A Genetic Model to Study Increased Hexosamine Biosynthetic Flux

Abstract

Recently, we identified harvest moon (hmn), a fully penetrant and expressive recessive zebrafish mutant with hepatic steatosis. Larvae showed increased triacylglycerol in the absence of other obvious defects. When we attempted to raise these otherwise normal-appearing mutants to adulthood, we observed a developmental arrest and death in the early juvenile period. In this study, we report the positional cloning of the hmn locus and characterization of the defects caused by the mutation. Using bulk segregant analysis and fine mapping, we find that hmn mutants harbor a point mutation in an invariant residue within the sugar isomerase 1 domain of the gene encoding the rate-limiting enzyme of the hexosamine biosynthetic pathway (HBP) glutamine-fructose-6-phosphate transamidase (Gfpt1). The mutated protein shows increased abundance. The HBP generates β-N-acetyl-glucosamine (GlcNAc) as a spillover pathway from glucose. GlcNAc can be O-linked to seryl and threonyl residues of diverse cellular proteins (O-GlcNAc modification). Although some of these O-GlcNAc modifications serve an essential structural role, many others are dynamically generated on signaling molecules, including several impacting insulin signaling. We find that gfpt1 mutants show global increase in O-GlcNAc modification, and, surprisingly, lower fasting blood glucose in males. Taken together with our previously reported work, the gfpt1 mutant we isolated demonstrates that global increase in O-GlcNAc modification causes some severe insulin resistance phenotypes (hepatic steatosis and runting) but does not cause hyperglycemia. This animal model will provide a platform for dissecting how O-GlcNAc modification alters insulin responsiveness in multiple tissues.

An activating mutation in the gfpt1 gene was identified in a zebrafish strain found to have complex phenotypes suggestive of altered insulin signaling.

Previously, we identified hmn, a zebrafish mutant with two distinguishing phenotypes: hepatic steatosis at a point in development when maternally deposited yolk lipids are exhausted, and developmental arrest in the late larval period (1). Although we were initially interested in identifying mutants with hepatic steatosis to isolate new genes involved in the liver’s regulation of the transition from fed to fasted states (2), we found that neither feeding nor fasting ameliorated or exacerbated the developmental arrest of hmn mutants (1).

In this study, we found that neither feeding nor fasting altered the hepatic steatosis phenotype of hmn mutants. This finding suggested that a permanent defect in metabolic regulation was present in the mutants. Thus, we positionally cloned and functionally characterized the hmn locus. We found a recessive activating mutation in the gfpt1 gene that increases enzymatic flux through the hexosamine biosynthetic pathway (HBP). Because it is well established that increased HBP flux causes multiorgan insulin resistance (IR), the gfpt1 mutant we have isolated could serve as a useful tool for revealing the molecular mechanisms of IR.

Materials and Methods

Zebrafish mutagenesis, screening, and histological analyses

Animal studies were approved by the Institutional Animal Care and Use Committee of the University of Utah. The hmn^z110 mutant was identified, as we described previously (1). In brief, F2 larvae from males mutagenized with N-ethyl-N-nitorosurea were fixed and stained with Oil Red O (ORO) 7 days postfertilization (dpf) and scored for the presence of hepatic steatosis. The agtpbp1^sa17482 allele was purchased from the Zebrafish International Resource Center (3). ORO staining and histological analyses were done exactly as described previously; samples were stained with Sudan Black and eosin (4).

Positional cloning

A map cross of heterozygous hmn carriers to the polymorphic WIK line was performed (5). A single pair of heterozygous progeny was bred to generate all the individuals used to positionally clone the locus. First, larvae were fixed and stained with ORO and then sorted into two pools. DNA from 21 wild-type (WT) and 21 hmn mutant larvae were examined by bulk segregant analysis using a standard set of simple sequence-length polymorphism microsatellite markers (z markers) on a meiotic map (6, 7). This allowed us to assign the mutated gene to the south subtelomeric region of chromosome 8. Examination of 51 additional progeny from this single pair allowed us to narrow the locus further to a single contig. Complementary DNAs (cDNAs) were cloned for all protein-coding genes in the locus and the indicated genes just outside the locus, and sequenced. Genomic DNA was obtained from individual 7-dpf larvae after fixation and staining in ORO by digesting at 65°C with proteinase K (10 μg/mL final) in 10 mM Tris, pH 9, 50 mM KCl, 1.5 mM MgCl₂, 0.3 Tween 20, and 0.3% IGEPAL-CA630. Following heat inactivation of proteinase K, DNA was diluted 1:10 in water and used in polymerase chain reaction–based genotyping (4).

cDNA synthesis and sequencing

To extract RNA, 7-dpf larvae were sorted following live staining with Nile Red (8). Total RNA was extracted by homogenization with a tissue grinder into Qiagen (Germantown, MD) RNeasy MINI lysis buffer (RTL with 2-mercaptoethanol) and purified, following the manufacturer’s protocol, for tissue samples. cDNA was prepared using RNA to cDNA EcoDry Premix (Takara Clontech, Mountainview, CA). Two forward and two reverse primers were used in all possible combinations to successfully amplify the gfpt1 message: forward primer, 5′-CGCCTTCTTGACCAATACCA-3′ and reverse primer, 5′- GGTGTTGGTCAGGAAGCATT-3′. After purification of the products, GoTaq (Promega) was added for one cycle to create A-overhangs for T-A cloning into pGEMT (Promega, Madison, WI). SP6 and T7 primers were used for Sanger sequencing of cloned cDNAs.

Immunoblot analyses

Immunoblot analyses were performed exactly as we described previously using dissected zebrafish livers (4). Rabbit anti-human GFPT1 was from Proteintech (Rosemont, IL). Mouse monoclonal IgG RL2 against O-β-N-acetyl-glucosamine (GlcNAc) modification (9) was purchased from Abcam (Cambridge, MA). Rabbit anti–β-tubulin (Tubb) was purchased from Abcam. Catalog numbers, dilutions, and RRIDs are shown in Table 1.

Table 1.

Antibodies Used

Peptide/Protein Target	Antigen Sequence (if Known)	Name of Antibody	Manufacturer, Catalog No., and/or Name of Individual Providing the Antibody	Species Raised in; Monoclonal or Polyclonal	Dilution Used	RRID
Gfpt1	Human GFPT1	Gfpt1	Proteintech, 14132-1-AP	Rabbit; polyclonal	1 to 1000	AB_2110155
O-GlcNAc modification	O-GlcNAc	RL2	Abcam, ab2730	Mouse; monoclonal	1 to 1000	AB_531200
β-Tubulin	Human β-tubulin	β-Tubulin	Abcam, ab6046	Rabbit; polyclonal	1 to 1000	AB_2210370

Abbreviation: RRID, research resource identifier.

Blood glucose measurements

Blood glucose was measured exactly as described previously in animals 3 months postfertilization (10).

Statistical analysis

The two-tailed Student t test was used to compare plasma glucose levels. Values are reported as mean ± standard error of the mean.

Results

Positional cloning of the hmn locus reveals a mutation in gfpt1

The hmn ^z110 mutant shows hepatic lipid accumulation 7 dpf [Fig. 1(a)]. This steatosis phenotype is not modified by prolonged fasting [Fig. 1(b)] or feeding [Fig. 1(c)]. These findings suggested that the mutation causes a metabolically inflexible change in liver metabolism. Importantly, the WT animals used are derived from a heterozygous cross of hmn^−/z110 adults: one-third is hmn^+/+ and two-thirds are hmn^−+/z110.

Figure 1.

Hepatic steatosis in hmn mutants is not modified by nutritional status. (a) Diffuse accumulation of neutral lipid droplets in never-fed 7-dpf livers stained with eosin following osmium impregnation is seen in hmn mutant livers. (b) Never-fed larvae were live-sorted at 7 dpf with Nile Red staining. They were fixed and stained with ORO at the indicated ages and scored for hepatic steatosis. (c) Larvae were live-sorted at 7 dpf and fed until fixation and staining with ORO at the indicated ages.

We used standard methods to isolate the mutated gene. Bulk segregant analysis established linkage to chromosome 8 [Fig. 2(a)]. Fine mapping narrowed the critical interval to four candidate genes: agtpbp1, naa35, gfpt1, and antxr1l [Fig. 2(b)]. Full-length cDNAs were cloned and sequenced from transcripts for all four candidates. There were no coding changes in naa35 or antxr11 cDNAs cloned from hmn^z110 mutants. Because the rs5020109055 polymorphism causes a L450H coding change in agtpbp1, we performed a complementation test with the null allele agtpbp1^sa17482. This agtpbp1 allele complemented the hmn ^z110 mutant, indicating agtpbp1 loss-of-function does not give rise to the hmn phenotype (data not shown).

Figure 2.

Positional cloning of the hmn mutant reveals a substitution mutation in an invariant residue of Gfpt1 protein within the first sugar isomerase domain. (a) A map cross of heterozygous hmn carriers to the polymorphic WIK line was performed. A single pair of heterozygous progeny was bred to generate all the individuals used to positionally clone the locus. First, larvae were fixed and stained with ORO and then sorted into two pools. DNA from 21 WT and 21 hmn mutant larvae were examined by bulk segregant analysis using a standard set of simple sequence-length polymorphism microsatellite markers (SSLP, z markers) on a meiotic map (6). This allowed us to assign the mutated gene to the south subtelomeric region of chromosome 8. (b) Examination of 51 additional progeny from this single pair allowed us to narrow the locus further to a single contig. cDNAs were cloned and sequenced for all protein-coding genes within the critical interval. (c) At position 987 of the gfpt1 coding sequence, a G-to-A transition mutation was identified in hmn mutants (and was confirmed to have been induced in the mutagenized male through sequencing genomic DNA). (d) The mutation changed codon 329 from an E to a K residue. (e) Comparison among multiple species from yeast to humans revealed that E329 (zebrafish numbering, shaded in red) is an invariant residue. Many neighboring residues are also invariant (yellow) or highly conserved among the species examined. (f) Inspection of the crystal structure of human GFPT1 (a homodimer, with both chains shown in different colors) shows that the invariant E239 residue is in close proximity (9.6 Å) to the fructose-6-phosphate substrate. (g) The HBP is shown, with the rate-limiting enzyme GFPT, catalyzing the conversion of fructose-6-phosphate to glucosamine-6-phosphate via a transamidation-requiring glutamine.

A mutagen-induced G to A transition was found in exon 11 of gfpt1, causing an E to K substitution at codon 329 [Fig. 2(c)]. Residue E329 of Gfpt1 falls in the start of the first isomerase domain [Fig. 2(d)], and is invariant among orthologs from Saccharomyces cerevisiae to Homo sapiens [Fig. 2(e)]. In the crystal structure of human GFPT1, residue E329 is within 10 Å of the fructose-6-phosphate substrate that receives the NH₄⁺ from the glutamine substrate [Fig. 2(f) and 2(g)]. The location of the mutated residue and the substitution of a negatively charged side chain with a positively charged side chain suggest that substrate binding, catalytic activity, competitive inhibition by glucosamine 6-phosphate, allosteric inhibition by uridine diphosphate, or a combination of these biochemical processes is altered (11). The last two possibilities are particularly attractive because this mutation is recessive: a homodimer of E329K mutants is most likely required for the observed phenotypes.

The gfpt1^E329K mutation increases Gfpt1 protein expression, global O-GlcNAc modification, and lower blood glucose

Gfpt1 is the rate-limiting enzyme of the HBP [Fig. 2(g)]. This nutrient-sensing pathway diverts excess sugars from glycolysis to the production of signaling molecules in the form of O-GlcNAc modification of diverse proteins that alter protein function (12, 13). To establish the consequence of the gfpt1^E329K mutation on Gfpt1 protein expression and O-GlcNAc modification (a direct measure of HBP flux), we performed immunoblot analyses for Gfpt1 protein and O-GlcNAc modification. Compared with WT siblings, adult homozygous gfpt1^E329K mutants showed increased Gfpt1 protein and global O-GlcNAc modification of proteins in their livers [Fig. 3(a)]. These two findings, and the observation that heterozygous carriers [two-thirds of the WT cohorts in Fig. 1(b) and 1(c)] do not develop hepatic steatosis, strongly suggest that the gfpt1^E329K mutation causes increased HBP flux (i.e., the mutation is a recessive, gain-of-function) by increasing Gfpt1 protein abundance, activity, or both. Finally, gfpt1 transcript abundance was unchanged in livers of hmn mutants (data not shown), indicating additional mutagen-induced gene expression changes are unlikely to account for the observed phenotypes (data not shown).

Figure 3.

Gfpt1^E329K mutants have higher Gfpt1 protein abundance and increased O-GlcNAc modification, but lower fasting blood glucose. (a) Livers from three adult WT and gfpt1^E329K mutants were dissected following an overnight fast. The livers were pooled and homogenized in protein extraction buffer. Proteins (15 μg per sample) were separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (molecular weight standards shown in kDa) and transferred to polyvinylidene fluoride membranes, and immunoblot analyses were performed with anti-GFPT1, anti–O-GlcNAc modification, and anti-Tubb IgGs. As expected, the anti-GFPT1 IgG detected a single band of nearly 80 kDa; multiple O-GlcNAc modified proteins were detected in both groups. (b) Animals were fasted overnight prior to measurement of blood glucose.

Ten to 15% of homozygous gfpt1^E329K mutants survive to adulthood (1). This observation allowed us to measure fasting blood glucose. In both sexes, blood glucose was lower in gfpt1^E329K mutants, although the difference was statistically significant in males only [Fig. 3(b)].

Discussion

Type 2 diabetes mellitus (DM2) has reached pandemic proportions (14, 15). This multifactorial disease has high direct and indirect burdens on individuals and societies (16). It is widely recognized that environmental factors that drive IR in multiple cell types are a prerequisite for the development of DM2 in genetically susceptible individuals (17–19). Although multiple IR mechanisms have been identified, an incomplete understanding of these pathogenic metabolic and signaling pathways remains. This state has hindered the development of novel diagnostic and therapeutic modalities to address DM2.

Increased HBP flux is a major driver of IR. First, increased flux through the HBP inhibits adipocyte glucose disposal (20). Second, skeletal muscle biopsies from humans with DM2 show increased flux through the HBP pathway, and this flux correlates with decreased glucose disposal rates (21). Third, transgenic overexpression of Gfpt1 in mouse liver increases O-GlcNAc modification and is sufficient to cause IR, marked by hyperglycemia, hypertriglyceridemia, and increased hepatic glycogen stores (22–26). Other O-GlcNAc modification pathway transgenic overexpression models (27) and pharmacological manipulations to increase HPF flux in vivo (28, 29) confirm that this pathway is sufficient to drive IR. Cell-based models confirm these results (12).

What remains unclear is whether increased HBP flux is necessary for the development of IR (30). This is an important question because its answer would reveal whether (and where) attempts to alter HBP flux could be used to reverse IR. At present, a viable GFPT1 loss-of-function model is not available. This reflects the key role of HBP in early skeletal development and in neuromuscular maturation (31, 32). Through phenotype-driven forward genetic screening, we have isolated a zebrafish mutant with increased HBP flux, and two distinguishing phenotypes of severe, lifelong insulin resistance: hepatic steatosis and runting (1, 33). The mutant protein we identified should be studied in vitro to establish the biochemical basis for global increase in O-GlcNAc modification. It is also possible that the mutation decreases the rate of protein degradation in vivo.

A minority of homozygous gfpt1^E329K mutants is viable to adulthood; we found that lifelong, global activation of HBP flux did not cause fasting hyperglycemia when animals are fed their normal diets. Indeed, the lower fasting blood glucose observed in gfpt1^E329K mutant males suggests that there are most likely multiple metabolic alterations present in animals with global increase in HBP flux. Because homozygous loss-of-function gfpt1^j23e1 mutants have lethal craniofacial defects (32), conventional genetic approaches to dissecting the function of the gfpt1^E329K mutation in adults might not be possible; however, if transheterozygous gfpt1^j23e1/ gfpt1^E329K mutants prove viable, then it would be informative to assess glucose homeostasis in animals carrying a single recessive gain-of-function Gfpt1 allele. Additionally, it is nominally possible that the gfpt1^E329K mutant harbors additional gene-regulatory lesions that contribute to the phenotype of increased HBP flux. Blood glucose should be evaluated at additional ages during adulthood under normal and high-fat feeding. Likewise, insulin tolerance tests could be performed on the few surviving mutants (34). In summary, the Gfpt1^E329K mutant affords the opportunity to examine the role of HBP in glucose homeostasis on an integrated physiological level and at organ- and cell-type resolution.