐9/LIN‐10/GLR‐1 pathway. Under conditions of either hypoxia or loss of EGL‐9 activity, there is a decrease in spontaneous reversal frequency for nematode locomotion off of food, consistent with depressed GLR‐1 function. Affected animals track for longer stretches, resulting in a greater pattern of exploratory behaviour. In the case of egl‐9 mutants, reversal behaviour is fully restored when EGL‐9 activity is provided back solely to the GLR‐1‐expressing interneurons. This would suggest that, like the AQR, PQR, and URX sensory neurons, the command interneurons also sense oxygen and modulate animal behaviour in response.
The EGL‐9/LIN‐10 pathway might also have a neuroprotective function. Ischaemic stroke results in nervous system damage due to local oxygen deprivation in the brain, as hypoxic neurons undergo glutamate receptor‐mediated excitotoxic necrosis (Khaldi et al, 2002; Limbrick et al, 2003; Arundine and Tymianski, 2004; Mehta et al, 2007). It is unclear how glutamate receptors respond to such hypoxic insult. Interestingly, neurons can use homeostatic ‘synaptic scaling’ mechanisms to downregulate surface AMPAR expression in response to increased neural activity (Turrigiano, 2008), raising the possibility that a similar response could be triggered in ischaemic neurons. We propose that neurons likely employ multiple adaptive responses to hypoxia, and that the EGL‐9/LIN‐10 pathway, and perhaps a PHD/Mint orthologous pathway in mammals, might represent one such underlying neuroprotective mechanism.
Materials and methods
Transgenes and germline transformation
Transgenic strains generated in this study were isolated after microinjecting various plasmids (5–50 ng/ml) using rol‐6dm (a gift from C Mello, UMass), ttx‐3∷rfp (a gift from O Hobert, Columbia Univ.), or lin‐15(+) (a gift from J Mendel, CalTech) as a marker. Plasmids containing the glr‐1, followed by the indicated cDNA construct, were generated using standard techniques. All resulting transgenes were introduced into the germline and followed as extrachromosomal arrays. The Phosphomutant LIN‐10(S/Tmut)∷GFP construct was a gift from P Juo (Tufts; termed LIN‐10(10Pm)∷GFP in his paper). The proline mutant LIN‐10(Pmut)∷GFP was created using a construct containing synthesized LIN‐10 cDNA (GeneScript); the following proline residues in LIN‐10 were mutated to alanine in LIN‐10(Pmut): P16, 27, 29, 88, 156, 202, 331, 397, 417, and 421. These are corresponding sites for LIN‐10(S/Tmut)∷GFP (Juo et al, 2007).
Animals were grown at 20°C on standard NGM plates seeded with OP50 E. coli. For hypoxia, animals were incubated in a hypoxia chamber (C‐174 chamber, Biospherix) for 24 h at 20°C and recovered in ambient oxygen for 12 h at 20°C. The oxygen level was automatically maintained with an oxygen controller (ProOx P110, Biospherix) supplied with compressed nitrogen gas.
GFP‐ and RFP‐tagged fluorescent proteins were visualized in nematodes by mounting larvae on 2% agarose pads with levamisole. Fluorescent images were observed using a Zeiss Axioplan II. A × 100 (NA=1.4) PlanApo objective was used to detect GFP and RFP signals. Imaging was done with an ORCA charge‐coupled device (CCD) camera (Hamamatsu, Bridgewater, NJ) using IPLab software (Scanalytics, Inc, Fairfax, VA) or iVision v4.0.11 (Biovision Technologies, Exton, PA) software. Exposure times were chosen to fill the 12‐bit dynamic range without saturation. In most cases, maximum intensity projections of z‐series stacks were obtained and out‐of‐focus light was removed with a constrained iterative deconvolution algorithm (iVision). For images, we captured the ventral cord dendrites in the region posterior to the RIG and AVG cell bodies.
The quantification of ventral nerve cord fluorescent objects (i.e., puncta and elongated compartments) was done using ImageJ (Collins, 2007) to automatically threshold the images and then determine the outlines of fluorescent objects in ventral cord dendrites. ImageJ was used to quantify both the shape and the size of all individual fluorescent objects along the ventral cord. This allowed us to distinguish between the small GLR‐1∷GFP puncta in wild‐type animals and the large, aberrant compartments (which have an elongated shape rarely observed in wild type) in hypoxic animals, as well as in egl‐9 and lin‐10 mutants. Object size was measured as the maximum diameter for each outlined puncta. Object number was calculated by counting the average number of puncta per 100 microns of dendrite length. The amount of a given fluorescent protein per puncta was calculated by summing all of the pixel values contained within each individual punctum to yield an IOD score for each punctum.
The reversal frequency of individual animals was assayed as previously described, but with some modifications (Zheng et al, 1999). Single young adult hermaphrodites were placed on NGM plates in the absence of food. The animals were allowed to adjust to the plates for 5 min, and the number of spontaneous reversals for each animal was counted over a 5‐min period. Twenty or more animals were tested for each genotype, and the reported scores reflect the mean number of reversals per minute, normalized as a percentage of the value of wild‐type controls. Nose‐touch mechanosensation was assayed by placing young adult hermaphrodites on NGM plates with food. Individual young adult animals were allowed to collide with a human hair 10 consecutive times within a 5‐min period. Activation of the reversal behaviour was scored immediately after each contact with the hair stimulus and the score summed over the 10 trials. Twenty or more animals were tested for each genotype, and the reported scores reflect the mean number of responses.
Patch‐clamp whole‐cell recording
In‐vivo whole‐cell recordings were carried out at room temperature with an EPC‐10 amplifier and Patchmaster software (HEKA) using a protocol described in a previous study (Wang et al, 2008; Ward et al, 2008). Briefly, the head of glued worms was dissected and the AVA neurons were exposed for patch‐clamp recordings in the bath solution. Recording pipettes were pulled from borosilicate glass. The pipette solution contained 115 mM K‐gluconate, 25 mM KCl, 50 mM HEPES, 0.1 mM CaCl2, 1 mM BAPTA, 5 mM MgATP, 0.5 mM NaGTP (315 mOsm, pH adjusted to 7.35 with KOH). The bath solution contained 150 mM NaCl, 5 mM KCl, 5 mM CaCl2, 1 mM MgCl2, 15 mM HEPES, 10 mM Glucose (325 mOsm, pH adjusted to 7.35 with KOH). Voltages were clamped at −70 mV. Current data were sampled at 22 kHz. In all, 1 mM glutamate was applied for 1 s by pressure ejection.
Immunolabelling of cell surface
Cell surfaces along the ventral nerve cord in transgenic animals expressing HA∷GLR‐1∷GFP (a gift from AV Maricq, Univ. of Utah) were immunolabelled as described (Gottschalk and Schafer, 2006). Briefly, AlexaFluor594‐conjugated mouse monoclonal anti‐HA antibodies (Molecular Probes) were 200‐fold diluted in injection buffer (20 mM K3PO4, 3 mM K citrate, 2% PEG 6000, pH 7.5) and injected into the pseudocoelomic space of transgenic animals. After a 5 h of recovery, animals were mounted and observed for epifluorescence as described above.
Immunodetection of LIN‐10
To measure LIN‐10∷GFP protein levels, 30 young adults of each wild type and mutants expressing the odIs22 transgene were dissolved in 1 × Laemmli buffer by fresh freezing and boiling for 10 min. To observe phosphorylation, the amino‐terminal half of wild‐type and proline‐mutant LIN‐10 (amino acids 1–574) was N‐terminally tagged with FLAG, expressed in rabbit reticulocyte‐based protein expression system (Promega). In vitro‐expressed proteins were incubated in kinase buffer (50 mM HEPES, pH 7, 10 mM MgCl2, 100 μm ATP, 1 mM dithiothreitol, 50 mM β‐glycerolphosphate) for 30 min at 37°C either with or without recombinant CDK5/p25 (Millipore). Reactions were stopped by adding Laemmli buffer and analysed onto 10% SDS–polyacrylamide gels containing 30 mM Phos‐tag AAL‐107 (Nard Institute, Japan). Western blotting was performed using mouse anti‐GFP (1:1000, Roche), mouse anti‐Actin (1:2000, MP Biomedicals), and mouse anti‐Flag (1:3000, Sigma).
Yeast two‐hybrid interactions
A yeast two‐hybrid bait plasmid containing just the catalytic domain of EGL‐9 (amino acids 347–608) was generated so as to screen a C. elegans cDNA library in pGAD‐GH (David Sherwood, Duke University). Transformed yeast were selected for interaction on –Trp –Leu plates; 3 470 000 transformed colonies were screened, resulting in 40 His+ positives. Candidates were tested for β galactosidase activity, yielding 19 positive clones with in‐frame cDNA inserts.
To confirm the interactions between LIN‐10 and EGL‐9, corresponding cDNAs were introduced into the Matchmaker™ GAL4 Two‐Hybrid System 3 (Clontech, Palo Alto, CA), which uses four reporter genes: HIS3, ADE2, MEL1 (α galactosidase), and lacZ (β galactosidase). Transformed yeast was selected on –His –Trp –Leu plates. Twenty colonies were selected for serial dilution analysis on –His –Trp –Leu –Ade plates containing X‐α galactoside, or were assayed for β galactosidase activity on filters. Strong growth was indicated by activation of all four reporters. Weak growth was indicated by activation of at least three reporters.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
Supplemental Figures and Methods
We thank Andy Fire, the C. elegans Genetics Center, Villu Maricq, Joshua Kaplan, Peter Juo, Shohei Mitani, and Richard Tsien for reagents and strains. We thank Maureen Barr, Barth Grant, Monica Driscoll, Natalia Morsci, Doreen Glodowski, and Indrani Chatterjee for comments on the manuscript. This work was supported by National Institutes of Health (NIH) R01 NS42023 and a State of New Jersey Brain Injury Research Grant (07‐3201‐BIR‐E‐1) to CR, NIH R01 GM78424 to JAP‐C, a State of New Jersey Spinal Cord Postdoctoral Research Fellowship (10‐2950‐SCR‐E‐0) to ECP, and an NIH F31 NS071741 Fellowship and Waksman Benedict Michael Fellowship to PG.
Author contributions: ECP and PG designed and performed the genetic, molecular, behavioural, and cell biological experiments under the direct supervision of CR, with each author making equal contributions. LK and SX performed the electrophysiological recordings. ZS, QY, and JAP‐C designed and performed the yeast two‐hybrid analysis of LIN‐10/EGL‐9 interactions. PG, ECP, and CR wrote the manuscript, with significant editorial input and discussion of the results and their implications at all stages of the process from JAP‐C.
- Copyright © 2012 European Molecular Biology Organization