Splicing regulation of GFPT1 muscle-specific isoform and its roles in glucose metabolisms and neuromuscular junction

Summary Glutamine:fructose-6-phosphate transaminase 1 (GFPT1) is the rate-limiting enzyme of the hexosamine biosynthetic pathway (HBP). A 54-bp exon 9 of GFPT1 is specifically included in skeletal and cardiac muscles to generate a long isoform of GFPT1 (GFPT1-L). We showed that SRSF1 and Rbfox1/2 cooperatively enhance, and hnRNP H/F suppresses, the inclusion of human GFPT1 exon 9 by modulating recruitment of U1 snRNP. Knockout (KO) of GFPT1-L in skeletal muscle markedly increased the amounts of GFPT1 and UDP-HexNAc, which subsequently suppressed the glycolytic pathway. Aged KO mice showed impaired insulin-mediated glucose uptake, as well as muscle weakness and fatigue likely due to abnormal formation and maintenance of the neuromuscular junction. Taken together, GFPT1-L is likely to be acquired in evolution in mammalian striated muscles to attenuate the HBP for efficient glycolytic energy production, insulin-mediated glucose uptake, and the formation and maintenance of the neuromuscular junction.


OPEN ACCESS
9][20][21][22] Glucosamine is converted to glucosamine-6-phosphate by hexokinase-1, and increases the production of UDP-GlcNAc without utilizing GFPT1 (Figure 1D).Increased production of UDP-GlcNAc by glucosamine also induces insulin resistance in skeletal muscle. 23,24Similarly, the enzymatic activity of GFPT1 is elevated in skeletal muscle in patients with non-insulin dependent diabetes mellitus (NIDDM). 25onversely, transgenic mice overexpressing Gfpt1 in skeletal muscle and fat develop insulin resistance, which is a hallmark of NIDDM. 26,27RSF1 is a ubiquitously expressed RBP of the serine-and arginine-rich (SR) protein family.SRSF1 primarily promotes exon inclusion by binding to an exonic splicing enhancer in both constitutively and alternatively spliced exons. 28The Rbfox proteins are brain-and muscle/heart-specific RBPs, and are comprised paralogous Rbfox1, Rbfox2, and Rbfox3.0][31] Rbfox proteins bind to the highly conserved (U)GCAUG element across all vertebrate species. 29,30,32,33The binding of Rbfox to an alternative exon or its upstream intron represses splicing, whereas the binding to its downstream intron activates splicing. 29,30HnRNP H and hnRNP F are closely related and ubiquitously expressed RBPs that belong to the heterogeneous nuclear ribonucleoprotein (hnRNP) family. 34,35HnRNP H and hnRNP F share highly conserved sequences (68% amino acid identity), and both bind to poly(G)-rich sequences (G-runs) located in a target exon and its flanking introns to regulate alternative splicing. 36,37][40][41] We here dissected the molecular mechanism of muscle-specific alternative splicing of exon 9 in human GFPT1.We showed that SRSF1, Rbfox1/2, and hnRNP H/F cooperatively regulated the alternative splicing of human GFPT1 exon 9. To understand the functional significance of GFPT1-L in skeletal muscle, we generated Gfpt1 exon 9 knockout (KO) mouse and found that lack of Gfpt1 exon 9 markedly increased GFPT1-S and UDP-GlcNAc, and compromised glycolytic energy production, insulin-mediated glucose uptake, and the formation and maintenance of the NMJ.

Muscle-specific GFPT1 exon 9 is alternatively spliced in skeletal muscle and heart in mammals
The inclusion of exon 9 in humans incorporates 18 amino acids close to the C-terminal end of the glutamine amidotransferase (GATase) domain of GFPT1 protein (Figure 1A) and alters its enzymatic activity. 8,9In agreement with previous reports, 8,9 RT-PCR of total RNA extracted from various human tissues showed that GFPT1 exon 9 was included only in skeletal muscle and heart (Figure 1B).Similar alternative inclusion of exon 9 was observed in the course of myotube differentiation of immortalized human KD3/Hu5 myoblasts (Figure 1C).
The phylogenetic tree of paralogous GFPT1 and GFPT2 by Ensembl showed that they were likely to have evolved after the appearance of vertebrates, because invertebrate GFPT1/2, vertebrate GFPT1, and vertebrate GFPT2 make distinct clusters (Figure S2).The HomoloGene database at NCBI also showed that invertebrate GFPT1/2 is likely to be ancestor(s) of vertebrate GFPT2, but not vertebrate GFPT1 (Figure S3).Multiple alignment of genomic sequences of GFPT1 exon 9 in 100 vertebrates using the UCSC genome browser showed that amino acid sequences encoded by exon 9 are highly conserved in mammals (Figure S4).We calculated tissue-specific percent spliced-in (PSI) of GFPT1 exon 9 in four mammals, and found that exon 9 was exclusively included in skeletal muscle and heart (red bars in Figure 1E).
RNA-binding proteins, SRSF1, Rbfox1/2, and hnRNP H/F, regulate alternative splicing of GFPT1 exon 9 Putative binding motifs of RNA-binding proteins (RBPs) on human GFPT1 exon 9 and its flanking introns were searched for using human splicing factor databases, ESE finder 3.0 42 and SpliceAid 2 43 (Figure S5A).Knocking down of candidate RBPs revealed that downregulation of SRSF1 causes skipping of exon 9 in differentiating human KD3/Hu5 myotubes (Figure 2A).Similarly, downregulation of Rbfox1 partially and Rbfox2 marginally enhanced skipping of exon 9, while simultaneous downregulation of both Rbfox1 and Rbfox2 almost exclusively skipped exon 9, indicating that these RBPs are functionally redundant to each other (Figure 2A).In contrast, individual downregulation of hnRNP H or F partially enhanced inclusion of exon 9, whereas simultaneous downregulation of both hnRNP H and F resulted in almost exclusive inclusion of exon 9 (Figure 2A).Downregulation of the other candidate RBPs (SRSF2, SRSF3, SRSF5, YB1, hnRNP A1) had marginal effects on splicing (Figure S5B).Taken together, SRSF1, Rbfox1, and Rbfox2 functioned as splicing enhancers, whereas hnRNP H and F functioned as splicing silencers for GFPT1 exon 9.
To dissect splicing cis-elements for GFPT1 exon 9, we constructed human GFPT1 minigene by inserting exon 9 and its flanking introns in the modified exon-trapping pSPL3 vector. 44,45As has been reported in other minigenes, 35,46,47 splicing efficiency of the pSPL3 minigene was low and exon 9 was not included in transfected KD3/Hu5 myoblasts/myotubes (Figures 2B-2D).ESE finder predicted two high-affinity binding motifs for SRSF1 in exon 9, with the sequences of CACACGG (score 5.86) and CACAGGG (score 5.25).RT-PCR showed that overexpression of SRSF1 partially increased exon 9 inclusion in the wild-type GFPT1 minigene (Figure 2B).In contrast, disruption of the first putative SRSF1-binding motif (CACACGG) compromised SRSF1-mediated exon 9 inclusion, whereas disruption of the second putative SRSF1 binding motif (CACACGG) had no such effect (Figure 2B), indicating that the first binding motif is the functional motif for SRSF1.Since disruption of the second putative SRSF1-binding motif had no effect on splicing, we did not make a construct that disrupted both putative SRSF1 binding motifs.It is well established that Rbfox proteins bind to the (U)GCAUG motif. 29,48An Rbfox motif of UGCAUG is present in intron 9. Overexpression of either Rbfox1 or Rbfox2 enhanced exon 9 inclusion in the wild-type GFPT1 minigene (Figure 2C).In contrast, disruption of the Rbfoxbinding motif abrogated exon 9 inclusion even with overexpression of Rbfox1 and/or Rbfox2, indicating that the intronic UGCAUG motif is the target binding site of Rbfox1/2 (Figure 2C).6][37] There are two G-runs in exon 9 and one G-run in intron 9. Disruption of the exonic G-runs resulted in exclusive inclusion of exon 9, whereas disruption of the intronic G-run had no such effect, indicating that the exonic G-runs are functional motifs for hnRNP H and F (Figure 2D).(D) RT-PCR of KD3/Hu5 myoblasts transfected with GFPT1 minigenes carrying wild type (WT MG) and mutant exonic (Ex-H/F BSmut MG) and intronic (In-H/F BSmut MG) binding sites for hnRNP H/F.(E) Schematic of the GFPT1 minigene carrying the MS2 hairpin-loop (hp) substituting for the first SRSF1-binding motif in exon 9 (SRSF1BS > MS2hp MG), while the second putative SRSF1-binding motif is mutated (red letters).SRSF1 is fused to the MS2 coat protein (SRSF1-MS2) (inverted U-shape) to directly tether SRSF1 to the MS2 hairpin-loop with or without knocking down of Rbfox1 and Rbfox2.(F) Schematic of the GFPT1 minigene carrying the MS2 hairpin-loop substituting for the Rbfox-binding motif in intron 9 (RbfoxBS>MS2hp MG).Rbfox1 and Rbfox2 are fused to the MS2 coat protein (Rbfox1-MS2 and Rbfox2-MS2) to directly tether Rbfox1 and Rbfox2 to the MS2 hairpin-loop, respectively, with or without knocking down of SRSF1.(A-F) In all the transfection studies, minigenes, siRNAs, and RNA-binding proteins were transfected into KD3/Hu5 myoblasts on differentiation day 1 and harvested on differentiation day 3.
To further identify the position-specific effects of the identified trans-acting splicing factors, we tethered each RBP to a specific RNA segment using the bacteriophage coat proteins MS2 or PP7.To this end, SRSF1, Rbfox1, and Rbfox2 were fused with MS2 coat protein (SRSF1-MS2, Rbfox1-MS2, and Rbfox2-MS2), and hnRNP H and F were fused with PP7 coat protein (hnRNP H-PP7 and hnRNP F-PP7) (Figure S6).We first introduced the MS2 coat protein-binding hairpin-loop sequence (MS2-HP) in place of the first SRSF1 binding motif (CACACGG) in exon 9, while the second putative SRSF1 binding motif was mutated (CCAAGGG).As expected, tethering of SRSF1 induced exon 9 inclusion (Figure S7A).We next replaced the Rbfox binding motif (UGCAUG) with the MS2-HP, and found that tethering of either Rbfox1 or Rbfox2 enhanced exon 9 inclusion (Figure S7B).We then replaced the two exonic G-runs in exon 9 (GGGG .GGG) with the PP7 coat protein-binding hairpin-loop sequence (PP7-HP).As predicted, tethering of either hnRNP H or hnRNP F suppressed exon 9 inclusion (Figure S7C).Thus, our data demonstrate that the splicing regulatory RBPs function through binding to their specific cis-regulatory motifs to regulate GFPT1 exon 9 splicing.
We next questioned whether SRSF1 and Rbfox1/2 cooperatively enhance exon 9 inclusion.As shown previously, tethering of SRSF1 enhanced exon 9 inclusion; however, individual knockdown of Rbfox1 or Rbfox2 partially suppressed the effect of SRSF1 (Figure 2E).In addition, simultaneous knockdown of both Rbfox1 and Rbfox2 markedly suppressed the effect of SRSF1.Similarly, tethering of either Rbfox1 or Rbfox2 enhanced exon 9 inclusion, but simultaneous knockdown of SRSF1 suppressed these effects (Figure 2F).Thus, SRSF1 and Rbfox1/2 depend on each other for efficient inclusion of GFPT1 exon 9.
SRSF1 and Rbfox1/2 enhance U1 snRNP recruitment at the 5 0 splice site, whereas hnRNP H/F suppress the recruitment We next examined the expression levels of the identified trans-acting RBPs in myotube differentiation of human KD3/Hu5 myoblasts.qRT-PCR showed that SRSF1 was highly expressed in undifferentiated myoblasts and was marginally increased upon myotube differentiation (Figure 3A).Similarly, the expression levels of Rbfox1 and Rbfox2 were markedly increased during myotube differentiation (Figure 3B).In contrast, the expression levels of hnRNP H and hnRNP F tended to decrease upon myotube differentiation (Figure 3C).Similar results were observed at the protein level by western blotting in the course of myotube differentiation of KD3/Hu5 myoblasts (Figure 3D).
As the alternative inclusion of GFPT1 exon 9 is specific to mammals, we next asked whether the binding sites of SRSF1, Rbfox1/2, and hnRNP H/F were conserved across species.Alignment of the genome sequences revealed that Rbfox1/2-binding site (UGCAUG) in intron 9, as well as the first SRSF1-binding site (CACACGG) and two hnRNP H/F-binding sites (GGG) in exon 9, were conserved in mammals but not in other species (Figure S4).Thus, the splicing regulation of GFPT1 exon 9 by SRSF1, Rbfox1/2, and hnRNP H/F is likely to be conserved across mammalian species.
We noticed that the 5 0 splice site at the boundary of exon 9 and intron 9 had a weak splicing cis-element with the GAG|gtggga sequence with a MaxEntScan score 49 of 4.3 and an SD score 50 of À3.333.We then made a minigene construct carrying the optimal 5 0 splice site sequence (GAG|gtaagt) with a MaxEntScan score of 11.08, and an SD score of À2.247 (Figure 3E).RT-PCR showed that exon 9 is exclusively included when the 5 0 splice site was optimally strengthened (Figure 3F).In addition, exon 9 was not skipped even when the SRSF1-or Rbfoxbinding site was mutated in the optimally strengthened minigene construct, indicating that the optimization of the 5 0 splice site made SRSF1 and Rbfox1/2 dispensable (Figure 3F).Thus, the weak 5 0 splice site makes GFPT1 exon 9 being alternatively spliced and being regulated by splicing trans-factors, SRSF1, Rbfox1/2, and hnRNP H/F.(E) Schematic of GFPT1 minigenes carrying wild type (WT) or artificially optimized 5 0 splice site (green letters) at the exon 9/intro 9 boundary.Additional mutations (shown in red letters) were introduced to disrupt the binding sites for either SRSF1 or Rbfox1/2.Respective MaxEntScan::score5ss 49 and SD scores 50 at the 5 0 splice sites are indicated on the right.(F) RT-PCR of minigene constructs shown in (E) transfected in KD3/Hu5 myoblasts.(G) Schematic of 3xMS2 hairpin-attached wild type and mutant RNA probes used for the pull-down of early spliceosome complex.In the mutant probes, the binding sites for SRSF1, Rbfox, and hnRNP H/F were disrupted each.Recombinant Rbfox1 and Rbfox2 were supplemented to the splicing-competent HeLa nuclear extract.Immunoblotting of the U1 snRNP components (U1-70K, U1A, and U1C) assembled on the indicated RNA probes are shown in the bottom panel.(H) Immunoblots of HeLa nuclear extracts depleted for Mock (DMock), SRSF1 (DSRSF1), and hnRNP H/F (DhnRNP H/F).U2AF65 was used as an internal control.(I) Schematic of 3xMS2 hairpin-attached wild-type RNA probe used for the pull-down of early spliceosome complex.As neither Rbfox1 nor Rbfox2 was detected in HeLa nuclear extracts, recombinant Rbfox1 and Rbfox2 were supplemented to the splicing-competent HeLa nuclear extract except for lane 3, which was labeled as DRbfox1/2.Immunoblotting of the U1 snRNP components assembled on the wild-type RNA probe with nuclear extracts depleted for the indicated proteins are shown in the bottom panel.(J) Model of coordinated tissue-specific alternative splicing regulation of GFPT1 exon 9 mediated by SRSF1, Rbfox1/2, and hnRNP H/F.In the undifferentiated myoblasts, expression levels of hnRNP H/F are high, with very low expression levels of Rbfox1 and Rbfox2.In the lack of Rbfox1 and Rbfox2, SRSF1 alone fails to recruit U1 snRNP at the weak 5 0 splice site (ss) at the exon 9/intron 9 border.Additionally, high expression levels of hnRNPs H and F suppress the recruitment of U1 snRNP at the weak 5 0 splice site and produce the GFPT1 short (GFPT1-S) transcript.On the contrary, Rbfox1 and Rbfox2 expression levels are high in differentiated myotubes, with lower expression levels of hnRNPs H and F. SRSF1 and Rbfox1/2 cooperatively enhance the recruitment of U1 snRNP at the weak 5 0 splice site, and thereby ensure the inclusion of exon 9 to produce GFPT1-L transcript in differentiated myotubes.We next analyzed the assembly of U1 snRNPs at the 5 0 splice site in the presence or absence of SRSF1, Rbfox1/2, and hnRNP H/F.We first made four different RNA probes, each harboring exon 9 and its flanking intronic regions attached to a 3xMS2 hairpin loop sequence for pulldown assays.The first probe had the wild-type sequence, whereas the three other probes carried mutations disrupting SRSF1-, Rbfox1/2-, and hnRNP H/F-binding sites, respectively (Figure 3G).We incubated the RNA probes on the MS2 coat protein-coated beads with splicingcompetent HeLa nuclear extract to assemble and pull down the early spliceosome complex at the 5 0 splice site.The HeLa nuclear extract was supplemented with recombinant human Rbfox1 and Rbfox2 (10 ng/mL each), since we could not observe any detectable expression for these proteins in the HeLa nuclear extract.We observed that the association of U1 snRNP proteins (U1-70K, U1A, and U1C) to the RNA probe was markedly reduced when SRSF1-and Rbfox1/2-binding sites were mutated, and markedly enhanced when hnRNP H/F-binding sites were mutated (Figure 3G).
To further dissect the mechanism of U1 snRNP recruitment at the 5 0 splice site, we next depleted each regulatory RBP from the HeLa nuclear extract (Figure 3H), and performed the U1 snRNP pull-down assay using these extracts.Since we could not observe any detectable expression of Rbfox1 and Rbfox2 in HeLa nuclear extract, we added recombinant Rbfox1 and Rbfox2 in the mock-depleted (DMock), SRSF1-depleted (DSRSF1), and hnRNP H/F-depleted (DhnRNP H/F) nuclear extracts.Mock-depleted HeLa extracts without adding recombinant Rbfox1 and Rbfox2 were considered as Rbfox-depleted (DRbfox1/2) nuclear extract.We incubated the wild-type RNA probe with the nuclear extracts and examined the assembly of early spliceosome complex at the 5 0 splice site.As expected, the recruitment of U1 snRNP molecules was markedly reduced in DSRSF1 and DRbfox1/2 nuclear extracts, whereas the recruitment was further enhanced in DhnRNP H/F extract, compared to DMock nuclear extract (Figure 3I).Taken together, SRSF1 and Rbfox1/2 enhance, and hnRNP H/F suppresses, U1 snRNP recruitment at the 5 0 splice site.

Lack of GFPT1-L in skeletal muscle markedly increases GFPT1-S and UDP-GlcNAc, and attenuates glycolysis and TCA cycle, as well as glucose uptake in response to insulin, in skeletal muscle
To examine the roles of specific inclusion of Gfpt1 exon 9 in the mammalian skeletal muscle, we generated a mouse line deficient for Gfpt1 exon 9 by CRISPR/Cas9 system (Figures S8A and S8B).In wild-type mice, the ratios of Gfpt1-L mRNA were 83% in triceps brachii and 35% in heart (Figure S8C), whereas Gfpt1-L mRNA was not detected in either tissue in KO mice (Figure 4A).A marginal difference in predicted molecular weights (2.1 kDa) between Gfpt1-L and Gfpt1-S prevented us from quantifying the ratio of Gfpt1-L and Gfpt1-S at the protein level in wild-type mice.The amount of total Gfpt1 mRNA was increased 1.3-fold in triceps brachii in KO mice but not in the heart or kidney (Figure 4A).In contrast, total Gfpt1 protein in the triceps brachii in KO mice was 4.5-times higher than that in wild-type mice (Figure 4B).Similarly, total Gfpt1 protein in KO mice was 1.8-times higher in the heart, but was similar in the liver, compared to wild-type mice (Figure S8F).Analysis of the amount of total Gfpt1 protein in primary myoblasts from wild-type and KO mice revealed that the amounts of total Gfpt1 protein were similar between the two primary myoblasts (Figure S8E).However, in primary myotubes, the amount of total Gfpt1 protein in KO myotubes was increased $1.5-fold compared to that in wild-type myotubes (Figure S8E).A $4.5-fold increase of the GFPT1-S protein with only a 1.3fold increase of Gfpt1 mRNA could be accounted for by either higher translation efficiency of GFPT1-S than -L, or higher degradation speed of GFPT1-L than -S.We first predicted the structures of GFPT1-L and -S by AlphaFold2.AlphaFold2 appropriately predicted the dimeric structures of human and mouse GFPT1-S (Figure S9A).In contrast, a segment comprised 18 amino acids encoded by exon 9 in GFPT1-L was predicted to be disorganized or failed to be predicted by AlphaFold2 (loops in Figure S9A).This disorganized loop may make GFPT1-L more unstable than GFPT1-S.We thus analyzed the translation efficiencies and stabilities of GFPT1-S and -L, and found that GFPT1-S and -L were translated at similar levels and were degraded at similar speeds after adding cycloheximide (Figure S9B).Taken together, one of the (G) Intracellular energy status (adenylate and guanylate energy charges) and the redox status (NADPH/NADP + and NADH/NAD + ratios) were calculated using the metabolomic data (Table S2) (n = 3 mice each).(H) The malate/aspartate and lactate/pyruvate ratios, which are indirect indicators for NADH/NAD + , were calculated using the metabolomic data (Table S2) (n = 3 mice each).physiological roles of Gfpt1 exon 9 is likely to reduce the Gfpt1 protein level in skeletal muscle.However, neither translation efficiency nor protein stability accounted for the effect of Gfpt1 exon 9. Since total Gfpt1 protein was elevated in skeletal muscle in KO mice, we examined whether HBP was upregulated in KO mice by quantifying the final product of HBP, UDP-GlcNAc, in skeletal muscle by liquid chromatography/tandem mass spectrometry (LC/MS/MS).As predicted, KO mice had 2.8 times more UDP-HexNAc (UDP-GlcNAc plus UDP-GalNAc) than wild-type mice in the triceps brachii (Figure 4C).Dot blots with RL2 and CTD110.6 antibodies for detecting O-glycosylated proteins (Figure 4D) and both O-and N-glycosylated proteins (Fig- ure 4E), respectively, showed $2-times increases of these glycoproteins in skeletal muscle in KO mice.These results suggest that GFPT1-L downregulates O-GlcNAcylation and N-glycosylation in skeletal muscle by attenuating the HBP.
To further analyze the effects of enhanced HBP by lack of Gfpt1 exon 9, we quantified 116 metabolites in the glycolysis pathway and its downstream pathways, which play a pivotal role in energy metabolism, in skeletal muscle by capillary electrophoresis-mass spectrometry (CE-MS) (Table S2).We observed no significant differences between wild-type and KO mice in adenylate and guanylate energy charges, which are indicators of the intracellular energy status (Figure 4G).Although the NADH/NAD + ratio was not changed in KO mice (Figure 4G), its surrogate markers, the lactate/pyruvate and malate/aspartate ratios, were increased in KO mice by 20% and 50%, respectively (Figure 4H).Metabolites in the glycolysis pathway and the tricarboxylic acid (TCA) cycle tended to be decreased in Gfpt1 exon 9 KO mice, while statistically significant decrease was observed only in aspartate ($50% of wild-type mice) (Figure 4F).Asparagine ($70%), methionine ($80%), and threonine ($70%), all of which were downstream to aspartate, were also decreased in KO mice.The urea cycle was also downstream to aspartate, and their metabolites were decreased in KO mice, especially of urea ($50%) and citrulline ($70%) (Figure 4F).Thus, in skeletal muscle in KO mice, the activated HBP suppressed glycolysis and reduced aspartate and its downstream metabolites (asparagine, methionine, threonine, and urea cycle), and increased the malate/aspartate ratio.The physiological significance of the metabolomic changes will be addressed in the discussion.
As HBP and its downstream O-GlcNAcylation pathway serve as mediators for nutrient-sensing, and regulate glucose uptake in response to insulin, [51][52][53] we performed an oral glucose tolerance test.The test showed that the blood glucose level was rapidly increased and slowly decreased, with a significantly higher area-under-the-curve (AUC) in KO mice compared to wild-type mice (Figure 4I).Similarly, the insulin tolerance test showed a compromised response to insulin with a significantly higher AUC of blood glucose in KO mice compared to wildtype mice (Figure 4J).Thus, in Gfpt1 exon 9 KO mice, the compromised response to insulin impaired glucose uptake and increased the blood glucose levels likely through the upregulation of HBP in skeletal muscle.

Aged Gfpt1 exon 9 KO mice show impaired motor performances
As CMS patients with loss-of-function variants in GFPT1 demonstrate a limb-girdle pattern of myasthenia, 10,12,13,54 we analyzed motor functions and the NMJ structures in KO mice.Body weights of KO mice were similar to those of wild-type mice even with aging (Figure 5A).mCT showed that there was no difference in the areas of paravertebral skeletal muscle and fat between wild-type and KO mice at age 12 months (Figure S10A).Similarly, the wet weights of five skeletal muscles and three fat tissues showed no difference between the two groups of mice (Figure S10B).In contrast, motor performance of KO mice evaluated by the rotarod test was similar to wild-type mice up to age 6 months, but was gradually declined from the age 9 months with aging (Figure 5B).Similarly, an inverted screen test showed no difference at age 6 months, but was declined from age 9 months in KO mice, though statistical significance was not observed (Figure 5C).We next asked whether the impaired motor functions in aged KO mice were due to defects at the NMJ, as were observed in GFPT1-CMS patients.RNA-seq analysis of skeletal muscles at age 13 months showed that the expression levels of 22 essential genes at the NMJ, which are often defective in CMS, were similar between wild-type and KO mice (Figure S10C).Similarly, no tubular aggregates or centrally located nuclei were observed in myofibers in aged KO mice (Figures S10D and S10E).In contrast, in triceps brachii, gastrocnemius, and rectus femoris muscles of KO mice at ages 12 to 14 months, AChR clusters were fragmented (Figure 5D).Quantitative analyses showed that the AChR cluster areas were significantly reduced (Figure 5E), and the average number of fragmented AChR clusters per motor endplate was significantly increased (Figure 5F), in all the three skeletal muscles in KO mice.The ultrastructures of the NMJ at ages 12 to 14 months showed small nerve terminals with simplified junctional folds (Figure 5G).Quantitative analyses revealed that both the number of junctional folds and the length of the postsynaptic membrane were markedly reduced in KO mice compared to wild-type mice (Figures 5H and 5I).[14][15][16]

DISCUSSION
The striated muscle-specific splicing isoform of GFPT1 (GFPT1-L) was first reported in human and mouse in 2001, 8,9 and later in pig in 2010. 55owever, the splicing regulation of GFPT1-L in skeletal muscle remained to be elucidated.The 5 0 splice sites are recognized by direct basepairing with the 5 0 end of the U1 small nuclear RNA (snRNA), although recognition is not fully dependent on this pairing. 56The 5 0 splice site is composed of the last three nucleotides of an exon (positions À3 to À1) and the first six nucleotides of an intron (positions +1 to +6).The consensus sequence of the 5 0 splice site is (C/A)AG|gu(a/g)ag(u/g)a, where a vertical line represents an exon-intron boundary. 50,56As  (A, B, C, E, F, H, and I) Mean and SEM are indicated.Unpaired Student's t test was applied to (C, H, and I).two-way ANOVA with posthoc Tukey test was applied to (A, B, E, and F).*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.observed in other alternatively spliced exons, the 5 0 splice site of GFPT1 exon 9 (GAG|guggga) has a weak splicing signal with non-consensus nucleotides at positions À3, +4, and +6 (Figure 3).Complex interactions between splicing cis-elements (enhancers and silencers) and trans-acting RBPs determine how and when the alternative splice sites are activated. 57Splicing-enhancing and suppressing activities of RBPs are often determined in a binding position-specific manner. 58In general, SR proteins enhance and suppress splicing at the 5 0 splice site, when they are recruited to an exon and an intron, respectively. 58In contrast, hnRNPs enhance and suppress splicing at the 5 0 splice site, when they are recruited to an intron and an exon, respectively. 58Thus, the position-dependent effects are opposite between SR proteins and hnRNPs.For GFPT1 exon 9, the splicing-enhancing effect of SRSF1 bound to exon 9 was consistent with the general position-dependency of SR proteins (Figure 3J).Similarly, the splicing-suppressive effect of hnRNP H/F bound to exon 9 was also consistent with the general rule of hnRNPs (Figure 3J).Splicing activation and suppression by SR proteins and hnRNPs are likely to be substantiated in the early spliceosome complex.SRSF1 simultaneously recognizes an exonic splicing enhancer and U1-70K to recruit U1 snRNP at the 5 0 splice site, where SRSF1 directly interacts with U1-70K by their respective RNA recognition motif (RRM) domains. 59We indeed showed that U1 snRNPs including U1-70K were present in the early spliceosome complex at the 5 0 splice site of GFPT1 exon 9 (Figures 3G-3I).Disruption of the SRSF1-binding site and depletion of SRSF1 protein abrogated, whereas disruption of the hnRNP H/F-binding sites and depletion of hnRNP H/F enhanced, the formation of the early spliceosome complex at the 5 0 splice site of GFPT1 exon 9 (Figures 3G-3I).
Rbfox proteins are neuron-and muscle-specific RBPs that either enhance or suppress splicing.1][62][63][64][65] A representative example of splicing enhancement by Rbfox1/2 is observed in exon 16 of EPB41 encoding the erythrocyte membrane protein band 4.1 in erythroid differentiation, in which binding of Rbfox2 to (U)GCAUG elements in intron 16 recruits U1 snRNP and promotes splicing of exon 16 33,65 .The authors show that Rbfox2 directly interacts with U1 snRNP-associated proteins, U1C and U1-70K but not U1A.U1C is required for a stable interaction between the pre-mRNA 5 0 splice site and U1 snRNP, 66 although U1C doesn't directly bind to U1 snRNA.The binding of U1C to the U1 snRNP core domain is mediated by U1-70K through its N-terminal domain. 67,68The N-terminal cysteine/histidine-rich zinc finger-like domain of U1C interacts with the duplex between pre-mRNA and the 5 0 -end of U1 snRNA, even though it makes no base-specific contacts with pre-mRNA. 69The very same domain of U1C is necessary for its interaction with the C-terminal domain of Rbfox2. 65U1C thus fine-tunes the affinity to stabilize the binding of U1 snRNA to non-canonical nucleotides at the 5 0 splice site.Our findings support the notion that Rbfox1/2 enhances the role of U1C to stabilize the association between the 5 0 splice site and U1snRNP (Figure 3).Rbfox1/2 is preferentially expressed in skeletal and cardiac muscles, as well as various neuronal regions. 70In addition, Rbfox1/2 is alternatively spliced in muscles to promote muscle-specific splicing. 70-72SRSF1 and hnRNP H/F are abundantly expressed across different cell and tissue types. 73Our data showed that the expressions of Rbfox1/2 were upregulated while those of hnRNP H/F were downregulated during myogenic differentiation (Figures 3A-3D).These results suggest that simultaneous upregulation of Rbfox1/2 and downregulation of hnRNP H/F contribute to the specific inclusion of GFPT1 exon 9 in skeletal muscles.
The role of GFPT in biosynthesis of glucosamine-6-P from fructose-6-P was first reported by Ghosh et al. 4 The phylogenetic tree indicates that the first speciation of GFPTs occurred between prokaryotes and eukaryotes, suggesting that prokaryotic and eukaryotic GFPTs can have different functional properties.One such example is that mammalian GFPT is more sensitive to a feedback inhibition by UDP-GlcNAc than prokaryotic GFPT. 74Eukaryotic GFPT was split into the invertebrate and vertebrate orthologs, and the vertebrate GFPT was then split into GFPT1 and GFPT2 (Figure S2).Invertebrate GFPTs were more similar to vertebrate GFPT2s than vertebrate GFPT1s (Figure S3), suggesting that GFPT1 was likely to have evolved to exert novel functions in vertebrates.GFPT1 and GFPT2 are $76% identical at the amino acid level, 5 but their tissue expression profiles are different.GFPT1 is highly expressed in the heart, skeletal muscle, placenta, pancreas, and testis, whereas GFPT2 is more expressed in the central nervous system where GFPT1 is less expressed. 5,7GFPT1 exon 9 and its flanking introns are conserved in mammals (Figure S4), and GFPT1-L is uniquely expressed in skeletal muscle and heart only in mammals (Figure 1D).GFPT1-L has a lower maximum enzymatic activity ($two-times lower Vmax) and a lower substrate affinity ($two-times higher K M ) than GFPT1-S, and is more susceptible to UDP-GlcNAc inhibition ($five-times lower Ki) than GFPT1-S. 8,9he primary role of UDP-GlcNAc in the HBP is for N-and O-glycosylation of biological molecules, but UDP-GlcNAc also exerts the other essential roles in cell metabolisms.Continuous intravenous administration of glucosamine under an euglycemic hyperinsulinemic clamp increased the UDP-GlcNAc concentration $4-fold in skeletal muscle in rat, and downregulated the expression of gene sets in fatty acid oxidation, mitochondrial substrate shuttles, and TCA cycle. 75The authors showed that glucosamine significantly decreased whole-body oxygen consumption and energy expenditure.The suppression of the glycolysis pathway and the TCA cycle in Gfpt1 exon 9 KO mice (Figure 4F) might also be accounted for by a $2.8-fold increase of UDP-GlcNAc concentration in skeletal muscle (Figure 4C).In addition to the suppressed glycolysis pathway and TCA cycle, we observed that aspartate and its downstream metabolites were markedly reduced (Figure 4F), which was not addressed in the glucosamine-administered rats. 75Cytoplasmic aspartate is converted to malate by accepting electrons from NADH.Malate is then transferred to mitochondria via the malate-aspartate shuttle, and serves as an electron source for the mitochondrial electron transport chain and a carbon source for the TCA cycle.Activation of the malate-aspartate shuttle was thus likely to compensate for the suppressed glycolytic energy production.Gfpt1 exon 9 KO mice increased UDP-GlcNAc $2.8-fold (Figure 4C), but adenylate and guanylate energy charges were preserved (Figure 4G).7][78] In contrast, administration of glucosamine increased UDP-GlcNAc $4-fold and reduced the ATP level. 75This was likely because a feedback inhibition to the GFPT1 activity was operational in Gfpt1 exon 9 KO and the overexpression of GFPT1. 75In contrast, glucosamine enters the HBP downstream of GFPT1 (Figure 1D), and a feedback inhibition on GFPT1 should have no effect on the production of UDP-GlcNAc.Indeed, glucosamine increased UDP-GlcNAc $4-fold, whereas our model and the overexpression of GFPT1 [76][77][78] increased it $2.8and $2-fold, respectively.Mammals were thus likely have evolved to minimize the production of UDP-GlcNAc by making GFPT1-L for sufficient energy production via glycolysis and TCA cycle in skeletal muscle.
Loss-of-function variants of GFPT1 cause limb-girdle CMS with tubular aggregates in skeletal muscles, which is characterized by slowly progressive limb-gridle muscle weakness with minor or no involvement of facial, ocular, and bulbar muscles. 10,12,13,54GFPT1-CMS patients show decreased GFPT1 protein levels in skeletal muscle, 10,54 decreased O-GlcNAcylated proteins in skeletal muscles, 10,12 and defective neuromuscular signal transmission with decreased AChR. 10,12In accordance with GFPT1-CMS patients, 10,12,13,54 muscle-specific Gfpt1 KO mice exhibit simplified NMJ with small and fragmented AChR clusters. 17In contrast to GFPT1-CMS patients 10,12,13,54 and muscle-specific Gfpt1 KO mice, 17 however, Gfpt1 exon 9 KO mice increased the amounts of GFPT1, as well as of O-and N-glycosylated proteins (Figures 4B-4D).Although the directions of deviations of GFPT1 activities were opposite, simplified NMJs in aged Gfpt1 exon 9 KO mice were similar to, but were less severe than, those in GFPT1-CMS patients 10,12,13,54 and muscle-specific Gfpt1 KO mice 17 (Figure 5D).Similarly, the inverted screen test showed that muscle-specific Gfpt1 KO mice showed markedly shortened dwell time as early as age 6 weeks, 17 whereas Gfpt1 exon 9 KO mice remained normal at age 6 months and tended to decline from age 9 months (Figure 5C).Muscle-specific Gfpt1 KO mice showed muscle fatigue by in situ force measurement of TA muscle at age 3 months. 17Although we did not directly measure muscle fatigue in situ, the evaluation of the muscle strength and muscle fatigue by the accelerated rotarod test and the inverted screen test showed that abnormal muscle fatigue is unlikely to be present at age 6 months in Gfpt1 exon 9 KO mice (Figures 5B and 5C).Tubular aggregates and muscle regeneration with centrally located nuclei, which are frequently observed in GFPT1-CMS patients 10,12,13,54 and muscle-specific Gfpt1 KO mice, 17 were not observed in Gfpt1 exon 9 KO mice (Figures S10D and S10E).Additionally, unlike muscle-specific Gfpt1 KO mice, 17 AChR subunit genes (Chrna1, Chrnb1, Chrnd, and Chrne) were not downregulated and Musk was not upregulated in Gfpt1 exon 9 KO mice (Figure S10C).All these differences point to the notion that Gfpt1 exon 9 KO mice have milder NMJ and muscle phonotypes than muscle-specific Gfpt1 KO mice. 17n GFPT1-CMS, four pathogenic variants affecting GFPT1 exon 9 have been reported.First, NM_001244710.2:c.686-2A> G at the intron 8/exon 9 boundary activates a cryptic splice site and eliminates the first four nucleotides of exon 9 predicting 56 missense amino acids followed by a stop codon. 12Second, NM_001244710.2:c.686dupCpredicts NP_001231639.1:p.R230X. 16Third, NM_001244710.2:c.706> T predicts NP_001231639.1:p.K236X. 79Fourth, NM_001244710.2:c.719G> A predicts NP_001231639.1:p.W240X. 10 All the four pathogenic variants generate a stop codon, and O-and N-glycosylated proteins were markedly reduced in a patient with c.686-2A > G. 12 In contrast to Gfpt1 exon 9 KO mice, the four patients are likely to have reduced GFPT1 in skeletal muscle.If a pathogenic splicing variant causes skipping of GFPT1 exon 9, the patient phenotypes are predicted to be similar to those in Gfpt1 exon 9 KO mice.
In 1996, increased enzymatic activities of GFPT1 were reported in skeletal muscle in patients with NIDDM. 25Similarly, in insulin-resistant obese mice, the enzymatic activity of GFPT1 was elevated in the skeletal muscle but not in the liver. 80Conversely, overexpression of GFPT1-S in skeletal muscle, 26 fat, 26,81 liver, 77 and pancreatic b cells, 78 impaired insulin-mediated glucose update in mice.In accordance with these mouse studies, glucose uptake in response to insulin was impaired in aged Gfpt1 exon 9 KO mice that had $4.5-fold more GFPT1-S in skeletal muscle (Figure 4B).Preservation of the insulin responsiveness is likely to be one of the essential roles of inclusion of Gfpt1 exon 9 in skeletal muscle.Additionally, type 1 diabetic mice and non-obese diabetic mice show NMJ dysfunctions with reduced acetylcholinesterase (AChE) and AChR fragmentations at the NMJ, [82][83][84] as observed in Gfpt1 exon 9 KO mice.Thus, the abnormal NMJ phenotypes in Gfpt1 exon 9 KO mice may involve a pathway similar to that observed in the diabetic mice.In both mouse models, impaired glucose metabolisms may cause the abnormal NMJ phenotypes, although the exact mechanisms still remain unknown.Taken together, GFPT1 is likely to act as a doubleedged sword in skeletal muscle, and the finely tuned HBP activity in skeletal muscle is essential for glucose metabolisms, as well as for the formation and maintenance of the NMJ.

Limitations of the study
We have shown cooperative splicing regulation of GFPT1 exon 9 by SRSF1, Rbfox1/2, and hnRNP H/F, as well as the roles of GFPT1-L in glucose metabolisms and the NMJ.The primary benefit of acquisition of GFPT1-L in evolution in skeletal muscle in mammals is likely to suppress the HBP for efficient glycolytic energy production in skeletal muscle, as well as the formation and maintenance of the NMJ.However, we could not identify the mechanisms of $4.5-fold increase of GFPT1 in the skeletal muscle in Gfpt1 exon 9 KO mice.Increased UDP-GlcNAc compromised the glucose uptake and the energy production via glycolysis and TCA cycle in skeletal muscle in Gfpt1 exon 9 KO mice.Similar associations have been repeatedly reported in diabetes mellitus in humans 25 and mice. 26,27,77,78,80,81However, the exact underlying mechanisms remain to be elucidated in either study.Additionally, our study was limited to male mice, leaving the comparison to female mice unaddressed.Abnormal NMJ formation is observed in diabetic mice [82][83][84] and in Gfpt1 exon 9 KO mice.However, the causal relations remain to be solved in both diabetic and our mice.Further studies are required to elucidate the hidden scenarios played by GFPT1-L in skeletal muscle.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Immunoblotting of cultured cells and muscle tissues
KD3/Hu5 cells were washed in PBS with 13Protease Inhibitor Cocktail (Thermo Fisher Scientific), followed by centrifugation at 2,000 3g for 5 min.The pellets were resuspended in the lysis buffer [10 mM HEPES-NaOH (pH 7.8), 0.1 mM EDTA, 10 mM KCl, 1 mM DTT, 0.5 mM PMSF, 0.1% NP-40, 13 Protease Inhibitor Cocktail] and kept on ice for 10 min.Following sonication, samples were centrifuged at 20,000 3g for 10 min to remove cell debris, and supernatants were collected as total cell lysates for immunoblotting.Muscle, heart, and liver tissues were homogenized in GFPT buffer (50 mM KH 2 PO 4 , 10 mM EDTA, 5 mM reduced L-glutathione, 12 mM D-glucose-6-phosphate Na 2 , 10 mM PMSF, 1 mM pepstatin A, and 10 mM phosphatase inhibitor [pH 7.6]) 10 using a disposable homogenizer (BioMasher II, Funakoshi).Samples were mixed for 30 min on the rotator at 4 C, sonicated 4 times for 10 s using an ultrasonic disruptor (UR-21P, Tomy Digital Biology), and centrifuged at 17,900 3g for 20 min.The supernatant was collected, and protein concentration was measured with the Pierce 660 nm Protein Assay Reagent (22660, Thermo Fisher Scientific) according to the manufacturer's instructions.The supernatant was incubated with an equal volume of 23SDS Sample Buffer [0.125 M Tris-HCl (pH 6.8), 4% (w/v) SDS, 20% (v/v) glycerol, and 0.01% (w/v) bromophenol blue] at 95 C for 5 min, and subjected to immunoblot analyses.Coomassie gel staining was performed to normalize the loading amount in Western blots.
Mouse primary myoblasts and HEK293 cells were washed in PBS with 13Protease Inhibitor Cocktail (Thermo Fisher Scientific), followed by centrifugation at 2,000 3g for 5 min.The pellets were resuspended in the lysis buffer (50 mM HEPES, 150 mM NaCl, 1.5 mM MgCl2, 1 mM EGTA, 100 mM NaF, 10 mM sodium pyrophosphate, 10% glycerol, 1% Triton X-100, 10 mM PMSF, 1 mM pepstatin A, and 10 mM phosphatase inhibitor [pH 7.0]) and mixed for 1 h on the rotator at 4 C. Samples were centrifuged at 20,000 3g for 10 min to remove cell debris, and supernatants were collected as total cell lysates for immunoblotting.
Proteins were separated by electrophoresis on a 7.5% SDS-polyacrylamide gel in Tris-Glycine and transferred onto a polyvinylidene difluoride membrane (PVDF, Immobilon-P, 0.45 or 0.2 mm, Merck Millipore).The membrane was washed in Tris-buffered saline containing 0.05% Tween 20 (TBS-T) and blocked for 1 h at room temperature in TBS-T with 3% bovine serum albumin (BSA) or 5% Non-Fat Dry Milk (NFDM).The membrane was incubated overnight at 4 C with primary antibody.The membranes were washed 3 times for 10 min with TBS-T and incubated with a secondary antibody for 1 h at room temperature.The blots were visualized using Amersham ECL Western blotting detection reagents and quantified with ImageJ software.

LC/MS analysis for quantification of HexNAc
Muscle tissues (20-40 mg) were weighed and quickly frozen in liquid nitrogen.Cellular extracts for nucleotide sugar analysis were prepared as reported previously. 94Hydrophilic interaction liquid chromatography and electrospray tandem mass spectrometry (HILIC-ESI-MS/MS) was performed on an LCMS-8060 (Shimadzu) coupled with a Nexera HPLC system (Shimadzu).Chromatography was performed on a BEH-amido column (2.1 mm i.d.x 150 mm, 3 mm; Waters). 95,96Analysis of nucleotide sugars was conducted in the multiple reaction monitoring mode using specific precursor ion [M-H] -and product ions pairs as follows: UDP-HexNAc, m/z 606.1/384.7.The nucleotide sugar levels were normalized as pmol/mg tissues.

Immunodot blot assay of O-GlcNAc of skeletal muscle
The Dot Microfiltration apparatus (SCIE-PLAS, DHM48) was used to analyze the O-GlcNAc level in the lysates of skeletal muscle.A nitrocellulose membrane (0.45 mm pore size, BIO-RAD) was placed in the apparatus, and protein extracts (0.8 mg in 2 mL) were transferred to the membranes by water vacuum for 20 min.The membrane was incubated with blocking solution containing 3% BSA in Tris-buffered saline

Figure 1 .Figure 2 .
Figure 1.GFPT1 exon 9 is specifically spliced-in in striated muscles only in mammals (A) Schematic of the genomic structure of human GFPT1 gene and the domain structure of GFPT1 protein.Exons are shown in boxes and introns are shown in solid lines.The 5 0 and 3 0 untranslated regions (UTRs), constitutive exons, and alternatively spliced exon 9 are indicated in black, blue, and yellow, respectively.GATase, glutamine amidotransferase; SIS, sugar isomerase.(B) RT-PCR of endogenous GFPT1 in human tissues.Splice variants are schematically shown on the right side.(C and D) Representative phase-contrast images of differentiating immortalized human KD3/Hu5 myoblasts.RT-PCR showing alternative splicing of endogenous GFPT1 exon 9 at different days (D) of differentiating KD3/Hu5 myoblasts.Expressions of myogenin (MyoG), myogenic differentiation 1 (MyoD1) and GAPDH are shown as internal controls.(D) Schematic of the hexosamine biosynthetic pathway (HBP).A key regulator of the HBP, GFPT, catalyzes the first and rate-limiting step.OGT, O-GlcNAc transferase; OGA, O-GlcNAcase, DPAGT1, dolichyl-phosphate N-acetylglucosaminephosphotransferase 1; and ALGs, asparagine-linked glycosylation homologs.(E) Calculation of percent spliced-in (PSI) of GFPT1genes in 11 tissues in 4 mammals in RNA-seq in the public database.Note that GFPT1 exon 9 and its flaking introns are conserved only in mammals (Figure S4).Skeletal muscle and heart are shown in red.Each PSI in public database is indicated by a circle, and the circle size represents the number of reads in RNA-seq.Mean and SD are indicated.

Figure 3 .
Figure 3. SRSF1 and Rbfox1/2 enhance, and hnRNP H/F suppress, the recruitment of U1 snRNP components at the 5 0 splice site of GFPT1 exon 9 (A-C) qRT-PCR to quantify transcripts for (A) SRSF1, (B) RBFOX1 and RBFOX2, and (C) HNRPH1 and HNRNPF in the course of myogenic differentiation of KD3/ Hu5 myoblasts.Mean and SD are indicated (n = 3).p values by one-way ANOVA are indicated below the gene names.*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 by Sidak posthoc test.(D) Immunoblots of endogenous SRSF1, Rbfox1, Rbfox2, hnRNP H, and hnRNP F in the course of myogenic differentiation of KD3/Hu5 myoblasts.(E)Schematic of GFPT1 minigenes carrying wild type (WT) or artificially optimized 5 0 splice site (green letters) at the exon 9/intro 9 boundary.Additional mutations (shown in red letters) were introduced to disrupt the binding sites for either SRSF1 or Rbfox1/2.Respective MaxEntScan::score5ss49 and SD scores50 at the 5 0 splice sites are indicated on the right.(F) RT-PCR of minigene constructs shown in (E) transfected in KD3/Hu5 myoblasts.(G) Schematic of 3xMS2 hairpin-attached wild type and mutant RNA probes used for the pull-down of early spliceosome complex.In the mutant probes, the binding sites for SRSF1, Rbfox, and hnRNP H/F were disrupted each.Recombinant Rbfox1 and Rbfox2 were supplemented to the splicing-competent HeLa nuclear extract.Immunoblotting of the U1 snRNP components (U1-70K, U1A, and U1C) assembled on the indicated RNA probes are shown in the bottom panel.(H) Immunoblots of HeLa nuclear extracts depleted for Mock (DMock), SRSF1 (DSRSF1), and hnRNP H/F (DhnRNP H/F).U2AF65 was used as an internal control.(I) Schematic of 3xMS2 hairpin-attached wild-type RNA probe used for the pull-down of early spliceosome complex.As neither Rbfox1 nor Rbfox2 was detected in HeLa nuclear extracts, recombinant Rbfox1 and Rbfox2 were supplemented to the splicing-competent HeLa nuclear extract except for lane 3, which was labeled as DRbfox1/2.Immunoblotting of the U1 snRNP components assembled on the wild-type RNA probe with nuclear extracts depleted for the indicated proteins are shown in the bottom panel.(J) Model of coordinated tissue-specific alternative splicing regulation of GFPT1 exon 9 mediated by SRSF1, Rbfox1/2, and hnRNP H/F.In the undifferentiated myoblasts, expression levels of hnRNP H/F are high, with very low expression levels of Rbfox1 and Rbfox2.In the lack of Rbfox1 and Rbfox2, SRSF1 alone fails to recruit U1 snRNP at the weak 5 0 splice site (ss) at the exon 9/intron 9 border.Additionally, high expression levels of hnRNPs H and F suppress the recruitment of U1 snRNP at the weak 5 0 splice site and produce the GFPT1 short (GFPT1-S) transcript.On the contrary, Rbfox1 and Rbfox2 expression levels are high in differentiated myotubes, with lower expression levels of hnRNPs H and F. SRSF1 and Rbfox1/2 cooperatively enhance the recruitment of U1 snRNP at the weak 5 0 splice site, and thereby ensure the inclusion of exon 9 to produce GFPT1-L transcript in differentiated myotubes.

Figure 4 .
Figure 4. Gfpt1 exon 9 is required for down-regulation of O-GlcNAcylation and N-glycosylation in skeletal muscle to regulate glucose uptake in response to insulin (A) Representative and quantitative RT-PCR of endogenous Gfpt1 in indicated tissues of wild type and Gfpt1 exon 9 KO mice at age 12 months (n = 4 mice each).Splice variants are schematically shown on the right side.(B) Representative immunoblots and quantification of endogenous Gfpt1 (both short and long isoforms, S/L) in triceps brachii of wild type and Gfpt1 exon 9 KO mice at age 12 months (n = 7 mice each).(C) LC/MS/MS analysis of UDP-HexNAc (UDP-GlcNAc plus UDP-GalNAc) in triceps brachii of wild type and Gfpt1 exon 9 KO mice (n = 6 mice each).(D) Triplicated blots and quantification of O-GlcNAcylated proteins probed with RL2 antibody in triceps brachii of wild-type and Gfpt1 exon 9 KO mice (n = 4 mice each).(E) Triplicated blots and quantification of O-GlcNAcylated and N-glycosylated proteins probed with CTD110.6 antibody in triceps brachii of wild type and Gfpt1 exon 9 KO mice (n = 4 mice each).(F) Metabolomic analysis of triceps brachii of wild type and Gfpt1 exon 9 KO mice at age 13 months (n = 3 mice each).Pathway diagram of glycolysis, TCA cycle, and urea cycle is indicated with the concentration of each metabolite.(G)Intracellular energy status (adenylate and guanylate energy charges) and the redox status (NADPH/NADP + and NADH/NAD + ratios) were calculated using the metabolomic data (TableS2) (n = 3 mice each).(H) The malate/aspartate and lactate/pyruvate ratios, which are indirect indicators for NADH/NAD + , were calculated using the metabolomic data (TableS2) (n = 3 mice each).(I and J) Glucose tolerance test of wild-type mice (n = 4) and Gfpt1 exon 9 KO mice (n = 5) at age 12 months.The area under the curves (AUCs) are plotted on the right.(J) Insulin tolerance tests of wild-type mice (n = 6) and Gfpt1 exon 9 KO mice (n = 5) at age 12 months.The AUCs are plotted on the right.(A-J) Mean and SEM are indicated.Unpaired Student's t test was applied to (B, C, D, E, H, I) (AUC), and J (AUC).Welch's t-test was applied to F. two-way ANOVA with posthoc Tukey test was applied to (G and I) (temporal profile), and J (temporal profile).*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Figure 5 .
Figure 4. Gfpt1 exon 9 is required for down-regulation of O-GlcNAcylation and N-glycosylation in skeletal muscle to regulate glucose uptake in response to insulin (A) Representative and quantitative RT-PCR of endogenous Gfpt1 in indicated tissues of wild type and Gfpt1 exon 9 KO mice at age 12 months (n = 4 mice each).Splice variants are schematically shown on the right side.(B) Representative immunoblots and quantification of endogenous Gfpt1 (both short and long isoforms, S/L) in triceps brachii of wild type and Gfpt1 exon 9 KO mice at age 12 months (n = 7 mice each).(C) LC/MS/MS analysis of UDP-HexNAc (UDP-GlcNAc plus UDP-GalNAc) in triceps brachii of wild type and Gfpt1 exon 9 KO mice (n = 6 mice each).(D) Triplicated blots and quantification of O-GlcNAcylated proteins probed with RL2 antibody in triceps brachii of wild-type and Gfpt1 exon 9 KO mice (n = 4 mice each).(E) Triplicated blots and quantification of O-GlcNAcylated and N-glycosylated proteins probed with CTD110.6 antibody in triceps brachii of wild type and Gfpt1 exon 9 KO mice (n = 4 mice each).(F) Metabolomic analysis of triceps brachii of wild type and Gfpt1 exon 9 KO mice at age 13 months (n = 3 mice each).Pathway diagram of glycolysis, TCA cycle, and urea cycle is indicated with the concentration of each metabolite.(G)Intracellular energy status (adenylate and guanylate energy charges) and the redox status (NADPH/NADP + and NADH/NAD + ratios) were calculated using the metabolomic data (TableS2) (n = 3 mice each).(H) The malate/aspartate and lactate/pyruvate ratios, which are indirect indicators for NADH/NAD + , were calculated using the metabolomic data (TableS2) (n = 3 mice each).(I and J) Glucose tolerance test of wild-type mice (n = 4) and Gfpt1 exon 9 KO mice (n = 5) at age 12 months.The area under the curves (AUCs) are plotted on the right.(J) Insulin tolerance tests of wild-type mice (n = 6) and Gfpt1 exon 9 KO mice (n = 5) at age 12 months.The AUCs are plotted on the right.(A-J) Mean and SEM are indicated.Unpaired Student's t test was applied to (B, C, D, E, H, I) (AUC), and J (AUC).Welch's t-test was applied to F. two-way ANOVA with posthoc Tukey test was applied to (G and I) (temporal profile), and J (temporal profile).*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Figure 5 .
Figure 5. Continued (H and I) Blinded morphometric analyses of the number of postsynaptic folds (H) and the length of postsynaptic membrane normalized to the nerve terminal area (I) in wild type and Gfpt1 exon 9 KO mice (n = 20-26 NMJs in 4-5 mice each).(A, B, C, E, F, H, and I) Mean and SEM are indicated.Unpaired Student's t test was applied to (C, H, and I).two-way ANOVA with posthoc Tukey test was applied to (A, B, E, and F).*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

TABLE
d RESOURCE AVAILABILITY B Lead contact B Materials availability