Gene targeting of GAN in mouse causes a toxic accumulation ......INTRODUCTION Giant axonal neuropathy (GAN), ﬁrst described in 1972 (1), is a severe, autosomal recessive motor and

Gene targeting of GAN in mouse causes a toxicaccumulation of microtubule-associated protein8 and impaired retrograde axonal transport

Jianqing Ding1,{, Elizabeth Allen1,{, Wei Wang1, Angela Valle1, Chengbiao Wu1,

Timothy Nardine1, Bianxiao Cui4, Jing Yi5, Anne Taylor6, Noo Li Jeon6, Steven Chu4, Yuen So1,

Hannes Vogel3, Ravi Tolwani2, William Mobley1 and Yanmin Yang1,*

1Department of Neurology and Neurological Sciences, 2Department of Comparative Medicine, 3Department of

Pathology, Stanford University School of Medicine 1201 Welch Road, Stanford, CA 94305-5489, USA, 4Department of

Physics, Stanford University, Stanford, CA, USA, 5Department of Cell Biology, Shanghai Second Medical University,

Shanghai, China and 6Department of Biomedical Engineering, University of California, Irvine, CA, USA

Received January 24, 2006; Revised and Accepted March 15, 2006

Mutations in gigaxonin were identified in giant axonal neuropathy (GAN), an autosomal recessive disorder.To understand how disruption of gigaxonin’s function leads to neurodegeneration, we ablated the geneexpression in mice using traditional gene targeting approach. Progressive neurological phenotypes andpathological lesions that developed in the GAN null mice recapitulate characteristic human GAN features.The disruption of gigaxonin results in an impaired ubiquitin–proteasome system leading to a substantialaccumulation of a novel microtubule-associated protein, MAP8, in the null mutants. Accumulated MAP8alters the microtubule network, traps dynein motor protein in insoluble structures and leads to neuronaldeath in cultured wild-type neurons, which replicates the process occurring in GAN null mutants.Defective axonal transport is evidenced by the in vitro assays and is supported by vesicular accumulationin the GAN null neurons. We propose that the axonal transport impairment may be a deleterious conse-quence of accumulated, toxic MAP8 protein.

INTRODUCTION

Giant axonal neuropathy (GAN), first described in 1972 (1), isa severe, autosomal recessive motor and sensory neuropathyaffecting both the peripheral nervous system and the centralnervous system (CNS). Onset of symptoms usually occursbefore the age of 4 years, with rare cases occurring up to theage of 10. The typical clinical manifestations of GAN for thefirst 7 years of the life are clumsiness, followed byprogressive and predominant distal weakness and sensory losswith a gait disturbance. CNS involvement includeselectroencephalographic abnormalities, mental retardation, sei-zures and defective upper motor neuron function (2,3). Curly orkinky hair is not a constant finding but appears in most of thepatients (4,5). Death nearly always occurs by the third decadeand is common by late adolescence (3). Pathologically, GAN

is characterized by giant axonal swellings, fiber loss and neuro-degeneration (6,7). At the ultrastructural level, GAN is charac-terized by generalized cytoskeletal abnormalities. In addition todensely packed neurofilaments (NFs) with displacement or seg-regation of other axonal organelles, reduced microtubuledensity and accumulation of intermediate filaments (IFs)outside of the nervous system are also noticed in biopsiesfrom GAN patients (4–6,8–12). The recessive inheritanceimplies that GAN is a loss of function disorder.

The disease locus of GAN was assigned to a single locuson chromosome 16q24.1 (13–15) and the GAN gene encodinga protein, called gigaxonin, was identified (16). To date, 24distinct mutations have been identified in 20 GAN patientfamilies (16–18). The mutations were evenly distributedthroughout the coding sequence of 11 exons. They includedone insertion mutation at seventh amino acid, which leads to

# The Author 2006. Published by Oxford University Press. All rights reserved.For Permissions, please email: [email protected]

{The first two authors contributed equally to the work.

*To whom correspondence should be addressed. Tel: +1 6507361032; Fax: +1 6504986262; Email: [email protected]

Human Molecular Genetics, 2006, Vol. 15, No. 9 1451–1463doi:10.1093/hmg/ddl069Advance Access published on March 24, 2006

a premature termination, two deletions in the BTB domain,seven nonsense mutations in the kelch repeat domain and 14missense mutations scattered over the entire coding region.The gene discovery represents an important step towardsunderstanding the pathogenesis of the disorder. However,the mechanism by which disruption of the gene leads todevastating neurodegeneration was not fully elucidated.

Gigaxonin belongs distantly to BTB–kelch-repeat super-family (16,19). Much has been learned in recent years aboutthe functions of BTB-containing proteins that involve inmany aspects of cellular processes, such as actin cytoskeletoninteraction (20), cell morphology and growth (21), cyto-plasmic sequestration of transcription factors (22) and possiblyviral pathogenesis (23). Recent works suggest that the BTB–kelch proteins are involved in ubiquitin–proteasome system(UPS) and may define a new class of substrate-specificadaptors of Cul3-based E3 ligase complexes (24–26). It hasbeen showed that BTB protein Keap1 targets the transcriptionfactor Nrf2 for ubiquitination (27–29). The devastating clini-cal features occurring in GAN patients indicate an essentialrole of gigaxonin in axonal structure and neuronal function.Determining the molecular pathways that involve gigaxoninmay contribute significantly to understanding the pathogenesisof GAN as well as other neurodegenerative disorders.

We have previously reported that gigaxonin can bind directlyto the neuronal-specific microtubule-associated protein,MAP1B, and control its degradation (30,31). In this study, wereport GAN null mice as a model for human GAN. We showhere that gigaxonin interacts with MAP8, a recently identifiedmicrotubule-associated protein (32). Functional analysisdemonstrates that gigaxonin negatively regulates the protein

level of MAP8 through UPS, a function which is disrupted byGAN-associated mutations. The survival of GAN null neuronscould be substantially improved by reducing MAP8 levelusing specific siRNA. Interestingly, in transfected cells, accu-mulated MAP8 protein caused alterations in the microtubulenetwork and trapped dynein motor protein into insoluble struc-tures. Ultrastructural analysis of GAN null axons revealed amarked accumulation of vesicular organelles and mitochondria,pointing to impaired axonal transport which is evidenced by thein vitro assays. Taken together, these findings identify theimportance of gigaxonin’s function to axonal transport, whichis essential for neuronal maintenance.

RESULTS

The GAN null mouse is a model of human GAN

The recessive inheritance of GAN and the disruptive mutationsidentified in human GAN patients imply that the disorder isdue to a loss of gigaxonin function. To elucidate molecularmechanisms underlying the disorder, we ablated the GANgene using traditional gene targeting. The targeting vectorwas constructed as depicted in Figure 1A and a 5 kb fragmentcontaining exons 3, 4 and major portion of 5 was replaced bythe 1.8 kb Neo gene cassette. Heterozygous offspring derivedfrom independent ES clones were mated to produce homozy-gous mutants, which were genotyped for identification(Fig. 1B). To confirm unequivocally that the expected targetingevent had taken place, we analyzed gigaxonin expression toreveal that both the mRNA and the protein were completelyabsent in the homozygous null mutants (Fig. 1C and D).

Figure 1. Gene targeting of GAN ablates the expression of gigaxonin. (A) The top drawing schematically depicts the mouse WT GAN genomic locus. The bottomdrawing depicts GAN targeting vector. The 1.5 kb 30-arm extends from primers GANYA1 to GANYA2. The 7.3 kb 50-arm extends from the SpeI site and con-tains exon 2, whereas exons 3–5 are replaced by the 1.8 kb Neo gene cassette. An EcoRV restriction digestion site is deleted in the targeted allele. (B) Genotyp-ing of the targeted mice. The genomic DNAs were extracted from mouse tails. WT3, Neo1 and mGigDiag32 primers were used to amplify a PCR product of328 bp for mutant (mut) and 436 bp for wild-type (wt). 2/2, homozygous null; þ/2, heterozygous; þ/þ, wild-type. (C and D) Brains and spinal cords of30-day-old wt and KO mice were used to prepare protein lysates and mRNAs. (C) RT–PCR analysis was performed using primers to N-terminal gigaxoninand actin to confirm the null samples. (D) The protein samples were immunoblotted with antibodies to gigaxonin and actin.

1452 Human Molecular Genetics, 2006, Vol. 15, No. 9

These data confirmed that the gene targeting in mouse hadsuccessfully resulted in ablation of gigaxonin expression.

A prominent feature of most GAN null mutants is aprogressive deterioration in motor function with onsetvarying from 6 to 10 months. A substantial reduction of spon-taneous movement is the initial sign of neurological abnorm-alities. The subclinical motor deficits were evidenced byestablished criteria utilizing specialized testing apparatus.Footprint tests recording locomotion displayed comprisedmotor functions and gait disturbance of the null mice at theearly phenotypic stage of age 6–10 months (Fig. 2). Shortlyafterward, most of the null mutants could be distinguishedfrom their littermate controls by bizarre limb positioning andoverall weakness (Fig. 3B, C and E) occasionally with signsof spasticity and seizure. The involvement of weaknessstarted from hind limbs (Fig. 3C) and progressed to the fore-limbs. By the age of 12 months and beyond, some of thenull mice exhibited muscle wasting and progressive weightloss (Fig. 3B and E). Later during progression, the effort to

walk for the phenotypic null mice becomes more severelycompromised. Surprisingly, only very few GAN null mutantsdied prior to the age of 12 months. In addition to neurologicalproblems, many human GAN patients display strikingly kinkyor curly hairs. Likewise, hair phenotypes could be also foundin some of the GAN null mice; some had no whiskers (Fig. 3E)and others displayed abnormal or fragile hairs (Fig. 3F). Theheterozygote mice are indistinguishable from their wt litter-mates over the entire life course. Thus, the phenotypic featuresof GAN null mice bear a striking resemblance to the clinicalmanifestations of human GAN patients, although withmilder course and significantly slower progression.

Pathological lesions in the null mice include axon lossinvolving both myelinated and unmyelinated nerve fibers,with a variable severity ranging from �18–27% in theperipheral nerves at the age of 9–12 months. On average,the axons are enlarged, but the featured giant axons found inhuman GAN patients could not be seen in the GAN nullmice (Fig. 4G–I). Ultrastructural studies on phenotypicalnull mice revealed enlarged axons distended with NFs(Fig. 4E). The densely packed NFs displaced or segregatedother axonal organelles (Fig. 4C), a feature characteristic ofhuman GAN pathology. Pronounced axonal swelling anddegeneration with thin myelin sheath is observed in bothperipheral nervous system (Fig. 4B and F) and CNS(Fig. 4C). At later stages, axons in the phenotypic nullanimals are marked by vesicular accumulation (Fig. 4F),suggesting the impairment of axonal transport.

We conclude that the phenotypes and the pathologicalfeatures of GAN null mice validate this null mouse as amodel of human GAN.

Gigaxonin interacts directly with MAP8

To elucidate the molecular mechanisms underlying the dis-order, we set out to understand the physiological functionsof gigaxonin by identifying its binding partners using yeasttwo-hybrid screen. Using full-length gigaxonin as bait, anovel protein was also identified in addition to MAP1B-LC(30,31). Functional characterization of this protein revealedtwo microtubule-binding domains with distinctive affinities,one at each chain (32). We named it as microtubule associatedprotein 8 (MAP8). The C-terminal microtubule-bindingdomain of MAP8 was also identified by another group intransfected cells and reported as MAP1S (33). Post-natally,the protein expression was primarily found in the nervoussystem, with barely detectable levels in most tissues outsidethe nervous system (32). The specific interaction betweengigaxonin and MAP8 was further tested in mammalian cells.When co-expressed in COS7 cells with MAP8, the gigaxoninprotein no longer displayed a diffuse distribution (30), butco-localized with MAP8 along microtubules (Fig. 5C–E).The presence of the two proteins in co-immunoprecipitationsalso confirmed their association (Fig. 5B, lane 3). To preciselydefine the domains responsible for the protein interaction,we mapped their binding sites. Gigaxonin was dissected intothe C-terminal kelch-repeat and N-terminal BTB domains.The yeast two-hybrid results indicated that the kelch repeatdomain contains the direct binding site for MAP8 (Fig. 5A).Moreover, in co-transfected COS7 cells, the C-terminal

Figure 2. Footprint analysis indicates a compromised motor function in theGAN null mice. (A) The paw prints of mice were recorded while ambulatingon blotting paper. Footprint analysis indicated a decreasing average stridelength (SL) and increasing average distance between feet (DBF) as themouse ambulatory ability decreased. (B) The gait functional index, calculatedas SL/DBF, was greatest in wild-type (þ/þ) mice and lowest in GAN homo-zygous deficient (2/2) mice, quantitatively indicating the decreased ambulat-ory ability of GAN deficient mice. Error bars represent SE and asterisksrepresent significance (P, 0.05).

Human Molecular Genetics, 2006, Vol. 15, No. 9 1453

kelch repeat domain co-aligned with MAP8 (data not shown),supporting the notion that kelch-repeat motif is important forthe interaction. Next, we examined which domain of MAP8is the target for gigaxonin’s association. Furthermore, theco-localization of kelch-repeat domain with the C-terminusof MAP8 (MAP8-CT) on the actin filaments (Fig. 5F–H)was consistent with the results of the yeast two-hybridscreen. Thus, through its kelch-repeat domain, gigaxonininteracts directly with the C-terminus of MAP8. Theseresults also suggest that the binding of MAP8 to gigaxonindoes not affect its cytoskeletal associations.

To assess whether GAN-associated mutations disrupt theinteraction of gigaxonin with MAP8, we analyzed threemutations for a possible loss of the specific association ofgigaxonin with MAP8: the truncating mutation (18insA) andtwo nonsense mutations (R242X and R293X). As assessedby immunoblot analysis, all three mutations caused unstableprotein expression, either undetectable for the 18insAmutant or significantly reduced in the cases of R242X (datanot shown) and R293X mutants (Fig. 5B, lane 2). As revealedby co-immunoprecipitations, the R293X mutant failed to bepull-downed by the Flag-tagged MAP8-LC (Fig. 5B, lane 4),in contrast to its wild-type counterpart (Fig. 5B, lane 3).Furthermore, of the few detectable cells co-transfectedwith MAP8-CT and R242X (data not shown) or R293X(Fig. 5I–K), the mutant proteins were diffusely accumulatedin the cytoplasm and had completely lost their binding to

the C-terminus of MAP8/MAP1S (Fig. 5I–K), demonstratingthat an effective interaction of the two proteins requiresa functional kelch-repeat domain.

Gigaxonin negatively regulates the proteinlevel of MAP8

Gigaxonin is reported to bind directly with ubiquitin-activating enzyme E1 (UBE1) through its BTB domain andplays an important role in controlling MAP1B’s protein degra-dation (31). We next determined whether disruption of thegigaxonin–MAP8 interaction also leads to protein accumu-lation of MAP8. Total lysates of mouse brain, spinal cordand sciatic nerves of wt and the null were analyzed by immu-noblotting with antibodies to MAP8-HC and actin (Fig. 6A) orto dynein intermediate chain (DIC) and MAP8-LC (Fig. 6B).As judged by three independent assays, the absence of gigaxo-nin led to substantially increased levels of MAP8 proteinin comparison with unchanged DIC or actin controls in thenull animals (Fig. 6A and B). Densometric scans of westernblots, normalized with levels of DIC and actin, indicate anearly 1.5–3-fold increase of MAP8 level in the nullsamples at the age of 10–15 months (Supplementary Material,Fig. S1). To determine whether this increase was due toupregulated gene transcription, we analyzed MAP8 transcriptin wt and the null animals. In contrast to the increase inprotein, RT–PCR and real-time PCR revealed no change in

Figure 3. TheGANnull mice exhibit limb weakness and axonal degeneration. (A and D) Wild-type controls (WT). (B, C, E and F) GAN null mice (KO) at the ageof 9–14 months display limb weakness and bizarre positioning. Some of the null mice have no whiskers, indicated by arrows in (E) and the others displayabnormal fragile hairs as depicted in (F).


transcription of MAP8 in comparison to actin in the nullsamples (Fig. 6C and D). Thus, the increased level of MAP8was due to protein accumulation.

The accumulation of MAP8 protein in the gigaxonin nullmice prompted us to investigate whether gigaxonin playsan important role in controlling the degradation of MAP8.PC12 cells, which express endogenous gigaxonin andMAP8, were selected for assays. As the initial assessment,the protein stability of MAP8 was evaluated by utilizingcycloheximide (CHX) to suppress protein translation. Asdetermined by immunoblot assays using anti-MAP8-LC(Fig. 7A) or anti-MAP8-HC (data not shown), the endogenousMAP8 is quite stable; the majority of the protein was stillpresent 8 h after CHX treatment (Fig. 7A, lanes 1–7). More-over, the protein level remains constant when both proteasomeinhibitor (PI) and CHX are present (Fig. 7A, lanes 8–12),suggesting that a functional UPS is required during MAP8degradation. Following the protein stability assays, the effectof gigaxonin on the protein levels of endogenous MAP8 wasexamined in Flag-gigaxonin or GFP-transfected PC12 cells.

A transfection efficiency of �15–20% was achieved forboth Flag-gigaxonin and GFP constructs. In contrast to GFP(Fig. 7B, lanes 3 and 4), overexpression of gigaxonin wasaccompanied by a consistent reduction of MAP8-LC proteinlevel 20–50 h post-transfection in PC12 cells (Fig. 7B, lanes1 and 2). The moderate reductions detected at 24 h could beaccounted by the fact that the reduction only occurred in15–20% of the gigaxonin-overexpressing cells. Importantly,this gigaxonin-mediated reduction could be prevented byaddition of PI to the gigaxonin overexpressing cells (Fig. 7B,lanes 5 and 6). These findings demonstrate that gigaxoninpromotes destabilization of MAP8 protein and, in view ofthe fact that PI reverses MAP8 turnover, suggest the involve-ment of UPS in its degradation.

If the gigaxonin-controlled protein degradation of MAP8 ismediated by the UPS, the protein should become covalentlyconjugated with poly-ubiquitin when the degradation is inhib-ited. We tested this hypothesis on both exogenous (Fig. 7C,lanes 1 and 2) and endogenous (Fig. 7C, lanes 3 and 4)MAP8 proteins. To facilitate the assays of ubiquitination, an

Figure 4. Pathological lesions were found in the null animals. (A–F) Ultrastructural analysis (EM) on wt (A and D) and the null axons (B, C, E and F). Cross-section (A–C) and longitudinal (D–F) section of wt (A and D) and KO (B, C, E and F) samples revealed swollen axons (B), densely accumulated NFs (E), thinmyelin sheath (indicated by arrows in B, C and F), completely disorganized cytoarchitecture and heavily accumulated membranous structures (C, from brain andF, from sciatic nerve) in the null animals. (G–I) Semi-thin sections of sciatic nerves from wt (G) and GAN null mice (H and I) at the age of 28 weeks showaveragely enlarged axons with signs of degeneration in the null samples. Bar represents 500 nm for (A–D), 750 nm for (E and F) and 12 mM for (G–I).


excess of ubiquitin was supplied. Indeed, co-transfection ofHA-tagged ubiquitin and Flag-MAP8-LC led to formationof an apparently typical ubiquitin ladder in the presence ofPI, equivalent to the conjugation of poly-ubiquitin tags onthe Flag-MAP8-LC (Fig. 7C, lane 2). Similarly, endogenousMAP8-LC was also poly-ubiquitinated when an excess of

ubiquitin and PI were supplied, displaying the high molecularshifts of MAP8 concentrated by immunoprecipitations(Fig. 7C, lane 4). When a similar analysis was applied on cul-tured cortical neurons isolated from wt and the null animals,endogenous MAP8 in the GAN null cells failed to be poly-ubiquitinated in the absence of gigaxonin (Fig. 7D, lane 2).Significantly, when HA-gigaxonin was re-introduced into theGAN null neurons by transient transfections (Fig. 7D, lane30), the characteristic high molecular shifts of MAP8 poly-ubiquitination, co-immunoprecipitated by anti-Flag-ubiquitinand detected with anti-MAP8-LC, was not only restored butalso enhanced (Fig. 7D, lane 3), compared with that in wtneurons (Fig. 7D, lane 1). In contrast, the gigaxoninmutant-associated GAN, HA-R293X (Fig. 7D, lane 40), exer-cised no recovering effect (Fig. 7D, lane 4), providingadditional evidence that MAP8 ubiquitination requiresfunctional gigaxonin. We conclude that, in contrast to itsmutant counterpart, gigaxonin has essential role in linkingpoly-ubiquitin tags to MAP8. Thus, mutations in gigaxonincould cause failure in UPS function leading to substrateoverstabilization which may have toxic consequence onaffected cells.

Accumulation of MAP8 is toxic to neurons in vitro

Gigaxonin apparently functions to control protein degra-dations of its binding partners including MAP1B (31),TBCB (34) and MAP8. To address whether MAP8 accumu-lation also contributes significantly to the pathogenesis ofGAN, we tested the effect in cultured neurons. The corticalneurons isolated from wt or GAN null mice were cultured inparallel. During the initial 4–6 days in vitro (DIV), abundantneurites were seen in all cultures and formed a dense neuritenetwork indistinguishable between wt and null neurons (datanot shown). However, after �10–12 days in culture, neuritesof the GAN null neurons became progressively shorter, sparserand beaded (Fig. 8A and B); in contrast, wt neurites continu-ously grew denser and longer in the following weeks(Fig. 8C). By 17 days, there was a marked reduction (up to90%) in cell density in cultured null neurons (SupplementaryMaterial, Fig. S2) with apparent protein accumulation ofMAP8 (Fig. 8A and B).

Next, we overexpressed MAP8 protein in wt neurons atDIV 2–3 using transient transfections to elucidate whetherMAP8 accumulation alone is sufficient to cause neurodegen-eration. A transfection efficiency of �5–10% was achievedfor both the GFP control and MAP8. Within the first 24–48 h post-transfection, no significant differences across allcultured neurons were observed. However, filamentousbundles (insets in Fig. 8E) were seen in a subset of MAP8overexpressing neurons after 48–72 h post-transfection;ranging from 15–20% of transfected neurons in five indepen-dent experiments. By days 4–6 post-transfection, �60–70%of the MAP8 overexpressing neurons began to show beadedand shortened neurites with aggregates in cell bodies(Fig. 8F–I). As judged by DAPI staining, all MAP8 trans-fected neurons displayed extensive nuclear disintegration orcondensation 4–6 days after transfection. Shortly thereafter,an overwhelming majority of dying neurons completely lostany sign of normal morphology, a phenotype recapitulating

Figure 5. The kelch-repeat domain of gigaxonin binds directly to theC-terminus of MAP8. (A) pGBKT7-BTB domain or pGBKT7-kelch wasmated to pGADT7-MAP8-C, showing no binding with BTB, but a positivereaction for the kelch-repeat domain of gigaxonin in yeast. (B)Co-immunoprecipitations were conducted on total lysates of COS7 cellsco-transfected either with full-length gigaxonin (HA-Gig-F, lane 3) ormutant gigaxonin (HA-R293X, lane 4) and MAP8 light chain (Fl-M8-LC).The cells were co-IPed with anti-Flag beads and probed with anti-HAantibody. The arrows indicate the wt gigaxonin band in lane 3 and theabsence of the mutant band in lane 4. The total lysates of co-transfectedcells were loaded as controls (lanes 1 and 2). Asterisk indicates IgG chains.(C–E) Co-transfection of HA-gigaxonin (HA-Gig-F) and Flag-MAP8(Fl-M8-F) in COS7 cells showed co-localization of the two proteins. (F–H)The kelch-repeat domain of gigaxonin (HA-Gig-C) and the GFP tagged C-terminus of MAP8 (GFP-M8-CT) displayed identical staining patterns in theco-transfected COS7 cells. (I–K) R293X mutant gigaxonin (HA-R293X)cannot bind the cytoskeletal associated C-terminus of MAP8 in COS7 cells.Bar represents 12 mm.


degeneration and cell death occurring in the culturedGAN null neurons. In contrast, GFP-expressing and untrans-fected wt neurons were indistinguishable over the entiretest period, displaying long axonal processes with signifi-cant branching (Fig. 8D). These findings demonstrate thatoverexpression/accumulation of MAP8 contributed impor-tantly to the neurodegeneration and cell death occurringin GAN null neurons. The formations of filamentousbundles and aggregates were observed to precede devastatingneuronal death.

To further evaluate the pathogenic contribution of accumu-lated MAP8, scrambled or specific MAP8 interfering RNAs(siRNA) co-expressing GFP were introduced into gigaxoninnull neurons at DIV 3–5 to prevent MAP8 accumulation. Asrevealed by immunofluorescence microscopy, no significantchanges could be observed within the first 24 h. However,3 days after transfection, an apparent reduction of MAP8protein level could be seen in the MAP8 siRNA expressingnull neurons. By days 7–10 post-transfection, the MAP8siRNA expressing neurons display substantially improvedsurvival rate (Supplementary Material, Fig. S3) in comparisonwith untransfected null neurons (Fig. 8L and M). In contrast,the null neurons transfected with scramble siRNA begin toundergo degeneration with MAP8 protein accumulating andaggregating in the cells (Fig. 8J and K). Fourteen days aftertransfection, survived MAP8 siRNA expressing neurons stilldisplayed numerous elaborate neurite structures (Fig. 8N),whereas, the scrambled siRNA expressing neurons weremorphologically undetectable. Thus, we conclude that theaccumulation of MAP8 contributes directly to the pathogen-esis of neurodegeneration in GAN. Consistently, althoughmuch improved, MAP8 siRNA expressing neurons could notbe maintained for as long as wt neurons, strongly suggesting

that accumulated MAP8 protein is not the only pathogenicfactor; other toxic factors including accumulated MAP1B(31) are involved in neurodegeneration and cell deathoccurring in GAN.

MAP8 overexpression alters the microtubule networkand traps motor protein in aggregates

How the toxicities mediated by MAP8 accumulation requiredfurther investigation. One possibility is that aberrant MAP8accumulation alters microtubule networks or dynamics thatmay interfere with the movement of motor proteins leadingto impaired axonal transport. We tested this possibilityin cells by overexpressing MAP8 over a prolonged period.In COS7 cells, overexpression of MAP8 caused alterationsof the microtubule network exhibiting a wavy and bundledpattern (Fig. 9A and B) or forming aggregates (Fig. 9Cand D). Most interestingly, the accumulation of MAP8, butnot the control protein (Supplementary Material, Fig. S4),trapped dynein motor protein into insoluble aggregates,as revealed by immunostaining using anti-DIC (Fig. 9Eand F). This finding was further substantiated by immuno-fluorescent analysis on cultured GAN null neurons, display-ing a similar pattern of aggregates positive for DIC andMAP8 antibodies (Fig. 9I and J). We propose that MAP8protein accumulation causes an aberrant microtubulenetwork that interferes with motor protein function leadingto disrupted axonal transport; this explanation could accountfor the apparent vesicular accumulation in the null neurons(Fig. 4F).

Retrograde axonal transport is impaired inthe gigaxonin null neurons

We wish to explore the posited link between abnormaldistribution and accumulations of vesicles in GAN nullaxons and impaired axonal transport. The presence of aggre-gated DIC and Golgi dispersion prompted us to assess retro-grade transport in neurons of dorsal root ganglia (DRG)isolated from post-natal day 0 animals of wt and GAN null,and cultured in a microfluidic chamber (35). At DIV 6–15,the DRG neurons were incubated with BtNGF-QD605, astreptavidin-Quantum dot 605 conjugated to biotinylatedNGF (36), to label a subfraction of vesicular structuresarising from endocytosis. Retrograde transport was subsequen-tly monitored using time-lapse confocal microscopy. Themovements of BtNGF-QD605-labeled vesicles within axonswere analyzed for both their frequency and the persistence.Indeed, in comparison with that in wt neurons at DIV 6–8,retrograde movement of BtNGF-QD605-associated vesicleswas reduced and discontinued in �30% of the null samples(Fig. 10A) (Supplementary Material, Movie S1; Fig. S5). AtDIV 13 and thereafter, the traced vesicles in wt neuronsexhibit processive or sustained motion towards cell bodies(Fig. 10C) (Supplementary Material, Movie S3; Fig. S5).In contrast, in �81% of GAN null neurons, the frequency ofthe labeled vesicle movements were nearly diminished inneurites (Fig. 10B and S2 movie), revealing severely impairedretrograde axonal transport.

Figure 6. Ablation of gigaxonin causes the protein accumulation of MAP8.(A and B) About 25 mg of total lysates of brain (Br), spinal cord (SP) andsciatic nerve (SC) from wild-type (WT, lane 1s) or gigaxonin knockout(KO, lane 2s) mice were analyzed using antibodies to MAP8-HC (M8-HC)and actin in (A) or to DIC and MAP8-LC (M8-LC) in (B). The substantialaccumulation of MAP8 in the null sample. (C and D) RT–PCRs (C) and real-time PCRs (D) were conducted on mRNAs of WT or KO using multipleprimer pairs to MAP8-LC and actin. In contrast to the protein levels, MAP8mRNA levels are unchanged.


Taken together, our data demonstrate that gene targeting ofGAN results in toxic accumulation of proteins includingMAP8 leading to aberrant cytoskeletal network and defectivevesicle transport in the null animals. Impaired retrogradeaxonal transport preceded the observable signs of degenerationmay directly and importantly contribute neurodegenerationand cell death in human GAN.

DISCUSSION

In this report, we describe gigaxonin as a distinct cytoskeletonregulatory protein, important for maintaining a physiologicalprotein level of MAP8 through ubiquitin-mediated degra-dation. Loss of gigaxonin’s function leads to a substantialincrease in MAP8 protein level that has detrimental effects

Figure 7. Gigaxonin enhances protein degradation of MAP8 via UPS. (A) Stability assays of endogenous MAP8-LC. PC12 cells were treated with 50 mg/mlCHX for 0–15 h (lanes 1–7). Eight hours after CHX addition, majority of the proteins still remained. The protein level stayed constant over the testedperiod when both CHX and MG132 (PI) were present (lanes 8–12). (B) Overexpression of gigaxonin in PC12 cells correlates with a decreased level ofMAP8-LC (compare lane 2 with 1), whereas GFP overexpression has no such effect (lanes 3 and 4). Addition of PI, MG132, to the gigaxonin-overexpressingPC12 cells reverses the reduction of MAP8-LC protein level (compare lane 6 with 5). Asterisk indicates non-specific bands. (C) Lanes 1–2: exogenousMAP8-LC (Flag-MAP-LC) protein was poly-ubiquitinated in the presence of MG132 (lane 2) and an excess of HA-ubiquitin. Lanes 3–4: ubiquitination ofendogenous MAP8. Total cell lysates of PC12 cells transfected with HA-ubiquitin were immunoprecipitated by anti-HA agarose beads and immunoblottedwith anti-MAP8-LC, showing poly-ubiquitin conjugations of endogenous MAP8-LC in the presence of MG132 and excess ubiquitin (lane 4). Conjugationsof poly-ubiquitin tags could not be seen in the absence of MG132 for both exogenous and endogenous MAP8 (lanes 1 and 3). (D) Ubiquitination assays oncultured primary neurons. Cortical neurons isolated from wt (lane 1) and the null animals (lanes 2–4) were transfected at DIV 3–5 (days in vitro culture)with Flag-ubiquitin (Fl-Ubi) only (lanes 1–2) or co-transfected either with HA-gigaxonin (HA-Gig) (lane 3) or with HA-gigaxonin mutant (HA-R293X)(lane 4). The total cell lysates prepared from MG132 pre-treated neurons were co-immunoprecipitated with anti-Flag agarose beads (Sigma) and immunoblottedwith anti-MAP8-LC. The poly-ubiquitination in the null samples were restored by introducing gigaxonin’s expression (lane 3), but not by its mutant counterpart(lane 4). A duplicate blot (lanes 20 –40) was probed with anti-HA to confirm the expression of HA-gigaxonin (lane 30) and HA-R293X mutant (lane 40).


on neurons, leading to progressive neuronal degeneration. Theaccumulations of vesicles, mitochondria and other membra-nous structures in GAN null neurons correlate with a markedimpairment of retrograde axonal transport. Our study pointsto impaired retrograde axonal transport as a contributingfactor on pathogenesis of GAN.

GAN is a severe neurodegenerative disorder; this class ofdiseases is among the most puzzling and devastating illnessesin medicine. The causes of neuronal degeneration and deathare still largely unknown. Because of the scarcity of diseasetissues and the inability to conduct rigorously controlledinvestigations with human subjects, the pathogenesis ofGAN was difficult to investigate prior to the development ofanimal models. On the basis of its recessive inheritance andthe types of mutations in gigaxonin, it was proposed that theneurodegeneration occurring in GAN patients is the conse-quence of a loss of gigaxonin function(s). Using conventionalgene targeting, we generated GAN null mice which developedprogressive motor weakness as well as abnormalities innon-neuronal tissues. Although bearing a striking resemblanceto the clinical manifestations of human GAN, the phenotypic

features displayed by GAN null mice were less severe. Theonset of observed motor deficit in the null mice was muchlater (3–7 years old in human patients versus 6–10 monthsin mouse) and the phenotypes progressed more slowly andmoderately. Also GAN null mice appear to have a relativelynormal life span because most of them survived over theage of 20 months. Cytopathologically, the prominent featuresfound in GAN null mice are similar to those in human GANpatients, including NF accumulation, an abnormal microtubulenetwork, mitochondrial swelling and thinned myelin sheaths.However, the giant axons in peripheral nerves that are charac-teristic of human GAN could have been hardly seen in GANnull mice. The later disease onset, slower progression, lesssevere axonal pathology and longer life span in the mousemodel are unexplained, but could result from a number offactors including reduced motor stress for peripheral nervesin the rodent, which walks on four relatively short limbs.A portion of the null animals display no observable or verysubtle neurological phenotypes suggesting that some geneticmodifiers may exist. Nevertheless, our studies providecompelling evidence that the GAN null mice generated

Figure 8. Overexpression of MAP8 provokes neurites degeneration and cell death. (A–C) Cortical neurons from KO (A–B) and WT (C) day 16 embryos werecultured for 15 days. The neurons were stained by anti-MAP8-LC after culture. (D–I) wt cortical neurons were cultured for 48–72 h prior to transfections ofGFP (D) or Flag-MAP8 (E–I). The transfected neurons were processed for immunofluorescence microscopy using mouse anti-Flag (E–I) or GFP (D). TheFlag-M8 transfected neurons show bundle formation (insets in E), neurite fragmentation (F), intracellular aggregates or practically no neurites (G–I). (J–N)Cortical neurons (DIV3-5) transfected with scrambled siRNA (J–K) or MAP8 siRNA (L–N) were stained with rabbit anti-MAP8-LC and mouse anti-GFPat DIV 14 (J–M) or at DIV 19 (N). The bar in (A) represents 50 mm for (A–C and J–N), 60 mm for (D) and 10 mm for (E–I).


herein are an animal model of human GAN. This model cannow be used to observe disease progression, to conduct acomprehensive examination throughout the disease courseand to analyze the pertinent biological pathway(s) and thepathology of the disorder well beyond what is available inpatients. Of more general interest, this genetic model ofneurological diseases in mice may help to elucidate thepathogenesis of other more prevalent neurological disordersassociated with cytoskeletal aberration and UPS deficits.

In addition to MAP1B-LC, gigaxonin can directly bind toMAP8, a new member of microtubule-associated proteinfamily (32,33). It is noteworthy that an efficient binding of

gigaxonin to MAP8 requires functional kelch-repeat domain.At least six out of 24 distinct mutations identified in humanGAN patients (16–18) result in a complete loss of the kelchrepeats, which may consequently lead to disruption of thegigaxonin–MAP8 interaction. Our results on the three testedmutations, 18insA, R242X and R293X, provided directevidence for a disruption. The interaction between gigaxoninand MAP8 is apparently required for the degradation ofMAP8 protein. The overexpression of gigaxonin in transfectedcells significantly increased UPS-mediated MAP1B degra-dation, suggesting an involvement of gigaxonin’s functionin the UPS pathway (31,34). Over the years, the ubiquitinsystem of intracellular protein degradation has been impli-cated in the control of many fundamental cellular processes.Defects in this system have been directly linked to thedevelopment of human diseases, including cancer and neuro-degenerative diseases (37–39). Our results provide additionalevidence linking a defect in UPS-mediated protein degra-dation with a human neurodegenerative disorder.

MAP8 joins a growing family of classical microtubule-associated proteins. It has a generalized pattern of tissue distri-butions and contains direct binding sites to microtubules andactin microfilaments. Emerging evidence indicates that MAPexpression levels are under tight control in response tomorphological and functional requirements of neuron (40,41).Regardless of the details of how gigaxonin overexpressionaccelerates protein degradation in transfected cells, the signi-ficant protein accumulation of MAP8 in GAN null animalsverifies that gigaxonin is a critical player in controlling theprotein levels of MAP8 in addition to MAP1B and TBCB(31,34). Significant protein reduction of MAP8 throughRNAi decreased but did not entirely eliminate death of GANnull neurons in culture. This suggests that accumulation ofMAP8 is sufficient to cause neurodegeneration, but does notserve as the only factor. Thus, gigaxonin may be involved indegradation of proteins and/or in other important pathways.The mechanisms underlying cell death mediated by MAP8accumulation must be elucidated. Several possibilities havebeen proposed. One possibility is that MAP8 accumulationmay interfere with the movement of motor proteins by directlycompeting for or blocking adjacent binding sites on themicrotubule surface. Several studies suggest this possibility,as MAP2 can inhibit the motility of dynein or kinesin invitro (42,43). Another possibility is that aberrant elevationof MAP levels affects transport processes indirectly by alter-ing microtubule dynamics. Our data suggest that accumulatedMAP8 protein alters microtubule network and traps the dyneinmotor protein into aggregates. The market accumulations ofmembrane structures in the GAN null neurons and theresults of our transport assays point to impaired retrogradeaxonal transport as a potential factor causing neurodegenera-tion. Our study speaks to the pathogenesis of human GAN.It is now possible to envision that gigaxonin disruption,either directly or indirectly, represents a central pathologicalstep in several neurodegenerative diseases that feature cyto-skeletal abnormalities and aberrant dysfunction of axonaltransport. Whether the featured demyelination of axons inGAN is a primary defect directly involving gigaxonin functionor is a pathogenic consequence requires to be elucidated.Whether the densely packed NFs are a compensatory response

Figure 9. MAP8 accumulation alters the microtubule network and traps DICaggregates. COS7 cells transfected with Flag-MAP8 (Fl-M8-F) were culturesfor 72–96 h and subjected to double staining using rabbit anti-Flag (A–F)with mouse anti-tubulin (A–D) or with mouse anti-DIC (E–F). Not onlymicrotubules (D) but also the DIC protein (F) was trapped in the insolubleaggregate structures. The double staining of anti-DIC and anti-MAP8 oncultured cortical neurons shows an even distribution in wt (G and H) andaggregated pattern in the null samples (I and J). The bar in (A) represents12 mm for (A–D) and 10 mm for (E–J).


to the altered microtubule network or whether gigaxoninprimarily involves in IF network through a yet unknownfactor awaits further investigations.

MATERIALS AND METHODS

Generation of antibodies

Gigaxonin antibodies were generated and purified as describedpreviously (30). A purified GST fusion protein encompassingN-terminal portion of the light chain or a synthetic peptidelocated within the heavy chain was used for raisinganti-M8-C or anti-M8-N antibodies in rabbits, respectively.The antisera were affinity purified and tested against purifiedrecombinant protein or total cell lysates from Flag-M8-HCor Flag-M8-LC transfected cells prior to use.

Generation of the GAN knockout mice

A 16 kb mouse genomic DNA fragment was cloned from themouse 129Sv/Ev lambda genomic library. The targetingvector was constructed as depicted in Figure 1A. We madeour 30-arm relatively short to make PCR screening andmouse genotyping easier, whereas the long 50arm was toenhance homologous recombination. A 1.5 kb PCR DNA frag-ment amplified with primer pair of GANYA1 and GANYA2was used as the 30-short arm. Primer GANYA1 is located1.65 kb downstream of exon 5, whereas primer GANYA2 islocated 150 bp downstream of exon 5. The short arm wasinserted into the 50 of the Neo gene cassette using the MluIsite. The long arm was a 7.3 kb SpeI genomic fragment.In this strategy, exons 3–5 containing �5 kb genomicregion were replaced by the 1.8 kb Neo gene cassette. Forthe screening strategy, as depicted in Figure 1A, the primerpair of Neo1 and mGigDiag31 was designed for assuring 30

arm homologous integration and the pair of Neo 2 and mGig-Diag52 for confirming 50 arm integration. Of a total of �300ES clones screened with sequential PCRs for both arms, threetargeted ES clones were scored positive for desired homolo-gous recombination, which were confirmed by Southern blotanalysis. Sufficient numbers of high-contribution chimericswere generated from these three independent ES cloneswhich successfully contributed to the germline transmission.Heterozygous offspring derived from all three independentES clones were mated to produce homozygous mutants, andtheir offspring were genotyped to determine whether theycarry the gigaxonin null mutants (Fig. 1B). To confirm unequi-vocally that the expected targeting event had taken place, weanalyzed gigaxonin’s expression at both mRNA and proteinlevels, confirming that the gene targeting successfullyresulted in ablation of gigaxonin’s expression in mouse(Fig. 1C and D).

Electron microscopy

WT and gigaxonin null animals were sacrificed by intravenousperfusion with 2% paraformaldehyde and 0.05% glutaral-dehyde. The dissected brain and sciatic nerves samples werepost-embedded as described previously (30). The sectionswere analyzed under a Philips CM10 microscope.

Cortical neuron culture

Cortical neuron cultures were prepared from mouse embryosat days 15–18 (E15–E18). The procedures were similar tothose described in Current Protocol of Neurobiology.Briefly, the mouse embryo brains were removed andenzymatically dissociated with 0.25% trypsin (Sigma) indissociation buffer (15 mM HEPES, 1% sodium bicarbonatein 1 � Hanks buffer). The dissociated cells were then platedon 15 mm diameter cover slips coated with poly-L-lysine(50 mg/ml) at a density of 100–160 cells/mm2. The corticalneurons were then grown in neurobasal medium containingB27, 0.5 mM L-glutamine and 1% Penicillin–Streptomycinsupplements suggested in Gibco protocol (Gibco). Thecultures could be maintained up to 4 weeks.

Figure 10. Retrograde axonal transport is impaired in the null neurons. TheDRG neurons isolated from wt (C) and KO (A and B) and cultured in amicrofluidic chamber for 8 (A) or 15 days (B and C) were assessed fortheir retrograde transport using streptavidin-Quantum dot 605 conjugatedNGF (BtNGF-QD605) as the tracer. Time-lapse images of movement ofBtNGF-QD605 within the axon in the microgroove were captured usingconfocol microscopy and the transport velocity was analyzed using Image J.The arrows point to a moving vesicle associated with BtNGF-QD605,which migrated retrogradely along the axon toward the cell body at anaverage velocity of �0.9 mm/s for wt and �0.4 mm/s for the null samplesat DIV 8.


For transfections, the cortical neurons were seeded at adensity of 1–1.5 � 104 cells/well in a 12-well plate. Theexpression vectors of Flag-MAP8 or pEGFPC2 were trans-fected at culture days 3–5 using OptiMEM Iw and LIPOFEC-TAMINE 2000TM according to manufacturer’s instructions(Invitrogen, Rockville, MD, USA). The cultures were termi-nated for immunofluorescent microscopy analyses at varioustime points for assays.

For siRNA experiments, the pZ-OFF EGFP vector was thegenerous gift from Dr Craig Garner’s laboratory. The follow-ing oligos for the siRNA constructs (Genscript, USA) werecloned into the pZ-OFF EGFP vector using BglII and HIIIsites. MAP8 siRNA: CTGCAACTGTAGCTGCCAAGA;scrambled siRNA: CGGCTTGACCATCATGAGCTA.

Real-time PCR

Real-time PCR was performed using Mx3000P QPCR system(Stratagene) and SYBR Green Master Mix (Stratagene600548). The MAP8 primer sequences were as follows.Forward: 50-CAGCGCCCGCAGTGAGCCTG-30 and reverse:50-GGTAGGCCAGGTC CAGGTATAC-30. Master Mix,primers and mouse cDNA were mixed according tomanufacturer’s instructions. After initial incubation at 958 for10 min, the samples were amplified for 40 cycles at 958 for30s, 588 for 1 min and 728 for 30s. Data were analyzed usingthe comparative quantification method by the 22DDCT equation,DDCT ¼ DCT (gigaxonin-null) 2 DCT (wild-type), where DCT

is the threshold cycle (CT) value of the housekeeping gene(GADPH ) subtracted from the CT value of the targetgene (MAP8).

SUPPLEMENTARY MATERIAL

Supplementary Material is available at HMG Online.

ACKNOWLEDGEMENTS

We thank Dr Susanna M. Grzeschik for technical advises onreal-time PCR and other colleagues of Yang Lab for scientificdiscussions. The pZ-OFF EGFP vector was the generous giftof Dr Craig Garner. We also thank Prasanthi Bhagavatulafor technical assistance. This work was supported by MuscularDystrophy Association Research Grant and NIH grantsNS42791 and NS43281 for Y.Y.

Conflict of Interest statement. None declared.

REFERENCES

1. Berg, B.O., Rosenberg, S.H. and Asbury, A.K. (1972) Giant axonalneuropathy. Pediatrics, 49, 894–899.

2. Igisu, H., Ohta, M., Tabira, T., Hosokawa, S. and Goto, I. (1975) Giantaxonal neuropathy. A clinical entity affecting the central as well as theperipheral nervous system. Neurology, 25, 717–721.

3. Ouvrier, R.A. (1989) Giant axonal neuropathy. A review. Brain Dev., 11,207–214.

4. Peiffer, J., Schlote, W., Bischoff, A., Boltshauser, E. and Muller,G. (1977) Generalized giant axonal neuropathy: a filament-formingdisease of neuronal, endothelial, glial, and schwann cells in a patientwithout kinky hair. Acta Neuropathol., 40, 213–218.

5. Treiber-Held, S., Budjarjo-Welim, H., Reimann, D., Richter, J.,Kretzschmar, H.A. and Hanefeld, F. (1994) Giant axonal neuropathy: ageneralized disorder of intermediate filaments with longitudinal groovesin the hair. Neuropediatrics, 25, 89–93.

6. Asbury, A.K., Gale, M.K., Cox, S.C., Baringer, J.R. and Berg, B.O.(1972) Giant axonal neuropathy—a unique case with segmentalneurofilamentous masses. Acta Neuropathol., 20, 237–247.

7. Tandan, R., Little, B.W., Emery, E.S., Good, P.S., Pendlebury, W.W. andBradley, W.G. (1987) Childhood giant axonal neuropathy. Case report andreview of the literature. J. Neurol. Sci., 82, 205–228.

8. Prineas, J.W., Ouvrier, R.A., Wright, R.G., Walsh, J.C. and McLeod, J.G.(1976) Gian axonal neuropathy–a generalized disorder of cytoplasmicmicrofilament formation. J. Neuropathol. Exp. Neurol., 35, 458–470.

9. Koch, T., Schultz, P., Williams, R. and Lampert, P. (1977) Giant axonalneuropathy: a childhood disorder of microfilaments. Ann. Neurol, 1,438–451.

10. Donaghy, M., King, R.H., Thomas, P.K. and Workman, J.M. (1988)Abnormalities of the axonal cytoskeleton in giant axonal neuropathy.J. Neurocytol., 17, 197–208.

11. Sabatelli, M., Bertini, E., Ricci, E., Salviati, G., Magi, S., Papacci, M. andTonali, P. (1992) Peripheral neuropathy with giant axons andcardiomyopathy associated with desmin type intermediate filaments inskeletal muscle. J. Neurol. Sci., 109, 1–10.

12. King, R.H., Sarsilmaz, M., Thomas, P.K., Jacobs, J.M., Muddle, J.R. andDuncan, I.D. (1993) Axonal neurofilamentous accumulations: acomparison between human and canine giant axonal neuropathy and2,5-HD neuropathy. Neuropathol. Appl. Neurobiol., 19, 224–232.

13. Ben Hamida, C., Cavalier, L., Belal, S., Sanhaji, H., Nadal, N., Barhoumi,C., M’Rissa, N., Marzouki, N., Mandel, J.L., Ben Hamida, M. et al.(1997) Homozygosity mapping of giant axonal neuropathy gene tochromosome 16q24.1. Neurogenetics, 1, 129–133.

14. Flanigan, K.M., Crawford, T.O., Griffin, J.W., Goebel, H.H.,Kohlschutter, A., Ranells, J., Camfield, P.R. and Ptacek, L.J. (1998)Localization of the giant axonal neuropathy gene to chromosome 16q24.Ann. Neurol., 43, 143–148.

15. Cavalier, L., BenHamida, C., Amouri, R., Belal, S., Bomont, P., Lagarde,N., Gressin, L., Callen, D., Demir, E., Topaloglu, H. et al. (2000) Giantaxonal neuropathy locus refinement to a ,590 kb critical interval.Eur. J. Hum. Genet., 8, 527–534.

16. Bomont, P., Cavalier, L., Blondeau, F., Ben Hamida, C., Belal, S., Tazir,M., Demir, E., Topaloglu, H., Korinthenberg, R., Tuysuz, B. et al. (2000)The gene encoding gigaxonin, a new member of the cytoskeletal BTB/kelch repeat family, is mutated in giant axonal neuropathy [see comment].Nat. Genet., 26, 370–374.

17. Kuhlenbaumer, G., Young, P., Oberwittler, C., Hunermund, G.,Schirmacher, A., Domschke, K., Ringelstein, B. and Stogbauer, F. (2002)Giant axonal neuropathy (GAN): case report and two novel mutationsin the gigaxonin gene [see comment; erratum appears in Neurology 2002,58, 1444]. Neurology, 58, 1273–1276.

18. Bomont, P., Ioos, C., Yalcinkaya, C., Korinthenberg, R., Vallat, J.M.,Assami, S., Munnich, A., Chabrol, B., Kurlemann, G., Tazir, M. et al.(2003) Identification of seven novel mutations in the GAN gene.Hum. Mutat., 21, 446.

19. Timmerman, V., De Jonghe, P. and Van Broeckhoven, C. (2000) Of giantaxons and curly hair [comment]. Nat. Genet., 26, 254–255.

20. Hernandez, M.C., Andres-Barquin, P.J., Martinez, S., Bulfone, A.,Rubenstein, J.L. and Israel, M.A. (1997) ENC-1: a novel mammaliankelch-related gene specifically expressed in the nervous system encodesan actin-binding protein. J. Neurosci., 17, 3038–3051.

21. von Bulow, M., Heid, H., Hess, H. and Franke, W.W. (1995) Molecularnature of calicin, a major basic protein of the mammalian sperm headcytoskeleton. Exp. Cell Res., 219, 407–413.

22. Itoh, K., Wakabayashi, N., Katoh, Y., Ishii, T., Igarashi, K., Engel, J.D.and Yamamoto, M. (1999) Keap1 represses nuclear activation ofantioxidant responsive elements by Nrf2 through binding to theamino-terminal Neh2 domain. Genes Dev., 13, 76–86.

23. Adams, J., Kelso, R. and Cooley, L. (2000) The kelch repeat superfamilyof proteins: propellers of cell function. Trends Cell Biol., 10, 17–24.

24. Pintard, L., Willis, J.H., Willems, A., Johnson, J.L., Srayko, M., Kurz, T.,Glaser, S., Mains, P.E., Tyers, M., Bowerman, B. et al. (2003) The BTBprotein MEL-26 is a substrate-specific adaptor of the CUL-3ubiquitin-ligase. Nature, 425, 311–316.


25. Geyer, R., Wee, S., Anderson, S., Yates, J. and Wolf, D.A. (2003)

BTB/POZ domain proteins are putative substrate adaptors for cullin 3ubiquitin ligases. Mol. Cell, 12, 783–790.

26. van den Heuvel, S. (2004) Protein degradation: CUL-3 and BTB—partners in proteolysis. Curr. Biol., 14, R59–R61.

27. Cullinan, S.B., Gordan, J.D., Jin, J., Harper, J.W. and Diehl, J.A. (2004)

The Keap1-BTB protein is an adaptor that bridges Nrf2 to a Cul3-basedE3 ligase: oxidative stress sensing by a Cul3-Keap1 ligase. Mol. Cell.

Biol., 24, 8477–8486.

28. Zhang, D.D., Lo, S.C., Cross, J.V., Templeton, D.J. and Hannink,M. (2004) Keap1 is a redox-regulated substrate adaptor protein for aCul3-dependent ubiquitin ligase complex. Mol. Cell. Biol., 24,

10941–10953.

29. Furukawa, M. and Xiong, Y. (2005) BTB protein Keap1 targetsantioxidant transcription factor Nrf2 for ubiquitination by the Cullin

3-Roc1 ligase. Mol. Cell. Biol., 25, 162–171.

30. Ding, J., Liu, J.J., Kowal, A.S., Nardine, T., Bhattacharya, P., Lee, A. and

Yang, Y. (2002) Microtubule-associated protein 1B: a neuronal bindingpartner for gigaxonin [erratum appears in J. Cell Biol. 2003, 163, 189].J. Cell Biol., 158, 427–433.

31. Allen, E., Ding, J., Wang, W., Pramanik, S., Chou, J., Yau, V. and Yang,Y. (2005) Gigaxonin-controlled degradation of MAP1B light chain iscritical to neuronal survival. Nature, 438, 224–228.

32. Ding, J., Valle, A., Allen, E., Wang, W., Nardine, T., Zhang, Y., Peng,L. and Yang, Y. (2006) Microtubule-associated protein 8 contains twomicrotubule binding sites. Biochem. Biophys. Res. Commun., 339,

172–179.

33. Orban-Nemeth, Z., Simader, H., Badurek, S., Trancikova, A. and Propst,

F. (2005) Microtubule-associated protein 1S, a short and ubiquitouslyexpressed member of the microtubule-associated protein 1 family. J. Biol.Chem., 280, 2257–2265.

34. Wang, W., Ding, J., Allen, E., Zhu, P., Zhang, L., Vogel, H. and Yang,Y. (2005) Gigaxonin interacts with tubulin folding cofactor B and controlsits degradation through the ubiquitin–proteasome pathway. Curr. Biol.,15, 2050–2055.

35. Taylor, A.M., Blurton-Jones, M., Rhee, S.W., Cribbs, D.H., Cotman,C.W. and Jeon, N.L. (2005) A microfluidic culture platform for CNSaxonal injury, regeneration and transport. Nat. Methods, 2, 599–605.

36. Bronfman, F.C., Tcherpakov, M., Jovin, T.M. and Fainzilber, M. (2003)Ligand-induced internalization of the p75 neurotrophin receptor: a slowroute to the signaling endosome. J. Neurosci, 23, 3209–3220.

37. Watts, R.J., Hoopfer, E.D. and Luo, L. (2003) Axon pruning duringDrosophila metamorphosis: evidence for local degeneration andrequirement of the ubiquitin–proteasome system. Neuron, 38, 871–885.

38. Zhai, Q., Wang, J., Kim, A., Liu, Q., Watts, R., Hoopfer, E., Mitchison,T., Luo, L. and He, Z. (2003) Involvement of the ubiquitin–proteasomesystem in the early stages of wallerian degeneration. Neuron, 39,217–225.

39. Ehlers, M.D. (2004) Deconstructing the axon: Wallerian degenerationand the ubiquitin–proteasome system. Trends Neurosci., 27, 3–6.

40. Tucker, R.P. and Matus, A.I. (1987) Developmental regulation of twomicrotubule-associated proteins (MAP2 and MAP5) in the embryonicavian retina. Development, 101, 535–546.

41. Lu, R., Wang, H., Liang, Z., Ku, L., O’Donnell, W.T., Li, W., Warren,S.T. and Feng, Y. (2004) The fragile X protein controlsmicrotubule-associated protein 1B translation and microtubule stability inbrain neuron development. Proc. Natl Acad. Sci. USA, 101,15201–15206.

42. Paschal, B.M., Obar, R.A. and Vallee, R.B. (1989) Interaction of braincytoplasmic dynein and MAP2 with a common sequence at the C terminusof tubulin [see comment]. Nature, 342, 569–572.

43. Lopez, L.A. and Sheetz, M.P. (1993) Steric inhibition of cytoplasmicdynein and kinesin motility by MAP2. Cell Motil. Cytoskeleton, 24, 1–16.


Documents

Gene targeting of GAN in mouse causes a toxic accumulation ......INTRODUCTION Giant axonal neuropathy (GAN), ﬁrst described in 1972 (1), is a severe, autosomal recessive motor and