Decoding microtubule detyrosination: enzyme families, structures, and functional implications

Microtubules are a major component of the cytoskeleton and can accumulate a plethora of modifications. The microtubule detyrosination cycle is one of these modifications; it involves the enzymatic removal of the C‐terminal tyrosine of α‐tubulin on assembled microtubules and the re‐ligation of tyrosine on detyrosinated tubulin dimers. This modification cycle has been implicated in cardiac disease, neuronal development, and mitotic defects. The vasohibin and microtubule‐associated tyrosine carboxypeptidase enzyme families are responsible for microtubule detyrosination. Their long‐sought discovery allows to review and summarise differences and similarities between the two enzymes families and discuss how they interplay with other modifications and functions of the tubulin code.

A major component of the cell's cytoskeleton is the microtubules.Superficially, all microtubules look the same; they consist of polymerised a-b tubulin heterodimers.Longitudinally assembled dimers are called protofilaments; typically, 13 protofilaments assemble to form hollow cylindrical tubules.Microtubules are for a large part responsible for keeping the cell's shape, rigidity, and internal compartmentalisation.In addition, they facilitate intracellular transport and separate the chromatids during cell division.Microtubules are also involved in more complex intricate structures like axonemes or muscle tissue.To accommodate the microtubule's diverse array of functions, cells can make small variations in microtubules.
In analogy to the histone code, variations in microtubules are collectively being referred to as the tubulin code [1][2][3].Variations can be substantiated by incorporating different tubulin isoforms into the microtubule.There are several isoforms for both aand b-tubulin; 12 and 11 genes for aand b-tubulin respectively, in humans.While the sequences between the isoforms are very similar on the globular domain, the disordered C-terminal tail carries most of the sequence diversity.Posttranslational modifications are another way of creating microtubule variations.Modifications can accumulate both on the globular domain and the flexible C-terminus, and can alter properties of the microtubule, such as (de)polymerisation rate [4][5][6], and shearing likelihood [7,8], as well as the recruitment and activity of microtubule associated proteins [9,10].Modifications on tubulin can range from chemically well-defined changes, like acetylation at Lys-40 on a-tubulin [11,12], to rather heterogeneous variations like polyglutamylation [13][14][15][16], where (sometimes branched) chains of variable length are added to the disordered C-terminal tail of aor b-tubulin [17].Other modifications, including but not limited to, are methylation, polyglycylation, phosphorylation, detyrosination, and the combinations thereof [1][2][3]9,10].
In this review, we will talk about the detyrosination-retyrosination cycle, focusing on the recent discovery of the enzymes that remove the genetically encoded carboxy-terminal tyrosine, and attempt to give a mechanistic insight into the function and importance of the detyrosinating enzymes based on structural and biochemical data.

The tubulin detyrosination-retyrosination cycle
Detyrosination is an unusual post-translational modification.Rather than adding a chemical group on a side chain of a residue, it involves changing the amino acid composition of a-tubulin.The C-terminal tyrosine residue that is encoded in most of the a-tubulin isoforms is cleaved off.Detyrosination takes place preferentially on assembled microtubules [18][19][20][21].The opposite reaction, re-ligation of the tyrosine (which is most often referred to simply as "tyrosination", but we will refer to as "retyrosination" to define it as an event that restores the genetic encoding), preferentially takes place on the disassembled building blocks of the microtubule, the tubulin heterodimers [22] (Fig. 1).The detyrosination-retyrosination cycle is therefore not simply a back-and-forth equilibrium reaction mechanism, but rather a cyclic enzymatic event that is tied to microtubule polymerisation and depolymerisation.
In the early 1970s, a number of papers was dedicated to tubulin modifications.Tyrosination was initially described in 1973 [23] and detyrosination in 1977 [24].This latter study first incorporated radioactive tyrosine in (detyrosinated) rat brain tubulin preparations, to then measure its release.Remarkably, it discusses that detyrosination correlates to the assembly of microtubules; the authors found that addition of guanine triphosphate (GTP) and MgCl 2 (known to promote microtubule polymerisation) also promoted the release of tyrosine.Similarly, the addition of colchicine (an inhibitor of microtubule polymerisation) also inhibited the release of radioactive tyrosine from tubulin.The authors tentatively considered the enzyme activity found as a carboxypeptidase, although at that time they could not determine its identity.

Discovery of the tubulin tyrosine ligase
The enzyme that can retyrosinate detyrosinated tubulin was readily identified and characterised.Tubulin tyrosine ligase (TTL) is a particular protein that can ligate amino acids in a tRNA-independent manner [25].Years later, it was shown that TTL binds specifically to a-b-tubulin dimers (but not microtubules), due to their curved state [22].TTL binds to the tubulin dimer in such a way that it even actively prevents microtubule polymerisation [22].

Discovery of two distinct families of tubulin tyrosine carboxypeptidases
The identity of the detyrosinating enzymes remained unknown until 2017 when two groups, using a genetic approach in haploid human cells and chemical proteomics [19,20], independently uncovered that the vasohibins, VASH1 and VASH2, and their co-factor Small Vasohibin Binding Protein (SVBP), which were originally known as extracellular angiogenesis regulators [26], act as microtubule detyrosinases.While the genetic approach identified SVBP as the main regulator of microtubule detyrosination [19], the proteomic approach directly identified the vasohibins using irreversible inhibitors [20].Both studies show that overexpression of the vasohibins together with SVBP results in a substantial increase in detyrosination [19,20].
Cell lines without functional vasohibins, however, still contained detyrosinating activity.Using haploid genetic screens, but this time in a vasohibin-deficient background, another detyrosinating enzyme could be identified [18].This was called Microtubule-Associated Tyrosine CarboxyPeptidase (MATCAP).Later, it was shown that a paralog of MATCAP, called tubulin metallocarboxypeptidase (TMCP) 2 in the original publication, acts mostly as a b-tubulin tail modifying enzyme [21].Two variants of the TMCP2 protein (transcript 3 and 5, as listed on the NCBI database) harbour detyrosinating activity, but its physiological relevance remains uncertain.The same study showed also that MATCAP has secondary activity as de-glutaminase and Δ2 carboxypeptidase.We note that the HUGO Gene Nomenclature Committee (HGNC) designated the official symbols MATCAP1 for MAT-CAP/KIAA0895L (HGNC:34408) and MATCAP2 for TMCP2/KIAA0895 (HGNC:22206), which we use hereafter.

Structure and evolution of the vasohibins and MATCAPs
The two enzyme families, vasohibins and MATCAPs, are evolutionary unrelated.The vasohibins have a catalytic domain that belongs to the cysteine peptidases family [19,20], while the MATCAPs catalytic domain belongs to the Gluzincin clan of zinc metalloproteases, bearing a degenerate version of the family signature motif (HExxxH instead of HExxH) (Fig. 2) [18].
Superposition of the human VASH1 and VASH2 crystal structures shows that the structure of the catalytic domain of the two vasohibins is well conserved (Fig. 2A).The SVBP-holding helices are missing in VASH2 from some species, but are highly conserved in VASH1.Further, one of the human isoforms of VASH2 lacks the SVBP-holding helices, but a similar isoform does not exist for VASH1 [27] (Fig. 2B).Recombinantly expressed VASH1 and VASH2, in the absence of SVBP, both retain some enzymatic activity in vitro [19,27].An interesting question remains if some isoforms of VASH2 can retain physiologically relevant activity independently of SVBP, in vivo.
Comparing the crystal structure of MATCAP1 with the structure of MATCAP2 (isoform 3) predicted by AlphaFold [28], the overall fold of the catalytic domain is very similar (Fig. 2C).MATCAPs, similar to vasohibins, have different regions that are differentially conserved in their homologues.At the N-terminal end of the catalytic domain there is a helix and a disordered loop that are missing in some species from MATCAP1, but are conserved in MATCAP2.A stretch of un-conserved residues in MATCAP2, largely overlapping with the alternative exon 5 [21], intriguingly contains the microtubule-binding helix.Notably, a small helix at the C-terminus of both MATCAPs, is missing in some species (Fig. 2D).
The mode of binding of the two enzyme families is fundamentally different (Fig. 3).Vasohibins bind between two protofilaments, explaining the microtubule substrate preference.The vasohibins catalytic domain interacts both with the a-tubulin monomer whose tail will be cleaved, and with a-tubulin of the adjacent protofilament [29] (Fig. 3A).Although VASH1 and VASH2 generally recognise the microtubule in a similar manner, intercomparison between VASH1 and VASH2 binding to the microtubule reveals some differences in binding, including a 24 degrees tilt [30].
In contrast to vasohibins, MATCAP1 binds to only one protofilament.The main interface is formed by a MATCAP1 helix that interacts with the C-terminal helix preceding the a-tubulin tail harbouring the tyrosine [18].Interestingly, only two isoforms of MAT-CAP2 contain this helix; MATCAP2 isoforms that do not have this helix, do not display any cleavage of the C-terminal a-tubulin tail upon overexpression in human embryonic kidney (HEK) 293 cells [21].Α second interface between MATCAP1 and microtubules is a small interdimeric loop, that reaches out to interact with b-tubulin on the next dimer in the same protofilament [18] (Fig. 3B), and could explain some of MAT-CAP1's preference for microtubules.Vasohibins contain both N-and C-terminal intrinsically disordered regions (IDRs) [27,31].The N-and C-terminal IDRs of VASH1 and VASH2 are of similar length, but the N-terminal IDR is less conserved than its C-terminal counterpart.In both vasohibins, the well-conserved C-terminal IDR is positively charged, presumptively interacting with the negatively charged microtubule.Notably, the un-conserved N-terminal regions have opposite charges in the two vasohibins: while VASH1 has a highly negatively charged patch, the N-terminal region of VASH2 is overall positively charged, in line with the C-terminal domains (Fig. 2B).A comparative cell-based study between VASH1 and VASH2 has shown diffuse detyrosination on microtubules by VASH1, and more localised punctuated detyrosination profiles by VASH2.The authors attribute this difference as a consequence of the differently charged disordered domains [30].
Unlike vasohibins, MATCAPs only contain a positively charged N-terminal IDR, which encompasses regions of lower and higher conservation and predicted disorder (Fig. 2D).When stretches of the MATCAP1 IDR are increasingly truncated, the protein loses more and more localisation to microtubules in cells [18].This could be understood by a decreasing interaction surface between the negatively charged microtubule and MATCAP's positively charged disordered Nterminus.
It should be noted that TTL, which retyrosinates tubulin dimers but not microtubules, does not contain IDRs (Fig. 3D).In contrast, TTL-Like enzymes like the polyglutamylase and polyglycylase family, which have a similar fold to TTL, preferentially act on microtubules and contain IDRs, similar to detyrosinating enzymes.Intriguingly, while the tubulin-contacting residues are conserved amongst TTL orthologues, they are not conserved in different TTL-Like enzymes [22,32].That evokes the question whether the IDRs contribute to distinguish between tubulin dimers and assembled microtubules.

Vasohibins and MATCAPs recognise different residues of the a-tubulin tail
The differences between the two families extend to the way they recognise the residues adjacent to the scissile tyrosine of the a-tubulin tail.VASH1 is specifically sensitive to changing the second last glutamate in the sequence [31], while MATCAP1 relies on all four residues before the tyrosine [18] (Fig. 4A).Although none of these residues completely abrogate activity when mutated, a-tubulin isotype 3E, specifically enriched in heart and testis (www.proteinatlas.org), might not get modified by MATCAP due to its alanine on the second-last position (Fig. 4B).Three experimental structures of VASH1 to a-tubulin tail peptides are available.The first structure (protein data bank (PDB): 6J8F, [33]) has the penultimate glutamate substituted for a cysteine, that forms an erroneous cysteine bridge with the active site Cys169.The second structure (PDB: 6J8O) has a native a-tubulin tail peptide bound to inactive VASH1 (Cys169Ser) mutant, to avoid cleavage.The third structure (PDB: 6JZD, [34]) is of inactive VASH2 (Cys158Ala) bound to an a-tubulin tail.In all three structures, the scissile bond is not in position for cleavage, possibly due to the mutations necessary to obtain the experimental structures.A computational docking model compatible with mutagenesis data and the scissile bond in position for cleavage, has also been published [31,35].The orientation of the a-tubulin tail in the VASH1 structures varies, but it aligns with the model where the vasohibin catalytic domain interacts with the atubulin monomer whose tail will undergo cleavage, Fig. 4. Enzymatic preferences and catalytic site of the vasohibins and MATCAPs.(A) Relative activity of VASH1 (top row, in yellow) and MATCAP1 (bottom row, in blue) determined by alanine scanning mutagenesis in the last six residues of the C-terminal tail of a-tubulin; green (> 80% activity retained), light green (50-80%), light red (25-50%), red (< 25% retained); e.g., Y to A mutation retains less than 25% of detyrosinating activity for VASH1 compared to wild type, while MATCAP1 retains > 80% of activity.(B) Alignment of different Cterminal tail sequences of a-tubulin isotypes.(C) Experimental (PDB: 6J8O) and docking ( [31]) models of VASH1 bound to a C-terminal tubulin peptide, superposed on the cryo-EM model of VASH1 bound to a microtubule (PDB: 6WSL) (top left panel), and a zoom-in indicating the entry route of the a-tubulin tail into the enzyme (top right panel); an overlay of these peptide-binding structures in the bottom left panel indicates the catalytic site residues (ball and stick model) and the direction (arrowhead) of the peptides; in the bottom right two additional experimental structures of VASH1 with peptides (6JZD, 6J8F) that are not in a position amenable to catalysis.(D) A docking model ( [18]) of a C-terminal tubulin peptide bound to MATCAP1, superposed with the MATCAP1-microtubule cryo-EM structure (PDB: 7Z6S), indicating the last modelled residue of a-tubulin (top) and zoom-in indicating the catalytic residues (ball and stick model) and the direction of the peptide (arrowhead).(E) Indication of the ability of vasohibins (VASH1 yellow/green, VASH2 grey) and MATCAPs (MATCAP1 blue, MATCAP2 grey) to cleave different terminal residues in the context of an a-tubulin chain; tyrosine, phenylalanine and tyrosine-containing endings are indicated in purple; negatively charged glutamic acid in red; other residues in dark grey.Ability of cleavage is indicated in green (high cleavage ability), light green (low cleavage), or grey (no cleavage); conditions with no available data are indicated with a dashed line; note that while MATCAP2 can cleave the last residue, it also cleaves the second-last residue leaving atubulin with a D2 modification.proposed by the cryo-EM data [29,30].The orientation of the bound tail in the VASH2 complex, however, is not compatible with this model (Fig. 4C).
Structures of the covalent detyrosination inhibitors epoY and parthenolide bound to the catalytic residue C169 in VASH1 offer additional information [33,36].There are no published structures of peptide-bound MATCAP1, but a computational docking model agrees with the suggested a-tubulin tail entry site by cryo-EM [18] (Fig. 4D).
Experiments that swap the tyrosine for other amino acids indicate that the vasohibins strongly prefer to cleave aromatic residues, while MATCAP1 is more lenient and could also cleave other residues from the C-terminus [18,19,21] (Fig. 4E).While isoform 5 of MATCAP2 can cleave any carboxy-terminal residue tested, it can also cleave the penultimate glutamate, resulting in the irreversible D2 modification.It can also act as an exopeptidase on some of the b-tubulin isotypes [21], exposing a EExE C-terminus, similar to the a-tubulin D2 modification.D2 activity is also present for MATCAP1 (Ref.[21] and our unpublished data), but its physiological significance remains unclear.
It should be noted, that no clear preference for different a-tubulin isotypes was found for either vasohibins or MATCAP1 (3E not tested) [18,19].Interestingly, end-binding protein (EB) 1, a plus-end tracking protein, coincidentally also terminates in an acidic C-terminal tail (PQEEQEEY).However, no evidence has been found that EB1 can be detyrosinated [20,37].

Importance of the detyrosination-retyrosination cycle to life
The detyrosination-retyrosination cycle has been implicated in several pathologies [38][39][40][41], including neuronal development [20,[42][43][44].The discovery of vasohibins allowed to directly relate their function to radial migration of cortical neurons in newborn mice [20].Knockout of the SVBP gene in mice showed a reduction in total brain volume and numerous other structural defects [44], reminiscent to pathologies observed in human families with biallelic inactivation of SVBP [43,44].Loss of the MATCAP1 gene in mice resulted in a smaller cerebellum, olfactory bulb as well as total brain volume [18].Double knockouts of SVBP and MATCAP1 resulted in viable mice, with almost no detectable detyrosination, and a small-brain phenotype [18].Specifically, quantitative proteomics on brain samples did not indicate any detyrosination on tubulin 1A/B isotypes [18].However, immunoblot analysis showed a faint band for detyrosinated tubulin, which could be due to low amounts of the a-tubulin 4a isotype, which does not encode the terminal tyrosine [18].It cannot be excluded that this is due to SVBP-independent activity of vasohibinsalthough this has only been shown in vitro [19,27], or MATCAP2 isotypes that contain the microtubule-binding helix [21].
It is noteworthy, that the TTL knockout mice die within a few days after birth [42].There are two immediately apparent explanations for these phenotypes.In the absence of TTL, detyrosination levels build up [42], which could be toxic in a developing brain.Alternatively, tyrosinated tubulin could be essential.While ribosomal (tyrosinated) tubulin synthesis could in theory counteract the loss of TLL activity, in practice there is a substantial increase in detyrosinated tubulin upon TTL loss [42].In addition, one could speculate that retyrosinated tubulin could accumulate several other modifications as well, and thus be distinctly different from newly synthesised tubulin.Therefore, it is a possibility, that the combination of accumulated modifications on retyrosinated tubulin, rather than (re)tyrosinated tubulin, is essential.

Conclusions and perspectives
How is the tyrosination state of microtubules regulated, in light of the two newly discovered families of tyrosine carboxypeptidases?Individually, together, or perhaps in cooperation with other microtubule modifications?Detyrosination might be regulated through microtubule lattice spacing.Stabilised microtubules with expanded lattices, either through GTP (mimics) or paclitaxel binding, promote vasohibin-mediated detyrosination [45].Calmodulin-regulated spectrin-associated proteins (CAMSAPs), microtubule minus-end binders that stabilise microtubules through an expanded lattice [46], also promote detyrosination.Microtubule modifications might also affect detyrosination: the activity of the vasohibins seems to be promoted by microtubule polyglutamylation [47].MATCAP1 and MATCAP2 have shown to contain deglutamylating activity [21], which could be an interesting regulative mechanism.Further, detyrosination is known to co-occur with and potentially regulated by acetylation [48,49].It is unclear whether combinations of tubulin modifications have antagonistic effects, or specialised roles on microtubule behaviour and function.Calcium signalling has also been implicated in detyrosination through calpains [50].While the disordered N-terminus of vasohibins has a calpain cleavage site, this does not seem to affect detyrosination [50].
Ultimately, it would be interesting to better understand the importance and function of tubulin detyrosinases, not only in model organisms but also in humans.Generally speaking, alterations or mutations in the tubulin code are, amongst others, linked to several neuronal diseases [43,44,[70][71][72] and cancer [41].Specifically, upregulated detyrosination is associated with increased tumour aggressiveness, while TTL levels are often decreased [73], which might also be linked to the angiogenesis functionality of vasohibins [26].Vasohibins have also been linked to cardiomyocyte dysfunction [38,66].A working model is that ischaemic stress promotes detyrosination via microtubule affinity regulating kinase (MARK) 4, which removes microtubule associated protein (MAP) 4 from the microtubule surface [39].In line with the small-brain phenotype and lethality of detyrosination and retyrosination deficient mice, respectively, detyrosination has been shown to be important for neuronal development.The relation between a disturbed tyrosination cycle and neuronal and cardiac disorders is extensively reviewed here [74].
As detyrosination has been related to pathologies, inhibiting the microtubule detyrosinases could be an interesting therapeutic strategy.Because the detyrosinating enzymes are not essential in mice (unlike the retyrosinating TTL), inhibiting them should in principle be tolerated in patients.Epoxide-based warheads (epoY) can be used to irreversibly inhibit the function of vasohibins (PDB: 6J7B) [20,33].Parthenolide, an anti-cancer drug, also has been shown to covalently bind to the catalytic site of VASH1 (PDB: 6OCH) [36,69], inhibiting its function.While these molecules inhibit detyrosination, paclitaxel, a commonly used anti-cancer drug that stabilises microtubules, has been shown to promote detyrosination amongst other microtubule modifications [18,19,69].Treatment with paclitaxel can cause severe side effects like neuropathy.Therefore, paclitaxel's side effects could beat least in partattributed to promoted microtubule detyrosination.Collectively, it appears that targeting detyrosination could offer a therapeutic window to neuronal and cardiac disorders as well as counteracting adverse effects of microtubule-targeting cancer drugs; structural insights available can be guiding this process.

Fig. 1 .
Fig. 1.The detyrosination cycle.Big circle: assembled microtubules can undergo detyrosination (1) and detyrosinated tubulin heterodimers can be retyrosinated (2); detyrosinated and retyrosinated tubulin heterodimers can be repolymerised into microtubules resulting to microtubules that might be of a different composition than the original microtubule.Inlet circle: tubulin dimer and chemical model of the last two C-terminal residues of a-tubulin (-EY); cartoon scissors indicate the peptide bond that is broken during detyrosination, exposing a new C-terminal residue.

Fig. 2 .
Fig.2.Conservation and structure of the vasohibins and MATCAPs.(A) Superimposed VASH1 (yellow, PDB: 6NVQ) and VASH2 (grey, PDB: 6BQY) experimental structures bound to SVBP (green).(B) vasohibins conservation (filled area plot) and an estimate of structure disorder (black line) plotted along their respective sequence; positively charged residues (Lys, Arg) and negatively charged residues (Asp, Glu) are indicated with blue and red circles respectively; the conservation score was calculated by JALVIEW from a multiple sequence alignment (MSA) containing sequences from the UniProtKB reference proteomes target database; note that low conservation regions might appear not only because of mutations but also because some species do not contain the full sequence; the predicted local distance difference test (pLDDT) score from AlphaFold[28] was used as a proxy for estimating structure disorder.(C) Superimposed experimental structure of MATCAP1 (blue, PDB: 7Z5H) and the predicted structure of MATCAP from AlphaFold[28] (grey, AF-B4DYR3).(D) Same as (B) for the MATCAP1 and MATCAP2 proteins.

Fig. 3 .
Fig. 3. Microtubule binding properties of the tubulin detyrosinases: catalytic domain and disordered regions.(A, B) VASH1 (left) and MATCAP1 (right) bound to a microtubule, side view (top) and minus-end view (bottom).Microtubule interacting tubulin monomers and protofilaments are indicated in green.Microtubule plus (+) and minus (À) end are indicated as well as the location of the seam.The figure was created by superposing the tubulin dimers of experimental structures (PDB: 6WSL and 7Z6S for VASH1 and MATCAP1, respectively) to a full microtubule structure (PDB: 6O2S), creating a visual impression of how the enzymes bind to a microtubule.(C) Table indicating the isoelectric point (pI) of disordered termini of the vasohibins and MATCAPs, and the approximated reach of the disordered termini of the vasohibins (left) and MATCAP (right) calculated using a radius of 3 A per residue, assuming 65 and 60 N-and C-terminal residues for VASH1 and 137 residues for MATCAP, shown along the microtubule (left) and around the microtubule (right).(D) TTL molecular structure (green, PDB: 4IIJ) bound to curved dimers, supported by stathmin (pink).Note that TTL also interacts with b-tubulin on the same dimer.

1458FEBS
Letters 598 (2024) 1453-1464 ª 2024 The Author(s).FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

1459FEBS
Letters 598 (2024) 1453-1464 ª 2024 The Author(s).FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.