The UBR-1 Ubiquitin Ligase Regulates Glutamate Metabolism to Generate Coordinated Motor Pattern in C. elegans

UBR1 is an E3 ubiquitin ligase best known for its ability to target protein degradation by the N-end rule. The physiological functions of UBR family proteins, however, remain not fully understood. We found that the functional loss of C. elegans UBR-1 leads to a specific motor deficit: when adult animals generate reversal movements, A-class motor neurons exhibit synchronized activation, preventing body bending. This motor deficit is rescued by removing GOT-1, a transaminase that converts aspartate to glutamate. Both UBR-1 and GOT-1 are expressed and critically required in premotor interneurons of the reversal motor circuit to regulate the motor pattern. ubr-1 and got-1 mutants exhibit elevated and decreased glutamate level, respectively. These results raise an intriguing possibility that UBR proteins regulate glutamate metabolism, which is critical for neuronal development and signaling. Author Summary Ubiquitin-mediated protein degradation is central to diverse biological processes. The selection of substrates for degradation is carried out by the E3 ubiquitin ligases, which target specific groups of proteins for ubiquitination. The human genome encodes hundreds of E3 ligases; many exhibit sequence conservation across animal species, including one such ligase called UBR1. Patients carrying mutations in UBR1 exhibit severe systemic defects, but the biology behinds UBR1’s physiological function remains elusive. Here we found that the C. elegans UBR-1 regulates glutamate level. When UBR-1 is defective, C. elegans exhibits increased glutamate; this leads to synchronization of motor neuron activity, hence defective locomotion when animals reach adulthood. UBR1-mediated glutamate metabolism may contribute to the physiological defects of UBR1 mutations.


Introduction 1
In eukaryotic cells, the ubiquitin-proteasome system mediates selective protein 2 degradation (1, 2). E3 ubiquitin ligases confer substrate specificity via selective 3 interaction with the degradation signals in substrates (3)(4)(5)(6). 4 UBR1 acts not only as an E3 ligase for the N-end rule substrates, whose metabolic 5 stability is determined by the identity and post-translational status of their N-terminal 6 moiety, but also for substrates that do not harbor the N-terminal degrons (6, 7). The UBR 7 family proteins exist from yeast to man, and have been implicated in multiple cellular 8 processes (reviewed in 7). Yeast UBR1 is not essential, but ubr1 mutants exhibit less 9 efficient chromatin separation, and mildly increased doubling time (8). Yeast UBR1 also 10 participates in protein quality control, potentiating the degradation of mis-folded proteins 11 by ER membrane ligases (9). The loss of C. elegans UBR-1 results in delayed 12 degradation of a regulator for post-embryonic hypodermic cell division, but does not 13 cause obvious hypodermic defect (10). In mammalian cell lines, the N-end rule pathway 14 targets pro-apoptosis fragments for degradation, affecting the efficacy of induced 15 apoptosis (11). The simultaneous loss of two rodent UBR homologues results in 16 embryonic lethality with severe developmental defects in the heart and brain (12). In 17 human, loss-of-function mutations in one of several UBR family proteins, UBR1, cause 18 the Johanson-Blizzard Syndrome (JBS), a genetic disorder with multi-systemic symptoms 19 including pancreatic insufficiency, growth retardation, and cognitive impairments (13). 20 To date, a unifying physiological function of UBR proteins in animal models and human 21 is lacking. In fact, whether UBR1's role as an N-end rule E3 ligase is relevant for the JBS 22 pathophysiology remains elusive (14). 23 neighboring sequences interact with the substrates and E2 ubiquitin-conjugating enzyme 1 of the N-end rule (22), whereas the RING finger is the hallmark motif utilized by a large 2 class of E3 ligases to recruit non-N-end rule substrates (23)(24)(25). 3 To further verify that the motor phenotypes that we observed in hp684 mutants 4 result from the functional loss of UBR-1, we generated multiple ubr-1 deletion alleles, 5 frequency and amplitude of the calcium waveforms for A-MNs varied for each reversal 1 event (47), their calcium profiles exhibited phase relations that are consisted with the 2 expected temporal activation of muscle groups that they are predicted to innervate. 3 Specifically, VA10 and VA11, two A-MNs that innervate adjacent ventral muscles 4 exhibited asynchrony in activation (red and blue traces in Fig. 3C; left panel in Fig. 3C'), 5 with variable lags ( Fig. 3G; Methods), as expected from the sequential contraction of 6 adjacent muscles during reversals at different velocities. DA7 likely innervates dorsal 7 muscles that appose VA10 and V11's ventral targets ( dorsal-ventral opposition between VA10 and DA7's muscle targets. 14 We observed a striking difference in A-MN's activation pattern in ubr-1 mutants. 15 While they also exhibited calcium changes during reversals, all three A-MNs' activation 16 exhibited synchrony (blue, red and green traces in Fig. 3D; Fig. 3D'). This led to a drastic 17 reduction in the mean lag times between VA10 and VA11 (Fig. 3G), and between VA10 18 and DA7 (Fig. 3H), whereas the short lags between DA7 and VA11 remained statistically 19 unchanged (Fig. 3I) between wildtype and ubr-1 mutants. 20 Hence, reduced bending in ubr-1 mutants was caused by increased synchronization, 21 not lack of A-MNs' activities. Importantly, when UBR-1 was restored in the premotor 22 interneurons of the reversal circuit, A-MN's phasic relationships were restored in ubr-1 23 (Table 2). lf mutations in other metabolic enzymes or transporters that may be involved in 1 glutamate and aspartate metabolic pathways (Table 2), including the alanine 2 aminotransferase, glutamine synthetase, and glutaminase, did not rescue ubr-1 mutant's 3 motor defects either. Henceforth, we refer to got-1.2 as got-1. 4 A critical requirement of GOT-1 activity in premotor interneurons to affect bending 5 To determine the endogenous expression pattern of GOT-1, we generated a functional 6 GOT-1 reporter by inserting GFP at the endogenous got-1 locus. GOT-1::GFP exhibited 7 cytoplasmic expression in all somatic tissues, including broad expression in the nervous 8 system ( Fig. 4B; Fig. S2B). To determine the critical cells that mediate the genetic 9 interactions between ubr-1 and got-1, we restored cell-type specific GOT-1 expression in 10 ubr-1; got-1 mutants, and assessed their effect on the animal's motor pattern. Restoring 11 GOT-1 in all neurons (Prgef-1), but not in all muscles (Pmyo-3), fully reverted the motor 12 pattern of ubr-1; got-1 to the reduced bending as in ubr-1 mutants (Table 1); therefore, 13 both UBR-1 and GOT-1 function through neurons to regulate motor patterns. 14 Because GOT-1 is more broadly expressed in the nervous system than the UBR-15 1::GFP reporter (Fig. 4B), we examined whether GOT-1 functions through UBR-1-16 expressing neurons to regulate bending. Similar to our observation for neuronal sub-type 17 UBR-1 rescue (Fig. 2), restoring GOT-1 in premotor interneurons of the reversal circuit 18 (Table 1) was required for reversion of ubr-1; got-1's bending pattern to that of ubr-1. 19 Similarly, restoring GOT-1 in the same subset of these premotor interneurons, including 20 AVE and RIM, exerted the most significant, partial reversion of ubr-1; got-1's motor 21 defects ( Fig. 4D; Fig. S4). 22 Therefore, not only does a predicted metabolic enzyme GOT-1 exhibit genetic 1 interaction with UBR-1, but also both proteins exhibit similar prominent requirement in 2 premotor interneurons to regulate the reversal motor pattern. These results raise the 3 possibility that metabolic dys-regulation may underlie ubr-1's motor defects. 4 GOT-1 synthesizes glutamate using aspartate 5 The catalytic activity of GOT-family transaminases enables reversible conversions 6 between aspartate and glutamate (left panel in Fig. 5A). Their in vivo activity and 7 physiological function, however, have not been examined in animal models. 8 To determine the metabolic changes associated with the loss of GOT-1, we 9 quantified the amino acid levels of synchronized wildtype and got-1 adults by high 10 performance liquid chromatography (HPLC). As previously reported (50), glutamate is 11 an abundant amino acid, whereas aspartate is maintained at a low abundance in C. 12 elegans (upper panel in Fig. 5B). In got-1 mutants, glutamate level exhibited a decrease 13 of 27.6 ± 7.0% (mean ±SEM, n=4, P=0.0477 against wildtype animals), whereas the 14 aspartate level exhibited a massive increase of 27.3 ± 4.76 folds (n=4, P=0.0076 against 15 wildtype animals) (lower panel in Fig. 5B; Fig. 6A, B). These results implicate that in 16 vivo, GOT-1 preferentially synthesizes glutamate from aspartate (right panel in Fig. 5A). 17 Among the C. elegans GOTs, GOT-1 appears to be the key glutamate-synthesizing 18 enzyme, because removing its homologue, GOT-2, did not lead to glutamate reduction 19 (Fig. 6A). 20 We noted that compared to the massive aspartate accumulation, the reduction of 21 glutamate was mild in got-1 mutants. This may result from compensatory activation of 22 glutamate synthesis using other amino acids. As reported (50), alanine is the most 23 abundant amino acid in C. elegans (upper panel in Fig. 5B). In got-1 mutants, alanine 1 level exhibited a decrease of 48.5 ± 12.4% (n=4, P=0.0050 against wildtype animals) 2 (lower panel in Fig. 5B; Fig. 6C), supporting the notion that alanine becomes the 3 compensatory source for glutamate when conversion from aspartate is blocked. 4 Such a drastic shift in equilibrium of the three key amino acids -aspartate, 5 glutamate, and alanine -in got-1 mutants must exert indirect metabolic consequences. To 6 determine whether GOT-1's loss affects the global metabolic state, we performed liquid 7 chromatography-mass spectrometry (LC-MS) analyses on whole worm lysates. Indeed, 8 got-1 mutants exhibited increased AMP/ATP and NADP/NADPH, and decreased 9 GSH/GSSG glutathione ratios (Fig. 6E, F), two hallmark features for increased cellular 10 toxicity and metabolic stress (51). 11 We conclude that in C. elegans, GOT-1 synthesizes glutamate, and maintains 12 glutamate level using aspartate. The loss of GOT-1 leads to glutamate reduction, 13 aspartate accumulation, and potential compensatory glutamate synthesis using other 14 amino acids. 15

Some premotor interneurons exhibit defective morphology in ubr-1 adults 15
Glutamate is a neurotransmitter, as well as an abundant amino acid partaking in 16 other metabolic pathways including amino acid synthesis, energy production and urea 17 cycle. An elevation of glutamate level may result in not only increased neuronal signaling 18 in glutamatergic neurons, but also cellular stress in all UBR-1-expressing cells. 19 Because most of the critically required premotor interneurons for UBR-1's role in 20 motor pattern are cholinergic (52), the developmental effect of increased glutamate in 21 these neurons may play a prominent role. To explore this possibility, we examined these 22 interneurons using a transcriptional reporter (Pnmr-1-RFP) (18) that allows visualization 23 of their somata ( Fig. 7; Fig. S6), and a translation reporter (GLR-1::GFP) (18) that labels 1 synapses of the AVA and AVE premotor interneurons(18) (Fig. 7). 2 In adult ubr-1 mutants, both reporters exhibited prominent and moderate decrease 3 in fluorescence intensity in premotor interneurons AVA and AVE, respectively ( Fig. 7A-4 D for Pnmr-1; E-F for GLR-1::GFP). Such a decrease was not observed for other 5 interneurons including RIM (Fig. 7A-D). We further noted that unlike the smooth and 6 round somata in wildtype animals (Fig. S6, wildtype), in ubr-1 mutants, the AVA and 7 AVE somata exhibited rough edges and short branches in L4 stage larvae and adults (Fig. 8 S6A, L4 and Adult panels), whereas other premotor interneurons such as RIM were 9 appeared as in wildtype animals (Fig. S6B). The late onset of reversal bending defects 10 coincides with the morphological change in AVA and AVE becoming prominent by the 11 end of the larval stage (Fig. S6A). Their fluorescent intensity and morphology defects 12 were significantly rescued by restoring UBR-1 expression in premotor interneurons ( These results support the notion that an increased glutamate level may lead to 15 developmental defects in ubr-1 mutants, and premotor interneurons, in particular AVA 16 and AVE, may be more susceptible to such a perturbation than other neurons or cells. 17

GOT activity but not total protein level is increased in ubr-1 mutants 18
Our analyses attributed the motor defect of ubr-1 mutants to altered glutamate 19 metabolism, most critically from premotor interneurons. To address how UBR-1 may 20 negatively regulates glutamate level through GOT-1, we first assayed the total GOT 21 activity of the whole worm lysates. We observed a moderate increase of GOT activity in 22 ubr-1 mutants (Fig. 8A). Consistent with the presence of multiple GOT homologues, the 23 total GOT activity was drastically reduced, but not abolished in got-1 mutants. The 1 increased GOT activity in ubr-1 mutants was attenuated in ubr-1; got-1 double mutants. 2 The attenuation effect was also specific to GOT-1: the loss of GOT-2 did not reduce 3 GOT activity in ubr-1 mutants (Fig. 8A). 4 An increased GOT activity is consistent with an elevated glutamate level in ubr-1 5 mutants. However, we did not observe changes in GOT-1 protein level in ubr-1 mutants. 6 The level of GOT-1::GFP, expressed either from its endogenous locus or from an 7 exogenous panneuronal promoter (Prgef-1), was similar between wildtype and ubr-1 8 mutants (Fig. 8B). These results suggest that GOT-1 is not a direct substrate of UBR-1. 9 UBR-1-mediated regulation of glutamate metabolism may involve targeting other 10 pathway components. 11

Discussion 1
Dys-regulated E3 activities have been implicated in neurodevelopmental and 2 neurological disorders (53). Using the C. elegans model and its motor output as the 3 functional readout, we reveal a previously unknown role of UBR-1 in glutamate 4 metabolism. The absence of UBR-1 elevates glutamate level, which promotes 5 simultaneous activation of A-class motor neurons, leading to reduced bending during 6 reversal movements. This defect is compensated by removing a glutamate-synthesizing 7 enzyme and reducing glutamate level in premotor interneurons of the reversal motor 8 circuit. Aberrant glutamate metabolism may underlie ubr-1 mutant's physiological 9 defects. 10

ubr-1's motor phenotype is associated with glutamate level change 11
Coinciding with an elevated GOT activity, ubr-1 mutants exhibit increased glutamate 12 level. Removing a key glutamate-synthesizing enzyme GOT-1 reduces the glutamate 13 level in ubr-1 mutants. These metabolic changes parallel those in motor behaviors: 14 removing the activity of GOT-1 enzyme restores bending in ubr-1 mutants. These results 15 imply that elevated glutamate level is associated with ubr-1's defective bending. 16 Genetic mutations in key metabolic enzymes can exert effects beyond their 17 primary substrates and immediate metabolic pathways. For example, upon the loss of 18 GOT-1, the reduction of glutamate synthesis from aspartate likely induces compensatory 19 utility of other amino acids, thus the reduction of alanine. When converting aspartate to 20 glutamate, GOT-1 may indirectly affect the equilibrium of other substrates, OAA and α-21 KG, metabolites that take part in multiple metabolic functions, including energy 22 production, amino acid homeostasis, nucleotide synthesis, and lipid synthesis. 23 Our metabolomics analyses of got-1 mutants revealed no consistent or correlated 1 changes in the level of OAA, α-KG (Fig. 6C), and all other TCA cycle metabolites from 2 wildtype controls (Fig. S5). Because the same samples exhibited consistent and 3 correlated changes between the aspartate, glutamate and alanine levels, the variability in 4 other metabolite levels likely reflects flexibility of compensatory or adaptive changes in 5 response to the primary metabolic dysfunction upon the loss of GOT-1. Indeed, the only 6 consistent and anti-correlated metabolic change that we observed between ubr-1 and got-7 1 (and ubr-1; got-1) mutants was the glutamate level. Cumulatively, these results strongly 8 support that the primary metabolic change -the level of glutamate -causes the motor 9 pattern change in ubr-1 and ubr-1; got-1 mutants. 10 Our analyses measured glutamate level in whole animals, not the specific neuron 11 groups that are critically required for both UBR-1 and GOT-1's effect on reversals. 12 Examples of genes with broad expression, but cell-type specific functions have been 13 reported (32, 54, 55). These and other results -such as the presence and requirement of 14 both UBR-1 and GOT-1 in premotor interneurons in mediating their effects on the 15 reversal motor pattern -suggest that the glutamate level change in premotor interneurons 16 significantly contribute to the motor pattern change in ubr-1 and ubr-1; got-1 mutants. 17 We reckon that other UBR-1-expressing neurons likely also contribute to such an 18 alteration. Furthermore, the glutamate change in other neurons and cells may cause other 19 physiological effects that were not examined in this study, where we only followed the 20 most obvious motor defect. 21

Increased glutamate may cause defective premotor interneuron development 22
Glutamate has multiple functions. Being a key neurotransmitter itself, glutamate 1 also serves as the sole precursor of another neurotransmitter GABA. Glutamate is also 2 one of the abundant amino acids partaking in other metabolic pathways including amino 3 acid synthesis, energy production and urea cycle. An elevation of glutamate not only 4 affects signaling in glutamatergic and likely GABAergic neurons, but also increases 5 metabolic stress in all cells. 6 Because both UBR-1 and GOT-1 exert strong effects on the reversal motor pattern 7 through premotor interneurons, most of which cholinergic (52), the developmental effect 8 of increased glutamate may play a prominent role in these interneurons. Indeed, we 9 observed morphological changes of premotor interneurons that coincided with the onset 10 of motor pattern change in ubr-1 mutants. Interestingly, within the reversal motor circuit, 11 premotor interneurons that exhibited the prominent changes, AVA and AVE, provide 12 direct synaptic inputs to the A-class motor neurons. These observations suggest that 13 compared to other cells and neurons, the development and function of premotor 14 interneurons, in particular AVA and AVE, may be more susceptible to glutamate 15 increase, likely through both increased glutamatergic inputs and cellular stress. 16

GOT-1 is a key enzyme for glutamate homeostasis in neurons 17
Despite its broad expression, GOT-1 can only influence ubr-1's motor defects through 18 neurons. These results indicate that maintaining glutamate level in neurons, especially the 19 premotor interneurons, is critical for the proper motor output of ubr-1 mutants. 20 In the mammalian brains, glutamate has to be locally synthesized because it cannot 21 cross the blood-brain barrier (56). The activity of transaminases, mainly by GOT and 22 ALT, establishes and maintains homeostasis of three abundant amino acids, glutamate, 23 alanine, and aspartate (57, 58). A recent study suggests that GOT also contributes to 1 glutamate synthesis at synapses (59). 2 The main sources of glutamate synthesis and homeostasis in C. elegans neurons 3 have remained elusive (60). Our study establishes GOT-1 as a key enzyme for glutamate 4 synthesis. Further, it provides the first evidence of a critical role of GOT-1 in glutamate 5 metabolism in the nervous system. 6 Direct substrates through which UBR-1 regulates glutamate level are unknown 7 How UBR-1 negatively regulates glutamate level remains to be elucidated. UBR family 8 proteins are often addressed in the context of N-end rule E3 ligases (6, 7, 10). However, 9 UBR proteins have non-N-end rule substrates, and N-end rule substrates do not account 10 for all described functions of UBR proteins (21, 61-64), and their physiological relevance 11 remains to be clarified. 12 For example, in mammalian cells, UBR1's N-end rule substrates include pro-13 apoptotic fragments (11). In C. elegans, the CED-3 caspase promotes apoptosis. During 14 postembryonic development, it interacts with UBR-1 to expose the N-degron of LIN-28 15 (10), a regulator of seam cell patterning (65) to promote its degradation. However, neither 16 UBR-1 nor LIN-28's N-degron were essential for LIN-28's degradation (10). Removing 17 either apoptotic regulators CED-3 or CED-4, which generates pro-apoptotic fragments, or 18 LIN-28, neither mimic nor rescue ubr-1 mutants' bending defects (Table 2; 19 Supplementary Movie 3). Substrates for neuronal function of UBR-1 remain to be 20

identified. 21
Regardless of the nature of UBR-1 targets, our results show that UBR-1 is 22 unlikely to directly regulate GOT-1's protein stability: we did not observe a change in the 23 level of GOT-1::GFP in ubr-1 mutants. We speculate that UBR-1-mediated regulation of 1 glutamate level may involve targeting other components that subsequently affect 2 glutamate metabolism. For example, the activity of many rate-limiting metabolic 3 enzymes in glycolysis, fatty acid synthesis, cholesterol synthesis, and gluconeogenesis 4 are regulated by phosphorylation and de-phosphorylation (66, 67). UBR-1 may mediate 5 the ubiquitination of kinases or phosphatases that modify GOT-1's activity. A large-scale 6 ubr-1 suppressor screen may be required to yield insights on UBR-1's direct substrates. 7 The C. elegans motor behavior as a model to investigate UBR1 and JBS 8 The UBR family proteins have been extensively examined in yeast, cultured cells, and 9 mouse models, but only recently in C. elegans. In yeast, UBR1 affects cell cycle 10 progression, but is non-essential (6). Mouse models revealed the functional redundancy 11 of multiple UBR homologues, where combinatorial knockout of UBR1 and UBR2 results 12 in embryonic lethality (12). C. elegans UBR1 affects the stability of LIN-28, a regulator 13 of postembryonic hypoderm seam cell division (10); both UBR1 and LIN-28 are non-14 essential for viability. The simplicity and viability of the C. elegans ubr-1 model has 15 allowed us to reveal, and genetically dissect a previously unknown physiological role of 16 UBR-1 in glutamate homeostasis. We note that aberrant glutamate metabolism may cause 17 systemic cellular and other unexamined neuronal defects in ubr-1 mutants. However, the 18 simplicity and sensitivity of the C. elegans premotor interneuron circuit, and the 19 prominent motor pattern change provided a quantifiable functional readout to afford 20 genetic pathway dissection. 21 Our study provides the first demonstration of a UBR protein's physiological role in 22 glutamate regulation. We noted with interest that in a case study of a JBS patient with 23 severe cognitive impairments, GOT activity levels, used as a biomarker of inflammation, 1 were increased (68). In addition to its role in development, there is a growing body of 2 evidence for glutamate signaling in non-neuronal tissues (48,69,70). It will be of interest 3 to examine whether glutamate level and signaling in other JBS animal models and JBS 4 patients are aberrant, and whether they contribute to JBS pathophysiology. 5

Strains and constructs 2
Strains: See Tables 1, 2 and S1 for a complete strain list generated in this study. All C. 3 elegans were cultured on standard NGM plates seeded with OP50, and maintained at 4 hp684 and hp731 were isolated by EMS mutagenesis. They were mapped using 6 SNP mapping and whole genome sequencing (71, 72), followed by behavioral rescue 7 using fosmids and genomic fragments that harbor ubr-1 and got-1.2, respectively. Placing 8 hp731 over a deficiency mnDf1 fully recapitulated the rescuing effect in the ubr-1 mutant 9 background, confirming that hp731 being a lf allele of got-1.2 (Table 2). hp820, hp821, 10 hp820hp833 and hp865 were generated by CRISPR-Cas9-mediated genome editing, 11 following protocols described in (26). The rest of genetic mutants were obtained from 12 CGC; all strains were backcrossed at least four times against N2 by genotyping. 13 Transgenic strains include those with extra-chromosomal multi-copy arrays 14 (hpEx), integrated multi-copy arrays (hpIs), single-copy integrated arrays (hpSi), and a 15 got-1.2 GFP knock-in allele (hp). Transgenic animals carrying extra-chromosomal arrays 16 (hpEx) were created by co-injecting the DNA plasmid of interest and a co-injection 17 marker at 5-30 ng/µL. Extra-chromosomal arrays were integrated into the genome using 18 UV irradiation to create stable, transgenic lines (hpIs) (73). We found that GOT-1 19 overexpression rendered sick animals. Hence, all got-1 tissue-specific rescue lines were 20 generated using Mos1-mediated single copy insertion (hpSi) as previously described (74). 21 All integrated strains were outcrossed several times against N2 prior to phenotypic 22 analyses. 23 GOT-1.2::GFP knock-in: The in-frame GOT-1 C-terminal GFP fusion allele (hp881) was 1 generated by CRISPR-Cas9 mediated homologous recombination, following protocols 2 described in (75). The replacement template for GOT-1::GFP (pJH3629) includes 1kb 3 sequence upstream to the got-1 start codon, the entire got-1 coding sequence, and 1.5kb 4 sequence downstream of the got-1 stop codon. GFP was inserted in-frame after the last 5 amino acid of GOT-1. A LoxP-Prps-27-NeoR-loxP cassette was inserted between the 6 stop codon and 3' UTR. After injection, animals were allowed to lay eggs overnight at 7 25 0 C before G418 selection. Animals with extra-chromosomal arrays were selected 8 against as described. Candidates for insertion were confirmed by genotyping and 9 sequencing. Confirmed insertion lines were injected with a Pelt-3::Cre plasmid to 10 remove the LoxP cassette. The resulting hp881 allele was outcrossed against N2 wildtype 11 animals twice before crossed into the ubr-1 mutants. 12

Locomotion Analysis 13
Plate conditions: When transferred to a new, thinly seeded plate, C. elegans typically 14 spend most of the time moving forward, with brief interruptions of backward movement. 15 As previously described (42), 35 mm NGM plates with limited food (lightly seeded OP50 16 bacteria) were used for automated tracking and behavioral analyses. Here we quantified 17 body curvature, initiation frequency, and duration on one-day old young adults using 18 copper chloride (CuCl 2 ), an aversive stimulus for C. elegans (76). Briefly, immediately 19 before transferring worms onto the NGM plates, a ~10-20 µl of 100 mM CuCl 2 solution 20 was pipetted into a small circle (~20 mm in diameter) in the middle of the plate for the 21 animal to roam. One animal was placed in the center of the circle and allowed to 22 habituate for one minute prior to recording for three minutes. For the next recording, 23 CuCl2 was pipetted onto the same circle before another animal was placed on the plate 1 and recorded. For all data presented in the same graph, animals were recorded on the 2 same day, with at least one animal from each genotype recorded on the same plate. We 3 altered the order of recording for animals of different genotype. 4 Tracking and data analysis: Behavior was recorded using a Zeiss Axioskop 2 Plus 5 equipped with an ASI MS-40000 motorized stage and a CCD camera (Hamamatsu Orca-6 R2). Tracking and analysis were performed using Micromanager and Image J software 7 plug-ins developed in-house (courtesy of Dr. Taizo Kawano). Image sequences were 8 sampled at 100-msec exposure (10 frames per second). For the post-imaging analysis, an 9 Image J plug-in was used to skeletonize the worm and extract its centerline. The 10 centerline was divided into 29 equal segments, and the angle between each segment and 11 its tangent line was calculated to quantify the curvature of the animal. These segments 12 were binned into four groups to capture the most consistent curvature trends across the 13 entire length of the animal. The directionality of movement (forward vs. backward) was 14 determined by first identifying the anterior-posterior axis or the "head" and "tail" points, 15 which was manually defined at the first two frames and verified throughout the recording. 16 To calculate directionality of movement, the displacement of the midline point in relation 17 to the head and tail for each worm was determined based on its position in the field-of-18 view and the stage coordinates. Image sequences wherein animals touched the edge of the 19 recording field and crossed over on themselves were not processed. 20 Quantification: Analyses of the output data were carried out using an R program-based 21 code developed in-house (courtesy of Dr. Michelle Po). From a group of recordings 22 (N≥10 for each genotype per experiment), we quantified the following parameters, 23 among other behavioral parameters analyzed by the program: 1) Curvature, the average 1 curvature of each animal at each segment; 2) Initiation (defined as the frequency of 2 directional change for each animal; 3) Duration (defined as the time spent moving in the 3 same direction for >3 frames or 300msec, calculated for each bout of forward or 4 backward initiation). Body curvature, initiations, and durations were calculated for 5 forward and backward locomotion separately. 6 The A-class motor neuron calcium imaging and analysis 7 Animals of various genotypes were crossed into a reporter hpIs460 (Punc-4-8 GCaMP6::wCherry), using a calcium reporter GCaMP6 fused in-frame at its C-terminus 9 with wCherry for both tracking and ratiometric correction for motion artifacts in moving 10 animals during recordings (46). A-MN imaging in moving young adults was carried out 11 as described previously (32). Briefly, C. elegans were placed on a 2.5% agar pad, 12 immersed in a few drops of M9 buffer, and covered by a coverslip to allow slow 13 crawling. Each recording lasted for 3 minutes. Images were captured using a 40x 14 objective on a Zeiss Axioskop 2 Plus equipped with an ASI MS-40000 motorized stage, a 15 dual-view beam splitter (Photometrics) and a CCD camera (Hamamatsu Orca-R2). The 16 fluorescence excitation light source from X-CITE (EXFO Photonic Solution Inc.) was 17 reduced to prevent saturation of imaging field. The fluorescent images were split by 18 Dual-View with a GFP/RFP filter set onto the CCD camera operated by Micromanager. 19 The 4x-binned images were obtained at 50-msec exposure time (10 frames per second). 20 Regions of interest (ROIs), containing the soma of DA7, VA10 and VA11, were defined 21 and simultaneously tracked using an in-house developed Image J plug-in (32). The ratio 22 between GFP and RFP fluorescence intensities from individual ROI was plotted over 23 time to produce index for calcium profile for each neuron. Displacement for the DA7 1 soma, which exhibited the strongest RFP signal, was plotted over time to generate the 2 example instantaneous velocity profiles. For example velocity profiles, we also manually 3 annotated videos to verify the reversal periods. Reversal events longer than 5 seconds 4 were used for cross-correlation analyses shown in the example traces. 5 The pair-wise phase relationships between the activity patterns of A-MNs during 6 each reversal event were assessed by cross-correlation analyses. In each example trace, 7 cross-correlation during the entire period of reversal was calculated to determine the 8 phase lags between the calcium signals of VA11, DA7, and VA10. The maximum of the 9 cross-correlation function denotes the time point when the two signals are best aligned; 10 the corresponding argument of the maximum correlation values denotes the lag between 11 the two neurons. The absolute time lags were used to represent the synchrony of VA11, 12 DA7 and VA10's activation. N (≥10) refers to the number of reversal events analyzed for 13 each genotype per experiment. 14

C. elegans metabolomics analyses by HPLC and LC−MS/MS 15
Synchronized last-stage larval worms were grown on 100 mm NGM agar plates seeded 16 with OP50 bacteria. Worms were collected using M9 buffer and were washed thoroughly 17 to remove bacteria. Worm pellets were snap-frozen in liquid nitrogen and pulverized 18 using a cell crusher. Amino acids and other metabolites were extracted by addition of ice-19 cold extraction solvent (40% acetonitrile, 40% methanol, and 20% water) and incubated 20 on dry ice for an hour with occasional thawing and vortexing. Samples were then moved 21 to a thermo mixer (Eppendorf) and shook for an hour at 4°C at 1400 rpm. These samples 22 were centrifuged at 14000 rpm for 10 min at 4°C. Supernatant was transferred to fresh 23 tubes and lyophilized in a CentreVap concentrator (Labconco) at 4°C. Samples were 1 stored at −80 °C until used for HPLC or LC-MS/MS analyses. 2 Amino acid quantitation was performed using the Waters Pico-Tag System 3 (Waters). After hydrolysis and pre-column derivatization of the sample by PITC, samples 4 were analyzed by reverse phase HPLC (Amino acid facility, SPARC BioCentre, Sick 5 Kids, Toronto, Canada). LC-MS/MS metabolite analysis was performed as described 6 previously (77). Raw values were normalized against the total protein concentration as 7 determined by a Bradford Protein Assay (Bio-Rad). Results were compared from at least 8 three sets of independent experiments (N≥3), with all samples collected and analyzed in 9 parallel in each replica. 10

Measurement of total GOT Activity from the C. elegans lysate 11
Synchronized late L4 larval stage worms were grown on 100 mm NGM agar plates 12 seeded with OP50 bacteria. Worms were collected using M9 buffer and were washed 13 thoroughly to remove bacterial contamination. Samples were homogenized by sonication 14 in 100-200ul ice-cold AST buffer. Samples were centrifuged at 13,000g for 10 minutes to 15 remove insoluble materials. The supernatant was used for the AST assay and pellets were 16 used to quantify the total protein in the samples (as described above). 17 GOT (also referred to as AST) activity was measured using a colorimetric assay 18 with AST Activity Assay Kit (Sigma-Aldrich) according to the manufacturer's 19 instructions. Data were normalized by the total protein content of the whole worm lysate 20 as determined by a Bradford Protein Assay. 21

C. elegans Biochemistry 22
Mixed stage C. elegans were grown on 100 mm NGM plates seeded with OP50 and 1 collected using M9 buffer. Lysates were prepared as described previously (78). For 2 western blot analyses, total protein concentration was determined using a Bradford 3 Protein Assay (Bio-Rad). Anti-GFP antibodies (Roche) were used to probe for GOT-4 1::GFP and tubulin was used for the loading control. 5

Statistical Analysis 6
For bending curvature and calcium imaging analyses, statistical significance was 7 determined using the Kruskal-Wallis test and the two-way repeated measures (RM) 8 ANOVA and subsequent post-hoc analysis. For metabolite analyses, two-tailed Student's 9 t-tests were applied to determine statistical differences. p< 0.05 were considered to be 10 statically significant. All statistics were performed using Prism software (GraphPad). 11

Acknowledgements 12
We thank Asuka Guan, Josh Kaplan, Gene Knockout Consortium and National 13   in animals of respective genotypes. The asynchrony between VA10 and VA11 (G), and 8 between VA10 and DA7 (H) are significantly reduced in ubr-1 mutants compared to 9 wildtype animals, and restored in ubr-1 mutants by both UBR-1 expression in premotor 10 interneurons, and the got-1 mutation. The activation of DA7 and VA11 (I), with higher 11 synchrony than the other two pairs in wildtype animals, was not significantly altered in 12 ubr-1 mutants. *P<0.05, ** P<0.01, ***P<0.001by the Kruskal-Wallis test. Horizontal 13 lines represent mean values. 14 alanine (blue arrows), and increased aspartate, asparagine and cysteine (red arrows) 6 compared to wildtype animals (Top panel). In order to present data of a wide range, the 7 Y-axis utilizes three different scales at different concentration values. Consistent with 8 glutamate being the sole precursor of GABA synthesis, GABA level was also increased 9 in ubr-1 mutants (blue arrow). *P<0.05, **P<0.01 by the Student T test, N: 5 replica. 10 Data are represented as mean ± SEM. 11

Figure 6. Glutamate level is elevated in ubr-1 mutants 12
A-C) Free amino acid measured by HPLC, normalized against total protein in the lysate. 13 Levels in mutants were normalized to that of wildtype. A) Glutamate level was increased 14 in ubr-1, but decreased in got-1 mutants; the increase in glutamate was reversed in ubr-1; 15 got-1, but not in ubr-1 got-2 mutants. B) Aspartate exhibited a modest increase in ubr-1, 16 but massive accumulation in got-1 and ubr-1; got-1 mutants. The Y-axis utilizes two 17 different scales (0.5 to 4.0, 4.0 to 40, respectively) to accommodate vastly different 18 values. C) Alanine was significantly decreased in both got-1 and ubr-1; got-1 mutants. D-19   Table 1

S3 Fig. ubr-1 function is required in premotor interneurons to regulate reversals. 1
A-B) In ubr-1 mutants, the reversal duration is increased, while reversal initiation 2 frequency is decreased. Both the parameters were rescued by restoring the expression of 3 UBR-1 in premotor interneurons, but not in GABAergic or cholinergic motor neurons. C-4 D) Expression of UBR-1 in multiple premotor interneurons, which include AVE/RIM 5 exhibited significant rescue of ubr-1 for reversal duration and initiation frequency. 6 Expression of UBR-1 in the RIM alone or in AVA alone did not result in rescue.  Metabolite levels measured by LC-MS, normalized against the total protein level in the 1 lysate. All mutants were normalized to that of wildtype animals. Metabolites of the TCA 2 cycle did not show any coordinated changes among these mutants. ubr-1 animals, which were labeled with cytosolic RFP, from the L1 juvenile larva to 7 adult stages. Depending on the focal planes, some images contain that of nuclei, which 8 were devoid of RFP signals. Left panels: images from wildtype animals. The surface of 9 both somata was round and smooth throughout larval development and in young adults. 10 Right panels: images from ubr-1 mutants. In young (L1 to L3) larva, somata appeared 11 similar to those in age-matched wildtype animals. In the L4 larva and adult animals, 12 somata developed rough surface and short branches, denoted by arrows. B) The RIM 13 soma exhibited normal morphology in adults, similar to wildtype animals. C) The round 14 morphology AVA and AVE somata in ubr-1 adults were restored in ubr-1; got-1 double 15 mutant adults, and in ubr-1 adults with restored UBR-1 expression in multiple premotor 16 interneurons including AVE and RIM. Asterisks (*) mark the AVA or AVE axons, which 17 could not be shown in some panels when they were in a different focal plane as the 18 somata. Scale bar, 2µm (for the L1 and L2 panels) 5 µm (for the L3 to adult panels). 19 Supplementary movie captions 20 S1 Movie: Reversal behaviors exhibited by the wildtype (N2) (Part 1) and ubr-21