A missing enzyme-rescue metabolite as cause of a rare skeletal dysplasia

Abstract

Living cells depend on an intricate network of chemical reactions catalysed by enzymes, which sometimes make mistakes that lead to their inactivation. Here we report a metabolite-based mechanism for preserving enzyme function in an unfavourable environment. We found that the enzyme TGDS produces UDP-4-keto-6-deoxyglucose, a mimic of the reaction intermediate of the enzyme UXS1, which regenerates the essential cofactor NAD⁺ within the catalytic pocket of UXS1 by completing its catalytic cycle. Thus, the production of an 'enzyme-rescue metabolite' by TGDS represents a mechanism for maintaining the activity of an enzyme in a subcellular compartment where NAD⁺ is scarce. Using a combination of in vitro and in vivo studies, we demonstrate that the inability to produce sufficient amounts of this enzyme-rescue metabolite leads to the inactivation of UXS1, impairing the synthesis of specific glycans that are crucial for skeletal development. This provides an explanation for the development of the hereditary skeletal disorder Catel-Manzke syndrome in individuals with TGDS deficiency. Defects in similar protective layers might contribute to metabolic changes in other diseases that cannot be explained with common concepts in metabolic biochemistry.

Fig. 1. TGDS inactivation leads to a context-dependent inactivation of UXS1.

a–d, UDP-xylose (UDP-Xyl) was measured by LC–MS in parental cells and two TGDS-KO clones generated in U2OS (a), HCT116 (b), 293T (c) or HAP1 (d) cells. Rel., relative. e, Schematic representation of the biosynthetic pathway of UDP-xylose. UXS1 produces UDP-xylose from UDP-glucuronate (UDP-GlcA) in the Golgi and endoplasmic reticulum (ER). In turn, UDP-xylose inhibits production of UDP-glucuronate by UDP-glucose dehydrogenase. UDP-Glc, UDP-glucose. f–i, UDP-glucuronate was quantified by LC–MS in the same cell lines as in a–d. j,k, UDP-xylose (j) and UDP-glucuronate (k) were quantified in parental, TGDS-KO and UXS1-KO HAP1 cells transduced with a lentivirus driving expression of TGDS or an empty control. Results support the hypothesis that loss of TGDS impairs UDP-xylose production and secondarily leads to accumulation of UDP-glucuronate. Data are normalized area under the curve for the indicated metabolites (mean ± s.d. of 3 (U2OS, HCT116, 293T) or 4 (HAP1) independent experiments, each containing 3 biological replicates) and are presented relative to wild-type (WT) control cell lines. Paired two-tailed Dunnett (a–d,f–i) or Sidak (j,k) post hoc testing of log-transformed data after one-way ANOVA. *P < 0.05, **P < 0.01, ***P < 0.001; NS, not significant. For exact P values see Source Data. Source data

Fig. 2. Functionally inactivated UXS1 requires reactivation by a UDP-4-keto sugar when H6PD is highly active.

a, Working hypothesis: an abortive catalytic cycle of UXS1 leads to the inactivation of the enzyme, which is counteracted by UDP-4-keto sugars. The decarboxylation of UDP-glucuronate by UXS1 depends on the oxidation of the C4 hydroxyl group using a tightly bound NAD⁺. This generates NADH and a UDP-4-ketoxylose intermediate. Normally, formation of UDP-xylose regenerates NAD⁺, preparing the enzyme for another cycle. Infrequently, the intermediate dissociates from the catalytic pocket, leaving UXS1 bound to NADH and inactive. We hypothesized that UDP-4-keto sugars restore activity by facilitating the oxidation of enzyme-bound NADH to NAD⁺. Loss of the cofactor can also yield an inactive apoenzyme, but reactivation by NAD⁺ binding is limited by the low NAD⁺/NADH ratio in the endoplasmic reticulum maintained by H6PD. b, Schematic of the experiment to test whether UDP-4-ketoxylose produced by ArnA can reactivate UXS1. c–f, ArnA expression rescues UDP-xylose synthesis in TGDS-KO cells, but not in UXS1-KO cells. UDP-xylose (c,e), UDP-glucuronate (d,f) were quantified by LC–MS in parental 293T (c,d) and HAP1 (e,f) cells, and in TGDS-KO or UXS1-KO clones transduced with a lentivirus expressing ArnA or an empty vector. ND, not determined. g, Experiment exploring the role of H6PD in UXS1 dependency on TGDS. h–k, H6PD modulates UXS1 dependency on TGDS. In HAP1 cells, CRISPR–Cas9-mediated H6PD knockdown (using two different guide RNAs (gRNAs) versus control gRNA (−)) reduces the effect of TGDS deficiency on UXS1 function (h,i). In U2OS cells, H6PD overexpression (+) makes UXS1 activity dependent on TGDS (j,k). −, empty vector control. UDP-xylose (h,j) and UDP-glucuronate (i,k) were quantified by LC–MS. Data are mean ± s.d. of 3 independent experiments, each containing 3 biological replicates. Paired two-tailed Dunnett (c,d) or Sidak (e,f,i–k) post hoc testing of log-transformed data after one-way ANOVA. For exact P values see Source Data. Source data

Fig. 3. TGDS produces UDP-4-keto-6-deoxyglucose, an enzyme-rescue metabolite for UXS1.

a, Schematic of the suggested UDP-glucose-4,6-dehydratase activity of TGDS (bottom), and UDP-rhamnose or dTDP-rhamnose synthesis in bacteria (top). UDP-6dH, UDP-6-deoxyhexose. 4k6dG, 4-keto-6-deoxyglucose. b, TGDS activity was assessed at the indicated concentrations of UDP-glucose for 4 h at 30 °C. Formation of UDP-4-keto-6-deoxyglucose is presented in arbitrary units. c–f, Expression of human TGDS or B. cinerea UG46DH rescue the phenotype of TGDS-KO cells. UDP-4-keto-6-deoxyglucose (c), UDP-6-deoxyhexose (d), UDP-xylose (e) and UDP-glucuronate (f) were quantified in parental and TGDS-KO 293T cells upon transduction with recombinant lentiviruses driving expression of human TGDS or B. cinerea UG46DH. Metabolite levels in TGDS-KO cells in c,d were close to background. Levels of UDP-4-keto-6-deoxyglucose and UDP-6-deoxyhexose in TGDS-KO cells were unaffected by the inactivation of H6PD (Extended Data Fig. 4i–l). g, Experimental setup to assess the reactivation of UXS1 via the product of TGDS. h, The TGDS product rescues functionally inactivated UXS1. UDP-xylose, UDP-4-keto-6-deoxyglucose, UDP-6-deoxyhexose, NAD⁺ and NADH were quantified by LC–MS in reactions where UDP-glucuronate was incubated with 0.56 µM UXS1, sodium borohydride-inactivated UXS1 or inactivated UXS1 in the presence of a 1.5-fold excess of the product of TGDS (TGDS-P). Data are mean ± s.d. from three independent experiments, each containing three biological replicates in c–f,h. Paired two-tailed Sidak (c–f) or Dunnett (h) post hoc testing of log-transformed data after one-way ANOVA. For exact P values see Source Data. Source data

Fig. 4. Patient-derived cell lines and a mouse model of Catel–Manzke syndrome corroborate the molecular function of TGDS.

a, 3D reconstruction of the facial skeleton in individual 4. b–d, UDP-xylose (b), UDP-glucuronate (c) and UDP-6-deoxyhexose (d) were measured in fibroblasts from healthy controls and individuals with Catel–Manzke syndrome (ID1–5). Samples were analysed in two batches: individual 1 versus control 1; individuals 2–5 versus controls 2 and 3. e, Pathogenic variants of TGDS proteins. f, Western blot of Flag-tagged TGDS after transfection in U2OS cells. Quantification in Extended Data Fig. 6f and uncropped images in Supplementary Fig. 1. g,h, UDP-4-keto-6-deoxyglucose production by recombinant wild-type TGDS and indicated variants at 500 µM UDP-glucose for 24 h at 30 °C (g) or indicated UDP-glucose concentrations for 4 h at 30 °C (h). a.u., arbitrary units. i, Schematic of Tgds^A100S/− (KI/KO) mice carrying p.Ala100Ser and a frameshift deletion–insertion, causing loss of function. Created in BioRender. Lyubenova, H. (2025) https://BioRender.com/hsez57t. j, µCT sagittal and coronary images of E18.5 wild-type and KI/KO embryos showing brachycephaly, with shorter mandibles and snouts. Measurements that are significantly different between wild-type and KI/KO embryos are indicated in yellow. k,l, Skeletal preparations showing shortened hindlimb (k) and forelimb (l) long bones in KI/KO E18.5 embryos. m–p, Quantification of UDP-4-keto-6-deoxyglucose (m), UDP-6-deoxyhexose (n), UDP-xylose (o) and UDP-glucuronate (p) in organ lysates from 8-month-old wild-type and KI/KO mice. ND, not detectable for technical reasons. Data are mean of two (h) or mean ± s.d. from four (b–d,g) independent experiments; from 11 wild-type (j–l) and 9 KI/KO (j–l) mice; or from 4 wild-type and KI/KO mice (m–p), normalized to wild-type or control conditions. *, # and † denote groups that are significantly different by two-tailed Sidak (b–d) or Dunnett (g) post hoc testing of log-transformed data after one-way ANOVA (b–d,g), Holm–Sidak corrected multiple t-tests (j–l) or multiple t-tests after log transformation (m–p). For exact P values see Source Data. Source data

Fig. 5. TGDS deficiency leads to reduced heparan sulfate formation and reduced glycosylation of α-dystroglycan.

a, Schematic representation of the role of xylose in heparan sulfate. Xyl, xylose; Gal, galactose; GlcA, glucuronate; GlcNAc, N-acetylglucosamine. b–e, Representative histograms (b–d) and quantification of eight independent flow cytometry experiments (e) using an antibody against heparan sulfate in wild-type, TGDS-KO and UXS1-KO HAP1 cell lines and cell lines rescued with the indicated cDNAs. f, Schematic representation of the glycan of α-dystroglycan (matriglycan) and its detection by laminin overlay assay and flow cytometry. g, Representative laminin overlay analysis performed with samples from wild-type, TGDS-KO and UXS1-KO HAP1 cells. Western blot analysis for β-dystroglycan (βDG) on the same membrane is used as a control for dystroglycan abundance. For uncropped images see Supplementary Fig. 1. h, Quantification of three independent laminin overlay experiments. The signal of the laminin overlay (LO) was normalized to the β-dystroglycan western blot signal, and then to the wild type in each experiment. i–l, Representative histograms (i,j) and quantification of four (k) and three (l) flow cytometry experiments using an antibody against matriglycan in wild-type, TGDS-KO or UXS1-KO HAP1 (i–k) and 293T (l) cell lines as well as cell lines rescued with the indicated cDNAs. Data are mean ± s.d. from 8 (e), 3 (h), 4 (k) or 3 (l) independent experiments and are normalized to the mean of the value in wild-type cells. Paired two-tailed Sidak post hoc testing of log-transformed data after one-way ANOVA. MFI, mean fluorescent intensity; HepS, heparan sulfate. For exact P values see Source Data. Drawings in a,f Created in BioRender. Lyubenova, H. (2025) https://BioRender.com/hee241k. Source data

Extended Data Fig. 1. Inactivation of TGDS in 4 different cell lines does not affect most nucleotide sugar levels.

a, Sequence of the genomic locus and predicted TGDS proteins in TGDS-deficient HAP1, U2OS, 293 T and HCT116 cell line clones. b-g, CMP-sialic acid (b), GDP-mannose (c), UDP-hexose (d), UDP-GlcNac (e), GDP-fucose (f), and CDP-ribitol (g) were determined by LC-MS in parental cells and two TGDS KO clones from U2OS, HCT116, HAP1 and 293 T. Data present means ± SD of three (U2OS, HCT116 and 293 T) or four (HAP1) experiments performed with 3 independent samples each, and are normalized to levels observed in wild-type cells. * p < 0.05 in paired two-tailed Dunnett post-hoc testing of log-transformed data after one-way ANOVA. For exact p-values see source data file. Source data

Extended Data Fig. 2. Data complementing Fig. 2.

a-f, UDP-xylose (a,c), UDP-glucuronate (UDP-GlcA, b,d), UDP-4-ketoxylose (e) and UDP-arabinose (f) levels were determined by LC-MS in parental 293 T (a,b) and HAP1 (c-f) cells and in a TGDS KO or a UXS1 KO clone upon transduction with a recombinant lentivirus driving expression of ArnA (‘+’) or an empty vector control (‘-’). To allow comparison of the signal intensities obtained for different metabolites, data were only normalized to total ion current but not to the mean of the values observed in the parental cells within each experiment. g-h, Extracted ion chromatograms of the m/z of UDP-pentose (m/z 535.037) corresponding to panel a (g) and panels c,f (h). i-j, Extracted ion chromatograms of the m/z of UDP-GlcA (i, 579.027) or UDP-xylose (j, 535.037) in reactions containing 500 µM UDP-GlcA in the presence of 0.8 µM TGDS or 0.65 µM UXS1 for 24 h. k-l, Western blot for H6PD in wild-type and TGDS-deficient HAP1 (k) or 293 T (l) cell lines, transduced with lentiviral CRISPR/Cas9 constructs targeting H6PD (each ‘+‘ representing a different guide RNA) or a control gene (‘-‘), performed twice. See Supplementary Fig. 1 for uncropped images. m-n, UDP-xylose (m) and UDP-GlcA (n) were quantified in 293 T cell lines described in panel l. o-p, Western blot for H6PD in wild-type and a TGDS-deficient U2OS (o) or HCT116 (p) cell lines, transduced with recombinant lentiviruses driving expression of human H6PD cDNA (‘+‘) or an empty control (‘-‘), performed twice. q-r, UDP-xylose (q) and UDP-GlcA (r) were quantified in HCT116 cell lines described in panel p. Data are means ± SD of 3 independent experiments. * p < 0.05; ** p < 0.01; *** p < 0.001 in paired Sidak post-hoc testing of log-transformed data after one-way ANOVA. For exact p-values see source data file. Source data

Extended Data Fig. 3. TGDS and UXS1 localize to the ER and the Golgi apparatus, whereas H6PD is limited to the ER.

Immunofluorescence images of U2OS cells after transfection with pCMV-TGDS, pCMV-UXS1-FLAG and pCMV-H6PD-V5. TGDS localization was detected using anti-TGDS, UXS1 using anti-FLAG, and H6PD using either anti-H6PD or anti-V5 antibodies. Endogenous markers for the ER (PDIA1; CALR) and Golgi (GM130; GIANTIN) were used to determine the proteins localization. All panels except the costaining with GM130 (i.e. bottom three images in panel a) were performed with methanol-fixed cells. In the latter case, paraformaldehyde fixation was needed, which explains the slightly different morphology of the TGDS expression pattern. Scale bars, 10 µm.

Extended Data Fig. 4. Levels of the TGDS product UDP-4-keto-6-deoxyglucose and of UDP-6-deoxyhexose are lower in all TGDS KO cells and unaffected by inactivation of H6PD.

a, Recombinant purified human TGDS, B. cinerea UG46DH, E. coli ArnA and human UXS1 analysed by SDS-PAGE and Coomassie staining to demonstrate purity. b, Extracted ion chromatogram (EIC) of the m/z corresponding to UDP-4-keto-6-deoxyglucose (i.e., 547.037) in reactions of recombinant fungal UG46DH or human TGDS with UDP-glucose. c, UDP-4-keto-6-deoxyglucose, the reaction product of UG46DH and TGDS, is partially hydrated in a non-enzymatic reaction to form UDP-4,4-dihydroxy-6-deoxyglucose, which has the same mass as UDP-glucose. d-e, Extracted ion chromatograms of the m/z of UDP-glucose (d, 565.048) and UDP-4-keto-6-deoxyglucose (e, UDP-4k6dG, 547.037) in reactions containing 500 µM UDP-glucose in the presence of 0.8 µM TGDS or 0.65 µM UXS1 for 24 h. The trailing peak (arrow, d) in the ‘UDP-glucose’ chromatogram corresponds to the hydrated form of UDP-4k6dG previously reported in reactions for fungal orthologues. f-g, UDP 6-deoxyhexose (UDP-6dH, f) and UDP-4-keto-6-deoxyglucose (UDP-4k6dG, g) levels were determined by LC-MS in parental cells and 2 TGDS KO clones from U2OS, HCT116, HAP1 and 293 T. h, Tentative identification of UDP-6-deoxyhexose as UDP-6-deoxyglucose. Extracted ion chromatogram (EIC) of the m/z corresponding to UDP-6-deoxyhexose (i.e., 549.053) in a reaction where the TGDS product (TGDS-P) was chemically reduced with NaBH₄, leading to the formation of UDP-6-deoxyglucose and UDP-6-deoxygalactose (upper panel). Based on the catalytic cycle of UXS1, only UDP-6-deoxyglucose formation is expected when the TGDS product is incubated with UXS1 that has been inactivated using NaBH₄ beforehand (second panel). The UDP-6-deoxyhexose peak in cells coelutes with the one corresponding to UDP-6-deoxyglucose (third panel). Coelution is corroborated (lower panel) by the fact that the peak increases upon addition of a small amount of the sample presented in the upper panel. i-l, Inactivation of H6PD in TGDS-deficient cells does not lead to an increase in the rescue metabolite UDP-4k6dG (i,j) nor in UDP-6dH (k,l). LC-MS analysis for the indicated metabolites was performed in wild-type and a TGDS-deficient HAP1 (i,k) or 293 T (j,l) cell line, transduced with lentiviral CRISPR/Cas9 constructs targeting H6PD (each ‘+‘ representing a different guide RNA) or a control gene (‘-‘). Data are means ± SD of the means obtained in three (f,g) and two (i-l) independent experiments performed with three biological replicates. * p < 0.05; ** p < 0.01; *** p < 0.001 in paired two-tailed Dunnett (f-g) post-hoc testing of log-transformed data after one-way ANOVA. For exact p-values see source data file. Source data

Extended Data Fig. 5. Products from the bacterial and fungal orthologue of TGDS reactivate inactivated UXS1.

a, Extracted ion chromatograms for the m/z corresponding to UDP-xylose, UDP-4-keto-6-deoxyglucose (‘UDP-4k6dG’), UDP-6-deoxyhexose (‘UDP-6dH’), NAD⁺, and NADH, obtained by LC-MS in reactions where UDP-glucuronate was incubated with UXS1, with sodium-borohydride-inactivated UXS1, or with inactivated UXS1 in the presence of the product of human TGDS (‘TGDS-P’). Quantifications as in Fig. 3h, but including a condition where 0.4 µM NAD⁺ was added, demonstrating that low levels of free NAD⁺ present in the TGDS preparation do not suffice to rescue UDP-xylose production. b-c, Experiment described in panel a but using the Botrytis cinerea UG46DH product (‘UG46DH-P’, b) or E. coli ArnA product (‘ArnA-P’, c). d, Schematic representation of the formation of different NADH forms as well as the quantification of 1,6-NADH and 1,2-NADH in the experiment presented in Fig. 3h. e, Quantification of UDP-xylose production in an experiment as described in Fig. 3h, but including a condition where 10 µM NAD⁺ was added, demonstrating that UXS1 activity can be restored by sufficient amounts of NAD⁺. Data are means of two (e) or means ± SD of three independent experiments, each consisting of three (a,d) or one biological replicate (b,c). * p < 0.05, ** p < 0.01, *** p < 0.001 in two-tailed paired Dunnett post-hoc testing of log-transformed data after one-way ANOVA. For exact p-values see source data file. Source data

Extended Data Fig. 6. Characterization of clinical manifestations and TGDS variants in individuals with Catel-Manzke syndrome.

a-c, Schematic representation of the TGDS variant in affected individual 3, abolishing the splice acceptor site of exon 6 (a; created in BioRender. Lyubenova, H. (2025) https://BioRender.com/l6uhom8), leading to a transcript lacking exon 6 as demonstrated by RT-PCR (b) and (c) sequencing, performed twice with the same result. d-e, Alphafold model of TGDS containing NAD⁺ and dTDP-glucose modeled based on the structure of prokaryotic dTDP-glucose dehydratases with Alphafill (d) or Alphafold model obtained with the Alphafold Multimer algorithm (e), predicting a dimeric protein. Amino acid variants observed in affected individuals are highlighted in green. V239, A100 and T102, as well as the region deleted in affected individual 3 are lining the catalytic pocket, whereas E90 and E322 are more peripheral. f, Quantification of TGDS expression normalized to GAPDH expression in U2OS cells transfected with the indicated variants containing a FLAG tag. Data are means ± SD from three independent experiments with two-tailed Dunnett post-hoc testing after one-way ANOVA. * denotes the only significant change (p-value 0.032). g, Photographs, radiographs and CT scans with 3D reconstruction of individuals (‘I’) with Catel-Manzke syndrome. Top row, hand radiographs of I2, I3 and I4. Lower row, hand photographs of I2 and I5 as well as cranial frontal CT scans with 3D reconstruction of I4. Right side, frontal and lateral whole body radiograph of I5 at 24 weeks of gestation, after pregnancy termination. Note radial deviation of the 2nd fingers in I2 and I5 as well as an accessory ossification center at the base of the 2nd proximal phalanx in I2 (Manzke dysostosis). Manzke dysostosis is a variable feature in Catel-Manzke syndrome. I3, I4 and I5 have clinodactyly of the 5th finger, I4 has additional clinodactyly of the 2nd, 3rd and 4th finger as well as distally enlarged metacarpals of the 2nd 3rd and 4th ray. Cranial CT scan of I4 shows asymmetric micrognathia. Whole body radiograph of I5 shows mildly shortened long bones, clinodactyly of the 5th fingers as well as micro- and retrognathia. I, individual; y, years; w, weeks; m, months. Source data

Extended Data Fig. 7. Characterization of TGDS KI/KO mice.

a, Photographs of WT and KI/KO embryos at E18.5 show mildly shortened snouts and limbs in the mutants. Scale bar: 2 mm. b, µCT scans of 5-weeks-old WT and KI/KO mice. Measured parameters are shown in yellow. c, Quantification of skull measurements reveals tendencies similar to the ones observed in embryos, but they did not reach significance, likely due to smaller cohort sizes. Each measurement was normalized to the mean of the cohort of WT mice (littermates). Data are mean ± SD (n = 3). The differences between groups were analysed by two-tailed Welch’s t-test corrected for multiple testing according to Holm-Sidak. d, Skeletal preparations of front limb paws of WT and KI/KO embryos at E18.5 show normal digit morphology. e, Digit length was unaffected in the mutant animals. Each measurement was normalized to the mean of the WT littermates and data are mean ± SD (n = 8-10 for WT, and n = 8-9 for KI/KO). The differences between groups were analysed by two-tailed Welch’s t-test corrected for multiple testing by Holm-Sidak. L, left; R, right. Scale bar: 0.5 mm. For exact p-values see source data file. Source data