|Home | About | Journals | Submit | Contact Us | Français|
We report that eight heterozygous missense mutations in TUBB3, encoding the neuron-specific β-tubulin isotype III, result in a spectrum of human nervous system disorders we now call the TUBB3 syndromes. Each mutation causes the ocular motility disorder CFEOM3, whereas some also result in intellectual and behavioral impairments, facial paralysis, and/or later-onset axonal sensorimotor polyneuropathy. Neuroimaging reveals a spectrum of abnormalities including hypoplasia of oculomotor nerves, and dysgenesis of the corpus callosum, anterior commissure, and corticospinal tracts. A knock-in disease mouse model reveals axon guidance defects without evidence of cortical cell migration abnormalities. We show the disease-associated mutations can impair tubulin heterodimer formation in vitro, although folded mutant heterodimers can still polymerize into microtubules. Modeling each mutation in yeast tubulin demonstrates that all alter dynamic instability whereas a subset disrupts the interaction of microtubules with kinesin motors. These findings demonstrate normal TUBB3 is required for axon guidance and maintenance in mammals.
Nervous system development is highly dependent upon the microtubule cytoskeleton. Microtubules are copolymers assembled from tubulin heterodimers, which contain several different α- and β-tubulin isotypes encoded by separate genes (Lopata and Cleveland, 1987). Microtubule behavior varies according to isotype composition, suggesting each isotype may have properties necessary for specific cellular functions (Joshi and Cleveland, 1990; Luduena, 1993); however, precise functions for most tubulin isotypes remain poorly characterized. β-tubulin isotype III (TUBB3), one of at least six -tubulins found in mammals, is distinct because purified microtubules enriched in TUBB3 are considerably more dynamic than those composed from other β-tubulin isotypes (Panda et al., 1994), and because its expression is primarily limited to neurons (Katsetos et al., 2003). TUBB3 expression is greatest during periods of axon guidance and maturation; levels decrease in the adult central nervous system (CNS) but remain high in the peripheral nervous system (PNS) (Jiang and Oblinger, 1992). Thus, the unique dynamic properties and spatio-temporal expression pattern of TUBB3 suggest it could have a specific function for nervous system development and axon maintenance.
The development of human brainstem ocular motor neurons is particularly vulnerable to gene mutations that affect cytoskeletal proteins and axon guidance (Miyake et al., 2008; Yamada et al., 2003). Congenital fibrosis of the extraocular muscles type 3 (CFEOM3) is a rare ocular motility disorder in which affected individuals are born with blepharoptosis (drooping eyelids) and restricted eye movements (Doherty et al., 1999; Mackey et al., 2002). Using CFEOM3 as a marker for gene mutations that regulate human nervous system development and function, we now report that 8 different heterozygous missense mutations in TUBB3, also known as TuJ1, result in CFEOM3 in isolation or as a component of several previously undefined neurological syndromes. Neuroradiological and pathological findings in humans and a knock-in mouse model, respectively, demonstrate that TUBB3 is necessary for guidance of commissural fibers and cranial nerves. Furthermore, disease-associated mutations can alter microtubule dynamics, and a subset perturbs the interaction of microtubules with kinesin motor proteins. Thus, our work to define the TUBB3 syndromes establishes the requirement for a neuronal β-tubulin isotype in axon guidance and normal brain development.
CFEOM3 in the absence of additional neurological signs or symptoms (‘isolated CFEOM3’) is a ~90% penetrant autosomal dominant disorder that had previously been mapped to chromosome 16q in pedigrees BN and DP (OMIM#600638, Figure S1A, Table S1A, B) (Doherty et al., 1999; Mackey et al., 2002). The critical region for the CFEOM3 gene was 3.5 Mb and flanked by D16S498-16qter. To identify the CFEOM3 gene, we screened coding exons and intron-exon boundaries of positional candidates in probands from BN, DP, and additional families with isolated CFEOM3. We identified three heterozygous TUBB3 missense changes in 15 unrelated pedigrees: 784C>T (R262C) in 11 pedigrees, 904G>A (A302T) in three pedigrees, and 185G>A (R62Q) in one pedigree (Figure S1B, Table S1B, C, D).
We had ascertained study participants with CFEOM and additional neurological symptoms and, given the pan-neuronal expression of TUBB3 in humans (http://www.hudsen.org, HUDSEN Human Gene Expression Spatial Database, ID: 411), we next sequenced DNA from these probands. We identified five additional heterozygous TUBB3 missense changes in 13 unrelated pedigrees. 1249G>C (D417H) and 1249G>A (D417N) alter the same residue and co-segregate in a dominant fashion in one and four pedigrees, respectively. The remaining mutations, 1138C>T (R380C), 785G>A (R262H), and 1228G>A (E410K), were found in one, two, and six pedigrees, respectively, and each arose de novo as sporadic disease or from presumed germ-line mosaicism (Figure S1B, Table S1A, E). Each of the eight mutations segregated with the TUBB3 phenotype, was absent in parents of sporadic individuals, and was not present on over 1700 control chromosomes. The independent nature of the recurrent mutations is supported by de novo occurrences, ethnic and geographic diversity among probands, and multiple disease-associated haplotypes (Table S1C).
Congenital ocular motility defects resulting from R262C, A302T, R380C, and D417N amino acid substitutions ranged from mild to severe (Figure 1A–E), as previously described for pedigree BN (Doherty et al., 1999), whereas all participants with R262H, E410K, and D417H had severe CFEOM3 and congenital facial weakness (Figure 1G, H). Many subjects had aberrant eye movements and several had ptotic eyelid elevation associated with synkinetic jaw movements (Marcus Gunn phenomenon), clinical manifestations of aberrant innervation of cranial musculature by the trigeminal nerve. We conducted magnetic resonance imaging (MRI) of the intracranial motor nerves and orbital contents of affected members of four R262C or D417N pedigrees. Similar to imaging of individuals with KIF21A missense mutations which cause the isolated oculomotility disorder, CFEOM1 (Demer et al., 2005), we found hypoplasia of the oculomotor nerve and the muscles innervated by its superior division - the levator palpebrae superioris and superior rectus - as well as the medial rectus muscle innervated by its inferior division (Figures 1J-L, S1C). The oculomotor nerve also aberrantly innervated the lateral rectus muscle, normally innervated by the abducens nerve. Thus, ocular motility restrictions and/or synkinetic lid elevation with jaw movements could be explained by axon guidance defects.
All subjects harboring D417H or R262H were born with congenital wrist and finger contractures, suggesting maldevelopment of spinal motor neurons (Figure 1I), and developed lower extremity weakness and sensory loss in the first decade of life. Most subjects harboring D417N and the oldest patient harboring E410K developed lower extremity weakness and sensory loss in the second to third decade, all in the absence of congenital contractures (Figure 1F). Electromyography revealed chronic, generalized sensorimotor polyneuropathy that was predominantly axonal, and diagnosed as Charcot-Marie-Tooth Type 2 (CMT2) in some subjects. Several participants harboring D417N without CFEOM3 developed polyneuropathy (Table S1E), suggesting that TUBB3 mutations can cause an isolated CMT2-like disorder.
Brain imaging was reviewed from individuals with each TUBB3 mutation except D417H, and appeared normal only in the R62Q subject. Common findings were dysgenesis of the corpus callosum (CC), anterior commissure (AC), and internal capsule; generalized loss of white matter; and basal ganglia dysmorphisms that correlated with specific mutations (Figure 2). No images showed cortical dysplasia or evidence of cortical migration defects. Intellectual and behavioral impairments generally correlated with the severity of CC dysgenesis. Individuals with A302T, E410K, R262H, and R380C had more severe CC dysgenesis and mild to moderate intellectual, social, and behavioral impairments. By contrast, those with R62Q, R262C, or D417N substitutions had absent or mild CC dysgenesis and most were developmentally normal (Table S1).
To further examine the nature of the nervous system defects in humans, we generated a disease mouse model harboring the most common amino acid substitution (R262C) (Figure S2A). Wild-type (WT) and Tubb3+/R262C mice were born at the expected Mendelian frequencies; heterozygous mice appeared healthy, did not display external eye phenotypes, and had histologically normal appearing brains. Similarly, the R262C phenotype in humans can be non-penetrant, and when penetrant is limited to CFEOM3. By contrast, Tubb3R262C/R262C mice failed to breathe normally and died within hours of birth. Overall brain size was similar between WT and Tubb3R262C/R262C littermates, although basal ganglia asymmetries were sometimes present.
Histological analysis at E18.5 revealed normal cortical layer thickness and architecture in both homozygous and heterozygous mice versus WT littermates, and neocortical layer-specific markers confirmed that layering was preserved (Figures 3A–H, S2C). The hippocampus and dentate gyrus also appeared normal (Figure S2B). Thus, consistent with human MR findings, heterozygous and homozygous R262C mice do not show evidence of cortical cell migration defects.
Tubb3R262C/R262C mice showed defects in the guidance of commissural axons and cranial nerves. There was significant thinning and/or absent midline crossing of the AC throughout its anterior-posterior axis compared to WT mice; it appeared tortuous and often had aberrant fiber projections at the midline. The AC was also thinner at the midline in Tubb3+/R262C mice, similar to human patients (Figures 3I, K, S2D). Agenesis of the CC with bundles of stalled axons (Probst bundles) adjacent to the midline was observed in 2/5 homozygous mutant mice, both of which were on a mixed C57BL/6J:129S6 background (Figure 3J, L). By contrast, CC morphology was normal in all 18 mixed background WT mice analyzed. In the three homozygous mice in which the CC crossed the midline, it appeared abnormally thin in two and thick in one (Figure 3K). Whole-mount neurofilament staining at E11.5–E12 revealed numerous defects in the guidance and branching of cranial nerves (Figure 3M–P). The oculomotor nerve failed to reach the correct muscle anlage and instead projected toward the position of the superior oblique muscle that is normally innervated by the trochlear nerve. Trochlear nerve growth was often stalled, and the trigeminal nerve failed to grow and branch properly compared to WT littermates. These data strongly support a primary defect in the guidance of axons.
There was a decrease in the level of TUBB3 protein in Tubb3R262C/R262C and Tubb3+/R262C versus WT mice. Although protein levels in homozygous mutant mice were approximately 30% of WT values, remaining mutant protein incorporated into microtubules that were polymerized in vitro from brain extracts, as well as into microtubules throughout the cytoskeleton of dissociated cortical and hippocampal neurons (Figure 4A, B, D). Because Tubb3 is thought to be the most dynamic β-tubulin isotype, we asked if protein loss and/or the remaining mutant protein altered microtubule stability in the brain. We detected an approximate 30% increase in the levels of de-tyrosinated α-tubulin (Figure 4C), a post-translational modification that is indicative of stable microtubules (Webster et al., 1987). We also measured the steady-state level of tubulin polymerization in brain extracts from WT and homozygous mutant littermates. Although overall -tubulin levels were decreased in Tubb3R262C/R262C brain lysates, there was an increase in the amount of tubulin polymerization as detected by a larger microtubule pellet (Figure 4D). Taken together, these data suggest that microtubule stability is increased in Tubb3R262C/R262C mutants.
We next asked whether Kif21a microtubule interactions are altered in Tubb3R262C/R262C mutants because heterozygous mutations in this kinesin in humans cause isolated ocular motor dysfunction with clinical and neuroradiological signs of cranial nerve axon guidance defects (Demer et al., 2005; Yamada et al., 2003). We analyzed the ability of Kif21a to co-purify with microtubules isolated from the brains of Tubb3R262C/R262C mutant and WT littermates in the presence of ATP. A significant decrease of Kif21a was detected with microtubules from the brains of Tubb3R262C/R262C mice (Figure 4E). Thus, reduced Tubb3 microtubule incorporation and/or the residual mutant protein affect the ability of Kif21a to interact with microtubules and may contribute to oculomotor nerve misguidance in mice and humans. Commissural axon guidance defects and other TUBB3 disease-specific phenotypes not found in CFEOM1 patients suggest that additional motors and/or MAPs beyond Kif21a may also be affected.
A series of interactions with protein chaperones fold quasi-native tubulin monomers into functional αβ-tubulin heterodimers that can then polymerize into microtubules (Lewis et al., 1996). Point mutations in α- and β-tubulin can impair interactions with chaperone proteins, resulting in reduced heterodimer formation in mammalian cells (Jaglin et al., 2009; Keays et al., 2007), and could explain low protein levels observed in mutant mice. Thus, we asked whether R262C and the other seven substitutions affect the formation of TUBB3 heterodimers.
The coding sequences of WT and all eight TUBB3 mutants were fused to a C-terminal V5 epitope tag (TUBB3-V5) and expressed in rabbit reticulocyte lysate. This lysate is a cell-free system that contains necessary molecular chaperone proteins for de novo tubulin heterodimer formation (Cleveland et al., 1978); resulting heterodimers are comprised of rabbit α-tubulin and WT or mutant TUBB3-V5. WT and mutant proteins were expressed and translated at equivalent levels. Under native conditions, R62Q, R262C, A302T, and R380C showed significant reductions in heterodimer yield, whereas the remaining four mutants generated moderate yields (Figure 4F). This latter group was mixed with bovine brain tubulin and taken through two successive rounds of microtubule polymerization and depolymerization. The ability of these mutant heterodimers to co-cycle with native tubulin in vitro was equivalent to WT (Figure 4G).
WT and mutant TUBB3-V5 expression constructs were next transfected into HeLa cells that have well demarcated microtubules. Immunostaining against the C-terminal V5 tag and α-tubulin to detect the overall microtubule network revealed incorporation of WT and mutant heterodimers. Cells expressing R62Q and R262C TUBB3, however, had lower and more punctuate microtubule incorporation, whereas A302T, R380C, and the remaining mutants showed robust incorporation that was similar to WT (Figure S3). Interestingly, patients harboring R62Q or R262C have the mildest phenotypes, and this could result from lower amounts of mutant heterodimer incorporation into microtubules compared to the other mutants. Thus, although four disease-associated substitutions show scant heterodimer formation in vitro, all of the mutants can be incorporated into microtubules in mammalian cells, albeit at varying levels.
TUBB3-disease substitutions reside in regions of β-tubulin implicated in the regulation of microtubule dynamics, motor protein trafficking, and interactions with microtubule associated proteins (MAPs) (Li et al., 2002; Lowe et al., 2001) (Figure 5A–D, Movie S1). Residues R62 and A302 reside in regions proposed to mediate lateral interactions between longitudinal units of αβ-tubulin heterodimers, called protofilaments, which assemble to form cylindrical microtubules. Lateral protofilament interactions aid microtubule assembly and regulate dynamics (Nogales and Wang, 2006). By contrast, residues R380, E410, and D417 are found in paired alpha-helices H11 (R380) and H12 (E410, D417) on the external surface of microtubules that mediates interactions with numerous motor proteins and MAPs (Al-Bassam et al., 2002; Uchimura et al., 2006). Residue R262 is located in the loop between helix H8 and strand 7 of β-tubulin below helix H12, and forms a putative hydrogen bond with H12 through the carbonyl oxygen of residue D417. Upon mutation, this hydrogen bond is abolished, potentially affecting the tertiary protein structure and motor protein interactions with microtubules (Figure 5E).
The locations of the TUBB3 substitutions suggest that some could directly affect lateral protofilament interactions and microtubule dynamics, whereas others may also perturb motor protein interactions. Notably, these substitutions can result in diagnostically distinct phenotypes (Figure 5F), leading us to ask whether they could alter TUBB3 function and cause dominant effects upon microtubule behavior.
The sequence and protein structure of β-tubulin is conserved between different isotypes and across species (Luduena, 1993). TUBB3 shares considerable homology with the single β-tubulin isotype in budding yeast (Tub2p), including conservation of all disease-associated residues (Figure S4). Budding yeast provides advantages for studying the behavior of microtubules harboring TUBB3-disease substitutions (Reijo et al., 1994). Mutations can be introduced into TUB2 by homologous recombination, and a single β-tubulin isotype avoids the potential diminution of mutant phenotypes due to an abundance of other tubulin isotypes.
To examine potential dominant effects of disease associated mutations, we inserted each into the yeast TUB2 locus, and two independent heterozygous diploid strains for each mutation were isolated and grown on nutrient rich media. Heterozygous diploids were recovered at the expected frequency and did not display growth defects on rich medium, establishing that all of the mutant heterozygous strains were viable. However, only R62Q and R380C haploid spores were viable when present as the sole copy of TUB2, and they grew slowly compared to WT (Table S2).
To determine if mutant substitutions affect microtubule dynamics, WT, heterozygous, and surviving haploid TUB2 mutant strains were plated on media containing increasing amounts of benomyl, a compound that destabilizes microtubules and consequently inhibits cell division. Control strains grew normally on complete media without benomyl, but were modestly or completely inhibited by increasing drug concentrations. By contrast, all mutant diploid strains demonstrated varying degrees of benomyl resistance compared to WT (Figure S5A, B); this is in contrast to a previously reported TUB2 mutagenesis screen, in which alterations of conserved residues more often cause benomyl supersensitivity than resistance (Reijo et al., 1994). These results suggest that TUBB3 syndrome mutations in Tub2p increase the stability of microtubules by rendering them resistant to depolymerization. Finally, a heterozygous diploid strain was generated in which one copy of TUB2 was deleted. Unlike the TUBB3 substitutions, TUB2 haplo-insufficiency resulted in benomyl sensitivity (Figure S5A, B). Thus, the TUBB3 substitutions do not result in complete loss of TUB2 function and appear to cause dominant effects on microtubule behavior.
Using time-lapse microscopy, we then asked how TUBB3 syndrome mutations alter microtubule dynamics. We monitored astral microtubules in G1 cells since microtubules are most dynamic during G1 (Carminati and Stearns, 1997). GFP-Tub1p (α-tubulin) labeled microtubules from heterozygous strains R262C, R262H, A302T, and E410K and haploid strains R62Q and R380C were imaged because these residues reside within different regions of the tubulin dimer.
Substitutions A302T, R62Q, and R380C resulted in similar and significant changes to most measured parameters of microtubule dynamics (Figure 6A–C). Mutant microtubules were more stable and had longer lifetimes compared to WT, and spent the majority of time in prolonged paused states instead of growing and shortening. The frequency of transition to microtubule depolymerization (catastrophe) was reduced, whereas the frequency of transition to polymerization (rescue) was unaffected in A302T heterozygotes, and increased in R62Q and R380C haploids. The rates of both polymerization and depolymerization were significantly reduced. R380C microtubules appeared less dynamic than R62Q and A302T microtubules and spent more time in paused states, although WT tubulin in heterozygous A302T cells might dampen stability. Thus, A302T, R62Q, and R380C substitutions appear to increase the stability of microtubules and significantly diminish overall dynamics (Movie S2). A302 and R62 are found in loops within β-tubulin hypothesized to regulate lateral protofilament interactions and thus might be predicted to alter stability when mutated. In contrast, this finding is somewhat unexpected for R380C, given that this residue resides in H11, and to our knowledge this outer helix has not been previously implicated in the regulation of microtubule dynamics.
Diploid R262C, R262H, and E410K substitutions resulted in changes to dynamics distinct from those described above (Figure 6B, C). Microtubules in these heterozygous cells spent similar amounts of time growing compared to WT, but the average duration of individual growth events was prolonged and astral microtubules in these strains were longer on average (Figure S5C). The frequency of catastrophe events was only slightly less, whereas the frequency of rescue events was significantly less than WT. Finally, these mutants showed a decreased rate of polymerization and increased rate of depolymerization. Thus, mutant R262C, R262H, and E410K astral microtubules were often long and grew at reduced polymerization rates, followed by a more rapid and complete disassembly to the spindle pole body without recovery (Movie S2).
Amino acid residues E410 and D417 of yeast β-tubulin have been identified as microtubule binding sites for conventional kinesin (Kif5) in vitro (Uchimura et al., 2006), and we predict R262 forms a hydrogen bond with D417, which is abolished by R262C/H substitutions (Figure 5E). Perturbations to microtubule dynamics observed in R262C/H and E410K cells, including the faster rate of depolymerization and reduced rescue frequency, are similar to those found following the deletion of Kip3p (Kip3Δ). Kip3p is a plus-end directed kinesin motor found in yeast cytoplasm, and is necessary to position the mitotic spindle near the bud neck in dividing cells (DeZwaan et al., 1997). Kip3Δ, R262C/H, and E410K mutant cells often have long microtubules that result in mispositioning of the mitotic spindle away from the bud neck (Gupta et al., 2006).
Altered dynamic instability resulting from some TUBB3 substitutions might be explained by reduced Kip3p-microtubule interactions. Thus, we monitored the behavior of Kip3p in vivo fused to three tandem copies of YFP at its c-terminus (Kip3p-3YFP) in heterozygous mutant and WT cells containing CFP-Tub1p labeled microtubules. Localization of Kip3p-3YFP on WT microtubules was as previously reported, with strong intensity detected on the plus-end tips and discontinuous speckles along the length of cytoplasmic microtubules (Figure 7A, F) (Gupta et al., 2006). Time-lapse video microscopy revealed a strong reduction of Kip3p-3YFP along the lengths and tips of growing microtubules in R262C/H, E410K, and D417H/N cells (Movie S3). Overall, R262C/H, E410K, and D417H/N cells all had a ~70–80% decrease in signal intensity compared to WT (Figure 7B–J, U). We next examined the plus-end localization of Kip2p-3YFP, as Kip2p is the second major anterograde motor found in the budding yeast cytoplasm. Kip2p-3YFP localization was normal in WT cells; it was absent from spindle pole bodies, and was speckled along the lengths and highly localized to the plus-ends of astral microtubules (Figure 7K, P) (Carvalho et al., 2004). Mutant R262C/H, E410K, and D417H/N cells had a significant, albeit less dramatic and more variable, decrease in Kip2p-3YFP plus-end localization compared to Kip3p (Figure 7L-T, V). There were no significant differences in the protein levels of Kip3p-3YFP and Kip2p-3YFP between WT and mutant cells to account for the quantitative differences (Figure S6). Thus, the R262C, R262H, E410K, D417N, and D417H substitutions result in a significant decrease in Kip3p-3YFP and Kip2p-3YFP accumulation on microtubule plus-ends, suggesting that kinesin interactions on microtubules are reduced in humans harboring these amino acid substitutions. These data provide in vivo evidence to support previous in vitro studies (Uchimura et al., 2006) implicating E410 and D417, as well as identifying an additional residue, R262, as important amino acids for proper microtubule-kinesin interactions. These findings also support diminished KIF21A microtubule-interactions found in TUBB3R262C/R262C mice.
The localization of Kip3p-3YFP and Kip2p-3YFP on the plus-ends of cytoplasmic microtubules was not significantly altered between WT, R62Q and A302T heterozygous mutant cells (Figure 7U, V), and these residues are distal from known kinesin-microtubule interaction sites. Cytoplasmic microtubules in heterozygous R380C cells had an approximately 50% increase in the amount of Kip3p-3YFP at the plus-ends, but a non-significant increase in the amount of Kip2p-3YFP (Figure 7U, V). Thus, residues R62Q, A302T, and R380C all diminish microtubule dynamics in a similar fashion, and do not significantly reduce the levels of kinesin on the plus ends of microtubules.
We have identified TUBB3 as the mutated gene underlying a series of autosomal dominant disorders of axon guidance that we collectively call the TUBB3 syndromes. In 29 unrelated families, we identified eight unique heterozygous missense mutations that alter six amino acid residues. Phenotype-genotype correlations are emerging for these syndromes, and while the TUBB3 phenotypes limited to CFEOM3 and polyneuropathy often segregate in autosomal dominant families, those with more severe clinical findings typically arise de novo. Overall, most adults with isolated CFEOM3 have R262C or, less commonly, R62Q or A302T substitutions. Some children with isolated CFEOM3, however, harbor the D417N substitution and are at risk of developing a polyneuropathy in their teens or twenties. CFEOM3 with developmental delay, CC agenesis, and basal ganglia dysmorphisms in the absence of facial weakness may be predictive of the R380C substitution. Association of severe CFEOM3 with facial weakness, developmental delay, moderate to severe CC dysgenesis and likely a late-onset polyneuropathy predicts the E410K substitution, while the addition of finger contractures, basal ganglia dysmorphisms, and early onset polyneuropathy would suggest R262H or D417H substitutions.
Multiple findings suggest the mutations underlying the TUBB3 syndromes primarily alter microtubule function in a dominant fashion, although we cannot rule out partial loss of function for some. First, recurrent missense mutations in the absence of truncating mutations are most consistent with altered rather than loss of protein function as a primary genetic etiology. Second, facial paralysis and progressive sensorimotor polyneuropathy occur only in those individuals harboring R262H, E410K, D417H/N, the four mutations that permit heterodimer formation and efficient microtubule incorporation, and have dominant effects upon microtubule function by perturbing kinesin interactions in yeast. Third, among the four mutations that result in scant heterodimer formation, A302T and R380C have considerably more microtubule incorporation in mammalian cells and cause more severe phenotypes than R62Q and R262C. Finally, R262C and R262H result in relatively isolated CFEOM3 and a severe TUBB3 phenotype, respectively; these two substitutions are distinguished because R262H permits much more efficient heterodimer formation and microtubule incorporation than R262C both in vitro and in mammalian cells. Thus, more severe phenotypes and developmental disabilities that correlate with particular mutations might reflect a greater extent of mutant heterodimer incorporation and the specific nature of the dominant effect (i.e. dynamics, protein interactions).
Progressive axonal neuropathies can result from inactivating mutations in kinesin and dynein accessory proteins, underscoring the vulnerability of motor neurons and peripheral axons to protein trafficking defects (Chevalier-Larsen and Holzbaur, 2006). The TUBB3 syndromes now demonstrate that tubulin mutations resulting in secondary motor protein transport defects can also cause axonal neuropathies. TUBB3 expression is maintained at high levels only in the adult PNS (Jiang and Oblinger, 1992), supporting a role for TUBB3 in health and maintenance of peripheral motor and sensory axons. This role is in addition to the function of TUBB3 in cranial nerve development, and perturbations in protein trafficking caused by a subset of TUBB3 substitutions may also explain, in part, the cellular etiology of CFEOM3. Notably, recurrent dominant mutations in KIF21A result in CFEOM1 and, in rare families, CFEOM3 (Yamada et al., 2004). Our findings in mice suggest that mutations in TUBB3 can diminish KIF21A-microtubule interactions, possibly accounting for oculomotor nerve axon guidance defects.
Neuroimaging, clinical manifestations of cranial motor nerve misrouting, and the phenotypic analysis of a TUBB3 disease mouse model elucidate a critical role for TUBB3 in proper axon guidance. Throughout nervous system development, differentiating neurons require dynamic populations of microtubules in order to appropriately respond to growth and guidance cues (Gordon-Weeks, 2004; Kalil and Dent, 2005). Because TUBB3 is the most dynamic β-tubulin isotype and the only one with expression primarily restricted to the CNS and PNS (Katsetos et al., 2003), it has been hypothesized that the dynamic properties of TUBB3 could be required for specific developmental processes (Panda et al., 1994). We expand upon these earlier observations by demonstrating microtubule stability is increased in homozygous R262C mice, a finding that could result from the combined effects of TUBB3 heterodimer loss and the remaining mutant heterodimers. Dynamic changes could be exacerbated further by mutant heterodimers that diminish microtubule dynamics and incorporate at higher levels (R380C), resulting in more severe axon guidance phenotypes.
Certain aspects of the TUBB3 syndromes, including CC dysgenesis and basal ganglia dysmorphisms, converge with those resulting from heterozygous missense mutations in TUBA1A and TUBB2B in humans (Jaglin et al., 2009; Keays et al., 2007). Each of these three tubulin isotypes are highly expressed in post-mitotic differentiating neurons and the phenotypic similarities suggest they have important overlapping functions (Coksaygan et al., 2006; Jaglin et al., 2009; Liu et al., 2007). However, the primary brain malformations resulting from mutations in TUBA1A and TUBB2B are lissencephaly, pachygyria, and/or grey matter heterotopias that result from cell migration defects. By contrast, humans and mice with TUBB3 substitutions do not show signs of cortical cell migration defects, and a recent study of TUBA1A mutation-negative subjects ascertained on the basis of agyria and pachygyria failed to identify mutations in TUBB3 (Poirier et al., 2007) and instead identified TUBB2B (Jaglin et al., 2009).
Remarkably, TUBB2B and TUBB3 are the major β-tubulin isotypes expressed in the nervous system and share 90% protein sequence homology, including all disease-associated amino acids. These isotypes differ, however, in the sequence of their C-terminal tails, dynamic behavior in vitro, and unique post-translational modifications (Banerjee et al., 1990; Khan and Luduena, 1996), suggesting that functional differences between these isotypes may account for phenotypic distinctions in humans and mice. Our results now greatly expand upon previous observations that different β-tubulin isotypes may have evolved in higher vertebrates to serve specific cellular functions (Luduena, 1993), and support a critical role for TUBB3 in axon guidance.
Detailed experimental procedures can be found in Supplemental Data.
Probands were ascertained based on affection with CFEOM. Participants were enrolled by the Engle Laboratory at Children’s Hospital Boston or by collaborating laboratories following appropriate Institutional Review Board approval and informed consent. Clinical data were obtained from participants and medical providers. Genetic linkage and mutation detection by direct sequencing and DHPLC were performed as previously described (Doherty et al., 1999; Miyake et al., 2008) and using NCBI reference sequences NM_006086 and NT_010542.
Diagnostic MRI scans were reviewed from participants with each TUBB3 substitution except D417H. Images are T1 (B-E, G, B′, F′), T2 (A, F, H, J-N, A′-D′, G′, H′), Flair (I). Orbital MRI was performed as described previously (Demer et al., 2005).
E18.5 embryonic brains on mixed 129/B6 and pure 129S6 backgrounds were fixed in 4% paraformaldehyde and embedded for cryo- or paraffin sectioning. 10–12 μm sections were blocked with 5% normal goat/donkey serum and 0.1% Triton in PBS at room temperature for 1 hr, and incubated with primary antibody in 1% serum and 0.01% Triton at 4 C overnight. Secondary antibody was added at room temperature for 1 hr.
E18 whole brains from WT, heterozygous, and homozygous mutant mice were homogenized in BRB80 buffer and centrifuged at 50,000G at 4 C for 30 minutes. 1mM GTP was added to supernatant and microtubules were polymerized for 30 minutes at 37 C. Samples were layered on a 30% sucrose cushion and centrifuged for 20 minutes at 37 C at 100,000g to pellet the microtubules.
E18 brains from WT and homozygous mutant mice were homogenized with a glass homogenizer in BRB80 buffer containing protease inhibitors and PMSF and incubated on ice for 20 minutes. A tubulin rich fraction was obtained by centrifuging the lysates at 4 C for 30 minutes at 50,000g. 1mM ATP, 1mM GTP, and 10μM taxol were added to the supernatants and incubated at 37 C for 30 minutes. Samples were then layered on top of a 30% sucrose cushion and centrifuged for 20 minutes at 37 C at 100,000g. Pellets were resuspended in SDS buffer and run on SDS NuPAGE 8–12% Bis-Tris gels (Invitrogen).
1μg of plasmid DNA was added to 50μl of reticulocyte cocktail according to manufacturer’s instructions (Promega TNT T7 Coupled Reticulocyte Lysate), and the reactions were incubated at 30°C for 90 minutes. Following incubation, the products were chased with 300ng bovine brain tubulin and 1mM MgCl2 and GTP for 30 minutes at 37 C, and 2μl and 10μl from the overall reactions were run on SDS NuPAGE 8–12% Bis-Tris (Invitrogen) and NativePAGE 4–16% Bis-Tris Gels (Invitrogen), respectively. WT and mutant TUBB3 heterodimers were detected with a monoclonal V5 antibody (Invitrogen) against the C-terminal tag. Remaining products were cycled with bovine brain tubulin.
Each TUBB3 substitution was plotted on the solved protein structure for the αβ-tubulin heterodimer (pdb: 1JFF) using PyMOL software (1.1r1, http://www.pymol.org). Hydrogen bonds were predicted using the bond distance measurement function, and were also modeled using the Swiss-Pdb Viewer (DeepView). The protein model movie was created using PyMOL.
Cells were seeded overnight in SD-complete media (0.67% yeast nitrogen base without amino acids, 5% glucose, 0.5% casamino acids) and allowed to grow at 24°C. The next morning, cells were reseeded and grown at 24°C to mid-log phase. 1 ml of cells were pelleted, resuspended in SD-complete media, imaged at room temperature (~26°C) on an Axio Image MI (Zeiss) scope with a 63× Plan Fluor 1.4 N.A. objective, and captured using a Coolsnap HQ camera (Photometrics). The typical acquisition protocol acquired five z-series fluorescent images at 0.75 μm axial steps, and one differential interference contrast (DIC) image corresponding to the central fluorescent image. Time-lapse image series were acquired at 8 sec intervals.
Mutant and WT CFP-Tub1p-expressing cells containing either Kip3p-3YFP or Kip2p-3YFP were grown and imaged as described above in the in vivo microtubule dynamics section. Seven z-series images were merged into a single projection image (maximum) using deconvolution microscopy and the nearest neighbor algorithm function in Slidebook software (Intelligent Imaging Innovations, CO). In randomly selected fields, the plus-ends of all identifiable microtubules in multiple cells in all phases of the cell cycle were marked with a circle of equal radius, and the average plus-end intensity was then calculated using quantification software (Slidebook). At least two clones from WT and mutant strains were analyzed on at least two separate days, and net signal intensities from microtubule plus-ends and cell backgrounds were determined by averaging the values obtained from the total population on each day.
We thank the families for their participation; Michelle DeLisle and Carrie Wu for technical assistance, and members of the Engle lab for their thoughtful comments; A. Nurten Akarsu, Peter Kang, Lisa S Kearns, James Hoekel, Marijean Miller, Marilyn Miller, Peter Roggenkämper, and Sandra Staffieri for pedigree referrals and/or clinical exam data. This work was supported by NIH R01 EY012498, R01 EY013583, HD18655 [ECE], F32 EY016306 [HB], R01 GM061345-08 [DP], and VA Research Funds [TDB]. ECE and DP are investigators of the Howard Hughes Medical Institute.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.