A missense mutation in a patient with developmental delay affects the activity and structure of the hexosamine biosynthetic pathway enzyme AGX1

O‐GlcNAcylation is a post‐translational modification catalysed by O‐GlcNAc transferase (OGT). Missense mutations in OGT have been associated with developmental disorders, OGT‐linked congenital disorder of glycosylation (OGT‐CDG), which are characterized by intellectual disability. OGT relies on the hexosamine biosynthetic pathway (HBP) for provision of its UDP‐GlcNAc donor. We considered whether mutations in UDP‐N‐acetylhexosamine pyrophosphorylase (UAP1), which catalyses the final step in the HBP, would phenocopy OGT‐CDG mutations. A de novo mutation in UAP1 (NM_001324114:c.G685A:p.A229T) was reported in a patient with intellectual disability. We show that this mutation is pathogenic and decreases the stability and activity of the UAP1 isoform AGX1 in vitro. X‐ray crystallography reveals a structural shift proximal to the mutation, leading to a conformational change of the N‐terminal domain. These data suggest that the UAP1A229T missense mutation could be a contributory factor to the patient phenotype.

UDP-GlcNAc, the substrate of OGT, is the end product of the hexosamine biosynthetic pathway (HBP) (Fig. 1) that is present in all three kingdoms of life [24][25][26]. The first and rate-limiting step of the HBP is catalysed by glucosamine-fructose-6-phosphate aminotransferase 1 (GFAT) which transfers an amino group from glutamine to fructose-6-phosphate to produce glucosamine-6-phosphate. Glucosamine-6-phosphate then turns into GlcNAc-6-phosphate by the action of glucosamine-6-phosphate N-acetyltransferase (GNA1). The third step is catalysed by phosphoglucomutase 3 (PGM3) that converts GlcNAc-6-phosphate into GlcNAc-1-phosphate. Finally, the production of UDP-GlcNAc is reversibly catalysed by UDP-Nacetylhexosamine pyrophosphorylase (UAP1) which uses GlcNAc-1-phosphate and UTP as substrates in the forward reaction to produce UDP-GlcNAc and pyrophosphate.
There are several lines of evidence that HBP is essential for development. GFAT gene knockout cells cannot survive without supplementation of the enzyme product, GlcNAc [27]. GFAT gene mutations have been linked to congenital myasthenic syndromes [28], a group of conditions owing to the defect of signal transmission from nerve cells to muscles [29]. GFAT patients displayed reduced levels of GFAT protein and O-GlcNAcylation [28]. Inactivation of GNA1 in mouse embryonic fibroblasts led to defects in UDP-GlcNAc biosynthesis, O-GlcNAc modification and cell proliferation [30]. In addition, mutations of the PGM3 gene that resulted in compromised enzymatic activity were identified as a cause of immunodeficiency, skeletal dysplasia and neurocognitive impairment [31][32][33]. A study in Drosophila showed that mutation of nesthocker, the orthologue of the PGM3 gene, resulted in decreased levels of cytoplasmic UDP-GlcNAc, disrupted protein O-GlcNAcylation and finally blocked mesodermal development [34]. In Drosophila, the UAP1 gene orthologue mummy modulates decapentaplegic (Dpp) signalling in the embryonic epidermis and is required for epithelial morphogenesis and nervous system development [35][36][37]. These data collectively suggest a link between the perturbation of the HBP, protein O-GlcNAcylation and neurodevelopment. The human UAP1 gene encodes two different isoforms, named AGX1 and AGX2, with AGX1 being more abundant in testis and AGX2 in somatic tissues [38]. The two isoforms differ by a 17-amino acid insertion in the C terminus of AGX2. Compared to AGX2, AGX1 has similar efficiency in producing UDP-GlcNAc and 50 times more efficiency in producing UDP-GalNAc [39]. Structural characterization revealed that both AGX1 and AGX2 comprise three domains, that is N-terminal, catalytic and C-terminal domain, and the 17-amino acid insertion in the C-terminal domain of AGX2 ablated oligomeric assembly [40]. UAP1 genes, designated glmU in bacteria, have been reported to be essential in both prokaryotes and eukaryotes [41,42]. For instance, inactivation of GlmU thermosensitive mutant in Escherichia coli resulted in a 37% reduction in peptidoglycan content and cell lysis [43]. The fungal UAP1 gene orthologue UAP is essential in both Saccharomyces cerevisiae and Aspergillus fumigatus [44,45]. Because the action of OGT depends on the steady supply of cytoplasmic UDP-GlcNAc through the HBP, we considered whether UAP1 gene mutations would phenocopy OGT-CDG mutations.
As a clinical genetics programme, the Deciphering Developmental Disorders Study (DDD) was recently established to investigate undiagnosed developmental disorders using exome sequencing and array-based detection of chromosomal rearrangements [46]. A de novo heterozygous mutation in the UAP1 gene (NM_ 001324114:c.G685A:p.A229T, DDD entry: https://deci pher.sanger.ac.uk/ddd/research-variant/c9774b30226f59f 8c1f79a7578fe5fc3/overview), which encodes a missense variant UAP1 A229T , has been identified in a patient from the DDD study with reported cranial and skeletal abnormalities and intellectual disability. Here, we investigate possible effects of the A229T mutation on AGX1 using various in vitro approaches. First, clinical bioinformatics tools predict that the mutation is likely to be pathogenic, while biochemical analysis with recombinant AGX1 A229T reveals that the A229T mutation causes a reduction of protein thermal stability. Compared to wild-type AGX1, AGX1 A229T has lower activity in producing UDP-GlcNAc. X-ray crystallographic structural characterization demonstrates that the A229T mutation lies proximal to the active site. The mutation induces local structural shift which weakens the hydrogen bond network connecting the N-terminal and catalytic domain, leading to conformational changes of the N-terminal domain that explains changes in catalytic activity. Together, these in vitro data suggest that the UAP1 A229T missense mutation could contribute to the patient phenotype.

Construct cloning, protein expression and purification
The construct with full-length AGX1 (UniProt Q16222) cloned into pGEX6P4 (expressing GST-tagged protein) was obtained from our lab stock [47]. The A229T mutation was introduced using site-directed mutagenesis with forward primer ggtggtctttatcggAcacttgcagcccag and reverse primer ctgggctgcaagtgTccgataaagaccacc. All constructs were verified by sequencing.
E. coli (DE3) pLysS was used for recombinant expression of N-terminally GST-tagged AGX1 wt and AGX1 A229T . Protein expression was induced at 18°C by addition of 200 µM IPTG when absorbance at OD 600 reached 0.6 and incubated for a further 16 h. Cells were harvested by centrifugation at 4200 rÁmin À1 at 4°C for 30 min and then resuspended in lysis buffer (100 mM Tris, 150 mM NaCl, 0.5 mM TCEP pH 7.5, 0.1 mgÁmL À1 lysozyme, 0.1 µgÁmL À1 DNase, 1 mM benzamidine, 0.2 mM PMSF and 5 µM leupeptin). Cells were lysed using French Press with a pressure of 50 000 psi. Cell lysate was then spun down at 20 000 rÁmin À1 at 4°C for 30 min in an Avanti J-25 centrifuge (Beckman). The supernatant was incubated with Glutathione-Sepharose beads (GE Healthcare) at 4°C for 2 h, and beads were then washed with 4-5 column volumes of lysis buffer. Bound protein was cleaved overnight from GST beads by PreScission protease at a final concentration of 0.5 µM. Protein was pooled then concentrated to 5 mL before loading onto a pre-equilibrated Superdex 200 26/60 preparative column (AKTA Prime system used; GE Healthcare). Fractions containing pure protein were pooled and concentrated to 0.5 mL using a 50 kDa 20 mL Vivaspin concentrators. Protein was quantified using NanoDrop and then flash-frozen in liquid nitrogen.

Differential scanning fluorimetry assay
Assays were performed with 2 µM protein (1 mgÁmL À1 ), 5 9 SYPRO Orange dye (Sigma) and 1 mM ligand in the assay buffer (50 mM Tris pH 7.5, 150 mM NaCl, 2 mM MgCl 2 and 0.5 mM TCEP) in a total volume of 25 µL. Fluorescence (k ex = 530 nm, k em = 560 nm) was monitored with a Bio-Rad (CFX Connect TM ) real-time system, while system temperature was increased from 20 to 90°C in 1°C increments. Data were fitted to Boltzmann sigmoidal curve using GRAPHPAD PRISMÒ 6 to obtain T m , which is the inflection point of the curve. Assays were performed with six biological replicates, and each biological replicate contained three technical replicates.

Enzymatic activity assay
A coupled enzyme assay was deployed to study steady-state kinetics. In the assay, AGX1 wt utilizes UTP and GlcNAc-1P to produce UDP-GlcNAc and pyrophosphate; pyrophosphate is then degraded by pyrophosphatase (coupling enzyme) into inorganic phosphate. BIOMOL GreenÒ reagent (0.03% w/v malachite green, 0.2% w/v ammonium molybdate and 0.5% v/v Triton X-100 in 0.7 N HCL) was used to detect the inorganic phosphate produced [48]. To determine Michaelis-Menten parameters, the assay was performed in triplicate with each reaction in a total volume of 100 µL buffer containing 50 mM Tris, pH 7.5, 10 mM MgCl 2 , 10% (v/v) glycerol, 1 mM DTT, 0.05 units pyrophosphatase and 20 nM AGX1, and one of the two substrates was kept at an excess concentration (500 µM), while the other one varied from 0 to 500 µM. Reactions were initiated by the addition of 20 nM AGX1 wt / AGX1 A229T and then incubated at room temperature for 10 min. Reactions were terminated by the addition of 100 µL BIOMOL GreenÒ reagent. The assay plate was kept at room temperature for 30 min to allow colour development. The absorbance at 620 nm was measured using a Spectra max 340 PC. The assay was performed with five biological replicates, and each biological replicate contained three technical replicates. The turnover of substrates was < 10% under all conditions tested. Data were analysed using GRAPHPAD PRISMÒ to calculate K m , V max and P-values.
An XBZ (xanthene-based Zn) assay was developed to probe the activity of AGX1 wt and AGX1 A229T catalysing the reverse reaction. In the XBZ assay, the fluorescent probe XBZ is activated once it binds to UTP [49]. The high sensitivity and selectivity of XBZ for NTP over NDP-sugars and sugar-phosphates has been demonstrated previously [49,50]. Since pyrophosphate is one of the substrates of AGX1 in the reverse reaction and, in addition to UTP, it also induces XBZ fluorescence [49], we showed that 50 µM pyrophosphate was thoroughly degraded into inorganic phosphate after incubation with 0.05 U pyrophosphatase at room temperature for 30 min (Fig. S5A). A UTP standard curve was obtained using a UTP standard that was purchased from Sigma (catalog no. U6625) and serially diluted in the buffer containing 50 mM Tris pH 7.5, 0.5 mM MgCl 2 , 2% (v/v) glycerol, 20 µM pyrophosphate, 40 µM UDP-GlcNAc and 0.05 U pyrophosphatase in a total volume of 100 µL (Fig. S5B). Plates were placed at room temperature for 30 min to allow the development of background signals. For measuring the reverse reaction catalysed by AGX1 wt and AGX1 A229T , assays were performed in mixtures containing 50 mM Tris pH 7.5, 2% (v/v) glycerol, 0.5 mM MgCl 2 , 40 µM UDP-GlcNAc and 20 µM pyrophosphate in a total volume of 100 µL. Reactions were initiated by addition of 0.5 nM enzyme and then incubated at room temperature for 1 min. Reactions were terminated by boiling for 5 min. After cooling down to room temperature, reactions were then supplied with 0.05 U pyrophosphatase (in a volume of 5 µL) and incubated further at room temperature for 30 min to allow thorough degradation of unconsumed pyrophosphate. XBZ was prepared at 15 µM in a buffer containing 25 mM HEPES pH 7.5, 10 mM NaCl, 75 µM pyrocatechol violet and 50% (v/v) methanol. Two hundred microlitre XBZ solution was added to each assay, and fluorescence was measured using a Spectra max 340 PC with k ex and k em set as 485 and 530 nm, respectively. The turnover of both substrates was below 10%. Data analysis was performed using GRAPHPAD PRISMÒ.

Crystallization and structure solution
Before setting up crystal trays, AGX1 A229T (10 mgÁmL À1 ) was buffer exchanged into crystallization buffer containing 25 mM Tris, pH 7.5, which is similar to the crystallization buffer of wild-type AGX1 [40]. Crystallization was set up in sitting drop format with each drop containing 0.3 µL reservoir solution and 0.2 µL protein. Crystals appeared in 0.2 M MgCl 2 , 0.1 M Tris, pH 8.5 and 30% PEG 4000 within 3 days. Crystals were flash-frozen in liquid nitrogen and sent to the I24 beamline of the UK National Synchrotron (Diamond Light Source) for data collection. Data sets were indexed and integrated with iMOSFLM [51]. The phase problem was solved by molecular replacement using MOL-REP [52] and the PDB 1JV1 [40] as a phase donor. The structure was refined using REFMAC5 [53] and built manually using COOT [54]. PYMOL [55] was used to generate figures.

UAP1 A229T mutation is potentially pathogenic
In diploid organisms, haploinsufficiency is a phenomenon in which a single copy of a functional gene is not sufficient to produce the normal/wild-type phenotype. Since the patient is only heterozygous for the UAP1 A229T missense mutation (DDD entry: https://decipher.sanger.ac.uk/ddd/research-variant/c9774b 30226f59f8c1f79a7578fe5fc3/overview), we first investigated whether the UAP1 gene is associated with haploinsufficiency. Haploinsufficiency score (HI index) is the predicted probability of a gene to be haploinsufficient and over 80% of protein-coding genes in the human genome have been scored ranging from 0 (haploinsufficient) to 100% (haplosufficient) [59]. Based on the analysis of protein-coding genetic variation in 60 706 humans, the loss intolerance (pLI) is computed to predict the probability of a gene to be loss-of-function (LoF) intolerant [60]. High pLI scores (pLI ≥ 0.9) indicate LoF intolerance, whereas low pLI scores (pLI ≤ 0.1) indicate LoF tolerance. The predicted HI and pLI of the UAP1 gene are 42.14% and 0.91, respectively, which suggests that the UAP1 gene is potentially haploinsufficient and LoF intolerant.
To investigate whether the UAP1 A229 mutation is tolerated in an evolutionary context, primary sequences of eukaryotic UAP1 orthologues were aligned. The data show that A229 is conserved in all eukaryotes (Figs 2 and S1), suggesting an important structural and/or functional role. In addition to sequence analysis, the UAP1 A229T mutation was predicted to be pathogenic by in silico tools, such as SIFT [61] (score = 0.03) and Polyphen-2 [62] (score = 1) which evaluate effects of amino acid substitutions on proteins and MutationTaster [63] (P value = 1) which predicts the pathogenic potential of the DNA mutation. Together, these data suggest that the heterozygous UAP1 A229T mutation is potentially pathogenic.

The A229T mutation decreases AGX1 stability
We next investigated the possible effects of the A229T mutation on the biochemical properties of AGX1. AGX1 A229T was generated by site-directed mutagenesis, and GST-tagged AGX1 wt and AGX1 A229T were recombinantly expressed in E. coli and purified through glutathione-affinity and size exclusion chromatography. During protein purification, it was noticed that AGX1 A229T was prone to forming aggregates in solution, which may explain the lower protein yield (1.0 mgÁL À1 of culture) compared to AGX1 wt (4.5 mgÁL À1 of culture). To examine whether this was the result of intrinsic AGX1 A229T stability, we analysed the thermal stability of AGX1 wt and AGX1 A229T using a differential scanning fluorimetry assay (DSF). This revealed that, compared to AGX1 wt , AGX1 A229T was less stable as indicated by a reduction of the melting temperature (T m ) by approximately 5.3°C (Fig. 3A). DSF can also be used to observe protein-ligand interactions, which often increase the melting temperature [64,65]. The melting temperatures of both AGX1 wt and AGX1 A229T were increased by incubation with the substrates/products, that is UTP, GlcNAc-1P and UDP-GlcNAc (Fig. 3B), suggesting that AGX1 A229T retains the ability to bind these substrates. Of note, AGX1 A229T was less stable than AGX1 wt regardless of which ligand was supplemented (Fig. 3B). Hence, these data suggest that the A229T mutation decreases AGX1 stability in vitro.

The A229T mutation affects AGX1 activity
Given that A229 is conserved in eukaryotic UAP1 orthologues (Figs 2 and S1), and the A229T mutation resides in the catalytic domain, we next investigated the potential effects of the A229T mutation on the steady-state kinetics of AGX1 catalysing the forward reaction. The reaction was monitored through a colorimetric assay using pyrophosphatase as the coupling enzyme, which hydrolyses the AGX1 pyrophosphate product leading to free inorganic phosphate that can be detected with Biomol Green [48]. The results show that, compared to AGX1 wt , AGX1 A229T has a twofold Fig. 2. Sequence alignment of eukaryotic UAP1s. Sequences are named as organism abbreviation followed by sequence region. The phylogenetic tree was constructed based on full-length proteins using CLUSTALX [56] and then visualized with MEGA7 [57]. A229 is highlighted in a red box. DANRE, Danio rerio; XENTR, Xenopus tropicalis; DROME, Drosophila melanogaster; DICDI, Dictyostelium discoideum; ARATH, Arabidopsis thaliana; TRYB2, Trypanosoma brucei brucei (strain 927/4 GUTat10.1); CAEEL, Caenorhabditis elegans. increased K m (P < 0.01, t-test) towards both substrates (i.e. UTP and GlcNAc-1P) and decreased catalytic efficiency in utilizing UTP (P = 0.0382, t-test) (Fig. 3C,  D). To enable comparison of AGX1 wt and AGX1 A229T activity in catalysing the reverse reaction, we used a xanthene-based Zn (XBZ) fluorophore, which fluoresces upon binding of the reaction product UTP [49]. In this assay, AGX1 A229T is more active than AGX1 wt in consuming UDP-GlcNAc and pyrophosphate (P = 0.0359, t-test, Fig. 3E). Together, these data suggest that the A229T mutation affects AGX1 activity which may result in lower UDP-GlcNAc levels in vivo.

The A229T mutation induces structural changes
We next determined the crystal structure of AGX1 A229T to investigate the structural consequences of substituting the conserved A229 with a threonine.
We were unsuccessful in obtaining the AGX1 A229T crystals using the published AGX1 wt crystallization condition [40]. However, crystals of AGX1 A229T in complex with UDP-GlcNAc were obtained from screening commercial crystallization conditions, and synchrotron diffraction data were collected to 1.7 A ( Table 1). The structure was solved by molecular replacement using the published AGX1 structure (PDB: 1JV1 [40]) as a search model and refined to R = 0.187, R free = 0.226. AGX1 A229T crystallized with two molecules in a P1 unit cell. Structural superposition of AGX1 A229T and AGX1 wt revealed conformational differences (pairwise RMSDs of the A and B chains 0.4-1.0 A, Fig. S2) in the N-terminal domain of both chains (Fig. 4A,B). In addition, the uridine moiety of UDP-GlcNAc binding in the active site of the AGX1 A229T B chain points to the outside of the substrate binding pocket, which is different from the  (Fig. 4B). We suggest that the conformational changes in the AGX1 A229T B chain are due to an alkylation of Cys251 in the active site that displaces UDP-GlcNAc (Fig. 4B). We consider this alkylation to be artefactual as carbamidomethylation on multiple cysteines of both AGX1 wt and AGX1 A229T was observed with peptide mass fingerprinting (Table S1). Since our aim was to investigate the structural consequences of substituting the conserved A229 with a threonine, we focused on a detailed comparison of AGX1 wt A chain with its AGX1 A229T equivalent.
In line with the decreased catalytic efficiency of AGX1 A229T in consuming UTP (Fig. 3D), the A229T mutation is in the middle of the a-helix that coordinates the uridine moiety of UDP-GlcNAc (Fig. 4C). In AGX1 A229T structure, there is a shift of 1.2 A in the position of R228 (the residue next to A229) towards the N terminus and a shift of 1. 8 A in the C terminus of the helix encompassing the mutation to the opposite face of A229 (pushing effect) (Fig. 4C).
In the A and B chains of AGX1 wt , R228 interacts with E44 through two hydrogen bonds (3.0 and 3.1 A in A chain; 3.3 and 2. 6 A in B chain), which contributes to the interaction between the N-terminal and catalytic domain (Fig. S3). The R228-E44 interaction is abolished in the AGX1 A229T structure caused by the position shift of R228 (Fig. 4F). The pushing effect is likely due to the bulkier side chain of threonine compared to that of alanine. In AGX1 wt , the distances between the side chain of A229 and the main chain of Q195 and Q196 are 3.7 and 3. 6 A, respectively. However, In AGX1 A229T structure these distances are reduced to 2.7 and 2.2 A, respectively, which is below the C-O Van der Waals diameter (3.27 A [66]). Since alanine has high helix propensity while threonine has high b-sheet propensity [67], we then investigated whether the A229T mutation could cause distortion of this key active site a-helix. We aligned the AGX1 A229T A chain with eukaryotic UAP1 orthologues at the point of the mutation. The results showed that, in the aligned sequences, even though there are sequence variations at positions apart from A229 and L230, the AGX1 A229T A chain retains the a-helical conformation around position 229 (Fig. 4E). Together, the A229T mutation induced a local structural shift without disrupting the secondary structure of AGX1.
In a global structural view, the A229T mutation causes the N-terminal domain of AGX1 A229T A chain to adopt an open/relaxed conformation compared to that of AGX1 wt (Fig. 4A,F). In AGX1 wt B chain, a distant interaction between Q112 in the catalytic loop and M218 in the N-terminal domain is mediated by R169 through hydrogen bonds (3.3 and 3.1 A). The Q112-R169-M218 interaction was observed in the AGX1 wt B chain but not A chain (Fig. S3), which probably indicates the interaction is dynamic in solution or catalysis. Along with the conformational change of the N-terminal domain in the AGX1 A229T structure, is M218 shifted by 0.8 A away from R169, weakening the Q112-R169-M218 interaction (Fig. 4F). Together, the data suggest considerable conformational changes induced by the A229T mutation.

Discussion
UAP1 is the last enzyme in the HBP, reversibly converting UTP and GlcNAc-1P to UDP-GlcNAc and pyrophosphate. The steady supply of cytoplasmic UDP-GlcNAc is not only required for normal cellular physiology [43][44][45], but also essential for development [28,[31][32][33]. O-GlcNAcylation is one of the PTMs that depends on the steady supply of cytoplasmic UDP-GlcNAc. There is accumulating evidence that  [46], a de novo, heterozygous UAP1 gene missense mutation (NM_001324114:c.G685A:p.A229T) was identified in a patient with abnormalities of cranium, skeleton and nervous system. In this study, we demonstrated that UAP1 A229 is stringently conserved in eukaryotic UAP1 orthologues and the UAP1 A229T mutation is predicted to be pathogenic by clinical bioinformatic tools. Biochemical characterization revealed that, compared to AGX1 wt , AGX1 A229T has decreased stability and activity in producing UDP-GlcNAc while pyrophosphorolysis activity is increased. The stability and activity changes of AGX1 caused by the A229T mutation could collectively result in decreased UDP-GlcNAc and O-GlcNAcylation levels in vivo, giving a phenotype in the patient that resembles aspects of OGT-CDG. However, experiments with CRISPR/Cas9 knock-in cell lines or animal model systems are required to investigate this. Structural analysis revealed that the A229T mutation induced both local and global structural changes. R228 shifted by 1.2 A when A229 is mutated to threonine, followed by the disruption of the hydrogen bonds formed between R228 and E44. An allosteric inhibitor targeting Trypanosoma brucei UAP1 was discovered binding to the same site where R228 and E44 interact in human UAP1 (Fig. S4), and it was proposed that the inhibition was achieved by stabilizing the N-terminal domain in a conformation that prevents UTP binding [47], suggesting that the R228-E44 interaction in AGX1 could be contributing to regulation of activity. The R228-E44 interaction contributes to the interaction between the N-terminal and catalytic domains of AGX1 wt , and disruption in the AGX1 A229T structure leads to a shift of the N-terminal domain away from the catalytic domain. Accompanied by the shift of the N-terminal domain, is the disruption of the hydrogen-bonding interactions established through Q112-R169-M218 between the N-terminal domain and the key catalytic loop. Given both the R228-E44 and the Q112-R169-M218 interactions are human AGX1 specific [40,68,69] and AGX1 is over ten times more catalytically efficient than its fungal orthologues [45], they could be one of the factors that contribute to the high efficiency of AGX1, and disruption of them together in the AGX1 A229T structure could thus explain the stability and activity differences between AGX1 wt and AGX1 A229T . Threonine is one of the common substitutions of alanine in nonsynonymous coding single nucleotide polymorphisms (ncSNPs) database [70], and alanine-to-threonine substitutions have been associated with human conditions, such as amyloid diseases [70] and PD [71]. It seems that the mutation intolerance of UAP1 A229 in the evolutionary context is not determined by the R228-A229 interaction, as R228 is substituted to alanine in plants (Fig. 2), but more dependent on the interaction between the side chain of A229 and the main chain of Q195/Q196 (Fig. 4D). Together with the observation in fly models that the Drosophila UAP1 orthologue is required for epithelial morphogenesis and nervous system development [35][36][37], the UAP1 A229T missense mutation that results in considerable structure shifts that underpin the observed stability and activity changes of AGX1 A229T could be a contributory factor to the developmental delay observed in the patient.
Apart from being required for O-GlcNAcylation, UDP-GlcNAc is also the precursor for glycosylphosphatidylinositol anchor synthesis, heparan sulphate synthesis and N-linked glycosylation. Dysregulation of these pathways has also been associated with developmental disorders [72]. For instance, N-glycosylation is a common post-translational modification in all three kingdoms of life that regulates protein stability, protein processing and function [73]. DPAGT1/GPT (Dolichol phosphate N-acetylglucosamine-phosphotransferase) catalyses the first committed step of Nlinked glycosylation, using dolichol phosphate and UDP-GlcNAc as substrates [74]. Mutations in the DPAGT1 gene, which result in alteration of either protein activity, stability or structure, have been identified as causing congenital disorders [75][76][77][78]. Mutations of all the genes upstream of UAP1 in the HBP have been identified as causing developmental disorders, and decreases in O-GlcNAcylation levels are the common and predominant molecular phenotype observed in either patient-derived cells, mouse or fly models [28,[31][32][33][34]. N-linked glycosylation levels were unchanged in cells carrying patient GFAT mutations [33] and in mouse embryonic fibroblasts carrying a GNA1 gene disruption [30]; however, they were decreased in cells carrying patient PGM3 mutations [33]. The pathways affected by the UAP1 A229T mutation require further dissection using appropriate models. code 6Z2F. All remaining data are included within the manuscript and Supporting information.

Author contributions
DMFA conceived the study XC and ATF performed experiments. XC, OR and ATF analysed and interpreted the data. XC and DMFA wrote the manuscript with input from all authors.

Funding
XC is funded by the China Scholarship Council studentship. DMFA is funded by a Wellcome Trust Investigator Award (110061).

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Sequence alignment of eukaryotic UAP1s. Fig. S2. Heatmap showing the RMSDs of superposition between the two chains of AGX1wt and AGX1A229T. Fig. S3. Structural representation of AGX1 wt in complex with UDP-GlcNAc. Fig. S4. Structural superposition of AGX1 wt and TbUAP1 crystal structures. Fig. S5. Development of the XBZ assay to measure the reverse reaction catalysed by AGX1 A229T . Table S1. Analysis of AGX1 wt and AGX1 A229T carbamidomethylation by protein fingerprinting.