Post-translational modifications of tubulin in the nervous system

Authors


Address correspondence and reprint requests to Nobuyuki Fukushima, Research Institute for Science and Technology, Kinki University, 3-4-1 Kowakae, Higashiosaka 577-8502, Japan. E-mail: nfukushima@life.kindai.ac.jp

Abstract

J. Neurochem. (2009) 109, 683–693.

Abstract

Many studies have shown that microtubules (MTs) interact with MT-associated proteins and motor proteins. These interactions are essential for the formation and maintenance of the polarized morphology of neurons and have been proposed to be regulated in part by highly diverse, unusual post-translational modifications (PTMs) of tubulin, including acetylation, tyrosination, detyrosination, Δ2 modification, polyglutamylation, polyglycylation, palmitoylation, and phosphorylation. However, the precise mechanisms of PTM generation and the properties of modified MTs have been poorly understood until recently. Recent PTM research has uncovered the enzymes mediating tubulin PTMs and provided new insights into the regulation of MT-based functions. The identification of tubulin deacetylase and discovery of its specific inhibitors have paved the way to understand the roles of acetylated MTs in kinesin-mediated axonal transport and neurodegenerative diseases such as Huntington’s disease. Studies with tubulin tyrosine ligase (TTL)-null mice have shown that tyrosinated MTs are essential in normal brain development. The discovery of TTL-like genes encoding polyglutamylase has led to the finding that polyglutamylated MTs which accumulate during brain development are involved in synapse vesicle transport or neurite outgrowth through interactions with motor proteins or MT-associated proteins, respectively. Here we review current exciting topics that are expected to advance MT research in the nervous system.

Abbreviations used
Ac-T

acetylated tubulin

BDNF

brain-derived neurotrophic factor

CAP-Gly

cytoskeleton-associated protein Gly-rich

CCP

cytosolic carboxypeptidase

deTyr-T

detryrosinated tubulin

HD

Huntington’s disease

HDAC

histone deacetylase

MAP

MT-associated proteins

MT

microtubules

PGs

polyglutamylases

PTM

post-translational modifications

SIRT

sirtuin type

+TIP

MT plus-end tracking protein

TSA

trichostatin A

TTL

tubulin tyrosine ligase

TTLL

TTL-like

Tyr-T

tyrosinated tubulin

Microtubules (MTs), one component of the cytoskeleton, are composed of heterodimers of α- and β-tubulin. In mature neurons, MTs primarily act as ‘railways’ for cargo transport. In developing neurons, MTs not only serve as a scaffold for transport but also frequently undergo reorganization that is coordinated with neurite extension and retraction and growth cone advancement. Furthermore, MTs themselves are transported bidirectionally in neurites (Baas et al. 2006). These MT behaviors are highly regulated by the intrinsic GTPase activity of tubulins as well as MT-interacting proteins, including MT-associated proteins (MAPs), MT-severing proteins, MT plus-end tracking proteins (+TIPs), and motor proteins such as kinesin-1 (Dent and Gertler 2003; Rodriguez et al. 2003; Akhmanova and Steinmetz 2008; Jaworski et al. 2008).

Tubulin receives diverse PTMs which include acetylation, tyrosination, detyrosination, Δ2 modification, polyglutamylation, polyglycylation, palmitoylation, and phosphorylation (Westermann and Weber 2003; Verhey and Gaertig 2007; Hammond et al. 2008). The first implication of a tubulin PTM was the tyrosination of brain tubulin reported in 1974 (Barra et al. 1974). Since then, the existence of other tubulin PTMs has been demonstrated in neuronal and non-neuronal cells (Gozes and Littauer 1978; Gozes and Sweadner 1981; Cleveland and Sullivan 1985; Sullivan 1988). Earlier studies focused on the profiles and physiological relevance of modified tubulin and revealed that some PTMs are associated with MT age. Later studies further showed that tubulin PTMs are involved in interactions between MTs and certain MAPs or motor proteins. These observations have suggested that modified MTs play a role in neurite outgrowth and maturation as well as in maintaining neuronal morphology. Tubulin tyrosine ligase (TTL) was the first identified enzyme responsible for a tubulin PTM, in this case tyrosination (Ersfeld et al. 1993); since then, no enzymes involved in other tubulin PTMs have been cloned. Thus, the questions of how these PTMs are generated and regulated and what roles the PTMs play in the nervous system remained open. After 2002, the identification of tubulin deacetylase and discovery of its specific inhibitors have moved forward studies on the roles of acetylated MTs (Westermann and Weber 2003; Verhey and Gaertig 2007; Hammond et al. 2008). Another recent exciting advance was the discovery of TTL-like (TTLL) genes encoding polyglutamylase (PGs) and polyglycylase (Janke et al. 2005; Ikegami et al. 2006, 2008; van Dijk et al. 2007). Here we briefly describe the history of tubulin PTM research and review the recent progress of studies on enzymes involved in tubulin PTMs in the nervous system, particularly acetylation, deacetylation, tyrosination/detyrosination, polyglutamylation, and polyglycylation. Several kinases have been reported to phosphorylate tubulin (Verhey and Gaertig 2007). However, we will not discuss the significances of phosphorylation and palmitoylation of tubulins in this review because these modifications are less studied in the nervous system.

Tubulin in brain

Tubulin is encoded by multiple closely related genes. Seven α-tubulin genes and eight β-tubulin genes have been isolated in mice (Fig. 1a and b). Each α-tubulin shows more than 90% amino acid identity to other α-tubulins. β-Tubulins, except β1-tubulin, also show more than 90% amino acid identity to other β-tubulins and about 78% to β1-tubulin. The carboxyl-terminal (C-terminal) region consisting of ∼15 amino acid residues constitutes a major variable domain for β-tubulin, and to a lesser extent for α-tubulin as well (Fig. 1a and b). The C-terminal regions lie on the outer surface of the MTs and are the acceptors for most PTMs (tyrosination/detyrosination, Δ2 modification, polyglutamylation, and polyglycylation) that are involved in the binding of MT-interacting proteins (Luduena 1998; Westermann and Weber 2003; Hammond et al. 2008). Thus, variable sequences in the C-terminal regions may result in isoform-specific PTMs.

Figure 1.

 Alignment of the C-terminal amino acid sequences of mouse α-tubulin (a) and β-tubulin (b) isoforms. All sequences and gene names were derived from the NCBI and the Mouse Genome Informatics database, and aligned using MAFFT version 6 at http://align.bmr.kyushu-u.ac.jp/mafft/online/server/. Only the C-terminal regions are shown. Tubulin α-like 3 encoded by Tubal3 and tubulin α-related sequence 1 encoded by Tuba-rs1 are not included. α3A- and α3B-tubulin are encoded by different genes but have identical amino acid sequences. (§) α1–8 designations are from work by Villasante et al. (1986) and Stanchi et al. (2000). (#) The Mβ1–5 designations are from work by Wang et al. (1986). (+) The roman numerals are from work by Cleveland (1987) and Luduena (1993). Glutamate (E) residues highlighted in red and yellow are identified and potential polyglutamylated sites, respectively (see the Polyglutamylation section). Asterisks indicate identical amino acids. Colons indicate similar amino acids. A nomenclature for the mammalian tubulin gene family is available at http://www.genenames.org/genefamily/tub.php, http://www.informatics.jax.org/javawi2/servlet/WIFetch?page=pirsfDetail&key=4203450 or the report by the nomenclature committee (Khodiyar et al. 2007).

The expression of the tubulin isoforms differs among organs and cell types and is developmentally regulated. Brains express α1A-, α1B-, α1C-, α4A-, and α8-tubulin and β2A/B-, β3-, β4-, and β5-tubulin. Expression of α4A-, β3-, and β4-tubulin increases during postnatal brain development, while that of β2A/B- and β5-tubulin decreases (Lewis et al. 1985; Villasante et al. 1986). Any combination of α- and β-tubulin isoforms can copolymerize with each other into mixed MTs, suggesting that the combination of α/β-tubulin heterodimers varies during brain development. Thus, the changing expression of tubulin isoforms in developing brain might lead to changes in PTMs, resulting in differential utilization of MT-interacting proteins and changes in MT-based functions in neurons.

Other tubulin isoforms have been identified, which include γ-, δ-, and ε-tubulins (Chang and Stearns 2000; Inclan and Nogales 2001; Chang et al. 2003; Raynaud-Messina and Merdes 2007). These localize to centrosome and are involved in microtubule nucleation, centriole duplication, or other unknown cellular events. However, whether these isoforms are targets of PTMs remain unclear.

Acetylation

Acetylation occurs after MT assembly at the ε-amine of α-tubulin Lys40, which is preserved in all α-tubulin isoforms but not β-tubulin (Fig. 2a). Acetylated tubulin (Ac-T) exists in stable, long-lived MTs that are resistant to nocodazole, although acetylation does not cause MT stabilization (Palazzo et al. 2003). MTs containing Ac-T are non-uniformly distributed in neurons. In mature neurons in culture, they are enriched in the proximal site of the axon and present in lesser amounts in the cell body and growth cone (Baas and Black 1990; Ahmad et al. 1993; Brown et al. 1993) (Fig. 2b). In contrast, they are present in the proximal site of the neurite and the cell body in young neurons commencing neurite outgrowth (Smith 1994; Fukushima and Morita 2006). This distribution is distinct from that of tyrosinated tubulin (Tyr-T) (see section Tyrosination/Detyrosination/Δ2 modification), suggesting regional differences in MT stabilization within neurons. Recently, a quantitative study has demonstrated that the axon in polarized neurons and the future axon in unpolarized neurons show highest ratio of Ac-T versus Tyr-T compared with other neurites (Witte et al. 2008). Thus, MT stability might be a key factor for neuronal polarization.

Figure 2.

 Schematic figures illustrating features of tubulin modifications (a) and cellular localization of modified tubulins (b). (a) Microtubules (MTs) are composed of α- (light blue) and β-tubulin (dark blue) heterodimers. Here, α1A- and β3-tubulin are shown with part of the C-terminal regions lying on the outer surface of the MTs. Acetylation occurs at α-tubulin Lys40 located toward the lumen of the MTs by action of NAT1-ARD1 or other unidentified acetyltransferases. Ac-T binds kinesin-1 and dynein. Reverse deacetylation is mediated by HDAC6 or SIRT2. The Tyr residue of α-tubulin C-terminus is cleaved by Nna1 or other unidentified carboxypeptidases to yield deTyr-T which is retyrosinated by TTL. Tyr-T and deTyr-T binds CLIP170 and kinesin-1, respectively. Polyglutamylation occurs at α1A-tubulin Glu445 and β3-tubulin Glu438 by TTLLs. polyGlu-T interacts with MAPs and kinesins in a chain length-dependent manner. Enzymes responsible for depolyglutamylation are unknown. Attached and removed residues are highlighted in orange. Ac, acetyl residue; E1–E6, one to six glutamyl units. (b) Cellular localization of modified tubulins in a mature neuron. Ac-T and deTyr-T are enriched in the axon but not in the growth cone. Tyr-T is predominantly located in the distal region of the axon, growth cone and dendrites but is present in relatively lower amounts in the axonal shaft. polyGlu-T is distributed throughout a neuron. More detailed and quantitative analyses are needed to clarify the precise distribution of modified tubulins in a single neuron.

Although the enzyme acetylating tubulin has not been identified until recently, a recent report has shown that an N-acetyltransferase complex consisting of N-acetyltransferase 1 (NAT1) and arrest defective 1 (ARD1) mediates tubulin acetylation (Ohkawa et al. 2008) (Fig. 2a). Inhibition of this complex results in less dendritic branches in cultured Purkinje neurons, suggesting the involvement of Ac-T in dendrite growth. Further analyses for this enzyme complex are needed to understand its physiological roles.

On the other hand, two laboratories have described two enzymes responsible for the deacetylation of Ac-T: histone deacetylase 6 (HDAC6) (Hubbert et al. 2002) and the mammalian homolog of silent information regulator 2/sirtuin type 2 (SIRT2) (North et al. 2003) (Fig. 2a). HDAC6 is a member of the class II HDAC family and colocalizes with MTs, and has been recently found to show diverse functions through deacetylation of multiple targets, including histone, tubulin, cortactin, and heat shock protein 90 (Luxton and Gundersen 2007; Valenzuela-Fernandez et al. 2008). SIRT2 belongs to the class III HDAC family and requires NAD for its function. SIRT2 also shows cytoplasmic distribution and colocalization with MTs. Pharmacological inhibition or reduced expression of HDAC6 or SIRT2 increases the levels of Ac-T in cells. Whether there is an interdependency between the two enzymes remains to be elucidated.

The identification of tubulin deacetylase led to a better understanding of the roles of Ac-T in axonal transport. Treatment of neuronal cells with trichostatin A (TSA, a non-specific HDAC inhibitor) or tubacin (a specific HDAC6 inhibitor) stimulates kinesin-1-mediated anterograde transport of vesicles containing Jun-N-terminal kinase-interacting proteins into neurite tips (Reed et al. 2006; Bulinski 2007). These biochemical experiments further showed that kinesin-1 can bind to MTs only when Lys40 of α-tubulin is acetylated. Thus, MTs containing Ac-T is likely to provide a scaffold for kinesin-1 in neurons. However, because Lys40 is located toward the lumen of the MTs, how acetylation and deacetylation enzymes or kinesin-1 interact with this residue is unclear.

Recent reports have indicated that tubulin acetylation may be involved in neurodegenerative diseases such as Huntington’s disease (HD) and Parkinson’s disease (Dompierre et al. 2007; Outeiro et al. 2007; Suzuki and Koike 2007). HD is a neurodegenerative disorder characterized by cognitive and motor deficits. In HD, the MT-dependent transport of vesicles containing brain-derived neurotrophic factor (BDNF) is reduced, resulting in a reduction of trophic support and subsequent neuronal cell death. HDAC6 inhibition by TSA increased the amount of acetylated MT and concomitantly stimulated vesicular transport of BDNF in striatum-derived neuronal cell lines (Dompierre et al. 2007). This increased MT acetylation led to enhanced recruitment not only of kinesin-1 but also of dynein, which was likely to be responsible for the stimulation of BDNF transport. Interestingly, this study demonstrated that Ac-T levels were reduced in HD patients and that TSA rescued the transport defect in neuronal cells carrying a polyglutamine expansion, a model of HD. These observations imply that regulation of MT acetylation may be a therapeutic target in HD.

Parkinson’s disease is characterized by motor deficits, loss of dopaminergic neurons, and accumulation of α-synuclein in the midbrain. SIRT2 inhibition by the use of specific inhibitors or genetic engineering rescues α-synuclein–induced neuronal toxicity (Outeiro et al. 2007). SIRT2 inhibition also causes resistance to axonal degeneration in cerebellar granule neurons of mutant mice displaying slow Wallerian degeneration (Suzuki and Koike 2007). In both cases, the mechanisms by which SIRT2 inhibition protects neurons from cell death remain uncertain. In addition, because SIRT2 is highly expressed in oligodendroglial cells (Li et al. 2007), whether SIRT2 acts in neurons is unclear. Thus, further investigations are needed to clarify how SIRT2 is involved in neurodegeneration.

Tyrosination/Detyrosination/Δ2 modification

Only α-tubulin undergoes the tyrosination/detyrosination cycle. All α-tubulins, except for α4A- and α8-tubulin, contain a Tyr residue at the C-terminus immediately after translation (Fig. 1a). Tyr-T forms a heterodimer with β-tubulin and assembled into MTs (Gundersen et al. 1987). After assembly, Tyr-T in MTs is detyrosinated as time goes, resulting in production of detyrosinated tubulin (deTyr-T), which harbors a Glu residue at the C-terminus and is also called glutamylated tubulin (Gundersen et al. 1987). Following depolymerization of MTs containing deTyr-T, deTyr-T is retyrosinated to yield Tyr-T. Δ2-Tubulin (Δ2-T), a third pool of tubulin, lacks C-terminal Glu and Tyr residues and is generated by deglutamylation of deTyr-T (Paturle-Lafanechere et al. 1991; Lafanechere and Job 2000). However, Δ2-T is no longer converted back to either deTyr-T or Tyr-T. Whereas Tyr-T and deTyr-T are ubiquitously distributed in various organs, Δ2-T is abundant in brain, particularly in differentiated neurons rather than glial cells (Paturle-Lafanechere et al. 1994).

deTyr-T and Δ2-T are present in stable, long-lived MTs, as is Ac-T (Khawaja et al. 1988; Paturle-Lafanechere et al. 1994). In contrast, Tyr-T is observed in labile, newly formed MTs (Khawaja et al. 1988; Paturle-Lafanechere et al. 1994). Detyrosination is a consequence and not the cause of MT stabilization. Extensive studies in the early 1990 revealed that in axons, individual MTs are composed of a labile domain at the distal (plus) end which is highly tyrosinated and sparsely detyrosinated and acetylated, and a stable domain at the proximal (minus) end which is highly detyrosinated and acetylated and sparsely tyrosinated (Baas and Black 1990; Brown et al. 1992, 1993; Ahmad et al. 1993; Baas et al. 1993). These studies also demonstrated that Tyr-T is predominantly located in the distal region of the axon, contiguous with the growth cone, and in cell bodies but is present in relatively lower amounts in the axonal shaft (Baas and Black 1990; Ahmad et al. 1993; Brown et al. 1993; Shea 1999) (Fig. 2b). A few Tyr-T containing MTs penetrate into the peripheral regions of growth cones, which generally do not contain a MT network (Shea 1999; Dent and Gertler 2003; Fukushima and Morita 2006). These observations suggest that MT dynamics differ throughout the axon and growth cone. In dendrites where MTs display mixed polarity, Tyr-T exists at higher level than Ac-T and deTyr-T (Mansfield and Gordon-Weeks 1991; Witte et al. 2008) (Fig. 2b). However, the precise distribution and developmental changes of the modified tubulin remains unclear.

Given the differential localizations of Tyr-T and deTyr-T, each modified tubulin is likely to play a distinct role in MT function. Neither tyrosination and detyrosination nor acetylation essentially confers MT stability. Rather, these modifications regulate interactions between MTs and MT-interacting proteins. An example has been shown in fibroblasts where deTyr-T is able to preferentially bind kinesin-1 and transport cargos containing an intermediate filament vimentin (Kreitzer et al. 1999; Dunn et al. 2008). Although it is not known whether deTyr-T is involved in kinesin-mediated axonal transport in neurons, impaired interactions between deTyr-T and kinesin are accompanied by the collapse of vimentin and glial fibrillary acidic protein in astrocytes degenerating because of tau accumulation (Yoshiyama et al. 2003). However, deTyr-T is present in stable MTs that also contain Ac-T, raising the possibility that tubulin acetylation is associated with this preferential kinesin-1 binding.

Detyrosination has been long proposed to involve carboxypeptidase reactions. Recently identified metallocarboxypeptidases, including Nna1/cytosolic carboxypeptidase 1 (CCP1), might be involved in this processing (Kalinina et al. 2007) (Fig. 2a). Nna1/CCP1 is a member of the cytosolic carboxypeptidase family composed of CCP1–6 and was originally identified as a gene up-regulated during axonal regeneration. All CCP members are expressed in brain, although with different distribution profiles. Indirect evidence showing that CCPs are candidates for enzymes producing deTyr-T came from the finding that mitral cells in olfactory bulbs of pcd mutant mice lacking the CCP1 gene contained high levels of Tyr-T but not deTyr-T, compared with those in wild-type mice (Kalinina et al. 2007). However, detailed analyses of the enzymatic properties of CCPs are needed before we can conclude that CCPs are bona fide enzymes responsible for the detyrosination of tubulin.

The tubulin tyrosination reaction was suggested more than 30 years ago (Barra et al. 1974) but TTL was not identified until 1993 (Ersfeld et al. 1993). TTL was the first enzyme responsible for a PTM to be cloned and has been proven to be a member of a recently discovered TTLL family that includes 13 TTLL proteins in addition to TTL (Janke et al. 2005; Ikegami et al. 2006; van Dijk et al. 2007) (Fig. 2a and Table 1, also see the Polyglutamylation). Although TTL is expressed in the developing hippocampus, its exact distribution in the nervous system remains unclear (Fig. 3). However, the generation and analysis of TTL-null mice have highlighted the crucial roles of TTL and Tyr-T in neuronal development (Erck et al. 2005). These mice die perinatally because of abnormal brain organization, characterized by blurred layers of the cerebral cortex and a disrupted corticothalamic projection. No obvious malformation is detected in any other organs, suggesting a vital role for TTL in the nervous system.

Table 1.   Amino acid identity of core and extended TTL domains in mouse TTLL members Thumbnail image of
Figure 3.

 RT-PCR detection of TTL and TTLL members in the developing hippocampus. cDNAs were prepared from embryonic day 16 (E16), postnatal day 1 (P1), 7 (P7) or 15 (P15), or adult (Ad) mouse hippocampus and subjected to PCR analyses. The primers used were: for TTL, TTL-s1 (5′-ccaacttccggtcagtaactcc-3′) and TTL-as1 (5′-ccaggacccagcttcgaatg-3′); for TTLL1, TTLL1-s1 (5′-tggggaggactacaaccaca-3′) and TTLL1-as1 (5′-ccagggcttcagcttgtcatc-3′); for TTLL2, TTLL2-s1 (5′-tgccacggagaagtttgatctc-3′) and TTLL2-as1 (5′-ccaaaagccacggtttcagg-3′); for TTLL3, TTLL3-s1 (5′-tcccacgatgctaccgattg-3′) and TTLL3-as1 (5′-ggtagtagcgctgaaggaactg-3′); for TTLL4, TTLL4-s1 (5′-cccactgatccatacctctcact-3′) and TTLL4-as1 (5′-gcaccttatggtcaccaggttg-3′); for TTLL5, TTLL5-s1 (5′-gggaagttaggcggttctgtg-3′) and TTLL5-as1 (5′-gttcatgctctccacctttagctc-3′); for TTLL6, TTLL6-s1 (5′-ggctagaggaagtcaagggtttc-3′) and TTLL6-as1 (5′-gcacgactttgtctcttgcct-3′); for TTLL7, TTLL7-s1 (5′-tctctaccatgatgggcttgtg-3′) and TTLL7-as1 (5′-attgaccgcttgctgcctt-3′); for TTLL8, TTLL8-s1 (5′-cacagactggaatcccctaacc-3′) and TTLL8-as1 (5′-agatgatgctaccccacgttc-3′); for TTLL9, TTLL9-s1 (5′-agaccaccctcatgaacacac-3′) and TTLL9-as1 (5′-gggcatctcaaaggttttggg-3′); for TTLL10, TTLL10-s1 (5′-acccagagcattgaggatgac-3′) and TTLL10-as1 (5′-tccttcagcagcgtgtacag-3′); for TTLL11, TTLL11-s1 (5′-tcaggtggaaggagttcccatt-3′) and TTLL11-as1 (5′-attccatcaccctgacaaccac-3′); for TTLL12, TTLL12-s1 (5′-ccaggccattttggaggagaac-3′) and TTLL12-as1 (5′-cccggtgctggaaatagctg-3′); for TTLL13, TTLL13-s1 (5′-gctgtcgacactcaatgcct-3′) and TTLL13-as1 (5′-cgctccttgactcgacgtttg-3′). The upper products of TTLL9 have been yet unidentified.

In culture, TTL-null neurons show aberrant neurite growth and axonal differentiation (Erck et al. 2005). These neurons contain no Tyr-T but show a normal MT network and localization of MAPs or motor proteins, such as MAP1B, MAP2, and dynein. However, CLIP170, a +TIP containing a cytoskeleton-associated protein Gly-rich (CAP-Gly) domain, is mislocalized in neurites and growth cones in TTL-null neurons, whereas localization of end-binding 1 (EB1), another +TIP containing no CAP-Gly domain, is unaffected. +TIPs accumulate at the plus-ends of growing MTs and regulate MT dynamics and interactions with other intracellular molecules, leading to control of cellular locomotion, such as axonal growth cone advancement or neuronal migration (Galjart 2005; Akhmanova and Steinmetz 2008; Jaworski et al. 2008). Thus, the disrupted localization of CLIP170 might cause the disorganization of neuronal architectures observed in TTL-null mice. However, p150Glued, another CAP-Gly +TIP, is localized normally in TTL-null neurons (Erck et al. 2005). This is sharp contrast to the finding that three CAP-Gly +TIPs (CLIP170, CLIP-115, and p150Glued) are mislocalized in TTL-null fibroblasts (Peris et al. 2006). Thus, there are likely different molecular mechanisms for CAP-Gly +TIP-mediated regulation of MT dynamics in neurons and non-neuronal cells.

Polyglutamylation

Tubulin polyglutamylation, a unique polymodification, was first reported in 1990 (Edde et al. 1990). The researchers discovered that about half of neuronal α-tubulin in brain was polyglutamylated on the γ-carboxyl group of Glu445. This modification involves the addition of one to six glutamyl units and occurs on α1A- and α1B-tubulin, which contain identical amino acid sequences within their C-termini (Figs 1a and 2a). Later, α4A-tubulin was also reported to be polyglutamylated at two Glu residues, Glu443 and Glu445 (Redeker et al. 1998) (Fig. 1a). The second and third Glu of the polyglutamyl side chain are amide-linked to the α-carboxyl group of the preceding Glu unit, producing highly diverse polyglutamylation (Redeker et al. 1991). Three other groups extended this finding to β-tubulin isoforms and identified Glu435 of β2A/B-tubulin and Glu438 of β3-tubulin as polyglutamylated amino acids (Alexander et al. 1991; Redeker et al. 1992; Rudiger et al. 1992) (Figs 1a and 2a). Although other isoforms of α- and β-tubulin appear to contain potential polyglutamylation sites in their C-terminal regions (Kann et al. 2003) (Fig. 1a and b), detailed analyses remain to be carried out.

Further experiments demonstrated that polyglutamylated α- and β-tubulin (polyGlu-α-T and polyGlu-β-T) accumulate in neuronal culture and in brain during development (Audebert et al. 1993, 1994; Przyborski and Cambray-Deakin 1997; Okada et al. 2004). The polyGlu-α-T modification precedes the polyGlu-β-T, implying that the functions and polyglutamylation enzymes of polyGlu-α-T and polyGlu-β-T are different. Immunocytochemical studies using an antibody against polyGlu-T showed that polyGlu-T is distributed throughout a neuron (Przyborski and Cambray-Deakin 1997; Okada et al. 2004; Ikegami et al. 2006) (Fig. 2b). However, the precise localization of polyGlu-α-T and polyGlu-β-T with varying chain lengths remains unclear.

Polyglutamylation is detected in both Tyr-T and deTyr-T (Edde et al. 1992), indicating that this modification is not directly associated with stable MT structures nor the tyrosination cycle. Rather, as discussed for other PTMs, polyglutamylation has been suggested to play a role in the association of MTs with MAPs and motor proteins because the C-terminal domains of α- and β-tubulin are exposed on the outer surfaces of MTs. Indeed, biochemical experiments showed that tubulin without polyglutamylation failed to bind Tau, and its polyglutamylation restored Tau binding, reaching an optimum for a length of three glutamyl units then gradually decreasing for longer chains (Boucher et al. 1994). However, the polyglutamyl side chains do not represent the direct binding site for Tau because subtilisin-cleaved tubulin lacked the C-terminal regions carrying the chains, but still retained the ability to bind Tau. Rather, the polyglutamyl chains were thought to regulate Tau binding through modulation of the conformation of the C-terminus which contains Tau binding sites. (MAP1A, MAP1B, and MAP2) and a motor protein, kinesin, were also shown to bind to tubulin under the differential regulation by polyglutamylation (Larcher et al. 1996; Bonnet et al. 2001). Direct evidence showing multiple, physiological interactions between polyGlu-T and motor proteins has recently emerged from the analyses of mutant ROSA22 mice which exhibit a loss of polyGlu-α-T (Ikegami et al. 2007). MTs prepared from brains of ROSA22 mice showed lower binding affinity for KIF1A (kinesin-3), KIF3A (kinesin-2), KIF5 (kinesin-1), cytosolic dynein, and MAP1A than those from wild-type mice. However, only the distribution of KIF1A was disrupted in neurons. Because KIF1A movement requires ionic forces between a basic charged region of KIF1A and an acidic charged region of the tubulin C-terminus (Okada and Hirokawa 2000), loss of polyglutamylation in these mice might affect these ionic forces, resulting in the impaired distribution. However, these mice also show lower levels of Tyr-T, and further studies will be necessary to detail the role of polyGlu-T.

In contrast to the still unknown biological relevance of polyGlu-T, the enzymes catalyzing the polyglutamylation were recently identified. The researchers immunoprecipitated several polypeptides with tubulin PG1–5 activity from mouse brains (Regnard et al. 2003; Janke et al. 2005). Sequence analysis of these polypeptides identified PGs3 as an ortholog of the human TTLL1 protein, which shows 17% amino acid identity with TTL. A database search further revealed the existence of a large family of TTLLs in mouse that is composed of TTLL1–13 and TTL (Janke et al. 2005; Ikegami et al. 2006; van Dijk et al. 2007). All TTLL proteins except TTLL12 contain the conserved core TTL domain which is predicted to be essential for the ATPase reaction, suggesting the similar enzymatic reaction of these TTLLs (Table 1). Some TTLL proteins (TTLL1, 2, 4, 5, 6, 7, 9, 11, and 13) also contain a second conserved domain called the extended TTL domain, and bootstrap analyses show that these TTLL genes form a subfamily (van Dijk et al. 2007) (Table 1). Predicted regions that interact with substrates (i.e., tubulin and glutamate) are included in both the core and the second extended domain.

Functional studies demonstrated that TTLL4, 5, 6, 7, 11, and 13 have catalytic activity for polyglutamylation of tubulin when over-expressed in mammalian cells (Janke et al. 2005; Ikegami et al. 2006; van Dijk et al. 2007). This is consistent with higher percentage identities (over 40% in most cases) observed in both the core and the second domain which contain substrate-interacting regions of these TTLL members (Table 1). TTLL5, 6, 11, and 13 show a preference toward α-tubulin, whereas TTLL4 and 7 prefer β-tubulin (Fig. 2a). By contrast, the other TTLL members in the subfamily (TTLL1, 2, and 9) are inactive in the experiment, despite the presence of core and extended TTL domains (Janke et al. 2005; Ikegami et al. 2006; van Dijk et al. 2007), raising the possibility that co-factors might be needed for full PG activity. Indeed, TTLL1 (PGs3) was isolated from brain as part of components with PG activity, including PGs1–5 (Regnard et al. 2003; Janke et al. 2005). Further analyses revealed that TTLL4, 5, and 7 preferentially initiate glutamylation, whereas TTLL6, 11, and 13 preferentially elongate the side chain. All TTLL members (TTLL1–13) are expressed in the developing hippocampus, and some TTLL members with PG activity (i.e., TTLL6 and 11) seem to be developmentally regulated (Fig. 3). Determination of the detailed cellular loci for each TTLL member is further needed.

The PGs1 protein has no homology with any TTLL members and no catalytic activity but is associated with tubulin polyglutamylation (Janke et al. 2005). PGs1 is encoded by the mouse gene Gtrgeo22, highly expressed in brain (Campbell et al. 2002; Janke et al. 2005), and ROSA22 mice, which lack Gtrgeo22 function show a massive loss of polyGlu-α-T in the brain (Ikegami et al. 2007). However, monoglutamylation of α-tubulin is observed in these mice, suggesting that polyglutamylation (elongation of the side-chain) depends on PGs1, whereas monoglutamylation (initiation) does not. Thus, it is unlikely that other TTLL members could compensate for the loss of tubulin polyglutamylation in ROSA22 mice. As mentioned above, several motor proteins and MAPs bind less strongly to MTs in these ROSA22 mice, and KIF1A levels and synaptic vesicle density are lower. In addition, these mice have lower levels of Tyr-T which might cause mislocalization of CLIP170 in neurons as observed in TTL-null mice (Erck et al. 2005). Such abnormalities would explain the impaired synaptic transmission in ROSA22 mice (Ikegami et al. 2007). However, TTL-null mice show more severe phenotypes, perinatal death, and disorganized brains. Generation and analyses of TTLL-null mice are eagerly awaited.

TTLL7, a β-tubulin–specific PG, is involved in neurite extension through tubulin polyglutamylation (Ikegami et al. 2006). Suppression of TTLL7 in PC12 cells resulted in inhibition of neurite outgrowth as well as a reduction in polyGlu-β-T levels. This study also demonstrated that polyGlu-α-T and polyGlu-β-T are segregated within neurons; polyGlu-α-T is enriched within axons and polyGlu-β-T is enriched in the soma and dendrites. TTLL7 accumulated in the soma and dendrites where MAP2, a dendrite-specific MAP, is concentrated, suggest a functional interaction among TTLL7, polyGlu-β-T and MAP2. However, MAP2 binds to both β-tubulin and α-tubulin under the regulation by polyglutamylation (Bonnet et al. 2001). Thus, selective sorting of MAP2 to dendrites may require polyGlu-T and an additional regulator.

Polyglycylation

Polyglycylation is also a unique PTM of tubulin (Redeker et al. 1994). Like tubulin polyglutamylation, this PTM is known to occur on the γ-carboxyl group of multiple Glu sites of α- and β-tubulin, as shown for Tetrahymena tubulins (Redeker et al. 2005). Although purified bovine tubulin fractions contain polyglycylated populations that are specific for Δ2-tubulin (Banerjee 2002), the existence and role of tubulin polyglycylation in the nervous system is still under debate (Westermann and Weber 2003).

Extended studies of TTLL genes have identified TTLL10 as a polyglycylase (glycine ligase; Ikegami et al. 2008); however, its substrate is surprisingly not tubulin but a histone chaperone, nucleosome assembly protein 1 which is also polyglutamylated by TTLL4 (van Dijk et al. 2007). No other TTLL family members have shown glycylation activity on tubulin. Thus, the polyglycylase for tubulin and the existence and functions of polyglycylated tubulin in the nervous system are still unclear.

Future directions

As we reviewed above, recent advances in tubulin PTM research have revealed that PTMs regulate the interactions of MTs with MAPs and motor proteins and play an important role in neuronal network formation and synapse maturation. However, it is still unclear how modified tubulin or MTs are involved in changes of neuronal morphology and motility, such as neurite outgrowth and growth cone advancement, particularly in response to the extracellular guidance or repulsive cues essential for neuronal network formation. For instance, it is possible that extracellular cues alter the mode of tubulin PTMs directly or indirectly to affect MT dynamics through MAPs, resulting in neuronal shape changes. Indeed, lysophosphatidic acid, a potent neurite collapsing factor (Fukushima et al. 2002a,b; Fukushima 2004), is known to activate a small GTPase and stabilize MTs in fibroblasts, as characterized by an increase in deTyr-T levels (Cook et al. 1998), although whether this effect also occurs on MTs in neurons remains unclear.

Actin rearrangement has long attracted considerable interest as a major regulatory system in neuronal morphology and motility because actin filaments are enriched in the cell cortex or leading edge where the extracellular cues are received and transmitted into intracellular signals. Recently, there has been increasing evidence that MTs coordinate with actin and intermediate filaments through MAPs or cross-linking molecules (Leung et al. 2002; Dehmelt and Halpain 2004, 2005). In particular, the coordination between MTs and actin has been demonstrated to play a crucial role in growth cone motility and neurite outgrowth, and many proteins involved in the coordination are emerging (Schaefer et al. 2002, 2008; Dent and Gertler 2003; Rodriguez et al. 2003; Kalil and Dent 2005; Fukushima and Morita 2006; Burnette et al. 2007, 2008). Whether tubulin PTMs influence the activities of these newly identified proteins would be of interest.

Tubulin PTM research has also raised many other questions: How do modified PTMs physically influence the activity of motor proteins or MAPs? How is the activity of the PTM-associated enzymes regulated? What molecules catalyze deglutamylation to yield Δ2-T or depolyglutamylation? How are PTMs involved in the cytoskeletal network and dynamics? Further investigation of these questions would help to create a wide and comprehensive view of tubulin PTMs in the nervous system.

Acknowledgments

We thank Dr. T. Arai (Tokyo University of Science) for helpful discussion and Mr. S. Haga and T. Tamagawa for technical assistance. This work was supported by Kinki University (RK17-027).

Ancillary