A complex containing the O-GlcNAc transferase OGT-1 and the ubiquitin ligase EEL-1 regulates GABA neuron function

Inhibitory GABAergic transmission is required for proper circuit function in the nervous system. However, our understanding of molecular mechanisms that preferentially influence GABAergic transmission, particularly presynaptic mechanisms, remains limited. We previously reported that the ubiquitin ligase EEL-1 preferentially regulates GABAergic presynaptic transmission. To further explore how EEL-1 functions, here we performed affinity purification proteomics using Caenorhabditis elegans and identified the O-GlcNAc transferase OGT-1 as an EEL-1 binding protein. This observation was intriguing, as we know little about how OGT-1 affects neuron function. Using C. elegans biochemistry, we confirmed that the OGT-1/EEL-1 complex forms in neurons in vivo and showed that the human orthologs, OGT and HUWE1, also bind in cell culture. We observed that, like EEL-1, OGT-1 is expressed in GABAergic motor neurons, localizes to GABAergic presynaptic terminals, and functions cell-autonomously to regulate GABA neuron function. Results with catalytically inactive point mutants indicated that OGT-1 glycosyltransferase activity is dispensable for GABA neuron function. Consistent with OGT-1 and EEL-1 forming a complex, genetic results using automated, behavioral pharmacology assays showed that ogt-1 and eel-1 act in parallel to regulate GABA neuron function. These findings demonstrate that OGT-1 and EEL-1 form a conserved signaling complex and function together to affect GABA neuron function.

GABA neurons are a critical component of nervous systems across the animal kingdom from mammals (1,2) to simple invertebrates, such as Caenorhabditis elegans (3,4). They provide essential inhibitory activity within neural circuits. In humans, various dysfunctions in GABA neurons and the imbalance between excitatory and inhibitory neurotransmission contribute to neurodevelopmental disorders (5,6). Thus, understanding how GABA neuron function is regulated is critical for our understanding of nervous system function and disease.
Much remains unknown about molecular mechanisms that preferentially affect GABAergic transmission. Core presynaptic machinery, such as synaptotagmin, the SNARE complex, and active zone proteins, influence both glutamatergic and GABAergic transmission (7,8). A few post-synaptic regulators that preferentially or specifically affect GABAergic transmission are known, including Gephyrin, Neuroligin2, Slitrk3, and GARHLs (9 -13). In mammals, less is known about presynaptic GABA-specific regulators, but some proteins, such as synapsins, can differentially impact inhibitory transmission compared with excitatory transmission (14,15).
In C. elegans, core presynaptic components play conserved roles in neurotransmission in the motor circuit, a model circuit with balanced excitatory cholinergic and inhibitory GABAergic neuron function (4,16). Like mammals, relatively few proteins are known that preferentially regulate presynaptic GABA function in C. elegans. Nonetheless, the worm motor circuit has proven valuable for identifying molecules that regulate GABA neuron function. Examples include the NPR-1 neuropeptide Y receptor, the SEK-1 MAP2K, the F-box protein MEC-15, and the anaphase-promoting complex (17)(18)(19).
Recently, we showed the HECT family ubiquitin ligase EEL-1 (enhancer of EfL-1) is expressed broadly in the nervous system but preferentially affects GABAergic presynaptic transmission in the motor circuit of C. elegans (20). At present, it is unknown how EEL-1 regulates GABAergic presynaptic transmission. Our interest in exploring this question was heightened by extensive genetic links between the EEL-1 ortholog HUWE1 (HECT, UBA, and WWE domains containing protein 1) and intellectual disability. These include HUWE1 copy number increases (21) and missense loss-of-function mutations that cause Juberg-Marsidi-Brooks syndrome and non-syndromic X-linked intellectual disability (20,22,23).
In mammals, OGT is expressed in the brain and localizes to presynaptic terminals (27,28). Despite prominent OGT-mediated O-GlcNAcylation of synaptic proteins (29), the functional effects of OGT in the nervous system have only recently begun to be explored. OGT regulates mitochondrial motility in neu-rons (30) and has been implicated in neurodegenerative disease (31). In OGT conditional knockout mice, AgRP (agouti-related protein) and PVN (paraventricular nucleus) neurons are functionally impaired, leading to impacts on fat metabolism and feeding behavior, respectively (32,33). While glycosyltransferase activity is the most widely studied OGT activity, a much smaller body of work indicates that OGT can also act as a scaffold protein (25). At present, it is unknown whether this transferase-independent function of OGT has a role in the nervous system.
To explore the biological relationship between OGT-1 and EEL-1, we expanded upon our proteomic finding that OGT-1 was a putative EEL-1 binding protein with several independent experimental approaches. We biochemically validated the interaction between OGT-1 and EEL-1 and showed that it occurs in C. elegans neurons in vivo. Importantly, this interaction was conserved as it also occurred between HUWE1 and OGT, the orthologous human proteins. Similar to EEL-1, OGT-1 was broadly expressed in the nervous system, including the cholinergic and GABAergic neurons of the motor circuit, and localized to presynaptic terminals in GABA neurons. Results from genetic analysis using an automated behavioral assay and pharmacological manipulation of the motor circuit showed that OGT-1 affects GABA neuron function. Similar phenotypic defects in GABA neuron function were previously observed in eel-1 mutants (20). Furthermore, genetic results indicate that OGT-1 functions in parallel to EEL-1 in GABA neurons. Consistent with this, OGT-1 and EEL-1 also act in parallel to affect locomotion. Findings with point mutations that impair catalytic activity show that OGT-1 functions independently of glycosyltransferase activity to affect GABA neuron function, whereas EEL-1 ubiquitin ligase activity is required. Thus, our study reveals the discovery of an OGT-1/EEL-1 protein complex that regulates GABA neuron function and provides the first evidence of a nonenzymatic OGT-1 function in the nervous system.

Measuring C. elegans motor circuit function using an automated aldicarb assay
Previously, we used a combination of electrophysiology and behavioral pharmacology to show that EEL-1 regulates GABAergic presynaptic transmission (20). To determine how EEL-1 regulates GABA transmission, we wanted to use affinity purification proteomics to identify EEL-1 binding proteins. As the first step in this process, we developed an automated platform for evaluating motor circuit function using aldicarb pharmacology. Once established, this assay would allow us to rapidly and quantitatively evaluate whether EEL-1 reagents are functional in vivo and suitable for affinity purification proteomics.
The C. elegans motor circuit is composed of excitatory cholinergic and inhibitory GABAergic motor neurons that innervate body wall muscles to control contraction and relaxation, respectively (Fig. 1A). This balance of excitation and inhibition allows for coordinated sinusoidal movement of the body. A traditional pharmacological assay for assessing motor circuit func-tion relies upon aldicarb, an inhibitor of acetylcholine esterase (AchE). 2 By impairing AchE, aldicarb causes accumulation of Ach over time, which leads to excess muscle contraction and gradual paralysis (Fig. 1A). Traditionally, this is measured by assessing C. elegans paralysis while animals are on agar plates containing aldicarb. Aldicarb-induced paralysis on plates is usually assessed manually, but it has been automated (34,35). We developed an automated, liquid assay that uses MWT (Multi-Worm Tracker) to evaluate locomotion and aldicarbinduced paralysis (Fig. 1B). We simultaneously monitored 20 wells with 4 worms/well (Fig. 1B). We recorded 10 min of baseline movement, added a desired aldicarb dose, and recorded animal speed in response to drug. As expected, WT animals showed dose-dependent paralysis after aldicarb treatment (Fig.  1C). Compared with our experience with manual aldicarb assays on agar plates (20), this automated liquid assay increases throughput and has a large dynamic range that facilitates dose response analysis.
Mutants that have disrupted motor circuit function have altered aldicarb sensitivity (17,34). Mutants with impaired cholinergic function accumulate Ach more slowly at the synapse when treated with aldicarb, which results in slower paralysis and resistance to aldicarb compared with WT animals. This is also the case for mutants that affect cholinergic and GABAergic function equally. There are two scenarios that lead to aldicarb resistance. The first is mutants with increased cholinergic function. The second is mutants that have preferentially disrupted inhibitory GABA function, which results in loss of relaxation and faster paralysis in the presence of aldicarb (Fig. 1A).
To assess the performance of our automated aldicarb assay, we tested several aldicarb hypersensitive mutants: hypersensitive mutants with increased cholinergic transmission (goa-1 and slo-1), mutants that are defective in GABA biosynthesis (unc-25), and mutants that lack a GABA receptor subunit (unc-49) (17,36). We also evaluated eel-1 (zu462) deletion mutants, which we previously showed are hypersensitive to aldicarb due to defects in GABAergic presynaptic transmission (20). Consistent with prior studies, these mutants were hypersensitive in automated aldicarb assays ( Fig. 1D and Figs. S1 and S2A). Hypersensitivity in eel-1 mutants was rescued by an integrated transgene that expressed EEL-1 using the native eel-1 promoter ( Fig. 1D and Fig. S2A). It is unclear why the EEL-1 transgene did not fully rescue. This could be because the eel-1 promoter we designed is not ideal for EEL-1 expression, or EEL-1 is not expressed at optimal levels by the integrated multicopy transgene we used. Nonetheless, these results indicate that we have developed an automated liquid aldicarb assay that rapidly and quantitatively assesses motor circuit function. This approach allowed us to assess aldicarb hypersensitivity and rescue in eel-1 mutants. Thus, this assay is suitable for functional evaluation of EEL-1 constructs used for proteomics.

Functional assessment of EEL-1 affinity purification proteomic reagents
The next step toward EEL-1 affinity purification proteomics was to generate an eel-1 protein null allele and evaluate this mutant in automated aldicarb assays. A protein null allele is particularly important to ensure that transgenic EEL-1 used for affinity purification proteomics is not competing with endogenous EEL-1 protein or EEL-1 fragments. We generated bgg1, an eel-1 protein null, using Mos1-mediated deletion to eliminate the entire eel-1 coding sequence, including the HECT ubiquitin ligase domain (Fig. 2, A and B). Importantly, eel-1 (bgg1) mutants were hypersensitive to aldicarb (Fig. 2C), similar to eel-1 (zu462) mutants (Fig. 1D).
Next, we used automated aldicarb assays to evaluate rescue for EEL-1 constructs tagged with affinity purification tags. Sev-eral transgenes were generated that fused different constructs to a GS (protein G and streptavidin-binding protein) tag. GS was fused to WT EEL-1 or EEL-1 point-mutated at a critical residue (C4144A) required for E3 ubiquitin ligase activity ( Fig.  2B) (37). We refer to the catalytically inactive point mutant as EEL-1 LD (ligase-dead). GS-tagged GFP served as a negative control. All transgenes were driven by the native eel-1 promoter and expressed in the eel-1 (bgg1) protein null background. There were two reasons we included the EEL-1 LD transgene: 1) it remains unclear whether EEL-1 effects on aldicarb sensitivity and GABA transmission are mediated by EEL-1 ubiquitin ligase activity, and 2) we wanted to evaluate whether the EEL-1 LD can biochemically "trap" and enrich EEL-1 ubiquitination substrates in proteomic experiments.
As shown in Fig. 2C, GS::EEL-1 significantly rescued aldicarb hypersensitivity of eel-1 (bgg1) mutants. Similar to untagged EEL-1 (Fig. 1D), GS::EEL-1 only partially rescued eel-1. In contrast, rescue was not observed with GS::EEL-1 Figure 1. Automated behavioral assay shows that eel-1 mutants are hypersensitive to aldicarb. A, depicted is the C. elegans motor circuit composed of excitatory cholinergic and inhibitory GABAergic motor neurons. Balance of contraction and relaxation is required for normal movement (left). Application of aldicarb, an AchE inhibitor, leads to excess cholinergic transmission and paralysis (middle). Mutants with impaired GABAergic transmission, such as eel-1, are hypersensitive to aldicarb and paralyze faster (right). B, schematic showing automated aldicarb assay with MWT. 20 wells are monitored simultaneously. Shown is an image of a single well with four animals (right). MWT plots depict animal swimming over a 1-min timeframe with and without aldicarb treatment (below). C, MWT analysis of aldicarb dose response for WT animals. Shown is the mean of multiple wells for each dose (n ϭ 5-20 wells/dose); significance was determined using two-way ANOVA (dose versus time). D, eel-1 (zu462) mutants are hypersensitive to aldicarb, and transgenic expression of EEL-1 rescues aldicarb hypersensitivity. Shown are means (n ϭ 20 wells/genotype). Inset, mean speed (line) and speed in each well (circles) at the indicated time for each genotype. Comparisons between genotypes represent pairwise two-way ANOVAs. Comparisons in the inset represent Fisher's LSD post hoc test (see Fig. S2A for further statistical analysis). ***, p Ͻ 0.001.
Taken together, these results support several conclusions. First, the eel-1 protein null allele, bgg1, is hypersensitive to aldicarb and impairs locomotion in liquid. Second, the GS::EEL-1 affinity puri-

C. elegans proteomics identifies OGT-1 as a putative EEL-1 binding protein
Our strategy for EEL-1 affinity purification proteomics is portrayed in Fig. 2D. Transgenic animals expressing GS::EEL-1, GS::EEL-1 LD, or GS::GFP (negative control) on an eel-1 (bgg1) protein null background were grown in large-scale liquid culture, harvested, and frozen in liquid nitrogen. Frozen animals were cryomilled in liquid nitrogen-cooled cylinders to obtain micronscale grindates that facilitated rapid lysis and protein extraction. Whole worm lysates were applied to IgG-Dynabeads to affinity capture protein complexes containing GS-tagged target proteins. Sample quality was assessed and optimized using two parameters: 1) immunoblotting (1% of sample) to confirm GS-tagged target proteins were successfully purified ( Fig. S4) and 2) silver staining (9% of sample) to evaluate sample purity and estimate the total amount of purified target (Fig. S4). Using this approach, affinity purification procedures were extensively optimized to obtain as much GS target as possible, while also ensuring that the GS::GFP negative control was as clean as possible compared with GS::EEL-1 and GS::EEL-1 LD test samples.
The majority of each sample (90%) was run on SDS-PAGE and subjected to in-gel trypsin digestion, and peptides were identified by LC-MS/MS (Fig. 2D). The most prominent species present in all samples were the affinity purification targets, GS::EEL-1, GS::EEL-1 LD, or GS::GFP (Fig. 2, F and H). Across four independent proteomic experiments, we identified 23 other proteins that were exclusive to, or enriched Ն2-fold in, the GS::EEL-1 or GS::EEL-1 LD samples compared with GS::GFP ( Fig. 2F and Table S1). 13 proteins were either unique to WT GS::EEL-1 or present in both WT GS::EEL-1 and GS::EEL-1 LD samples (Fig. 2, F and G) and Table S1). These proteins represented putative EEL-1 binding proteins. 10 proteins were exclusive to or enriched in the GS::EEL-1 LD sample, which suggested that these proteins could be putative ubiquitination substrates ( Fig. 2G and Table S1).
In contrast to UBQ-1, the cytosolic and nuclear O-GlcNAc transferase OGT-1 was a novel, interesting putative EEL-1 binding protein (Fig. 2, E-H). OGT-1 spectra were found exclusively in GS::EEL-1 and GS::EEL-1 LD samples with sequence coverage of 27 and 41%, respectively (Fig. S5). To determine whether there was an increased number of OGT-1 peptides associated with EEL-1 LD compared with WT EEL-1, we normalized the total number of OGT-1 peptides to the total number of EEL-1 target peptides. Importantly, this was feasible because we performed four blinded, independent proteomic experiments. Normalization indicated that OGT-1 was present at similar levels and not significantly enriched in EEL-1 LD samples compared with WT EEL-1 samples (Fig. 2,  G and H). These results suggest that OGT-1 is more likely to be an EEL-1 binding protein than an EEL-1 ubiquitination substrate, although they do not rule out the possibility that OGT-1 is ubiquitinated by EEL-1. Because OGT-1 potentially forms a complex with EEL-1, we prioritized investigating the biological relationship between these two proteins further.
To determine whether OGT-1 is a conserved EEL-1 binding protein, we tested whether this interaction occurs between HUWE1 and OGT, the sole human orthologs of EEL-1 and OGT-1. We observed co-IP of FLAG-OGT with human GFP-HUWE1 from transfected HEK 293 cells (Fig. 3B). These results indicate that EEL-1 binds OGT-1 in neurons in vivo and that this interaction is conserved between human HUWE1 and OGT.

OGT-1 is expressed in the motor circuit and localizes to GABA presynaptic terminals
We previously showed that eel-1 is expressed broadly in the nervous system, including both cholinergic and GABAergic motor neurons (20). Because proteomic and biochemical results indicate that EEL-1 and OGT-1 form a protein complex (Figs. 2 and 3), we wanted to evaluate ogt-1 expression in the nervous system.
Previous work using an OGT-1 translational GFP reporter showed that OGT-1 is widely expressed in intestine, hypodermis, and neurons (44). To confirm this expression pattern and further characterize ogt-1 neuronal expression, we generated transgenic animals that express an ogt-1 transcriptional GFP reporter. We observed ogt-1 expression broadly in the nervous system, as well as in pharyngeal muscle, intestine, and vulva (Fig. 4A). To determine whether ogt-1 is expressed in the motor circuit, we coexpressed the ogt-1 transcriptional GFP reporter with markers for either cholinergic (P unc-129 mCherry) or GABAergic (P unc-25 mCherry) motor neurons. We observed coexpression of GFP with mCherry in both cholinergic (Fig. 4B) and GABAergic motor neurons (Fig. 4C). Thus, ogt-1 is expressed in both OGT-1/EEL-1 protein complex affects GABA neuron function cholinergic and GABAergic neurons of the C. elegans motor circuit, similar to eel-1.
We previously showed that EEL-1 is localized to presynaptic terminals in GABAergic motor neurons (20). Therefore, we evaluated where OGT-1 localizes in GABAergic motor neurons. To do so, we generated transgenic extrachromosomal arrays that use a GABA neuron promoter to express mCherry::OGT-1. This transgenic array was generated on a background that carried a second integrated transgene that expressed the synaptic vesicle marker SNB-1 (Synaptobrevin-1) fused to GFP (SNB-1::GFP) in GABA neurons. We found that mCherry::OGT-1 colocalized with SNB-1::GFP at presynaptic terminals of GABA motor neurons (Fig. 4D). This indicates that OGT-1 is present at presynaptic terminals in GABA neurons, similar to what was previously observed for EEL-1.

OGT-1 functions in GABA neurons to affect motor circuit function
Prior work using both electrophysiology and pharmacological approaches showed that EEL-1 regulates GABAergic presynaptic transmission, thereby impacting motor circuit function (20). Because OGT-1 binds EEL-1 (Figs. 2 and 3A and Fig. S6), is expressed in GABAergic motor neurons (Fig.  4C), and localizes to GABAergic presynaptic terminals like EEL-1 (Fig. 4D), we tested how OGT-1 affects motor circuit function using the automated aldicarb assay described earlier (Fig. 1). We evaluated two ogt-1 mutants, ok430 and ok1474. These are both large insertion/deletions that introduce premature stop codons upstream of the glycosyltransferase domain and are likely null alleles (45,46). Both ogt-1 (ok430) and ogt-1 (ok1474) mutants were mildly hypersensitive to aldicarb (Fig. 5, A and C and Figs. S2B and S7). Aldicarb hypersensitivity was rescued by Mos1-mediated singlecopy insertion (MosSCI) of OGT-1 using the pan-neuronal rab-3 promoter (Fig. 5A and Figs. S2B and S7). Single-copy expression of OGT-1 in GABA neurons using the unc-47 promoter also significantly rescued aldicarb hypersensitivity defects in ogt-1 (ok430) mutants ( Fig. 5A and Fig. S2B). Weaker rescue with the GABA neuron promoter compared with the pan-neuronal driver might occur for two reasons. The GABA-specific promoter we used might not express OGT-1 at appropriate levels for rescue. Alternatively, OGT-1 could function in other types of neurons to contribute to aldicarb hypersensitivity. Nevertheless, these results show that OGT-1 functions in GABA neurons to regulate motor circuit function in C. elegans.

ogt-1 and eel-1 function in parallel to regulate motor circuit function and locomotion
We showed that OGT-1 functions in GABAergic motor neurons to regulate motor circuit function as assessed by aldicarb pharmacology (Fig. 5A). Because this result is similar to prior findings with EEL-1 (20), we investigated the genetic relationship between ogt-1 (ok430) and eel-1 (bgg1) mutants using automated aldicarb assays. Interestingly, eel-1; ogt-1 double mutants showed enhanced aldicarb hypersensitivity compared with eel-1 single mutants at low aldicarb dose (8 M) (Fig. 5B). Enhanced hypersensitivity was observed across aldicarb dose response for eel-1; ogt-1 double mutants compared with eel-1 mutants (Fig. 5C). To ensure that enhanced aldicarb hypersensitivity was not due to effects on locomotion, speed was normalized to no drug controls for each genotype for these experiments. Enhanced hypersensitivity to aldicarb displayed by eel-1; ogt-1 double mutants was rescued by using MosSCI to express OGT-1 in the nervous system (Fig. 5D). These results indicate that eel-1 and ogt-1 function in parallel in neurons to affect aldicarb sensitivity.
To further assess how genetic interactions between ogt-1 and eel-1 influence motor circuit function, we analyzed locomotion using MWT to monitor swimming speed. ogt-1 mutants had a mild defect in swim speed compared with WT animals, and eel-1 mutants were moderately impaired (Fig. 5E). Importantly,

OGT-1/EEL-1 protein complex affects GABA neuron function
eel-1; ogt-1 double mutants showed enhanced decreases in swim speed compared with eel-1 single mutants (Fig. 5E). Genetic enhancer effects between ogt-1 and eel-1 were also observed when brood size and viability were evaluated (Fig. S8). Thus, OGT-1 and EEL-1 function in parallel to affect multiple phenotypes with prominent effects on motor circuit function and locomotion.

OGT-1 does not affect synapse formation in GABA motor neurons
Next, we evaluated a straightforward hypothesis that might explain how OGT-1 and EEL-1 function in parallel to regulate GABA neuron function: Do OGT-1 and EEL-1 function in parallel to affect synapse formation in GABA neurons? To test this, we analyzed the distribution and pairing of pre-and postsynaptic markers at GABAergic neuromuscular junctions (NMJs). As a presynaptic marker, we used an integrated transgenic array that expresses SNB-1::GFP in GABA motor neurons (Fig. 6A). To postsynaptically label GABA synapses, we used a singlecopy transgene that expresses the GABA receptor subunit UNC-49 fused to TagRFP (UNC-49::RFP) (Fig. 6A). Results from confocal microscopy indicated that pairing of pre-and postsynaptic terminals was normal in ogt-1 (ok430) mutants and eel-1 (bgg1) mutants (Fig. 6B). Similarly, pre-and postsyn-aptic pairing was normal in eel-1; ogt-1 double mutants (Fig.  6B). Quantitation of the number of SNB-1::GFP puncta showed no defects in single mutants or eel-1; ogt-1 double mutants (Fig.  6C). These results demonstrate that synapse formation is not impaired at GABAergic NMJs of eel-1; ogt-1 double mutants and rules this out as an explanation for enhanced defects in aldicarb hypersensitivity and locomotion in these double mutants.

OGT-1 glycosyltransferase activity is dispensable for motor circuit function
To begin deciphering how OGT-1 affects motor circuit function, we tested whether OGT-1 glycosyltransferase activity is involved. To do so, we evaluated whether point mutations in OGT-1 that impair glycosyltransferase activity can rescue aldicarb hypersensitivity of ogt-1 null mutants. We tested two point mutants, OGT-1 K957M and H612A, that affect conserved residues known to completely abolish and strongly impair glycosyltransferase activity of OGT-1 orthologs, respectively (47)(48)(49). Interestingly, we observed similar strong rescue when MosSCI was used to pan-neuronally express WT OGT-1, OGT-1 K957M, or OGT-1 H612A (Fig. 7, A and B). These results demonstrate that OGT-1 functions independent of

OGT-1/EEL-1 protein complex affects GABA neuron function
enzymatic glycosyltransferase activity to regulate motor circuit function.
As an independent approach to address this question, we asked whether loss of OGA-1 (O-GlcNAcase 1), which has opposing enzymatic activity to OGT-1 (46), affects motor circuit function. If motor circuit function is affected by O-GlcNAcylation, oga-1 mutants would be expected to have the opposite phenotype to ogt-1 mutants. In the case of auto-

OGT-1/EEL-1 protein complex affects GABA neuron function
mated aldicarb assays, this would be resistance to aldicarb paralysis rather than hypersensitivity displayed by ogt-1 mutants (Fig. 5, A and C). To ensure that our automated aldicarb assay can accurately evaluate aldicarb resistance, we tested rab-3 mutants, which are mildly resistant to aldicarb (34). Indeed, rab-3 mutants were resistant to aldicarb compared with WT animals in our automated liquid assay (Fig. 7, C and D). In contrast, oga-1 null mutants had similar aldicarb sensitivity to WT animals across a range of doses (Fig. 7, C and D). Taken together, these results with catalytically inactive OGT-1 point mutants and oga-1 mutants suggest that OGT-1 does not rely upon glycosyltransferase activity to affect motor circuit function.

OGT-1 forms a protein complex with EEL-1 and affects GABA neuron function
We previously showed that the EEL-1 E3 ubiquitin ligase, a gigantic protein broadly expressed in the nervous system, plays an important role in motor circuit function by regulating presynaptic GABAergic transmission (20). We now provide evidence that EEL-1 ubiquitin ligase activity affects aldicarb sensitivity and locomotion, and therefore is likely to be involved in this process (Fig. 2C and Fig. S3).
To begin further deciphering how EEL-1 regulates GABAergic transmission, we used affinity purification proteomics from C. elegans to identify EEL-1 binding proteins. This approach identified the OGT-1 O-GlcNAc transferase as a prominent EEL-1 binding protein (Fig. 2, E-H). Biochemical results from C. elegans indicate that OGT-1 binds EEL-1 in neurons in vivo ( Fig. 3A and Fig. S6). Like EEL-1, OGT-1 is expressed broadly in the nervous system, including the motor circuit, and localizes to GABAergic presynaptic terminals (Fig. 4). Results from an automated pharmacological assay showed that OGT-1 functions in GABAergic motor neurons to affect motor circuit function ( Fig. 5A and Fig. S2B). Interestingly, our genetic results show that defects in motor circuit function are enhanced in ogt-1; eel-1 double mutants (Fig. 5, B-D). Consistent with this, enhanced defects in locomotion were also observed in eel-1; ogt-1 double mutants (Fig. 5E). Taken together, our results and prior findings support the conclusion that OGT-1 and EEL-1 form a complex and function in parallel to affect GABA neuron function, thereby influencing the motor circuit and locomotion (Fig. 8). In the future, more extensive physiological and cellular studies will be useful to further test this model. Nonetheless, it is noteworthy that our findings are the first example of OGT-1 or its orthologs forming a complex with a ubiquitin ligase to affect a common process in any system.
Two observations suggest that OGT-1 may not be ubiquitinated by EEL-1. First, EEL-1 affinity purification proteomics was done with two EEL-1 constructs, WT EEL-1 and EEL-1 LD (Fig. 2). Because EEL-1 LD cannot ubiquitinate targets, it potentially enriches ubiquitination substrates. Consistent with this, we detected several proteins that were exclusive to or enriched in EEL-1 LD samples, which represent putative ubiquitination substrates (Fig. 2G and Table S1). In contrast, OGT-1 was not significantly enriched in normalized EEL-1 LD samples (Fig. 2, G and H). The second argument is based on genetics. We observed that aldicarb hypersensitivity was enhanced in ogt-1; eel-1 double mutants (Fig. 5, B-D), which indicates that ogt-1 and eel-1 function in parallel to regulate GABA neuron function. If EEL-1 ubiquitinates OGT-1, we would expect a genetic relationship reflective of same pathway genetics, such as suppression or no increased effects in double mutants, neither of which occurred. Thus, our findings with affinity purification proteomics and loss-of-function genetics do not rule out the possibility that OGT-1 is ubiquitinated by EEL-1 but do suggest that this is unlikely in the context of GABA neuron function.
At present, the function of C. elegans OGT-1 and mammalian OGT in the nervous system remains minimally explored. Our results show that OGT-1 is expressed broadly throughout the nervous system, including both excitatory cholinergic and inhibitory GABAergic neurons in the motor circuit (Fig. 4, A-C). This suggests that OGT-1 could have wide-ranging functions in the nervous system and could affect both excitatory and inhibitory motor neuron function. However, ogt-1 mutants are hypersensitive to aldicarb, and hypersensitivity is rescued by OGT-1 expression in GABA neurons ( Fig. 5A and Fig. S2B). Furthermore, aldicarb hypersensitivity in ogt-1 mutants is enhanced by eel-1, a known regulator of presynaptic GABAergic transmission (20). This suggests that the OGT-1/EEL-1 complex is likely to preferentially affect GABA neuron function. Consistent with this, a prior study in mammals showed that OGT regulates the activity of GABAergic AgRP neurons (32). Moving forward, it will be important to address whether the OGT-1/EEL-1 complex is a conserved regulator of GABA neuron function and determine how this complex regulates GABA neuron function.
As an initial foray into addressing this question, we show that OGT-1 does not require enzymatic glycosyltransferase activity to regulate motor circuit function. This is supported by results using two different conserved mutations that impair OGT-1 glycosyltransferase activity, one of which completely abolishes enzymatic activity (Fig. 7, A and B). Consistent with this, motor circuit function was not affected by OGA-1, which removes O-GlcNAc and opposes the activity of OGT-1 (Fig. 7, C and D). While prior studies have shown that OGT-1 has nonenzymatic activities, such as acting as a scaffold protein, these remain understudied (25). Importantly, it is unknown whether OGT has nonenzymatic functions in the nervous system. Our results using an in vivo model system now suggest that OGT-1 functions independent of glycosyltransferase activity and in a complex with EEL-1 to regulate GABA neuron function.
Interestingly, many of our observations on the OGT-1/EEL-1 complex share similarities with prior findings on the OGT/ mSin3A complex that represses transcription (50). In both cases, OGT glycosyltransferase activity is dispensable, and OGT acts in parallel to the enzymatic activity of its binding

OGT-1/EEL-1 protein complex affects GABA neuron function
partner. Thus, our work and this prior study highlight core biochemical principles at play in OGT complexes that act independently of glycosyltransferase activity.

Implications of the OGT/HUWE1 complex in cell biology and disease
Our discovery of the OGT-1/EEL-1 complex could have broad implications within the nervous system, outside the nervous system, and across species. We found that the physical interaction between OGT-1 and EEL-1 is conserved between the human orthologs, OGT and HUWE1 (Fig. 3B). Thus, the interaction between OGT and HUWE1 is conserved in human cells and could be functionally relevant across species. Consistent with our findings, two prior affinity purification proteomic studies with HEK 293 cells hinted at the possibility of an OGT/HUWE1 complex (51,52). We now show that an OGT/HUWE1 complex exists, the interaction between OGT and HUWE1 is evolutionarily conserved, and these molecules have functional genetic interactions in C. elegans that affect GABA neuron function and locomotion.
Consistent with our results, previous studies that examined OGT or HUWE1 independently indicate that these molecules converge on several cellular functions as well as during disease. Both OGT and HUWE1 (also called Arf-BP1/Mule) are implicated in oncogenesis and regulate the transcription factor Myc (53,54). In the nervous system, both HUWE1 and OGT affect neural progenitor proliferation (37,40,55,56) and mitochondrial function (30,(57)(58)(59). Moreover, genetic changes in both HUWE1 and OGT are linked to intellectual disability. Human genetic studies have shown that both increases in copy number and missense mutations in HUWE1 result in intellectual disability (21)(22)(23). Because some disorder-associated point mutations in HUWE1 are loss-of-function, it is likely that both increased and reduced HUWE1 function are associated with intellectual disability (20). In recent human genetic studies, three different mutations in OGT were suggested to cause intellectual disability (60,61). Thus, our discovery of a conserved, functionally relevant OGT-1/EEL-1 complex not only informs our understanding of how GABA neuron function is regulated in a model circuit, but could have important implications for oncogenesis, neural progenitor proliferation, mitochondrial function, and intellectual disability.
Construction details for transgenic alleles are described in Table S2 and were made using standard procedures (63,64). Plasmid details and sequences are available upon request.
Custom-written scripts calculated the mean speed every minute per well. For Figs. 1 and 2, speed was normalized to baseline (10 min of swimming in the absence of aldicarb) for each genotype. For Fig. 5, speed was normalized to no aldicarb controls for the length of the experiment to rule out effects of a given genotype on locomotion. This was particularly important for eel-1; ogt-1 double mutants that displayed enhanced defects in both aldicarb paralysis (Fig. 5, B-D) and swimming speed in the absence of aldicarb (Fig. 5E). Note paralysis threshold was defined as at or below 0.02 mm/s. For locomotion analysis in Fig. 5E, speed across all time points was normalized to the first 5 min of the assay for each genotype.
Volume of lysate required for 200 mg of total protein was incubated with 400 l of IgG-coated Dynabeads (Invitrogen) for 1 h at 4°C. Following affinity capture, beads were washed five times with lysis buffer. Purification quality was assessed by immunoblotting 1% of sample (anti-SBP antibody, Sigma-Aldrich) and silver staining 9% of the sample (Thermo Scientific). The remaining 90% of the sample was run on SDS-PAGE, subject to in-gel trypsin digestion, and run on LC-MS/MS. Prior to MS analysis, pooled peptides were acidified, desalted with a Zip-Tip C18 column, dried, and resuspended in 100 l of 0.1% formic acid. 13 l of sample was used per MS run. Samples were analyzed using an Orbitrap Fusion TM Tribrid TM mass spectrometer (Thermo Fisher Scientific) coupled to an EASY-nLC 1000 system. Peptides were eluted on an analytical RP column (0.075 ϫ 250-mm Acclaim PepMap RLSC nano Viper, Thermo
Spectra were analyzed using Mascot (Matrix Science) and Sequest (Thermo Fisher Scientific). Mascot and Sequest were set up to search the C. elegans proteome (UniProt, May 2017, 27,483 entries). Analysis parameters included fragment ion mass tolerance of 20 ppm (Mascot) or 0.02 Da (Sequest) and parent ion tolerance of 10.0 ppm. Carbamidomethyl of cysteine was specified as a fixed modification, and deamidation of asparagine and glutamine and oxidation of methionine were specified as variable modifications.
Scaffold (Proteome Software) was used to validate MS/MSbased peptide and protein identifications. Peptide identifications were accepted if they could be established at Ͼ5.0% probability to achieve a false discovery rate (FDR) Ͻ 1.0% by the Scaffold Local FDR algorithm. Protein identifications were accepted if they could be established at Ͼ98.0% probability to achieve an FDR Ͻ1.0% and contained at least two identified peptides. Protein probabilities were assigned by the Protein Prophet algorithm (67). EEL-1 binding proteins were identified using the following criteria: 1) the protein was detected in at least two of the four experiments; 2) the protein had 2ϫ or more total spectra in the GS::EEL-1 or GS::EEL-1 LD samples compared with GS::GFP negative control; 3) ribosomal proteins were removed. Spectra were normalized by subtracting spectra from negative control (GS::GFP) and dividing by the target (EEL-1).
Lysates (10 mg of total protein C. elegans, 0.65 mg of total protein 293 cells) were incubated with primary antibody for 30 min and precipitated for 1 h (C. elegans) or 4 h (293 cells) with 10 l of protein G-agarose (Roche Applied Science) at 4°C. Antibodies used for IP included mouse monoclonal anti-FLAG (M2, Sigma-Aldrich) and anti-GFP (3E6, Invitrogen). Precipitates were boiled in sample buffer and run on 3-8% Tris acetate gels (Invitrogen). Gels were transferred overnight to polyvinylidene difluoride membranes and immunoblotted with either rabbit polyclonal anti-FLAG (Cell Signaling) or mouse monoclonal anti-GFP (Roche Applied Science) antibodies. Proteins were visualized using horseradish peroxidase-conjugated secondary antibodies (GE Healthcare Life Sciences, Fisher Scientific) and ECL (Supersignal West Femto or Pico, Thermo Scientific). Blots were imaged with X-ray film for C. elegans experiments and digitally imaged for 293 experiments (KwikQuant TM Imager).

Microscopy
Epifluorescence microscopy was performed using a Leica CFR5000 with a ϫ40 magnification oil-immersion lens. Images were acquired using a CCD camera (Leica DFC345 FX). Confocal microscopy was performed using a Leica TCS SP8 MP confocal microscope system with a ϫ25 or ϫ40 water immersion lens. Settings that avoided bleed-through between different channels were confirmed when collecting coexpression and colocalization. For colocalization experiments, images were acquired in resonant mode.
Young adult animals were anesthetized and mounted on 2% agar for imaging. When colocalization was being evaluated, animals were anesthetized with 5 mM levamisole in M9 buffer. Epifluorescent microscopy was used to quantitate SNB-1::GFP puncta with animals anesthetized using 1% 1-phenoxy-2-propanol in M9 buffer.

Brood size and viability assay
Individual L4 larvae were placed on fresh plates every 12 h for 4 days at 20°C. Brood size was total eggs laid per animal over 4 days. Eggs were cultivated at ϳ23°C for 3 days, and viability was measured as the percentage of eggs that developed to L4 stage or older.

Statistical analysis
For aldicarb and swim speed experiments, data were derived from multiple wells acquired across multiple independent experiments. Comparisons were made using twoway ANOVAs. If a significant effect or interaction was observed (p Ͻ 0.05), Fisher's LSD post hoc tests were performed to evaluate differences. Individual time points that best represented differences between genotypes were chosen for presentation as insets, but post hoc comparisons at all time points were also done (e.g. see Fig. S2). For proteomics, comparisons were made using Student's t test. For quantitation of SNB-1::GFP puncta, images were collected for each genotype from three independent experiments, and significance was tested using ANOVA. For brood size and viability, significance was assessed using ANOVA and Fisher's LSD post hoc tests.