Four subcutaneous fibroblasts explants, MCH068, MCH070, WG0321 and WG0791, were obtained from the repository for mutant human cell strains at the McGill University Health Center. MCH068 and MCH070 cells were isolated from normal individuals while WG0321 and WG0791 cells were isolated from patients diagnosed with GAN. Both GAN patients experienced difficulty in walking and their electromyography showed diffuse axonal neuropathy. We sequenced the GAN cDNAs prepared from the fibroblast explants. No mutations were detected in MCH068 and MCH070 cells. We obtained two PCR products from WG0791 cells, a major product of ~1.8 kb and a minor product of ~1.7 kb (data not shown). Sequencing of the 1.8-kb product revealed that a missense mutation in exon 3 (c.545T>A). The mutation resulted in the substitution of the isoleucine at amino acid position 182 with an asparagine, I182N (Fig. 1A). The 1.7-kb fragment represented an mRNA product from the other GAN allele because it did not contain the I182N mutation. It was shorter than the wild-type message because it did not contain exon 2 (Fig. 1B). We then sequenced the first three exons and the intron-exon junctions of the GAN gene from WG0791 cells. While confirming the I182N missense mutation, we also discovered an A→C mutation near the exon 2-intron 2 junction (c.282+3A>C), which might account for the misspliced message (data not shown).

Sequencing of the GAN cDNA prepared from WG0321 cells revealed a deletion/insertion in the GAN message: nucleotides 1505–2056 were replaced with a 452-nucleotide-long sequence that was identical to a part of intron 9 of the GAN gene (Fig. 1C). We sequenced the 3' region of the GAN gene from WG0321 cells and discovered that the entire exon 10 and 446 base pairs of the exon 11 5'end were deleted in both alleles. The deletion caused exon 9 to be spliced into intron 9 (data not shown). A schematic representation of the mutated and normal GAN alleles is shown in Fig. 1D.

Because we were unable to screen additional healthy controls, the three novel GAN alleles should be considered putative disease-associated mutations.

Previous studies have shown that vimentin IFs form abnormal aggregates in GAN fibroblasts and that low-serum treatment enhances the aggregation. To determine whether WG0321 and WG0791 cells also contained IF aggregates, we performed immunocytochemical staining with antibodies against various IF proteins. In complete medium, 12% of WG0791 cells and 23% of WG0321 cells displayed compact vimentin aggregates. Upon low-serum treatment for 72 hours, the number of cells containing vimentin aggregates dramatically increased, up to 50% for WG0791 cells and 95% for WG0321 cells (Fig. 2A). Vimentin aggregates were never detected in normal fibroblasts, MCH068 and MCH070. Interestingly, although cytokeratins are usually not expressed in fibroblasts, a small percentage of both normal and GAN cells (~2%) exhibited positive staining for keratins (Fig. 2B–E). In GAN cells, some of the IF aggregates contained both vimentin and cytokeratin (Fig. 2D and 2E).

We studied the expression profiles of GAN and normal fibroblasts grown in low-serum medium by microarray (Affymetrix) analysis. To reduce the background noise, we performed a four-way comparison of MCH068, MCH070, WG0321 and WG0791 cells. We selected genes that showed consistent changes in WG0321 and WG0791 fibroblasts when compared to MCH068 and MCH070 cells. Genes that exhibited more than three-fold differences are grouped in Table 1 in the Supplemental Data according to their proposed functions. Gene products involved in lipid metabolism displayed the most dramatic changes. We confirmed the relative expression levels of these lipid metabolism genes by quantitative RT-PCR (Fig. 3). The results were in agreement with the microarray experiments. We also performed quantitative RT-PCR to determine the expression levels of gigaxonin in GAN fibroblasts (Fig. 3A). Compared to MCH068 and MCH070, GAN mRNA expression was dramatically upregulated in WG0321 (~26 fold) and WG0791 (~7 fold) cells.

We also detected significant changes in members of the ATP-Binding Cassette (ABC) protein family, ABCA6 and ABCB4. ABC transporters are multispan transmembrane proteins that translocate a variety of substrates. ABCA6 has been suggested to play an important role in lipid homeostasis 15; it was up-regulated in GAN fibroblasts. ABCB4 is also known as multidrug resistance P-glycoprotein 3 and functions as a translocator of phospholipids. Deficiencies in ABCB4 cause progressive intrahepatic cholestasis type III 16. ABCB4 was downregulated in GAN fibroblasts. In addition, Fatty Acid Binding Protein 5 (FABP5), Meltrin alpha, Complement C3 and Butyrylcholinesterase (BChE) were upregulated in GAN fibroblasts. FABP5 is involved in intracellular fatty-acid trafficking (reviewed in 17). Meltrin alpha is a member of the metalloprotease-disintegrin family and is involved in adipogenesis 18. C3 is a component of the complement system of innate immunity. It is also the precursor of an acylation-stimulating protein that can increase triglyceride synthesis (reviewed in 19). BChE is a serine hydrolase that exhibits increased activity in hyperlipidaemic patients 20. Acyl-CoA: Cholesterol Acyltransferase (ACAT) and Leptin were downregulated in GAN fibroblasts. ACAT is an enzyme that converts intracellular cholesterol into cholesteryl esters and promotes the storage of excess cholesterol in the form of cholesterol ester droplets 21. Leptin is a peptide hormone produced predominantly by white adipose cells; however, it is also expressed in non-adipocytes and is important in regulating fatty acid metabolism (reviewed in 22).

Because the microarray analysis revealed alterations in the expression of lipid metabolism genes in GAN fibroblasts, we studied the distribution of lipid droplets in the fibroblast explants by cytological staining. We used Oil Red O dye to label neutral lipid droplets of serum-starved GAN and normal fibroblasts. The proportions of cells containing lipid droplets were significantly higher in the mutant explants (52% in WG0321, 21% in WG0791) compared with the normal explants (6% in MCH068, 9% in MCH070; Fig 4A). In addition, GAN fibroblasts contained many more droplets per cell than did normal fibroblasts (Fig. 4B–D). These results were confirmed by Bodipy staining (data not shown).

Previously, it has been shown that vimentin can form cage-like structures around lipid droplets in adipocytes. We wondered whether the vimentin aggregates also surrounded the lipid droplets in GAN cells. We performed fluorescence microscopy on GAN cells co-stained for vimentin and lipid droplets. As shown in Fig. 4F–G, while some of the small vimentin aggregates appeared to encage lipid droplets, most of them did not (~80% in both cell lines).

Total RNA was isolated from fibroblasts using Trizol reagent (Invitrogen). First-strand cDNA was synthesized with oligo-dT primers and reverse transcriptase (Invitrogen). The procedures were performed according to the manufacturer's protocol. Gigaxonin cDNAs were amplified from the cDNA pool using forward primer, 5'-TTGATGGCTGAGGGCAGTGCCGTGTCTG-3' and reverse primer, 5'-TTCCTCCTCAAGGGGAATGAACACGAAT-3'. After electrophoresis, PCR products were purified by the GeneClean purification system (Q-Biogen) and were sequenced with the BigDye™ sequencing kit (Applied Biosystems). The shorter gigaxonin cDNA products from explant WG0791 were first cloned into pCR2.1-TOPO vector (Invitrogen) before sequencing. The conditions used for genomic PCR and the sequences of the PCR primers have been reported elsewhere 10.

Fibroblast explants were obtained from the repository for mutant human cell strains at McGill University. They were maintained at 37°C and 5% CO₂ in MEM Eagle medium (Earle's) with 10% fetal bovine serum. For low-serum treatment, cells were incubated in medium containing 0.1% fetal bovine serum for 72 hours. All experiments were performed on cells from passages 13–18. For immunocytochemical analyses, cells seeded on coverslips were fixed with 4% paraformaldehyde for 20 minutes and permeabilized with 0.1% Triton-X100 for 5 minutes. Fixed cells were incubated with primary antibodies at room temperature for one hour, followed by several washes with PBS and incubation with appropriate secondary antibodies for 30 minutes. The coverslips were then washed with PBS and mounted onto slides with Aquamount (Lerner Laboratories) for immunofluorescent microscopy. To co-stain lipid droplets and vimentin filaments, formaldehyde-fixed cells were permeabilized with 0.05% Saponin and incubated with an anti-vimentin antibody overnight. Before mounting onto slides, cells were stained for lipid droplets with 0.5% Oil Red O in propylene glycol (Poly Scientific). To stain the nuclei and the lipid droplets, cells were fixed in 4% paraformaldehyde and incubated with 0.5% Oil Red O in propylene glycol and Gill's Hematoxylin I (Poly Scientific) for 20 minutes. Antibodies used: monoclonal mouse anti-vimentin, clone V9 (Sigma), monoclonal mouse pan anti-keratin (Sigma), and polyclonal anti-vimentin 35.

Human 133A Genechips from Affymetrix were used to study the expression profiles. Biotin-labeled cRNA probes were prepared according to the manufacturer's protocol. In brief, cells were treated with low-serum medium for 72 hours, and total RNA was extracted. First-strand cDNAs were synthesized with oligo-dT-T7 primers and reverse transcriptase (Invitrogen). After second-strand cDNA synthesis, double-stranded cDNAs were used to produce biotin-labeled cRNA probes by T7 polymerase (Enzo Laboratory). The cRNAs were fragmented before being used for hybridization. Hybridization and scanning were carried out at the GeneChip analysis facility at Columbia University.

GeneChip results were analyzed by Affymetrix Microarray Suite (version 5.0). To reduce background noise, a four-way comparison of MCH068, MCH070, WG0321 and WG0791 cells was performed. Only genes that showed consistent changes in both WG0321 and WG0791 fibroblasts when compared to both MCH068 and MCH070 cells were selected. Genes that exhibited more than three-fold differences were considered as significantly altered.

To determine the relative expression levels of a gene, quantitative PCR was performed on cDNAs prepared from normal and GAN fibroblasts. Single-stranded cDNAs were generated from total RNA with Oligo-dT primers and reverse transcriptase (Invitrogen). Quantitative PCR was performed on a SmartCycler II PCR machine (Cephid) with gene-specific primers, SyBr Green fluorescence dye, and OmniMix HS PCR master mix (TakaRa). The sequences of the PCR primers and the PCR conditions are shown in Table 2. GAPDH was used as the internal control, and the fold difference of a gene-of-interest in MCH070, WG0321 and WG0791 relative to MCH068 was calculated using the ΔΔCt method 36.