To investigate the functions of UBA5 variants in vivo, we utilized Drosophila melanogaster as a model organism. Uba5 is the ortholog of human UBA5 in flies (UBA5 refers to the human gene; Uba5 refers to the fly gene). The two proteins share 64% identity and 75% similarity in amino acid sequence, and the Drosophila Integrative Ortholog Prediction Tool (DIOPT) score between UBA5 and Uba5 is 15/16, indicating a high degree of homology (Figure 1C) (Hu et al, 2021). The UBA5 protein has an adenylation domain, a UFM1-interacting sequence (UIS) and a UFC1-binding sequence (UBS), all of which are required for UFM1 activation and transthiolation (Bacik et al, 2010; Habisov et al, 2016; Kumar et al., 2021; Padala et al, 2017; Xie, 2014) (Figure 1C). Similarly, Uba5 has all three highly conserved functional domains, and all of the amino acid residues affected by the DEE44-associated variants reported so far are conserved in the fly protein (Figure 1C).

To study the variant-specific functions, we generated humanized fruit fly models in which the expression of the endogenous Uba5 gene is removed or severely suppressed and a human UBA5 cDNA is expressed under the control of the endogenous Uba5 enhancer and promoter. If the human reference UBA5 functions in flies and rescues the Uba5 severe LoF phenotypes, the DEE44- associated variants can be expressed and their functions can be assessed by the phenotypes of flies. To achieve this, we generated a Uba5^T2A-Gal4 allele using a CRISPR-Mediated Integration Cassette (CRIMIC) strategy (Lee et al, 2018). In the Uba5^T2A-Gal4 allele, an FRT-Splice Acceptor (SA)-T2A-GAL4-polyA-FRT cassette was inserted into a coding intron of the Uba5 gene. The SA causes the inclusion of the cassette during transcription, while the polyA sequence arrests the transcription generating a truncated transcript. The translation of the transcript is arrested at the viral ribosomal skipping site (T2A) and reinitiated after the site, producing an untagged GAL4 protein (Figure 1D) (Diao et al, 2015; Lee et al., 2018). The Uba5^T2A-Gal4 allele is likely a severe LoF allele (see below) (Lee et al., 2018). In addition, this allele also results in the expression of GAL4 under the control of the endogenous Uba5 enhancer and promoter, which enables the assessment of native gene expression pattern as well as the expression of human UBA5 cDNA (Figure 1D). We also generated a Uba5 null allele by CRISPR-induced indel formation (Uba5^{p.Arg55Profs*87}, named Uba5^KO) (Figure 1E).

We first tested the viability of the flies with the Uba5^T2A-Gal4 and the Uba5^KO alleles. The fly Uba5 gene is located on the X chromosome. For both alleles, homozygous female and hemizygous male flies are lethal at the embryonic stage, although a few Uba5^T2A-Gal4 escapers survive to the L1 larval stage (Figure 1F). The lethality is rescued by a genomic rescue (GR) construct that carries the Uba5 locus (P[acman] clone CH321-02B13) (Venken et al, 2010), indicating that the lethality in both lines is caused by the LoF of Uba5. Moreover, expression of flippase (FLP) using Uba5^T2A-Gal4 removes the insertion of the CRIMIC cassette and reverts the lethality of the Uba5^T2A-Gal4 hemizygous males, showing that the lethality is indeed caused by the Uba5^T2A-Gal4 allele (Figure 1F) (Lee et al., 2018). Finally, expression of reference human UBA5 cDNA using Uba5^T2A-Gal4 rescues the lethality of the mutants, showing that the functions of the fly and human proteins are evolutionarily conserved (Figure 1F).

Next, we examined the expression pattern of Uba5 by expressing a nuclear localized mCherry fluorescent protein (UAS-mCherry.nls) under the control of Uba5^T2A-Gal4. Uba5 is expressed in multiple tissues in L3 larvae and adult flies (Figure 2A), consistent with high-throughput gene expression profiling results(Leader et al, 2018). We next analyzed the expression of Uba5 in the central nervous system (CNS). We stained the Uba5^T2A-Gal4>mCherry.nls larval CNS and adult brain with anti-Elav and anti-Repo antibodies to mark the nuclei of neurons and glial cells, respectively. In both larval CNS and adult brain, the mCherry.nls signals are found in a subset of neurons and glia (Figures 2B and 2C), suggesting that Uba5 is expressed in the fly CNS but not in all cells. Intriguingly, Uba5 is expressed more widely in the adult brains than in the larval CNS. In larval CNS, it is expressed in many fewer neurons than Elav (Figures 2A and 2B). The expression pattern resembles that of the para gene, which encodes the sole voltage-gated sodium channel in Drosophila which is only expressed in differentiated, actively firing neurons (Ravenscroft et al, 2020). This suggests that UBA5 may be required for the activity of neurons.

Next, we sought to evaluate the function of UBA5 variants using the humanized fly model(Bellen & Yamamoto, 2015). We expressed reference or variant UBA5 cDNA and DEE44-associated UBA5 variants using Uba5^T2A-Gal4 and measured phenotypes including survival rate, developmental timing, lifespan, locomotor activity, and seizure-like activity following mechanical stimulation in Uba5^T2A-Gal4 hemizygous male flies. The variants we tested include all previously reported variants (Arnadottir et al., 2017; Briere et al., 2021; Colin et al., 2016; Daida et al., 2018; Low et al., 2019; Mignon-Ravix et al., 2018; Muona et al., 2016), as well as a novel variant from an individual we report in this study. This individual has compound heterozygous variants in UBA5, p.Met57Val and p.Gln312Leu (Table S3), and presents with hypotonia, generalized dystonia, lower extremity spasticity, global developmental delay, and failure to thrive. However, this individual is so far seizure free. Further clinical details of the individual are summarized in Supplemental Information.

We first assessed the ability of the variants to rescue the lethality of Uba5^T2A-Gal4 mutants. A synthetic enzyme-dead UBA5 mutant p.Cys250Ala failed to rescue lethality (Figure 3A). Similarly, four DEE44-associated variants (p.Arg55His, p.Gly168Glu, p.Leu254Pro, p.Cys303Arg) failed to rescue the lethality, indicating that they are severe LoF. Two variants (p.Tyr53Phe, p.Met57Val) partially rescued and are therefore likely hypomorphic alleles (Figure 3A). However, six variants (p.Arg72Cys, p.Val260Met, p.Gln312Leu, p.Ala371Thr, p.Asp389Gly, p.Asp389Tyr) fully rescued lethality, suggesting they are mild LoF or do not affect protein function. We then raised the flies at 18 °C as this causes a decrease in GAL4 activity and hence lowers the expression level of UBA5. At this temperature, three variants (p.Tyr53Phe, p.Met57Val, p.Val260Met) partially rescued lethality. Based on these results, we stratified the variants into three groups according to the activity of rescuing lethality: Group I, full rescue; Group II, partial rescue; Group III, failure to rescue (Table 1, Group I was further divided into IA and IB according to the other phenotypes described below).

The variants that survive to adults were next tested for the time it takes for animals to eclose as adults, and lifespan. The variants in Group II exhibited significant developmental delay as well as a shortened lifespan, showing that they are partial LoF (Figures 3B and 3C). In contrast, some variants in Group I (Group 1A: p.Ala371Thr, p.Asp389Gly, and p.Asp389Tyr) caused neither defect. Other Group I variants (Group 1B: p.Arg72Cys and p.Gln312Leu) caused a shortened lifespan but did not affect the timing of development (Figures 3B and 3C), indicating that they are also partial LoF variants but could cause milder defects than Group II variants.

To determine if the flies display features that are associated with dysfunction of the nervous system, we measured locomotor activity using a climbing assay and assessed susceptibility to seizures using a bang sensitivity assay(Song & Tanouye, 2008). Flies with variants in Groups IB and II displayed reduced climbing activity at Day 7 and more severe defects by Day 30, showing a progressive worsening of the defects (Figure 3D). Moreover, the variants in Group II exhibited a bang-sensitive phenotype by displaying seizure-like behavior and paralysis following mechanical stimulation (Figure 3E). However, the Group IA variants displayed neither a climbing defect nor bang sensitivity (Figures 3D and 3E). These results are consistent with our classification of variants based on other assays: Group IA, no obvious LoF or benign; Groups IB and II, intermediate LoF; and Group III, severe LoF.

In order to link our findings in the fly model with functional changes in UBA5, we first analyzed the potential structural changes caused by the UBA5 variants. Extensive structural analyses have been performed on UBA5, including its ability to bind ATP and homodimerize within the adenylation domain, its interaction with UFM1 in the process of activation, and its engagement of UFC1 prior to UFM1 transthiolation (Bacik et al., 2010; Kumar et al., 2021; Oweis et al., 2016; Padala et al., 2017; Soudah et al, 2019; Wesch et al, 2021). This structural detail of UBA5 function offers a unique opportunity to visualize the location of variants and categorize their predicted effects. We compiled a series of UBA5 structures to create a composite model that illustrates a) UBA5 homodimerization, b) ATP coordination, c) UFM1 binding, and d) UFC1 binding (Figure 4A). Focusing on the UFM1 C- terminal Gly and the active site Cys residues of UBA5 and UFC1, the movements that occur during UFM1 activation and transthiolation can be mapped (Figure 4A, yellow spheres).

With this structural model as a basis, we highlighted the location of all UBA5 variants (Figure 4A, red spheres). Variants at positions Asp389 and Ala371 lie outside of the regions with determined structure, so their locations are modeled based on all available data. All Group IA variants localize in proximity of the UBS in protein sequence. The p.Asp389Gly and p.Asp389Tyr variants affect Asp389 which is two amino acids upstream of the structurally-resolved UBS. For p.Ala371Thr, although there is no structure of this region, previous biochemical and crosslinking data have demonstrated its proximity to the UFC1 active site (Kumar et al., 2021). Four variants, p.Tyr53Phe, p.Arg55His, p.Met57Val, and p.Arg72Cys, affect residues near the ATP-binding site (Figure 4B). The side chain of Arg55 makes direct contacts to the bound ATP, while Met57 and Tyr53 make secondary contacts behind this site, and Arg72 makes more distant tertiary contacts. Substitutions at these positions, therefore, may affect the affinity of UBA5 toward ATP. Residue Leu254 is four amino acids downstream of the active site Cys250, contained within a loop region that must undergo conformational changes to support UFM1 activation and subsequent transthiolation (Figures 4A and 4B). The Pro substitution within the p.Leu254Pro variant may constrain the required flexibility of the Cys250 loop, thus impacting ATP binding, UFM1 activation, and UFM1 transthiolation. Val260 is buried within the UBA5 homodimeric interface, opposing residue Val260 of the second protomer (Figure 4B). The increased size of the p.Val260Met variant may cause a steric clash that reduces UBA5 dimerization, which would decrease ATP binding, UFM1 activation, and UFM1 transthiolation as a result. Positions Cys303 and Gly168 are fully buried, making structural contacts to support the fold of the adenylation domain (Figure 4C). The increased size and charge associated with the p.Cys303Arg and p.Gly168Glu variants would most likely cause significant defects in UBA5 folding and stability. Lastly, the site of the novel p.Gln312Leu variant reported here is partially buried, making structural contacts to a loop region underneath the UFM1-binding site, and thus substitutions at this site may also cause structural instability (Figure 4C).

To determine the functional capacity of the UBA5 variants, we expressed and purified reference and variant UBA5 proteins from Escherichia coli for biochemical assays. Two severe LoF (Group III) variants p.Gly168Glu and p.Cys303Arg were insoluble in the protein purification process, indicating that the variants influence the stability and/or folding of these proteins. We were able to obtain pure and homogeneous samples of all other UBA5 constructs (Figure 5A). For our first measure of UBA5 function, we measured the effects of each variant on protein stability using a thermal shift assay. Under our assay conditions, reference UBA5 demonstrated a melting temperature (Tm) of 46 °C. While most UBA5 variants exhibited little-to-no change in Tm, several showed a subtle destabilization by ≥1 °C, including p.Arg72Cys, p.Lys254Pro, p.Val260Met, and p.Ala371Thr (Figure 5B). Interestingly, the Group IB variant p.Gln312Leu exhibited a minor unfolding at 36 °C before fully melting at 48 °C. This could indicate destabilization in the local structure surrounding the Gln312Leu substitution, while leaving the remaining protein structure intact.

Ligand binding can stabilize protein structure and lead to a shift in Tm toward higher temperatures. This has been shown previously in the case of UBA5 binding to ATP, where it was determined to interact with a KD of ∼700 µM (Mashahreh et al., 2018). The ATP-dependent stabilization of UBA5 allowed us to assess mutational effects on ATP binding using the thermal shift assay. Whereas addition of 5 mM ATP led to a strong, 13 °C shift in the Tm of reference UBA5, many of the variants exhibited much weaker stabilization, indicative of a diminished capacity to bind ATP (Figure 5C). The other two Group III variants and one Group II variant p.Tyr53Phe showed the strongest defect, with only 4-5 °C shifts in Tm upon addition of ATP. The Group II variants p.Met57Val and p.Val260Met as well as one Group IB variant p.Gln312Leu showed a milder effect with a 6-8 °C shift in Tm. One Group IB variant p.Arg72Cys and all Group IA variants showed similar ATP-dependent stabilization to reference UBA5. Unlike the melting trend observed in the absence of ATP, in the presence of ATP the p.Gln312Leu variant displayed a single unfolding profile, suggesting that ATP binding corrected the local instability caused by the substitution.

These results show that not all tested variants strongly affect the stability of UBA5 protein. However, many variants impair the ATP binding capability of the protein. The levels of impairment correlate well with the phenotypic observations in vivo (Table 1 and 2), suggesting that the defect in ATP binding is a major contributor to the LoF associated with the variants. Consistent with their severe LoF in vivo, variants p.Gly168Glu and p.Cys303Arg exhibited impaired protein folding and could not be included in our in vitro analyses.

Previous functional characterization of UBA5 activity has relied upon gel-based assays that often lack kinetic information and are less sensitive to subtle changes. We sought to address this problem by developing a real-time, fluorescence polarization (FP) assay for UBA5 activity based upon a method we coined “UbiReal” (Franklin & Pruneda, 2019, 2023). The approach leverages the large changes in molecular weight that occur as UFM1 is activated by UBA5 and transferred to UFC1 (Figure 6A), which are read out as changes in FP of fluorescently-labeled UFM1. Indeed, upon addition of UBA5 we observed a large shift in FP of Alexa488-labeled UFM1, which reached a plateau over the course of ∼20 minutes under these conditions (Figure 6B). Addition of UFC1 to the reaction caused a concomitant downward shift in FP that reached a new plateau within ∼5 minutes. To validate the molecular species observed in this assay, we ran samples from each stage of the reaction on non-reducing SDS-PAGE and were able to visualize free UFM1, UBA5∼UFM1, and UFC1∼UFM1 as predicted (Figure 6C).

With the UFM1 UbiReal assay in hand, we proceeded to assess the effects of UBA5 variant on the first step of the reaction: UFM1 activation. As anticipated, addition of reference UBA5 caused a rapid shift in FP over time, whereas the enzyme-dead p.Cys250Ala variant remained at baseline (Figure 6D). We quantified these kinetic data with Area Under the Curve (AUC) analysis using data collected just before addition of UBA5 to establish a baseline. The panel of UBA5 variants displayed a wide range of effects. The two Group III variants p.Arg55His and p.Leu254Pro remained near the baseline, indicating a severe impairment in UFM1-activating function. An intermediate, statistically significant effect was observed with the p.Val260Met variant (Group II), while the remaining variants showed mild or, in the case of p.Ala371Thr (Group IA), no decrease in the rate of UFM1 activation (Figure 6E). Similar trends were observed at both 22 °C and 37 °C, though interestingly the effect of p.Val260Met substitution was less severe at higher temperature (Figure 6F). The results show that a defect in UFM1 activation contributes to the LoF in all variants except for p.Ala371Thr. The severity of LoF in UFM1 activation correlates with our phenotypic observations in vivo (Table 1 and 2).

Having observed defects in UFM1 activation for many of the UBA5 variants, we next analyzed those that exhibited mild or no effect for their ability to complete the second enzymatic role of UBA5: UFM1 transthiolation onto UFC1. After forming the activated UBA5∼UFM1 intermediates, we added UFC1 to the reactions and monitored the decay in FP over time (Figure 6G). Though many of the UBA5 variants showed similar trends for transthiolation, the p.Ala371Thr variant immediately stood out for having no effect on UFM1 activation (Figures 6D-6F), but a greatly impaired ability to transfer UFM1 onto UFC1 (Figure 6G). We quantified the kinetic data using AUC analysis of the inverted curve and observed a remarkable defect in the ability of the p.Ala371Thr variant to catalyze UFM1 transthiolation (Figure 6H). Interestingly, this defect is much more pronounced at 22 °C than at 37°C (Figure 6I). These results suggest that aside from a minor decrease in UBA5 stability (Figure 5B), UFM1 transthiolation is the only possible defect caused by the p.Ala371Thr variant. Transthiolation could be defective in other variants as well, such as p.Leu254Pro (Group III), but this effect is likely overshadowed by upstream effects on UBA5 stability, ATP binding, and/or UFM1 activation.

The proband was recruited through the Doernbecher Children’s Hospital and informed consent was obtained from legal guardians of the proband. The human study was approved by the Institutional Review Board at Oregon Health & Science University. The legal guardians of the proband consent to have the results of this research work published.

Trio exome sequencing was conducted by GeneDx using DNA extracted from blood. DNA was enriched using a proprietary capture system developed by GeneDx for next-generation sequencing. The enriched targets were simultaneously sequenced with paired-end reads on an Illumina platform. Bi-directional sequence reads were assembled and aligned to Human Genome Sequencing Center (HGSC) build 37, human reference genome 19. Reported variants were confirmed, if necessary, by an appropriate orthogonal method.

All fruit fly strains used in this study were cultured using standard fly food in a 25°C incubator unless a different culturing temperature was specifically indicated. The Uba5^KO mutant and the human UAS- UBA5 transgenic fly lines were generated in the Bellen lab (for methods, see below). The Uba5^{T2A- Gal4} (#78928), UAS-mCherry.nls (#38424), UAS-FLP (#4540), Uba5^GR (#30359), da-Gal4 (#5460), Act-Gal4 (#4414), elav-Gal4 (#8765), and repo-Gal4 (#7415) lines were obtained from the Bloomington Drosophila Stock Center (BDSC).

The Uba5^KO allele was generated using CRISPR/Cas9 genome engineering technology as previously described (Port et al, 2014). A TKO transgenic line expressing a Uba5-targeting single guide RNA (sgRNA) is available from the BDSC (BDSC #81448) (Zirin et al, 2020). The sgRNA (CAATCCGTACAGCCGCCTGA) targets the coding region in the Exon 1 of the only Uba5 transcript.

To generate indel variants, the TKO flies were crossed to nos-Cas9 transgenic flies (BDSC #78781) and then the first generation (F1) female progenies carrying both sgRNA and Cas9 were crossed with 1st chromosome balancer flies. Single F2 females were crossed with balancer flies again to establish ∼20 individual stocks with potential indel variants. Due to the lethality of Uba5 mutants, the stocks without unbalanced flies were screened for Uba5 indel variant by genomic PCR and Sanger sequencing. One mutant line with NM_132494.3 (Uba5):c.164_174del (p.Arg55ProfsTer87) varaint was isolated and designated as Uba5^KO allele in this study.

Human UAS-UBA5 transgenic fly lines were generated as previously described (Harnish et al, 2019). In brief, the human UBA5 cDNA sequences were cloned into the pGW-UAS-HA.attB vector (Bischof et al, 2013) using the Gateway Cloning system (Thermo Fisher) and validated by sequencing. The cDNA vectors were then injected into fly embryos and inserted into the VK37 (BDSC #24872) docking site by φC31-mediated transgenesis (Venken et al, 2006). The human UBA5 cDNA clone corresponding to Genebank transcript NM_024818.6 was obtained from the Ultimate ORF Clones library (Thermo Fisher). The UBA5 variants were introduced into the reference cDNA using Q5 site- directed mutagenesis (NEB) before the cDNA was cloned into the pGW-UAS-HA.attB vector.

Larval and adult flies were dissected in 1X PBS and the larval CNS and adult brain tissues were processed for immunostaining. Briefly, the tissues were fixed in 4% paraformaldehyde followed by normal horse serum blocking and incubation in the primary antibody (Rat anti-Elav, Developmental Studies Hybridoma Bank (DSHB) #7E8A10, 1:500; Mouse anti-Repo, DSHB #8D12 1:50). Cy5- conjugated secondary antibodies were used to detect the primary antibodies. Samples were mounted on slides using RapidClear (SUNJin Lab) and the images were captured using a confocal microscope (Zeiss 710).

For the measurement of lifespan, freshly eclosed flies were collected in separate vials and maintained at 25°C. Flies were transferred every day to fresh food in the first six days and every other day afterward. Survival was determined during every transfer. The results are represented as Kaplan-Meier curves.

To measure negative geotaxis, flies were transferred to a clean vial for at least 20 minutes prior to the experiment. During the test, flies were tapped to the bottom of the vial and their negative geotaxis climbing ability was measured. In each measurement, flies were allowed to climb for 30 seconds, after which the climbing distances were measured (18 cm is maximum). To perform bang-sensitive paralytic analyses, adult flies were transferred to a clean vial and vortexed at maximum speed for 10 seconds, after which the time required for flies to stand on their feet was counted (30 seconds is maximum).

The reference UBA5 gene and UFM1 cloned into pET15b and UFC1 cloned into pET32a were kind gifts from R. Wiener (The Institute for Medical Research Israel-Canada). The UBA5 p.Met57Val and p.Gln312Leu substitutions were cloned into this background by Quikchange PCR using Phusion DNA polymerase. All other UBA5 variant substitutions were subcloned into pOPIN-B using the constructs described above as templates. All of these constructs encoded N-terminal His-tags, and were purified in a similar manner. After plasmid transformation into E. coli Rosetta (DE3) cells, cultures were grown at 37 °C in Luria Broth containing 35 µg/mL chloramphenicol and 50 µg/mL of kanamycin. Once an optical density (600 nm) between 0.4-0.6 was reached, cultures were cooled to 18°C and protein expression was induced with 0.5 mM IPTG. Cells were harvested by centrifugation after 24 hours of expression and resuspended in 50 mM NaPO4, 500 mM NaCl, 2 mM ß-mecaptoethanol, pH 8.0 (Buffer A). The cell pellet was then subjected to a freeze-thaw cycle before adding DNase, PMSF, lysozyme and SigmaFAST Protease Inhibitor Cocktail (MiliporeSigma) and allowed to incubate on ice for 30 min. The cell pellets were then lysed by either french press or sonication, depending on the volume to be lysed. Lysates were then centrifuged at 35,000 xg, and the clarified lysate was added to a column containing HisPur Cobalt affinity resin (ThermoFisher), allowed to bind for 10 min, and washed with 1 L of Buffer A + 10 mM Imidazole. Proteins were then eluted with 5 mL of Buffer A + 350 mM Imidazole in a stepwise manner for a total elution volume of 25 mL. Purity of the fractions were analyzed via SDS-PAGE and those with highest purity were pooled and dialyzed overnight into 25 mM Tris, 100 mM NaCl, 2 mM DTT, pH 8.0 (Buffer B) at 4°C. After 24 hrs of dialysis, UBA5 proteins were further purified on a RESOURCE Q (Cytivia) anion exchange chromatography column equilibrated in Buffer B. The protein was eluted over a 20 column volume gradient against Buffer B + 1 M NaCl. Peak protein fractions were pooled and concentrated using Amicon centrifugal filters (MiliporeSigma) before being applied to a HiLoad Superdex 75 16/600 pg size exclusion chromatography column (Cytivia) equilibrated in Buffer B. Following affinity purification, the His-tags of UFM1 and UFC1 were removed by TEV cleavage during overnight dialysis into Buffer B at 4 °C. Proteins were then concentrated using Amicon centrifugal filters and applied to a HiLoad Superdex 75 16/600 pg as above. Peak fractions were evaluated for purity via SDS-PAGE, pooled, and concentrated before being quantified by absorbance at 280 nm and flash frozen above 10x their working stock concentration. All protein samples were stored at -80 °C.

UFM1-Alexa 488 substrates were prepared using an Alexa Fluor 488 TFP ester (ThermoFisher). Labeling was performed at room temperature for 1 hr in 0.1 M sodium bicarbonate buffer at pH 7.5, which directs labeling toward the N-terminus. Following labeling, excess fluorophore was quenched with addition of 150 mM Tris pH 7.4 and separated by size exclusion chromatography on a HiLoad Superdex 75 16/600 pg as above. Transfer and activation of UFM1 onto UBA5 and further transthiolation to UFC1 was monitored by fluorescence polarization (FP) based on published methods from the ubiquitin system (Franklin & Pruneda, 2019, 2023). FP was measured using a BMG LabTech CLARIOstar plate reader at an excitation wavelength of 482 nm, an LP 404 nm dichroic mirror, and an emission wavelength of 530 nm. Free UFM1 was used as a reference with a target FP of 190. All assays were performed in black Greiner 384-well small-volume HiBase, low protein-binding microplates. UFM1 and UBA5 stocks were prepared at 2X assay conditions in 25 mM NaPO4, 150 mM NaCl, 10mM MgCl2, pH 7.4 (Buffer C). UFM1 was prepared at 100 nM in Buffer C + 20 mM ATP (pH 7); UBA5 variants were prepared at 1 µM in Buffer C. UFC1 was prepared at 3 µM (10X assay concentration) in Buffer C. Final assay conditions in 20 µL volumes were 50 nM UFM1, 500 nM UBA5 variants, and 300 nM UFC1 in Buffer C + 10 mM ATP. FP data was first collected for the substrate only to establish a baseline before addition of UBA5 variants at 1:1 ratio to reach the described assay conditions. FP values were collected in 45-second intervals until the reference UBA5 readings plateaued, at which point UFC1 was added at 1:10 ratio to achieve final assay conditions. Reactions were performed in triplicate for each of three experimental replicates, both at 22 °C and 37 °C. AUC calculations were performed using Prism 9.5, with baseline values calculated from data collected prior to addition of UBA5/UFC1.

Conditions for the gel-based assay were identical to those described above for the 22°C FP assays. Parallel reactions were prepared, one to be read out by FP using the CLARIOstar plate reader and the other was left in a low light environment at room temperature. After 15 minutes of establishing a baseline FP reading for UFM1 alone, a 20 µL sample was taken from the parallel reaction and quenched using non-reducing sample buffer. FP readings were paused and reference UBA5 was added to both reactions before continuing FP data collection. The FP values were allowed to plateau before another 20 µL sample was collected from the parallel reaction and quenched with non- reducing sample buffer. FP readings were then paused and UFC1 was added to both reactions before continuing FP data collection. The FP values were allowed to plateau before taking a final 20 µL sample from the parallel reaction and quenching with non-reducing sample buffer. Gel samples were run on a TGX 4-20% SDS-PAGE gradient gel (Bio-Rad). The resulting gel was imaged using a Sapphire Biomolecular Imager (Azure Biosystems). This experiment was performed in triplicate.

The thermal shift assay was conducted in MicroAmp Fast 96-Well Reaction Plates (Applied Biosystems) with SYPRO Orange Protein Gel Stain (MiliporeSigma) using a QuantStudio 3 Real- Time PCR system (Applied Biosystems). Assays were performed in 20 µL volumes containing 5 µM UBA5 variants and 20X SYPRO dye (diluted from a 5000X stock) in 25 mM NaPO4, 150 mM NaCl, 10 mM MgCl2, pH 7.4 with or without 5 mM ATP. The protocol ramped temperature from 22 °C to 99°C over a gradient of 0.1 °C every 5 seconds, and fluorescence was monitored using an excitation wavelength of 580 ± 10 nm and an emission wavelength of 623 ± 14 nm.

Statistical analyses were carried out using the Student’s unpaired two-tailed t test for comparison of two groups or the Welch’s t test for data normalized to reference UBA5. Multiple comparisons within the group were tested against the corresponding control. Kaplan-Meier survival curves were analyzed using Gehan-Breslow-Wilcoxon test and log-rank test. Calculated p values of less than 0.05 were considered significant. All statistical analyses were performed using GraphPad Prism, version 9.5.0 (GraphPad Software).