Maintaining essential microtubule bundles in meter-long axons: a role for local tubulin biogenesis?

Axons are the narrow, up-to-meter long cellular processes of neurons that form the biological cables wiring our nervous system. Most axons must survive for an organism ’ s lifetime, i.e. up to a century in humans. Axonal maintenance depends on loose bundles of microtubules that run without interruption all along axons. The continued turn-over and the extension of microtubule bundles during developmental, regenerative or plastic growth requires the availability of α / β -tubulin heterodimers up to a meter away from the cell body. The un- derlying regulation in axons is poorly understood and hardly features in past and contemporary research. Here we discuss potential mechanisms, particularly focussing on the possibility of local tubulin biogenesis in axons. Current knowledge might suggest that local translation of tubulin takes place in axons, but far less is known about the post-translational machinery of tubulin biogenesis involving three chaperone complexes: prefoldin, CCT and TBC. We discuss functional understanding of these chaperones from a range of model organisms including yeast, plants, flies and mice, and explain what is known from human diseases. Microtubules across species depend on these chaperones, and they are clearly required in the nervous system. However, most chaperones display a high degree of functional pleiotropy, partly through independent functions of individual subunits outside their complexes, thus posing a challenge to experimental studies. Notably, we found hardly any studies that investigate their presence and function particularly in axons, thus highlighting an important gap in our understanding of axon biology and pathology.


The importance of axonal MT bundles
Axons are the slender, up-to-meter-long cellular processes of neurons that form the biological cables running through central nervous systems and peripheral nerves thus providing the wiring of neuronal networks; most of these delicate, structurally vulnerable extensions are irreplaceable yet have to survive for an organism's lifetime, and are therefore key lesion sites in injury, trauma, neural disorders and ageing (Conde and Caceres, 2009;Hahn et al., 2019;Prokop, 2020Prokop, , 2021. Loose bundles of staggered microtubules (MTs) run uninterruptedly all along axons (Fig. 1, top), with at least two fundamental roles (Cleveland and Hoffman, 1991;Prokop, 2020): firstly, to provide the essential highways for life-sustaining motor-driven axonal transport (Maday et al., 2014); secondly, to act as structural and/or functional components that are indispensable for and closely correlate with axonal morphology -also providing a pool from which MTs can emanate to drive the formation of new axon segments or branches (Bilimoria and Bonni, 2013;Datar et al., 2019;Kalil and Dent, 2014;Qu et al., 2019).

Mechanisms mediating the availability of tubulin heterodimers in axons
In non-neuronal cells, the logistics of distributing tubulins to sites of demand seems a relatively easy task. In contrast, sustaining MT nucleation and polymerisation in axons ('i' in Fig. 1e) up to a meter away from their genes in the neuronal cell body appears a logistical challenge. Accordingly, a reduction in tubulin levels negatively impacts axon growth: primarily through affecting nucleation as a process highly sensitive to tubulin depletion, whereas MT polymerisation can be sustained to lower tubulin levels (Consolati et al., 2020;Hahn et al., 2021;Zheng et al., 2017). Considering that the length of one tubulin heterodimer in a protofilament is ~8 nm ( Fig. 1; Krebs et al., 2005) and that the typical MT in mammalian axons has 13 protofilaments , the addition of 1 µm to a single MT would require about 1600 tubulin heterodimers, multiplied by the number of MTs existing in a respective axon section (typically several hundred MTs in larger diameter axons; Prokop, 2020). Consequently, the total amount of tubulins required will be enormous during times of axon growth, be it injury-induced regeneration, developmental axon extension or activity-induced branching. Several mechanisms are expected to cater for this need: First, tubulins can be recycled, thus reducing the need for de novo tubulin biogenesis. MTs undergo constant turnover through cycles of depolymerisation and repolymerisation (Akhmanova and Steinmetz, 2008;Bordas et al., 1983) which can drive MT network reorganisation and provides a means to prevent MT senescence. MT depolymerisation sets free α/β-tubulin heterodimers of low polymerisation competence due to the hydrolysation of β-tubulin's nucleotide when incorporated into MTs (GDP-tubulin; Lee and Timasheff, 1977). Furthermore, these tubulins may have been subjected to other post-translational modifications (PTM; Janke and Magiera, 2020) and motor-induced wear-and-tear (Andreu-Carbó et al., 2021), all affecting the functional and polymerisation competence of tubulins (Wu et al., 2022). This means that tubulins must undergo active recycling processes that involve reversal of PTMs, the exchange of GDP for GTP and a potential quality control that repairs heterodimers or targets them for degradation thus gradually depleting existing tubulin pools. Unfortunately, the underlying mechanisms are very poorly understood; components of the tubulin-specific chaperone complex (TBC) might be likely candidates to running uninterrupted from the cell bodies ('soma') to growth cones or synaptic terminals at the axonal tip ('nu', nucleus; arrow heads, synapses). The emboxed area is shown magnified below and illustrates key steps of tubulin biogenesis and regulation: (a) kinesin-mediated axonal transport of αand β-tubulin mRNA (left; 'RNPB', ribonucleoprotein bodies) or potentially heterodimers (right); (b) mRNA storage via RNPBs and their translation through ribosomes ('ribo'), with the gradually emerging nascent chain received by the co-translational prefoldin chaperone complex (Pfdn); (c) first folding steps via the chaperonin complex (CCT) generating quasi-native αand β-tubulin molecules; (d) final maturation steps via the dynamic involvement of TBCA-E ('A-E' in component-specific colour code) and the Arl2 GTPase ('G') as described in the text (here using the super-complex model from fission yeast; Nithianantham et al., 2015); note that tubulins need to acquire their GTPs (only E-site GTP on β-tubulin is shown) which is indicated as a speculative step (yellow circle); note that heterodimer release involves hydrolysis of β-tubulin's GTP ('P i ' and absence of yellow circle); (e) posttranslational regulation: (i) tubulin heterodimers feed nucleation ('γTuRC', gamma tubulin ring complex) and polymerisation ('XMAP' polymerase); note that GTP-tubulin is only at the tip of MTs ('GTP-cap') as a function of their polymerisation and subsequent hydrolysis; (ii) free tubulin heterodimers are subject to auto-inhibition as a co-translational event requiring the initial four residues of β-tubulin ('MREI'); (iii) hetero-dimer-binding proteins such as stathmins ('STMN') might provide a protected buffer or store; (iv) heterodimers (e.g. derived from MT depolymerisation; 'depol.', red arrow) can be disassembled, potentially as a means of quality control selecting for proteasomal degradation (red X) or recycling and storage (stippled green box). be involved ('iv' in Fig. 1e;Li and Moore, 2020).
Second, tubulins undergo long-range transport. For example, short MT fragments are transported in axons via slow transport at a rate of ~1 mm/day (Ahmad et al., 1998;Galbraith and Gallant, 2000;Galbraith et al., 1999;He et al., 2005;Keith, 1987;Reinsch et al., 1991). However, even if tubulins were moved as heterodi-or -oligomers via fast transport (Fig. 1a), their velocity would be in the range of a mere 40-80 mm/day (Twelvetrees, 2020), clearly posing a rate-limiting step in long axons.
Third, glia cells were reported to nurture axons by providing them with materials, for example via exosomes and extracellular vesicles (Rajendran et al., 2014;Sotelo-Silveira et al., 2006). Such supply mechanisms were reported to include tubulin proteins (Gurung et al., 2021;Jeppesen et al., 2019) and would provide a relatively rapid mechanism of provision (but see concluding section).
Fourth, the perhaps most flexible cell-autonomous means of supplying tubulins to axons would be through local translation (Fig. 1b;Jung et al., 2012): mRNAs could be provided through axonal transport or via glial delivery, they could be stored and, in times of acute need, made rapidly available serving as multiplicators by driving on-site biogenesis. However, this strategy would only work if the highly complex machinery of translation and subsequent protein maturation is locally available (Fig. 1b-e).
Here we will explore the possibility of local tubulin biogenesis in axons: we will summarise current knowledge of local translation and of the chaperone machineries involved in tubulin maturation, whilst exploring whether we have sufficient evidence for their presence and function in axons to propose local biogenesis as a means of local tubulin provision.
The assembly of ribosomes requires > 200 non-ribosomal factors and 75 small nuclear ribonucleoprotein particles (snoRNPs; Anger et al., 2013;Kressler et al., 2010). The assembly of pre-60S and − 40S particles occurs therefore in the nucleolus, followed by export to the cytoplasm for further maturation (Henras et al., 2008;Kressler et al., 2010). Consequently, this canonical biogenesis pathway requires that ribosomes are actively transported from the cell body into axons (Michaelevski et al., 2010;Noma et al., 2017), making the availability of ribosomes potentially rate-limiting for local translation, especially under conditions of sudden demand (for example upon axon injury; Gumy et al., 2010;Kalinski et al., 2015;Schaeffer and Belin, 2022;Taylor et al., 2009).
Unfortunately, we have only scarce knowledge of the mechanisms that axons can use to regulate ribosome numbers, distribution and activity (Holt et al., 2019;Schaeffer and Belin, 2022): first, it has been reported that axons which reach synaptic maturity, downregulate their mRNA levels and reduce ribosome numbers through proteasome-mediated degradation Shigeoka et al., 2016). Second, it was observed that ribosomes in growth cones undergo fast activation through actin-dependent coupling of their two subunits (Deng et al., 2021). Third, mRNAs for many ribosomal proteins are present and locally translated in axons, believed to mediate turnover to maintain the ribosome pool in a functional state (Shigeoka et al., 2019). Fourth, in vivo studies have shown that Schwann cells can deliver ribosomes and mRNA to injured axons (Court et al., 2008(Court et al., , 2011Qing et al., 2018).
The type of proteins encoded by axonal mRNA pools changes when external signals such as BDNF are applied, or as a function of development (e.g. a bias for growth cone guidance machinery during development versus synaptic components in maturity); furthermore, it appears that many of these mRNAs represent axon-specific splice versions that promote axonal transport and translation (e.g. Shigeoka et al., 2016;Sotelo-Silveira et al., 2006;Willis et al., 2005;Zivraj et al., 2010); examples are the splice-in of specific 3' UTRs in axonal mRNAs for β-actin and tau (Aronov et al., 1999;Eom et al., 2003). These mRNA splice versions are bound by specialised ribonucleoprotein bodies (RNPBs) containing factors such as ZBP1 or Vg1RBP, which silence them and link them to motor-driven axonal transport ( Fig. 1a; Baumann et al., 2012;Carson et al., 1997;Donnelly et al., 2011;Leung et al., 2006;Yoon et al., 2009). Also in their target areas, the translational activity of mRNAs is regulated by RNPBs including mRNA-processing bodies (P-bodies) and stress granules (Fig. 1b;Sfakianos et al., 2016); examples of RNPB components found in axons are G3BP1 (Sahoo et al., 2018) or ME31B (Wang et al., 2017;our unpublished observations).
In the late 1980 s, classical in situ hybridisation studies in the rat CNS clearly demonstrated that α-tubulin mRNA is strongly upregulated in neurons during times of axon growth, both during development and regeneration (Miller et al., 1987(Miller et al., , 1989, but the further processing and distribution of these mRNAs was not analysed. Whether axons contain tubulin mRNAs was doubted by some (Koenig, 1989;Olink-Coux and Hollenbeck, 1996), but many others report mRNAs for different αand β-tubulin isotypes to be present (Briese et al., 2015;Eng et al., 1999;Gioio et al., 2004;Hafner et al., 2019;Michaelevski et al., 2010;Ostroff et al., 2019;Saal et al., 2014;Willis et al., 2007), providing an important prerequisite for local tubulin translation in axons. [ 35 S]-methionine labelling followed by in situ autoradiography even suggested that tubulins might be actively translated in axons (Koenig, 1991). In addition, it was reported that also the mRNAs for components of the three chaperone complexes relevant for tubulin biogenesis ( Fig. 1b-d) are present in the axonal translatome of rat DRG neurons (Gumy et al., 2011). These three complexes are required for the posttranslational maturation of tubulin, as will be explained in the following.

Principal roles of chaperones
It is generally assumed that the three-dimensional organisation of a protein is written into its primary sequence, supported by the observation that small, denatured proteins can re-fold autonomously in vitro (Anfinsen, 1973). This ability is decreased in large proteins and is further complicated by macromolecular crowding of the cytoplasm (300-400 g/l) which favours unspecific aggregation rather than folding (Balchin et al., 2016). Furthermore, eukaryotic translation is slow (~25 s/100aa) and exposes the N-terminal end to a crowded environment whilst potential domains of the C-terminus that might be needed for proper folding are still in production; in addition, the ribosomal release channel is too narrow to allow higher order folding (Balchin et al., 2016). Preventing unproductive inter-domain interactions and achieving meaningful folding requires the functions of chaperones (Balchin et al., 2016).
As described elsewhere (Balchin et al., 2016), ~70% of newly translated proteins are passively assisted by either the nascent polypeptide-associated complex (NAC; with two subunits: NACA and NACB; Wiedmann et al., 1994) or the ribosome-associated complex (RAC; with two subunits: MPP11 and HSP70L1/HSPA14; Otto et al., 2005); the remaining 30% require active folding: some proteins are folded by the Hsp70/Hsp40 complex which, in 10% of cases, requires additional contributions by the Hsp60/chaperonin complex; other proteins circumvent Hsp70/Hsp40 through the action of the prefoldin complex which holds proteins in their native state until translation is completed, and then passes them on to the chaperonin complex for active folding.
Actins and tubulins belong to the last class of proteins that require prefoldin and chaperonin; in addition, tubulins need the tubulin-specific chaperone complex (TBC; Fig. 1b Cowan and Lewis, 2001;Lewis et al., 1997;Lundin et al., 2010). The importance of this three-step chaperone pathway is demonstrated by the fact that expression of αor β-tubulin in Escherichia coli, either alone or together, leads to the intracellular deposition of insoluble inclusion bodies (Tian and Cowan, 2013). If all required chaperones are present, this ensures that αand β-subunits take on their proper tertiary structure, heterodimerise and bind two molecules of GTP: one of them binds α-tubulin's N-site sandwiched between α and β (not exchanged nor hydrolysed); the other one binds β-tubulin's exposed E-site (Alushin et al., 2014). The E-site GTP can hydrolyse and exchange, which is an important event regulating MT polymerisation and depolymerisation dynamics ('GTP-cap' in Fig. 1; Alushin et al., 2014;Roostalu et al., 2020;Wang and Nogales, 2005). In the following, we will discuss current knowledge of the three chaperone complexes that mediate this structural outcome and what is known about their function in neurons.

The first step: prefoldin
The first step of the tubulin maturation pathway involves the group II hetero-hexameric prefoldin complex (also called GimC: Genes involved in MT biogenesis complex); it was discovered in 1998 as a chaperone required for the early steps of actin and tubulin maturation (Geissler et al., 1998;Vainberg et al., 1998). It is composed of 2 α and 4 β subunits that are evolutionarily well conserved, as shown for archaea, yeast, plants and animals; animal genomes contain two α subunit genes (PFDN3/VBP1, PFDN5) and four β subunit genes (PFDN1, PFDN2, PFDN4, and PFDN6) which all have close orthologues in plants (Cao, 2016;Siegert et al., 2000;Tahmaz et al., 2022).
All components jointly form a 'jelly fish-like' appearance composed of a disc from which 6 tentacles extend in one direction, able to bind and protect freshly translated nascent client proteins in their partly folded states (Fig. 1b); it does not actively fold its client proteins but acts as an ATP-independent co-chaperone ('holdase') that physically interacts with the chaperonin complex to transfer the proteins to chaperonin for catalysed folding (Hansen et al., 1999;Liang et al., 2020;Martín-Benito et al., 2002;Ohtaki et al., 2008;Sahlan et al., 2018;Siegers et al., 1999).
None of the prefoldin subunits has currently listed links to human disease on OMIM, but loss of prefoldin function has been studied in a variety of organisms. For example, mutations affecting prefoldin function in yeast show a range of phenotypes: reduced tubulin levels (Lacefield et al., 2006), impaired function of the actin and MT cytoskeleton (Vainberg et al., 1998), their MT networks are vulnerable to different external stressors or MT-destabilising drugs (Geissler et al., 1998) but display higher tolerance to over-expressed β-tubulin (which is usually toxic; Lacefield and Solomon, 2003). Individual loss of prefoldin components in the Thale cress Arabidopsis thaliana causes defects in the organisation of cortical MTs, increased sensitivity to MT-destabilising drugs, impaired cell elongation and division as well as reduced tolerance to salt stress and low temperature (Gu et al., 2008;Perea-Resa et al., 2017;Rodríguez-Milla and Salinas, 2009). A sextuple-mutant taking out all prefoldin genes in Arabidopsis was not lethal but showed reduced growth and premature flowering; its MT-related phenotypes were like those of the individual mutants, indicating that tubulins mainly depend on the prefoldin complex rather than additional individual functions of its components (Blanco-Touriñán et al., 2021).
Similar observations were made in animals. For example, individual knock-down of the 6 prefoldin genes in Caenorhabditis elegans caused impaired cell division leading to embryonic lethality due to reduced tubulin levels and MT polymerization; an additional cell migration phenotype was observed that was mirrored when knocking down tubulin or CCT, thus suggesting they share a common pathway (Lundin et al., 2008). Similarly, loss of PFDN3 function in Drosophila caused mitotic and meiotic phenotypes that are mimicked by loss of β-tubulin (Delgehyr et al., 2012). Mice displaying complete loss of PFDN1 were smaller, died within 5 weeks after birth and displayed hydrocephalus, neuronal loss as well as reduced and disorganised appearance of commissures and the cerebellar arbor vitae (Cao et al., 2008), which might suggest potential inhibition of axon growth. The PFDN5 L110R missense mutation in mouse was structurally predicted to slow down actin and tubulin folding: homozygous mutant mice were viable but displayed partial sterility, hydrocephalus, ataxia due to cerebellar malformation and apoptosis, and aberrant development and degeneration of the retina (Lee et al., 2011).
Although subcellular studies in the above loss-of-function analyses mainly focussed on MTs, it can be expected that also the biogenesis of actin and other potential client proteins is affected, but little appears to be known at this stage (Tahmaz et al., 2022). Instead, very different functions of prefoldin components have emerged that seem to partly occur outside the complex. For example, the above mentioned PFDN5 L110R mouse did show aberrant accumulations of poly-ubiquitinated proteins suggesting involvement in proteasome-or autophagy-dependent protein degradation . In this same vein, knock-down of PFDN2 or PFDN5 enhanced pathological protein aggregates in neuronal models of Huntington's and Parkinson's disease (Takano et al., 2014;Tashiro et al., 2013). Consistent with this, PFDN5 was found to be downregulated in Alzheimer's disease patients (Tao et al., 2020). Furthermore, the addition of Aβ 42 oligomers to primary neurons was less toxic when pre-incubated in vitro with human prefoldin complex . These studies suggest that prefoldin and/or its subunits can protect from harmful aggregates (Liang et al., 2020;Tahmaz et al., 2022).
Furthermore, the prefoldin complex and/or its individual components have nuclear functions in yeast, plants and humans; they were shown to bind and regulate proteins important for gene transcription, splicing and stress responses (Perea-Resa et al., 2017;Tahmaz et al., 2022). Such nuclear roles may also relate to the identified links of individual prefoldin subcomponents to cancer (Liang et al., 2020;Mo et al., 2020;Tahmaz et al., 2022).
In conclusion, a picture emerges where prefoldin and its components display a high degree of pleiotropy, thus complicating functional analyses. Some of the non-MT-related functions in aggregate prevention clearly are highly relevant for the nervous system. For the assembled prefoldin complex, tubulins seem to be a major client as revealed by prominent MT phenotypes upon loss-of-function. However, we could not find any descriptions of the physical presence of the prefoldin complex in axons; only commissural phenotypes in mouse provide a weak indication that there might be direct axonal requirements.

The second step: the CCT folding complex
The chaperonin complex, also referred to as CCT (chaperonin containing tailless complex polypeptide 1) or TRiC (tailless complex polypeptide 1 ring complex; Lewis et al., 1997;Lopez-Fanarraga et al., 2001) belongs to the Hsp60 family of heat shock proteins (chaperonins); they comprise the GroEL/GroES complex in E. coli, Cpn60/HSPD1 in chloroplasts and mitochondria (linked to neurodegenerative diseases; OMIM #118190), and CCT in the eukaryotic cytoplasm (Caruso Bavisotto et al., 2020). CCT was identified in the early 1990 s as a ring-shaped ATP-dependent chaperone complex able to fold tubulin, actin and luciferase (Frydman et al., 1992;Gao et al., 1992b;Lewis et al., 1992;Yaffe et al., 1992).
CCT is composed of 2 back-to-back rings, each containing 8 unique subunits (Fig. 1c) that are closely related to each other and evolutionarily well-conserved (CCT1-8 or α-θ, with CCT1 being synonymous to human T-Complex Protein 1/TCP1); in the complex, all subunits are present in equimolar amounts (Finka and Goloubinoff, 2013;Stoldt et al., 1996;Vallin and Grantham, 2019), and eliminating single CCT components leads to the disassembly of the whole complex (Amit et al., 2010;Brackley and Grantham, 2010;Kim and Choi, 2019). The structurally and functionally related bacterial chaperone GroEL requires the co-chaperonin GroES to form a lid that can close the cavity of its ring structure; in CCT, all components display flexible protrusions that jointly serve as lid (Kabir et al., 2011). Furthermore, GroEL is a homo-oligomer, whilst the hetero-oligomeric nature of CCT might provide a more complex inner structure adaptable to a broader range of substrates (Kabir et al., 2011).
CCT is involved in the folding of an estimated 5-10% of a cell's newly synthesised protein, of which actin and tubulin are highly abundant clients that absolutely require CCT (referred to as 'obligate'); each CCT is reported to bind up to 2 actins and 16 tubulins at a time, of which tubulins are released in a quasi-native state (requiring the TBC chaperone for finalisation; Finka and Goloubinoff, 2013; Llorca et al., 2000). For example, in Drosophila, CCT4 and CCT5 were shown to localise in dendrites of sensory neurons, required to promote MT organisation and dendrite growth .
The folding of various other proteins is catalysed by CCT, although some may use the complex without being absolutely dependent on it (opportunistic; Cowan and Lewis, 2001;Grantham, 2020;Kabir et al., 2011;Lopez et al., 2015). Examples of additional CCT clients are the tumour suppressor protein p53 (Trinidad et al., 2013) or the WD40-repeat-containing protein DCAF7/WDR68 involved in craniofacial development (Miyata et al., 2014). Furthermore, CCT functions do not restrict to protein folding: for example, CCT was shown in vitro to bind the actin-severing protein gelsolin in its fully mature state and inhibit its function (Svanström and Grantham, 2016).
To add to their pleiotropy, individual CCT subunits can have functions outside the complex. Correspondingly, CCT components were shown to be present in unequal amounts in mammalian and yeast cells, despite their equimolar presence in the complex (Finka and Goloubinoff, 2013;Matalon et al., 2014), and the individual knock-downs of the various components can cause very different phenotypes (Amit et al., 2010;Brackley and Grantham, 2010): For example, CCT5 in mammalian cells binds and activates myocardin-related transcription factor A (MRTFA), thus triggering actin network changes mediated by serum response factor (SRF), a MRTFA target (Elliott et al., 2015). Surprisingly, knock-down of CCT5 or CCT6 in mammalian cells inhibits re-polymerisation of F-actin whereas that of MTs was improved, suggesting independent MT-regulating roles of these CCT components (Brackley and Grantham, 2010). In support of this idea, individual CCT subunits were detected along MT lattices, even more so when tubulin C-termini were enzymatically removed (Roobol et al., 1999). In yeast, overexpression of CCT6 was able to overcome growth inhibition, irrespective of whether it was induced by inhibition of Tor signalling, Golgi sorting, cell division or endocytosis; this ameliorating function of CCT6 was abolished when mutating its ATP binding/hydrolysis site, suggesting energy dependence of its growth-promoting function (Kabir et al., 2005). Similarly, fly and mammalian CCT components work synergistically with the insulin/Tor pathway and are required for growth (Abe et al., 2009;Kim and Choi, 2019).
Further functions were suggested from studies of pathological protein aggregation, where CCT and/or its components can have ameliorating effects. For example, in yeast and in vitro, CCT was shown to cooperate with Hsp70 in preventing aggregation of polyQ-containing HTT fragments (Behrends et al., 2006). Systematic knock-down screens in C. elegans revealed that all 8 CCT components were among the ~10% of chaperone proteins able to suppress polyQ-or Aβ 42 -induced toxicity; this effect was reproduced in HeLa cells and appears consistent with the observation that CCT components are downregulated in brain pathology (Brehme et al., 2014; see also : Nollen et al., 2004). Studies in mammalian cells suggested that CCT might prevent pathological aggregates by promoting benign folding of nascent HTT fragments (Kitamura et al., 2006;Tam et al., 2006). In addition, CCT1 or even its substrate-binding apical domain alone, were able to protect neuroblastoma cells by preventing toxic aggregation in an ATP-independent manner, potentially by blocking the formation of pathological amyloidogenic conformations (Grantham, 2020;Kabir et al., 2011;Tam et al., 2006). Since CCT1's apical domain was shown to revert neuronal atrophy in mouse HD models (Zhao et al., 2016) and to penetrate cell membranes, it might have potential for therapeutic application (Sontag et al., 2013). The atrophy reversal through CCT1's apical domain was found to improve pathological transport deficits of BDNF, similarly achieved when expressing CCT3 or CCT5 (Chen, 2019;Chen et al., 2018;Zhao et al., 2016); for CCT5 this effect involved activation of CDK5 (cyclin-dependent kinase 5) leading to enhanced tau phosphorylation.
Despite these neuroprotective roles of CCT and several of its components, only CCT5 has a registered disease link on OMIM: the CCT5 H147R missense mutation is linked to recessive hereditary sensory neuropathy with spastic paraplegia (OMIM #256840; Bouhouche et al., 2006), indicating a role in axonal longevity. In addition, the CCT4 G1349A missense mutation was shown to cause early onset sensory neuropathy in the Sprague-Dawley rat strain (Lee et al., 2003). The sparse disease links among CCT components might reflect the enormous importance of this complex for cell viability (Kim and Choi, 2019), where mutations are more likely to cause abortion than postnatal disorders.
In conclusion, CCT is essential for tubulin biogenesis, but we did not find any data clarifying whether it is required locally in axons. Furthermore, any functional dissection of CCT in the context of tubulin biogenesis is obscured by its promiscuous roles in the folding of a wide range of proteins and further individual functions of CCT components outside the complex.
However, functional contributions of TBC components do not restrict to tubulin biogenesis ('iv' in Fig. 1e): they have also been shown to play roles in tubulin dissociation (Bhamidipati et al., 2000;Martín et al., 2000), storage/recycling (Abruzzi et al., 2002;Archer et al., 1998;Fanarraga et al., 1999), and degradation (Keller and Lauring, 2005). In the following, we will first outline current knowledge of the single TBC components in the canonical pathway.

TBCA takes care of β-tubulin
TBCA ('A' in Fig. 1) is the smallest TBC component, originally discovered as cofactor A required to permit the final folding of tubulins after they are released through an ATP-dependent step from CCT (Gao et al., 1994;Rommelaere et al., 1993). In vitro, mammalian TBCA was shown to form a stable complex with β-tubulin intermediates, but it was not required for β-tubulin folding (Tian et al., 1996). TBCA knock-down in HeLa and MCF-7 cells caused significant changes of MT networks and cell shapes, as well as G1 cell cycle arrest -all accompanied by a moderate decrease in the amount of soluble tubulin (Nolasco et al., 2005). Similar observations were made upon loss of the yeast orthologues Rbl2 A (Saccharomyces cerevisiae) or Alp31 A (Schizosaccharomyces pombe), which are both not important for β-tubulin folding but clearly required for proper MT maintenance, with Alp31 A and β-tubulin mutations displaying genetic interactions as an indication of their functional relationship; this said, Alp31 A loss is viable (Archer et al., 1995;Radcliffe et al., 2000aRadcliffe et al., , 2000b.
Increased TBCA levels observed in clear cell renal cell carcinoma, cause denser MT networks as part of the cancer phenotype (Zhang et al., 2013). In contrast, over-expression experiments with TBCA and Rbl2 A in other contexts failed to reveal clear phenotypes (Lopez-Fanarraga et al., 2001) suggesting that the primary role of TBCA is to maintain β-tubulin pools that can feed tubulin biogenesis (Lopez-Fanarraga et al., 2001) or tubulin recycling processes (see 2nd to last section). In agreement with this notion, Rbl2 A reduces the toxicity of over-expressed β-tubulin through direct interaction (Abruzzi et al., 2002;Archer et al., 1998), and a similar buffering role was observed in mouse (Fanarraga et al., 1999). Alp31 A seems to share these fundamental functions, since essential loss-of-function mutant phenotypes can be rescued by Rbl2 A ; but unlike Rbl2 A , its over-expression is lethal and causes MT network fragmentation, suggesting additional roles not displayed by its orthologues (Radcliffe et al., 2000a(Radcliffe et al., , 2000b. Potential roles in the nervous system seem not to have been reported so far. In Drosophila melanogaster, the closest TBCA orthologue CG1890 and a more distant orthologue CG9072 have been functionally assessed as part of nervous system-related genetic screens, but no phenotypes of importance seem to have been observed (Das et al., 2017;Neely et al., 2010).

TBCB/CKAP1 takes care of α-tubulin
Human and bovine TBCB ('B' in Fig. 1) were isolated in the mid 90s as proteins carrying a CAP-Gly domain in their C-terminal halves Watanabe et al., 1996). Apart from this domain, TBCB displays a DEI motif at its very C-terminus; this motif can bind the CAP-Gly motif of TBCE to mediate their physical interaction (Nolasco et al., 2021). It can also bind its own CAP-Gly, thus mediating auto-inactivation (Carranza et al., 2013). Since TBCB's CAP-Gly affinity for its own DEI motif is lower than for α-tubulin's C-terminal EEY motif (reduced by de-tyrosination), the presence of tubulin is believed to activate TBCB (Carranza et al., 2013). The auto-inactivation of TBCB seems to protect cells: overexpression of murine TBCB in cells leads to a fairly moderate MT depolymerisation effect, which is strongly enhanced when deleting the DEI motif (or when co-expressing its DEI-binding heterodimer partner TBCE; see 2nd to last section; Carranza et al., 2013). In addition, TBCB was shown to bind CCT components, suggesting that the transfer of α-tubulins from CCT to the TBC might involve close physical contact (Carranza et al., 2013). Alf1 B (S. cerevisiae) facilitates α-tubulin folding but is not absolutely required: its deficiency is not lethal, but it interacts with α-tubulin mutations and increases sensitivity to MT-destabilising drugs . In Drosophila oocytes, TBCB was shown to be required for the polarised localisation of axis-determining mRNAs and apico-basal polarity of the surrounding follicle cells (Baffet et al., 2012). Loss of Alp11 B in S. pombe is viable but causes a reduction in α-tubulin levels, whereas its overexpression is lethal (Radcliffe et al., 2000a(Radcliffe et al., , 2000bRadcliffe et al., 1999;. Similarly, knock-down of TBCB in MCF-7 cells has little impact on MTs; but upon activation through Pak1-mediated phosphorylation, it has a severe impact on MT networks and mitotic spindle formation relevant in cancer contexts (Vadlamudi et al., 2005). Therefore, elevated TBCB levels negatively impact on MT networks, likely through promoting α-tubulin degradation (see 2nd to last section). An alternative explanation is provided by the ability of TBCB to bind Eb1, suggesting that TBCB-overexpression destabilises MTs by competing Eb1 away from their polymerising plus ends (Carranza et al., 2013).
There are also some reports relating to the nervous system: TBCB is enriched in processes of cultured astrocytes, and TBCB knock-down caused the disruption of MTs and retraction of their processes (Zheng et al., 2022). Overexpression of TBCB in hippocampal neurons had little impact on MT networks but caused a reduction of MT-associated p150 (a dynein complex component) which it can bind directly (Kuh et al., 2012). In contrast, TBCB was shown to localise to MTs in growth cones of mammalian neurons where its knock-down enhanced axon growth and its over-expression caused axon retraction and degeneration -both consistent with the MT-destabilising roles mentioned in the previous paragraph (Lopez-Fanarraga et al., 2007). Also in zebrafish, overexpression of TBCB causes shortening of motoraxons accompanied by increased branching (Helferich et al., 2018). Furthermore, increased TBCB levels have been linked to at least two neurodegenerative disorders: ALS and giant axonal neuropathy (GAN): Firstly, ALS patient brains display a downregulation of the noncoding RNA MiR-1825, which is a direct negative regulator of TBCB; accordingly, levels of TBCB are increased correlating with decreased TUBA4A levels in ALS patient brains (Helferich et al., 2018). This observation is consistent with direct links of TUBA4A mutations to ALS22, another ALS class (OMIM #616208).
Secondly, gigaxonin seems to ubiquinate TBCB and trigger its degradation . Consequently, gigaxonin mutations in patients suffering from giant axonal neuropathy (OMIM #256850; Bomont et al., 2000) or in gigaxonin-mutant mouse models cause an increase in TBCB levels; this correlates with a decrease in αand β-tubulin levels and may explain the reduced number of axonal MTs observed in GAN disease Yang et al., 2007).
Taken together, there is currently little experimental evidence that TBCB is indispensable for tubulin biogenesis in cells, but there are various observations that its overexpression can cause pathology, likely through promoting α-tubulin degradation. Unlike most other factors involved in tubulin biogenesis, TBCB was shown to be present in axonal compartments (Lopez-Fanarraga et al., 2007; see also concluding section).

TBCE plays major roles in the nervous system
TBCE ('E' in Fig. 1) displays an N-terminal CAP-Gly motif, central leucin-rich repeats (LRR), and a C-terminal UBL (ubiquitin-like) domain (Serna and Zabala, 2016). Pac2 E of budding yeast contains a CAP-Gly motif required for MT association (Voloshin et al., 2010), whereas Alp21 E in fission yeast seems to lack this motif but can still bind Alp11 B or Alp1 D and suppress Alp11 B -mutant phenotypes (Radcliffe et al., 1999). In contrast, a TBCE-related protein in animals (LRRC35/TBCEL in vertebrates, Mulet/Mlt in Drosophila) which likewise lacks its N-terminal CAP-Gly motif cannot substitute for TBCE but destabilises MTs through direct disruption of the tubulin heterodimer, likely targeting it to the proteasome via its C-terminal UBL domain (Bartolini et al., 2005;Fabrizio et al., 2020;Nuwal et al., 2012).
There are consistent reports about the importance of TBCE for MT networks: Loss of Pfifferling E function in Arabidopsis causes severe MT depletion (Mayer et al., 1999). In Drosophila, TBCE is ubiquitously expressed localising to the cytoplasm; its loss is embryonic lethal and tissue-specific knock-down revealed requirements for MT polymerisation and synaptic growth, whereas overexpression had little effect (Jin et al., 2009). In Drosophila neural precursors, TBCE interacts through its CAP-Gly domain with nuclear envelope factors and tubulin, required for its localisation to the nuclear space and mitotic spindle organisation; spindle organisation was affected when deleting TBCE or the central motif of its CAP-Gly domain (Métivier et al., 2021). TBCE-overexpression in Drosophila neurons can rescue phenotypes of the α1-tubulin K394R mutation that inhibits acetylation of K394 and renders MTs less stable (Prokop, 2022;Saunders et al., 2022). In contrast, overexpression of TBCE in HeLa cells was shown to cause MT disassembly which required the CAP-Gly and LRR domains, but not UBL (Serna et al., 2015).
Knock-down of TBCE in mouse hippocampal neurons caused accumulation of polymerisation-incompetent tubulin (potentially through deficient tubulin recycling; see penultimate section) which triggered the phosphorylation and mislocalisation of tau (Fujiwara et al., 2020). Pmn (progressive motor neuropathy) mutant mice carry the TBCE W524G missense mutation and are used as a model for spinal muscular atrophy; the mutant TBCE W524G protein is less stable and correlates with lower MT numbers in motor axons in vivo (Martin et al., 2002), which occurs as a progressive retrograde loss (Schaefer et al., 2007). Also suralis and phrenic nerves in pmn-mutant mice showed reduced MT numbers and distal spheroids, and cultured dorsal root ganglion neurons revealed reduced MT polymerisation and retrograde transport (Schäfer et al., 2017). Furthermore, cultured pmn-mutant motorneurons grow short with severe swellings (Bömmel et al., 2002). Also the auditory nerve and outer hair cells are affected, leading to progressive hearing loss in pmn-mutant mice (Rak et al., 2013).
TBCE localises to the Golgi apparatus promoted by the small GTPase Arf1; its loss in pmn-mutant or TBCE-depleted motor neurons negatively impacts the formation of Golgi-derived MTs in the soma, which also impairs MT targeting into the axon initial segment (Bellouze et al., 2014;Haase and Rabouille, 2015;Schaefer et al., 2007). A further suggested mechanism in pmn mutant mice is increased neurofilament density which, in turn, negatively impacts on MT numbers (Yadav et al., 2016).
In humans, TBCE mutations are linked to progressive encephalopathy with amyotrophy and optic atrophy (OMIM 617207#); they cause instability and reduced levels of TBCE, negatively impacting α-tubulin levels, MT polymerisation and mitotic spindle organisation (Sferra et al., 2016). Deletion and truncation mutations of TBCE cause a reduction in MTs linked to the neurodevelopmental disorder Kenny-Caffey syndrome (OMIM #244460; Parvari et al., 2002).
Taken together, MT phenotypes seem to be caused primarily by loss of TBCE function, but the wide range of TBCE gain-and loss-of-function phenotypes observed across species makes it difficult to see a common mechanism through which this factor acts. Of all TBCs, TBCE appears the subunit most studied in the nervous system. Nevertheless, we could not find any studies reporting the presence of TBCE in axons (but see last section).

TBCD displays both beneficial and destructive potential
With over 1000 residues, TBCD ('D' in Fig. 1) is the largest TBC component composed exclusively of Armadillo and its closely related HEAT domains, which are frequently found in MT-interacting proteins, sometimes even mediating their association with MTs (Eng et al., 2017;Tewari et al., 2010). In the canonical pathway of tubulin biogenesis, TBCD is the β-tubulin-binding part of the super-complex and its functional loss is detrimental. However, bovine TBCD was also shown to disassemble α/β-tubulin heterodimers in vitro ; see penultimate section), and many other studies suggest roles in MT disassembly as detailed below.
In yeast, loss of Alp1 D curiously caused reduced αbut not β-tubulin levels; notably, loss of Alp21 E was compensated for by over-expression of Alp1 D but not vice versa, suggesting TBCD to be the more important player (Radcliffe et al., 1999). In parallel work, loss of Alp1 D caused severe MT fragmentation; overexpressed Alp1 D co-localised with MTs and was lethal at high levels, which was rescued upon co-expression of β-tubulin (Hirata et al., 1998); the same was true for mammalian TBCD (Martıń et al., 2000).
In Arabidopsis, loss of Pfifferling D caused severe MT depletion (Mayer et al., 1999). In Drosophila neuroblasts, TBCD overexpression caused a reduction in MT networks . In fly neurons, loss of TBCD or its overexpression equally caused axon degeneration, whereas dendrites showed overgrowth . In human cells, TBCD was found at midbodies and young centrioles, and its knock-down affected spindle MT dynamics and midbody abscission (Fanarraga et al., 2010). Further studies in HeLa cells demonstrated its ability to recruit γ-tubulin ring complex to centrosomes, and both loss-and gain-of-function impacted mitotic spindle defects; but tubulin levels were unaffected (Cunningham and Kahn, 2008). In non-neuronal Drosophila S2 cell lines, TBCD was reported to interact with the phosphorylation regulator Strip (Striatin interacting protein) potentially affecting MT stability .
OMIM lists so far only one TBCD-linked disorder called progressive early-onset encephalopathy with brain atrophy and thin corpus callosum (OMIM: 617193;Flex et al., 2016;Miyake et al., 2016). The two studies supporting this link found that the mutations reduced TBCD stability and its affinity for β-tubulin; one study demonstrated a reduced ability of mutant TBCD to rescue axon growth phenotypes in TBCD-depleted Drosophila neurons, consistent with the thin corpus callosum phenotype in human patients (Miyake et al., 2016); the other showed TBCD localisation to centrosomes and, when mutated, spindle aberrations that correlate with increased MT polymerisation rates (Flex et al., 2016). Further reported human TBCD mutations are not listed in OMIM: in one case they cause early onset progressive encephalopathy (Chen et al., 2021). In a further case, they are linked to atrophy with secondary microcephaly which correlated with reduced TBCD levels in patients; TBCD knock-down in mice caused reduced cell division and inhibited radial cell migration consistent with the human pathology (Edvardson et al., 2016). In another study, mutations were linked to brain atrophy with microcephaly and correlated with reduced TBCD and β-tubulin levels, but accelerated MT re-polymerisation; knock-down in zebrafish mimicked pathologies observed in patients, which were similarly caused by overexpression of TBCD (Pode-Shakked et al., 2017). A further study suggested links of TBCD mutations to spinal muscular atrophy (Ikeda et al., 2017).
Similar to TBCE, also TBCD has close links to the nervous system, but likewise a wide range of almost contradictory phenotypes. The stronger defects tend to be observed upon loss of function, but some instances report phenotypes caused by TBCD over-expression. Some axonal phenotypes were reported, but whether they are caused by aberrant TBCD function in the cell body or locally in the axons remains open.

Arl2 as the energiser
The small GTPase Arl2 ('G' in Fig. 1) was long considered an important regulator of the TBC complex, but it also plays roles in other cellular contexts, such as the regulation of mitochondrial dynamics (Newman et al., 2014). Severe MT depletion was observed upon deficiency of Hallimasch Arl2 in Arabidopsis (Mayer et al., 1999), or when Alp41 Arl2 in fission yeast was either lost or locked into its active GTP-or inactive GDP-bound state (Mori and Toda, 2013). Also GTP-locked versions of Arl2 in mammalian cells (Francis et al., 2017) and in budding yeast (Cin4 Arl2 ) decreased the number and stability of MTs (Nithianantham et al., 2015). Obviously, Arl2 is important, and its lossor gain-of-function manipulations seem to break the tubulin biogenesis cycle. This said, activity state-dependent differences were observed in Drosophila neural progenitors : Arl2 loss or GDP-locked versions of Arl2 led to reduced MT networks and polymerisation, whereas GTP-locked Arl2 versions caused enhanced MT polymerisation which seemed to involve recruitment/activation of the MT polymerase Mini spindles (orthologue of 'XMAP215' in Fig. 1e).
In budding yeast, Alp41 Arl2 was suggested to act upstream of Alp1 D in the tubulin biogenesis pathway (Radcliffe et al., 2000a(Radcliffe et al., , 2000b, suggested to bind and inactivate Alp1 D when in its GDP-bound state (Mori and Toda, 2013). Work in HeLa cells showed that GDP-Arl2 binds TBCD reducing its GAP activity and affinity for tubulin (Bhamidipati et al., 2000). Co-expression of mammalian Arl2 in its GDP-locked state (Bhamidipati et al., 2000;Tian et al., 2010) or of GDP-bound Alp41 Arl2 in yeast (Mori and Toda, 2013) was able to ameliorate MT loss induced by elevated TBCD levels. Jointly, these findings might suggest that GDP-Arl2 is a negative regulator of TBCD, not necessarily as part of the super-complex but potentially concerning MT-destabilising roles of TBC outside the complex.
More recent studies in budding yeast suggested a new model which proposes Arl2 as an inherent component of the super-complex (Nithianantham et al., 2015): detailed structural analyses revealed that Pac2 D , Cin2 E and GTP-bound Cin4 Arl2 form a stable complex that receives and embraces αand β-tubulin. In this model, the hydrolysis of Arl2's GPT provides the energy that can drive conformational change required for the final maturation of the α/β-tubulin heterodimer. The hydrolysis of Arl2 likely requires GAP activity of TBCC, as explained in the next section.

TBCC as the trigger
TBCC ('C' in Fig. 1) is related to the membrane-associated protein RP2 (Retinitis pigmentosa 2; linked to retinal degeneration; OMIM #312600) and centrosomal protein TBCCD1 (TBCC-domain containing 1; required for centrosome and Golgi positioning); all three proteins display GAP activity mediated by their C-terminal TBCC and CARP (CAP and X-linked retinitis pigmentosa 2 protein) domains; but of the three proteins, only TBCC displays tubulin-folding activity, likely because it also contains a ~30 residue-long flexible disordered N-terminus that interacts with tubulin, followed by a spectrin-like domain (Barrack et al., 2015;Bartolini et al., 2002;Feldman and Marshall, 2009;Garcia-Mayoral et al., 2011;Gonçalves et al., 2010;Grayson et al., 2002).
In fission yeast, loss of Tbc1 C caused severe tubulin depletion and loss of MTs, consistent with a role in tubulin biogenesis (Mori and Toda, 2013). In Arabidopsis, loss of Porcino C caused reduced immunostaining for tubulin correlating with reduced trichome branching, short growth linked to aberrant nuclear and cellular divisions, and infertility linked to meiotic defects (Kirik, Victor et al., 2002;Steinborn et al., 2002).
In HeLa cells, TBCC was found to be localised to the cytoplasm and centrosomes and its depletion caused multipolar spindles and mitotic failure (Garcia-Mayoral et al., 2011). However, these phenotypes do not necessarily involve direct roles of TBCC in the regulation of spindle MTs: for example, TBCC-deficient defects of MT network were suspected to cause aberrant transport of signalling complexes involved in growth regulation (Jang et al., 2020). Consistent with this notion, TBCC acts as a tumour suppressor (Hage-Sleiman et al., 2010).
In vitro experiments with purified proteins showed that TBCC is an important component of the super-complex with TBCD and TBCE: this complex was able to promote tubulin heterodimer dissociation and reformation, as deduced from the appearance of hybrid heterodimers when mixing 35 S-labelled and unlabelled heterodimers in the presence of the TBC components (Tian et al., 1999). An important role for TBCC in this process appeared to be that TBCC and TBCD act jointly as GAPs for β-tubulin, which was suggested to be an important step for the release of readily-folded tubulin heterodimers from the super-complex (Tian et al., 1999. In contrast, work in fission yeast suggested that Tbc1 C acts as a GAP for Alp41 Arl2 (Mori and Toda, 2013). This finding was supported by the work in budding yeast mentioned in the previous section showing that Cin2 C acts as a GAP for Cin4 Arl2 , thus acting as a trigger for energy-dependent conformational change of the super-complex ('super-complex' cycle in Fig. 1d); it was suggested that hydrolysis of β-tubulin's GTP which occurs upon its release from the complex (Fig. 1d), might be a function of its own GAP activity (Nithianantham et al., 2015).
Curiously, we could not find any studies addressing neural functions of TBCC although it appears to be a major player within the super-complex and relevant for MT networks across organisms. It contributes to the folding process but also provides an important trigger through its GAP activity. Whether its GAP activity acts on Arl2 or on tubulins might reflect species-specific differences, but studies outside yeast should address the possibility of TBCC-mediated Arl2 hydrolysis.

TBCB/E roles in disassembly and recycling
Besides the involvement of TBC components in tubulin biogenesis, we already mentioned examples of their roles also in MT destabilisation, α/β-tubulin dissociation and/or tubulin monomer or heterodimer recycling and storage. A prominent example is the joint role of TBCB and TBCE in heterodimer dissociation ('iv' in Fig. 1e): α/β-tubulin heterodimers are thermodynamically extremely stable, and their disassembly requires catalysis (Caplow and Fee, 2002). Dissociation is observed upon overexpression of TBCB or TBCE in cells (see above sections) or when TBCE is present in tubulin heterodimer solutions in vitro (Kortazar et al., 2006); this effect is strongly enhanced when both components are co-expressed Serna et al., 2015). TBCB and TBCE form a heterodimer via the DEI and CAP-Gly motives of TBCB and the CAP-Gly and LRR motives of TBCE (Carranza et al., 2013;Serna et al., 2015). Upon formation of a transient complex with an α/β-tubulin heterodimer, tubulin dissociation occurs as an energy-independent process resulting in a more stable α-tubulin/TBCB/TBCE complex, whereas β-tubulin is handed over to TBCA if available (Kortazar et al., 2006Nolasco et al., 2021;Serna et al., 2015). Only if TBCA is present can tubulin be recycled, otherwise it appears to be targeted for degradation (Li and Moore, 2020;Nolasco et al., 2021). Degradation likely involves TBCE's and TBCB's UBLs which protrude from the complex (Serna et al., 2015), available to interact with the proteasome.
These results illustrate how TBC components in different constellations can have very different functional outcomes, making their systemic or contextual analysis extremely difficult. Apart from deviations in cellular contexts, this modular behaviour of TBC components provides a further explanation for why functional studies provided in this review occasionally showed very different, almost contradictory results.

Concluding thoughts
Having written this review, we strongly feel that tubulin availability in axons is an important topic that is under-researched and underrepresented in the existing literature. This might partly be due to the surprisingly sparse neural disorder links displayed by the factors involved. This said, MT bundles are the essential highways for lifesustaining transport in axons and the crucial pillars sustaining axonal morphology. Tubulin deficiency in vitro or in axons negatively impacts MT nucleation and, at a higher threshold, also MT polymerisation; this closely correlates with decreased axon growth (Consolati et al., 2020;Hahn et al., 2021). Therefore, knowing how axons provide the tubulins that can sustain dynamic MT bundles, lies at the heart of understanding how axons can grow and branch, re-grow after injury, and be maintained long-term in a dynamic state. As discussed in this review, axonal transport of tubulin proteins (right in Fig. 1a) is unlikely to provide the solution for sudden high demand, especially in long axons. Glia delivery appears one existing solution and might even provide an explanation for why axons in culture appear to have more limited growth than in vivo; but important questions must be answered, such as: does glial delivery happen in different contexts of long axons, e.g. in Remak bundles, myelinated or unmyelinated axons, in the spinal cord as well as peripheral nerves? What are the signalling pathways that communicate tubulin demand from axons to glial partners? How is any axonal request for tubulins being dealt with in the narrow space of myelinating glial processes, or is it restricted to nodes of Ranvier? Can glia delivery cater for axon-, stage-or condition-specific provision of different tubulin isotypes which are known to influence context-specific MT properties (Chakraborti et al., 2016;Gasic, 2022;Hausrat et al., 2021;O'Hagan et al., 2022;Radwitz et al., 2022;Sirajuddin et al., 2014;Ti et al., 2018;Vemu et al., 2017)?
Even if there is extracellular glial supply of tubulins, it would still appear hugely advantageous if axons had cell-autonomous ways to deal with their tubulin logistics, and this would be best served by local biogenesis. Local biogenesis still requires supply of energy, amino acids, mRNAs, tRNAs, ribosomes and other proteins of the biogenesis machinery provided by axonal transport and/or glia; but once in place, this machinery would be a potent and flexible source to generate new tubulins on demand. As described in our review, local translation in axons is a well-proven phenomenon, although the flexible availability of ribosomes is still a rate-limiting factor that poses questions. The presence of tubulin mRNA is documented in various independent reports, but we only found very limited evidence for their local translation (Koenig, 1991). The biggest unknown is the presence of functional chaperone machinery in neuronal neurites, for which we found hardly any evidence apart from CCT4 and CCT5 localisation in sensory neuron dendrites of Drosophila , TBCB in mammalian growth cones (Lopez-Fanarraga et al., 2007), and the presence of chaperone-encoding mRNAs in axons of rat DRG neurons (Gumy et al., 2011). Further support comes from a very recent cryo-EM tomography study combined with mass spectrometry, which suggested the presence of TBCA, TBCD, TBDE and Arl2 within the lumen of axonal MTs or concentrated at MT lattice lesion sites (Chakraborty et al., 2022). However, the mere presence does not proof their involvement in tubulin biogenesis. Also the fact that almost every component of the prefoldin, CCT and TBC complexes was shown to be important in neurons, can still be explained through functions unrelated to tubulin biogenesis or occurring in somata rather than axons. The presence of chaperones mediating tubulin biogenesis in axons is still to be demonstrated.
A last aspect that emphasises the importance of on-demand tubulin biogenesis in axons, and needs therefore consideration, is the toxicity of high levels of tubulins, especially β-tubulins (Bhattacharya and Cabral, 2004;Burke et al., 1989;Katz et al., 1990;Weinstein and Solomon, 1990). This toxicity would suggest that it is not an option to cater for potential demand by amassing tubulins in axonal environments. Even more, extrapolating from work in non-neuronal cells, there is clear evidence that cells display auto-inhibitory mechanism that responds to over-availability of free soluble tubulins through destructions of β-tubulin mRNA and protein ('ii' in Fig. 1e); the co-translational regulation of β-tubulin requires its first four encoded residues ('MREI' in Fig. 1b) and regulatory factors such as tubulin-specific ribosome-associating factor (TTC5; linked to neurodevelopmental disorder: OMIM #619244; Bachurski et al., 1994;Ben-Ze'ev et al., 1979;Cleveland et al., 1981;Gasic, 2022;Gay et al., 1987;Lin et al., 2020;Yen et al., 1988). To deal with the problem of tubulin toxicity and auto-inhibition cells may develop mechanisms of protected tubulin storage, which may involve monomer storage via TBCA and TBCB ('iv' in Fig. 1e) or heterodimer storage via stathmins ('iii' in Fig. 1e; Chauvin and Sobel, 2015;Nouar et al., 2016;Steinmetz, 2007). In support of protective storage, loss of stathmins in Drosophila negatively impacted tubulin levels and axon growth, consistent with axon growth defects upon loss of the stathmin SCG10 also in mammalian neurons (Duncan et al., 2013;Grenningloh et al., 2004;Hahn et al., 2021;Manna et al., 2007;Ozon et al., 2002). But the key question is whether any storage acts as short-term buffer during peak times of local tubulin biogenesis or whether it occurs as long-term storage, which would be far more costly and require further logistics to overcome 'expiry date' issues.
Whatever the final answers to the various posed questions will be, gaining an understanding of the mechanisms that coordinate tubulin biogenesis, recycling and storage in axons will provide us with important further ideas for strategies that can be used in the important battle against axon degeneration as a key event in most neurodegenerative disorders.

Data Availability
No data was used for the research described in the article.