Microtubule polyglutamylation is an essential regulator of cytoskeletal integrity in Trypanosoma brucei

The shape of the human protozoan parasite Trypanosoma brucei is superbly adapted to life in various extracellular fluids of its mammalian hosts and in the tsetse fly vector (Heddergott et al., 2012). At the morphological level, this parasite has two characteristic features. First, it has a typical spindle-shaped cell body that is maintained throughout all stages of its complex life cycle, hydrodynamically well suited for a microswimmer (Kruger et al., 2018). Second, a single flagellum is attached to the cell body, providing the cell with the vigorous motility essential for transmission and survival within the host and vector (Broadhead et al., 2006; Engstler et al., 2007; Langousis and Hill, 2014; Rodriguez et al., 2009; Sun et al., 2018). Although the cell body is highly flexible to transmit flagellar beating forces, it is a resilient structure that must withstand the forces of the harsh environment. The main molecular components that determine flagellar function and cell body shape are microtubules (Sinclair and de Graffenried, 2019). The cell body is enveloped by a nematic array of subpellicular microtubules that are cross-linked to each other and to the adjacent plasma membrane by associated proteins (Gull, 1999; Hemphill et al., 1991). The microtubule lattice is extremely stable and does not undergo cycles of dynamic instability typical of cytoplasmic microtubule structures, e.g. in mammalian cells. This stability is most likely not an intrinsic property of the α- and β-tubulin subunits, as their sequence is very similar to that of tubulins in other eukaryotes. Instead, trypanosomes have evolved a large number of unique cytoskeleton-associated proteins that are essential for structuring and regulating the cytoskeleton (Robinson et al., 1991).

An additional layer of regulation is provided by a multitude of post-translational modifications (PTMs) of microtubules (Janke and Bulinski, 2011; Westermann and Weber, 2003). It has been known for decades that microtubules of many eukaryotes are subject to four main types of PTMs: a cycle of detyrosination and tyrosination of the C-terminal end of α-tubulin, acetylation of lysines, in particular lysine residue 40 on α-tubulin, and polyglycylation and polyglutamylation of the unstructured C-terminal tails of both α- and β-tubulin (Barra et al., 1973; Edde et al., 1990; L'Hernault and Rosenbaum, 1985; Raybin and Flavin, 1977). The latter modification is reversible and side-chain glutamates can be removed by cytosolic carboxypeptidases (CCPs) (Rogowski et al., 2010).

Combinations of these modifications have been shown to affect microtubule function in a variety of ways and the term ‘tubulin code’ has been coined, in analogy to the ‘histone code’, to describe the interdependencies of particular PTMs, interaction with associated proteins, differential expression of tubulin isotypes and microtubule function (Janke and Magiera, 2020; Verhey and Gaertig, 2007). With the exception of the absence of polyglycylation, all of the microtubule PTMs described above have been detected in trypanosomes using PTM-specific antibodies or mass spectrometry (Sasse and Gull, 1988; Schneider et al., 1997; Sherwin et al., 1987).

Polyglutamylation of microtubules is catalysed by enzymes of the tubulin tyrosine ligase-like (TTLL) family of proteins, named for the presence of a domain that is similar to a motif identified in the founding member of this family, the tubulin tyrosine ligase (Ersfeld et al., 1993; Janke et al., 2005). Eight TTLL proteins have been identified in T. brucei, similar to the nine TTLLs found in mammals (Janke and Magiera, 2020). The presence of multiple TTLLs can be explained by their differential activity in distinct cellular compartments (e.g. the cilium or flagellum, and the cytoplasm) and their specific catalytic activity. They can use either α- or β-tubulin as a substrate and act as initiases, adding a single glutamic acid to the primary protein chain via an isopeptide bond to a specific glutamic acid, or as elongases, adding additional glutamic acids to the first branched glutamate via a peptide bond (Mahalingan et al., 2020).

In most eukaryotic model systems, polyglutamylation mainly affects stable microtubule structures, such as the flagellar axoneme or microtubule bundles in nerve axons, whereas cytoplasmic microtubules are modified to a much lesser extent. Accordingly, the physiological roles of glutamylation are often associated with such structures. Using transgenic mouse models, it has been shown that polyglutamylation and/or polyglycylation deficiencies lead to fertility defects, impaired sperm motility and neurodegeneration, and are often associated with a complex of diseases known as ciliopathies. The underlying molecular defects are less well characterised, but polyglutamylation has been shown to contribute to the regulation of the activity of the microtubule-severing enzymes katanin, the microtubule affinity of tau, and interferes with the processivity of kinesin-1, a motor protein important for axonal cargo transport, leading to perturbed axonemal structures (Bedoni et al., 2016; Gadadhar et al., 2021; Genova et al., 2023; Giordano et al., 2019; Ikegami et al., 2010; Konno et al., 2016; Kubo et al., 2010; Lee et al., 2012).

We have recently shown that depletion of the enzymes TTLL6A and TTLL12B reduces glutamylation levels of both the subpellicular and axonemal microtubules (Jentzsch et al., 2020). Phenotypically, this leads to disruption of the microtubule architecture of trypanosomes by disintegrating the posterior cell pole through the formation of lobular extrusions containing microtubule bundles. In addition, quantitative tracking experiments revealed deficits in cell motility, i.e. the directed swimming capability was impaired.

Here, we extended the analysis of TTLL enzymes and used gene knockouts to analyse the function of TTLL1 in procyclic forms of T. brucei. We show that protein depletion and the associated reduction in polyglutamylation also affected the integrity of the cytoskeleton at the posterior cell pole and produced a phenotype that was clearly distinct from our previous observations. In particular, we observed the transition of the pointed tip of the cytoskeleton (Seebeck et al., 1990) into an extended ‘open-pipe’ configuration. Although the motility of the parasite in vitro did not appear to be affected at a biologically relevant magnitude, we observed significant changes in the glutamylation pattern of the flagellar axoneme.

Following our RNA interference (RNAi)-based analysis of the T. brucei polyglutamylases TTLL6A and TTLL12B (Jentzsch et al., 2020), we extended our analysis to another member of the T. brucei TTLL family, TTLL1. With the exception of TTLL6A and TTLL12B, in vivo activities of T. brucei TTLLs have not been demonstrated and their classification as putative TTLLs is based solely on phylogenetic inference and sequence homology (Casanova et al., 2015; Janke et al., 2005). We successfully generated a gene knockout of the TTLL1 gene by replacing the open reading frames of both alleles with antibiotic selection genes using homologous recombination (Fig. S1A,B). To exclude possible gene duplication events, we analysed the resulting ttll1^−/− cell line by real-time quantitative PCR (RT-qPCR) and no residual gene expression was detected (Fig. S1C).

We then examined the in vitro effect of TTLL1 depletion by western blotting of cytoskeletal preparations using antibodies that recognise the glutamyl branch point on α-tubulin (GT335) (Wolff et al., 1992), the branch point on β-tubulin (β-monoE) (Bodakuntla et al., 2022) and glutamyl side chains longer than three residues on either α- or β-tubulin (PolyE) (Shang et al., 2002) (Fig. 1A). We have previously shown that the GT335 and PolyE antibodies are specific for α- and β-tubulins and not for other cellular proteins of T. brucei on western blots (Jentzsch et al., 2020). This is in contrast to mammalian cells, in which other polyglutamylated proteins have been identified that cross-react strongly with these antibodies (van Dijk et al., 2008). Similarly, at the sensitivity level of chemiluminescent western blotting, the β-monoE antibody showed no cross-reactivity with proteins other than β-tubulin in T. brucei cell extracts (Fig. S2). With both the GT335 and PolyE antibodies, we observed a reduction in polyglutamylation mainly affecting α-tubulin. The PolyE and β-monoE signals were reduced by approximately by 50% in comparison to those in wild-type cells. (Fig. 1A). This suggests that TTLL1 has initiase activity and adds a first glutamic acid to the primary amino acid chain of α-tubulin, resulting in a reduction of the GT335 signal and, consequently, a reduction of the PolyE signal, because other polyglutamylases are no longer able to extend an initial glutamyl seed on the primary chain. However, the data do not exclude dual catalytic activity of TTLL1 as both an initiase and elongase. The antibody signals were not completely abolished, most likely owing to the fact that T. brucei has eight different polyglutamylases (Casanova et al., 2015), some of which are likely to generate products that are epitopes for the antibodies used.

To allow a direct comparison of the effect of TTLL1 deletion between wild-type and ttll1^−/− cells, we mixed the two populations in a ratio of 1:4 and processed the cells for immunofluorescence microscopy using the GT335 and PolyE antibodies (Fig. 1B). There was a clear difference in the signal strength of the two antibodies. Although wild-type cells reacted strongly with both antibodies, only weak signals were visible in the ttll1^−/− cells.

To obtain a higher resolution of cytoskeletal structures, we used a recently developed protocol for expansion microscopy in T. brucei (Fig. S1E) (Amodeo et al., 2021). Again, there was strong reduction of the PolyE signal (chains with more than three glutamates) (Fig. 1C; Fig. S3A). This difference also affected the flagellum, which was strongly stained with PolyE in wild-type cells but this staining was much weaker in ttll^−/− cells (see below). The reduction of the GT335 signal was less obvious here than by standard immunofluorescence microscopy, which might be caused by different fixation and processing protocols used in the two techniques.

The subpellicular cytoskeleton in T. brucei is formed by a helical, nematic array of microtubules. At the centre of the cells, this array comprises approximately 100 microtubules and decreases in number towards the anterior and posterior ends of the cell body (Gull, 1999). The posterior array terminates into a pointed structure. In transmission electron micrographs of detergent-extracted cells, this tip resolves into an open-pipe-like, tubular structure with a diameter between 0.5 and 1.0 µm, surrounded by approximately 25 microtubules (Hemphill et al., 1991; Seebeck et al., 1990).

In ttll1^−/− cells, this typical apical posterior structure was disrupted (Fig. 2A). Instead, more than 90% of the population had a blunt posterior end with an average diameter of 1.5 µm, sometimes up to 5 µm (Fig. 2B). The differences in the ultrastructure of the tip structure were further analysed by expansion microscopy (Fig. 2C). In contrast to wild-type cells, the cytoskeleton of ttll1^−/− cells did not form the typical curved terminal structure at their posterior end but remained straight, leading to a widening of the aperture. This structural perturbation did not affect the ability of the cell to terminate microtubule growth and to form a blunt end, albeit the end was wider than in wild-type cells. Therefore, the phenotype is unlikely to be caused by a dysregulation of microtubule assembly, but by the inability of the posterior microtubule corset to form a cohesive tip structure. This is different to the phenotype caused by TTLL6A and TTLL12B RNAi-mediated knockdown, where the posterior microtubules formed long, finger-like projections, leading to elongated cells (Jentzsch et al., 2020). In contrast, morphometric cell analysis revealed a statistically significant shortening of the cells, from approximately 20 µm for wild-type cells to on average of 17 µm for ttll1^−/− cells (Fig. 2D).

Next, we wanted to investigate possible reasons for this disorganisation. Was this a direct consequence of polyglutamylation deficiency or was the effect indirect, perhaps due to failed recruitment of microtubule-associated proteins? To address this question, we focused on two proteins that are associated with the posterior end of the cytoskeleton in T. brucei, the microtubule end-binding proteins EB1 and XMAP215. Using an antibody against EB1 and endogenously tagged mNeonGreen–XMAP215, we found that both proteins retained their localisation at the posterior end (Fig. 3A). However, in contrast to their localisation in wild-type cells (Sheriff et al., 2014; Wheeler et al., 2013), both proteins showed a broadened distribution that correlated with the increased diameter of the open tube structure of the posterior microtubule corset (Fig. 3B). This finding was surprising as we showed in a previous study that depletion of the polyglutamylases TTLL6A and TTLL12B abolished EB1 localisation at the cell tip (Jentzsch et al., 2020). In contrast, it is consistent with the observation that all microtubules terminate uniformly and are not frayed out as observed in TTLL6A- and TTLL12B-depleted cells.

To study the properties of microtubules at the posterior tip of the cell, we stained microtubules with the antibody YL1/2, which recognises tyrosinated α-tubulin (Fig. 3C, upper panel; Fig. S3B) (Kilmartin et al., 1982; Wehland et al., 1983). The presence of tyrosinated microtubules indicates newly assembled microtubules, whereas stable microtubules are progressively detyrosinated. In wild-type cells, the posterior end was only weakly stained, whereas in ttll1^−/− cells, the staining with YL1/2 was much stronger, indicating the presence of newly incorporated, not yet detyrosinated tubulin or, alternatively, a negative effect of TTLL1 depletion and lack of polyglutamylation on the ability of the tubulin tyrosine carboxypeptidase vasohibin (VASH) to catalyse detyrosination. To verify that this observation is not restricted to particular cell types (e.g. those in specific cell cycle stages) the increase in tyrosination after TTLL1 depletion was also demonstrated by imaging a wider field of view (Fig. S3B) and western blotting of cytoskeletal extracts (Fig. 3C, lower panel). Quantitation of blots revealed an approximately 2.5-fold increase of microtubule tyrosination in ttll1^−/− cytoskeletal extracts, concomitant with a decrease in polyglutamylation. A homologue of vasohibin has also been identified in T. brucei (van der Laan et al., 2019).

In addition to the morphological defects detectable in ttll1^−/− cells by microscopy, the cells exhibited an approximately 10-fold reduction in growth rate (Fig. 4A). Classification of the cell population into different stages of the cell cycle showed a decrease in the number of interphase cells, characterised by the presence of a single kinetoplast (the mitochondrial DNA, ‘K’) and a single nucleus (‘N’) (i.e. 1K1N cells), and an increase in the number of 2K2N cells, cells with more than two nuclei (>3N) and so-called zoids – cells with only one kinetoplast and no nucleus (1K0N) (Fig. 4B). The increase in the number of zoids and cells with more than two nuclei was also evident after flow cytometry analysis (Fig. S1D). Immunofluorescence staining for typical cell division markers (nuclei, flagella and basal bodies) showed abnormally high numbers of flagella and basal bodies and multiple nuclei within one cell (Fig. 4C–E). These observations are indicative of defects in cytokinesis, but not in nuclear or basal body/flagellar duplication.

The central concept of the tubulin code is the hypothesis that a combination of PTMs defines a functional readout (Janke, 2014). Therefore, a gene add-back experiment aimed at rescuing a mutant phenotype and thereby demonstrating the specificity of a gene deletion effect is not straightforward as different, largely unpredictable polyglutamylation patterns might be generated depending on protein expression levels, each with an individual phenotype. To address this potential problem to some extent, we designed two ectopic TTLL1 expression scenarios in ttll1^−/− cells. First, a cell line was generated that expresses ectopic TTLL1 driven by a strong, doxycycline-inducible promoter, resulting in ‘overexpression’ relative to TTLL1 levels in wild-type cells. Second, a cell line was generated that expressed a promoterless TTLL1 (myc-tagged) within the tubulin gene cluster, relying on polycistronic read-through transcription (Fig. 5A). This strategy would be more likely to result in TTLL1 levels that resemble endogenous expression. When we analysed the latter cells, we found that the mRNA levels, quantified by RT-qPCR, almost perfectly matched the TTLL1 transcript levels in wild-type cells (Fig. S1F). Analysis of these cells showed that the blunt-end phenotype was completely reverted to the typical pointed tip structure of the wild-type (Fig. 5B,C). The delayed growth defect of ttll1^−/− cells was also reversed to normal growth kinetics, identical to those of the wild-type (Fig. 5D).

A surprising result, but one that strongly supports the tubulin code hypothesis, was observed when TTLL1 was overexpressed (Fig. 6). Overexpression was evident from the RT-qPCR data (Fig. S1F) and the intensity of the anti-myc signal on western blots (compare Fig. 6A to Fig. 5A). In addition, the level of polyglutamylation, based on western blot analysis, was slightly higher than in wild-type cells (Fig. 6A). In TTLL1-overexpressing cells, the width of the posterior tip was still slightly wider than in wild-type cells (0.9 µm versus 0.6 µm), but smaller than in ttll1^−/− cells (1.5 µm) (Fig. 6B). The average cell length increased from 17 µm in wild-type cells to 23 µm (Fig. 6C,D). The flagellar length was similar to that of wild-type cells, but the range of lengths was much larger (Fig. 6E). Cell growth was significantly inhibited in induced cells, but in non-induced cells, it was identical to that in wild-type cells (Fig. 6F). Taken together, the data from the rescue experiments support the hypothesis that a balanced dosage of polyglutamylation is essential to maintain proper cellular architecture.

In mammalian model systems, PTMs of microtubules are preferentially found in stable microtubule structures of primary cilia, flagella or axons, rather than in highly dynamic cytoplasmic microtubules. In trypanosomes, both the subpellicular and axonemal microtubules form stable microtubule-based structures, and axonemes in T. brucei have been shown to be extensively modified by detyrosination, polyglutamylation and acetylation (Sasse and Gull, 1988; Schneider et al., 1997; Woods et al., 1989).

Having analysed the effect of TTLL1-mediated polyglutamylation in subpellicular microtubules, we next examined the effect of TTLL1 deletion on the trypanosome flagellum. By selectively depolymerising subpellicular microtubules in high-salt solution, we obtained highly enriched flagellar preparations. Analysis of the resulting protein extract by western blotting using both the GT335 and the PolyE antibodies revealed a reduction of polyglutamylation (Fig. 7A), similar to that observed in whole-cell cytoskeleton preparations (see above). We also measured an increase in the length of the flagellum from approximately 18 µm in wild-type cells to 21 µm in ttll1^−/− cells (Fig. 7B). However, this effect diminished with increasing time of the cells in culture, possibly mediated by yet unknown compensatory mechanisms.

We then analysed the extent of polyglutamylation in wild-type and ttll1^−/− cells using expansion immunofluorescence microscopy (Fig. 7C,D). When axonemes were double labelled with a general anti-α-tubulin antibody (TAT) and the PolyE antibody, both signals colocalised in wild-type cells. However, when we analysed ttll1^−/− cells, we observed a separation of the two signals. The PolyE signal was much more concentrated towards one side of the axoneme, whereas the opposite side was almost label free (Fig. 7C). Densitometric quantification of the fluorescence signals confirmed the colocalisation of the TAT and PolyE signals in wild-type cells and the separation into two peaks in ttll1^−/− cells (Fig. S4). To determine which region of the axoneme of the flagellum was unaffected by TTLL1 depletion, we double labelled flagella with the PolyE antibody and with an antibody that recognises the paraflagellar rod (PFR), a protein structure that runs along the axoneme and is located between the axoneme and the cell body (Fig. 7D,E). Here, we observed that the small line of PolyE-positive staining remaining in ttll1^−/− cells was restricted to microtubule doublets of the axoneme that are proximal to the PFR. Densitometric analysis revealed a broad PolyE peak in wild-type cells and a more focused peak in ttll1^−/− cells, located adjacent to the PFR signal (Fig. S4). Applying the standard numbering system of the nine microtubule doublets within an axoneme (Langousis and Hill, 2014), the region proximal to the PFR contained doublets 4–7 (Fig. 7E). Although the resolution obtained with expansion microscopy was insufficient to resolve individuals doublets within the axoneme, this observation correlates with the data from a recent study of intraflagellar transport (IFT) in T. brucei, showing that IFT is restricted to doublets 3 and 4, and 7 and 8 (Bertiaux et al., 2018) – the doublets that appeared to retain polyglutamylation after TTLL1 depletion. Doublets 5 and 6 are in close proximity to the PFR and are linked to the PFR by protein bridges (Kohl and Bastin, 2005). Our observations suggest intrinsic differences between doublets with respect to PTMs and differential usage as polyglutamylase substrates. Several studies have identified important roles of various PTMs for microtubule-based cargo transport in cilia and neuronal axons (reviewed in Janke and Magiera, 2020). We also analysed the flagellar ultrastructure by transmission electron microscopy but found no differences between wild-type and ttll1^−/− cells (Fig. S5).

Prompted by the observed changes in axonemal polyglutamylation in ttll1^−/− cells and following our previous approaches for cells in which TTLL6A and TTLL12B levels were reduced (Jentzsch et al., 2020), we next analysed the motility of individual cells. Specifically, we inserted small numbers of trypanosomes into custom-made polydimethylsiloxane (PDMS) chambers that restrict the swimming motion to a two-dimensional slab of 20 µm height (see Materials and Methods, and Fig. S6A,B). Chambers were coated hydrophobically to prevent unwanted adhesion of trypanosomes to the PDMS walls, resulting in large numbers of cells exhibiting unrestricted swimming motion. These features allowed us to use brightfield imaging to track cells over extended periods. The resulting image series were analysed with custom-made scripts to extract the mobility patterns (see Materials and Methods, the section ‘Analysis of the motility assay’ in the supplementary information and Fig. S6C–F). Briefly, we extracted the time series of cell positions from each image and dissected the resulting trajectories into phases of directed and tumble-like motion (hereafter referred to as run and tumble states). To this end, we determined the local straightness ‘S’ of the trajectories (determined within a period of 3 s; Fig. 8A) and used this dataset to define a threshold S₀ below or above which a position was classified as being in the tumble or run state. By definition, the local straightness varies between zero and unity, and typically a bimodal probability distribution function (PDF) of values, p(S), was observed (see, for example, wild-type kinetoplasts in Fig. 8A). Using this minimum position as a criterion for the classification (i.e. fixing S₀=0.4375) appeared to be the most appropriate and yielded a classification of trajectories that agreed well with a visual assessment (see example trajectories in Fig. 8B, inset); varying the threshold in the range 0.4≲S₀≲0.5 did not markedly alter the results. We then calculated for each position in each trajectory the instantaneous velocity ‘v’, defined via the distance travelled within a time period of τ₀=1 s. As images were taken with a frequency of ten frames per second, this velocity reflects jumps to positions ten frames downstream of the current position. With this definition, the local oscillatory features of trajectories resulting from the helical deformation of the cells during a propulsion cycle can be neglected (see the section ‘Analysis of the motility assay’ in the supplementary information for more details).

As a result of this analysis, we observed a rather broad PDF of velocities, p(v), for the ensemble of trajectories (black line in Fig. 8B). By dissecting the contributions from positions in the run and tumble states, we were able to retrieve a peaked distribution for positions in the run state with a mean of about 4 µm/s (Fig. 8B, red shaded fraction). In contrast, positions classified as being in the tumble state had velocities that were closer to zero, with a peak near 1 µm/s (Fig. 8B, blue shaded fraction). Analysis of the residence times that individual cells consecutively spent in the run or tumble state before switching to the opposite mode revealed approximately exponential PDFs with very similar shapes and mean values (Fig. S6G). This observation suggests that the switch is a memory-devoid (Markovian) process.

Comparing wild-type to ttll1^−/− cells revealed only minor changes (see Fig. 8C). In fact, there was a small but statistically significant difference in swimming velocities between wild-type and ttll1^−/− cells during run phases (tested by a two-sample Kolmogorov–Smirnov test at the 5% level). As the difference in the average swimming velocity was very small, we wondered whether this effect was also significant in a biological sense, i.e. whether it was indeed a consequence of TTLL1 depletion. As genetically manipulated cells are recent clonal, and hence genetically homogeneous, derivatives of a wild-type population with unknown genetic heterogeneity, we investigated the intrapopulation variability of cell motility by cloning wild-type 427 cells from their parental population. Surprisingly, the subsequent motility analysis revealed greater variability between clones of the wild-type population than between the non-clonal wild-type population and ttll1^−/− cells (see box plots in Fig. 8C). This result suggests that the observed differences in motility between wild-type and mutant cells, although statistically significant, are not caused by TTLL1 depletion, but rather emerge from the intrinsic heterogeneity at the population level. We therefore conclude that the motility of ttll1^−/− cells is not altered in a biologically significant way, while acknowledging considerable variation within an ensemble. The latter observation should also be taken as a caveat when quantifying or comparing trypanosome motility and other phenotypic variations in future studies.

The ribosome-independent addition of tyrosine to α-tubulin was discovered 50 years ago and tubulin polyglutamylation in 1990 (Barra et al., 1973; Edde et al., 1990). Studies in cultured cells gave only limited insight into the function of these PTMs. It was not until technologies such as RNAi or gene knockouts became available to study these tubulin PTMs at an organismal level that major progress was made in understanding the in vivo functions of these tubulin modifications (Erck et al., 2005; Ikegami et al., 2007). One reason why organismal studies were essential was the observation that detyrosinated and polyglutamylated microtubules are mostly associated with stable microtubule structures, such as axons, flagella or cilia, and are therefore relevant mainly for cells and tissues that are difficult or impossible to culture, such as neuronal cells, ciliated epithelia or sperm cells. Most cultured cells of mammalian origin have highly dynamic microtubule structures, making them less suitable for functional studies in this area.

However, in many protozoan organisms, the cytoskeleton is dominated by exceptionally stable microtubule filaments (Ferreira and Frischknecht, 2023). Most mechanisms of microtubule PTMs are highly conserved across organisms with cilia or flagella and, therefore, such organisms are excellent models to study the function of microtubule PTMs at an organismal level (Janke et al., 2005). The parasite T. brucei is ideally suited for such studies because an extensive toolbox for genetic manipulation is well established and readily applicable. In addition, the microtubule cytoskeleton of T. brucei has been intensively studied for more than half a century (Sinclair et al., 2021; Vickerman and Preston, 1970).

Experimentally induced imbalanced polyglutamylation leads to a structural and functional perturbation of stable microtubule structures. Phenotypes range from neurodegeneration (Rogowski et al., 2010; Wu et al., 2022), infertility due to deficient sperm motility (Bedoni et al., 2016; Giordano et al., 2019; Konno et al., 2016), impaired ciliary function (Hong et al., 2018; O'Hagan et al., 2017; Wang et al., 2022) and defects in microtubule-based transport processes (Bigman and Levy, 2020; Bodakuntla et al., 2019; Magiera et al., 2018). Our work shows that the activity of the tubulin tyrosine ligase-like protein TTLL1 in T. brucei is essential for the integrity of the subpellicular microtubule corset but not for cell motility. Deletion of the single TTLL1 gene led to a disorganisation of the microtubule structure at the posterior pole of the cell. The T. brucei cytoskeleton is highly polarised with almost all filaments pointing with their structural plus end towards the posterior end of the cell. The only exception is the so-called microtubule quartet: four parallel microtubules with a reversed polarity that originate at the basal body of the flagellum.

In mice, TTLL1 has been identified as an elongase that adds additional glutamyl residues to the initial branch point of the primary tubulin amino acid chain (Bodakuntla et al., 2021). Our data, based on the differential detection of initiation and elongation by the GT335 and PolyE antibodies, show a reduction in signal intensity for both antibodies on western blots, suggesting that TTLL1 in T. brucei functions as both an initiase and an elongase. However, by immunofluorescence, the reduction of the elongation marker PolyE was much more pronounced than the signal reduction of GT335, suggesting that the preferential activity of TTLL1 is indeed chain elongation. This is also supported by the observation that human initiase-type TTLLs have a highly conserved arginine residue at position 180 of their protein sequences, whereas the corresponding position in the protein sequences of human and T. brucei TTLL1 and in human TTLL6 (an experimentally verified elongase) is glutamine (Mahalingan et al., 2020).

The mouse TTLL1 enzyme is a multi-subunit protein consisting of probably five subunits, namely PGs1–PGs5, of which PGs3 has been identified as the catalytic subunit (Janke et al., 2005). The only other subunit characterised at the sequence level is PGs2 (Regnard et al., 2003), which has no homologue in the T. brucei gene database. In mice, TTLL1 appears to have a functional role mainly in the neuronal system, and TTLL1 knockout in mice is associated with neurodegeneration. In a mouse model lacking CCP1, a carboxypeptidase capable of truncating polyglutamyl side chains, TTLL1 causes excessive polyglutamylation and neurodegeneration. Deletion of TTLL1 protects against neurodegeneration in this model, probably by restoring the glutamylation and deglutamylation balance (Bodakuntla et al., 2021; Wu et al., 2022).

Depletion of T. brucei TTLL1 causes a disorganisation of the posterior pole of the microtubule lattice and shortening of the cells. This phenotype is similar, but not identical, to changes in cytoskeletal organisation following RNAi-mediated depletion of the microtubule-associated protein PAVE2 (Sinclair et al., 2021). The pointed posterior tip observed in wild-type cells disintegrates into small bundles of microtubules, leading to a widening of the tip aperture. In contrast to the PAVE2 RNAi phenotype, the microtubule structure in ttll1^−/− cells is not frayed but appears perfectly circular (Fig. 2A,C). In vitro studies on purified bovine microtubules with or without the C-terminal tails, which can be removed by subtilisin treatment, and thus with or without glutamyl side chains showed no differences in PAVE binding (Sinclair et al., 2021). Therefore, these two similar phenotypes are most likely not functionally related.

Previously, we have shown that RNAi depletion of the T. brucei polyglutamylases TTLL6A and TTLL12B also results in a severe disruption of the posterior cytoskeleton that is accompanied by changes in cell motility. However, in contrast to the TTLL1 and the PAVE2 phenotypes, the cells formed lobe-like extrusions and were longer than wild-type cells (Jentzsch et al., 2020). The observation that ttll1^−/− cells are still able to build a defined posterior cytoskeleton, albeit with a much wider aperture, indicates that there is no major effect on the regulation of the length of individual microtubules, but rather the inability to form a coherent, pointed microtubule end structure. This hypothesis is supported by the observation that two conserved microtubule end-binding proteins involved in length regulation, EB1 and XMAP215, are retained in ttll1^−/− cells. In contrast, we have shown in a previous study that EB1 is lost in TTLL6A- and TTLL12B-depleted cells and that these cells are deficient in microtubule length regulation (Jentzsch et al., 2020). However, ttll1^−/− cells showed a strong increase in tyrosinated microtubules at the posterior end. These data support recent in vitro observations of an interdependency between the tyrosination/detyrosination cycle and polyglutamylation. A study using semisynthetic tubulin with defined polyglutamyl side chains showed that polyglutamylation promotes detyrosination by the carboxypeptidase vasohibin (which is also present in T. brucei) (Ebberink et al., 2023). Another study provided data using synthetic α-tubulin tail peptides as substrates for recombinant mouse TTLL6, showing that TTLL enzyme activity was much higher on peptides lacking the C-terminal tyrosine (Mahalingan et al., 2020).

Defects in microtubule PTM patterns are associated with a number of ciliopathies, such as infertility due to abnormal sperm motility, retinal degeneration or complex diseases, such as Joubert syndrome (Lee et al., 2012; Yang et al., 2021). This prompted us to investigate the effect of TTLL1 depletion on the flagellum in T. brucei. TTLL1 depletion did not grossly affect cell motility. Characteristic features of the swimming motion, e.g. velocity and intermittent changes between directed and tumble-like phases, remained well within the appreciable heterogeneity of the wild-type ensemble. However, the polyglutamylation pattern along the axoneme of the flagellum was significantly altered. In wild-type cells, the pattern of PolyE antibody staining was fairly evenly distributed along the axoneme, whereas in ttll1^−/− cells, the staining pattern was strongly reduced, except along the microtubule doublets adjacent to the PFR structure. These doublets remained strongly positive for PolyE and appeared to retain their polyglutamyl side chains. This might indicate differential activity of distinct polyglutamylases along the axoneme. It has already been shown that the A and B tubules of the outer doublets of sea urchin axonemes are differentially modified with respect to tyrosination and polyglycylation, raising the possibility that these differences specify functionally or structurally distinct microtubules (Multigner et al., 1996). The contrasting polyglutamylation pattern between PFR distal and proximal doublets in TTLL1-deficient cells might be functionally related to the observation that IFT in T. brucei is almost exclusively restricted to two sets of axoneme doublets (3 and 4; 7 and 8), which are those proximal to the PFR structure (Bertiaux et al., 2018).

In conclusion, we have shown that microtubule polyglutamylation is essential for cytoskeletal integrity in T. brucei and that TTLL1 has differential activity at the axoneme, and we provide new evidence for a crosstalk between tubulin glutamylation and the tyrosination/detyrosination cycle.

Procyclic 427 T. brucei cells were cultivated in SDM-79 (Life Technologies, UK) supplemented with 10% fetal bovine serum (Capricorn Scientific, Germany) and 7.5 mg/l hemin (Merck, Germany) at 27°C. Knockout cells were selected with 10 µg/ml blasticidin and 1 µg/ml puromycin. Rescue cells were selected by the addition of 50 µg/ml hygromycin. Cells overexpressing TTLL1 (parental cell line PC449) were selected with 50 µg/ml phleomycin and 50 µg/ml hygromycin. Cell growth was measured with a CASY cell counter (Roche Innovatis AG, Germany). Growth curves were averaged using three biological replicates, each consisting of three technical replicates. Standard deviations for all growth curves shown are shown in Fig. S7A.

A modified version of the RNAi stem-loop vector pALC14 (Bochud-Allemann and Schneider, 2002) was used to make suitable knockout constructs. The stuffer element was removed with the enzymes XbaI and XhoI and replaced with a resistance cassette. The plasmids pNAT_BSD_6Myc (blasticidin cassette) (gift from David Horn, University of Dundee, UK) and pEnG0P (puromycin cassette) (gift from Samuel Dean, University of Warwick, UK) were used as templates for the amplification of the selection markers. To enable homologous recombination, approximately 100 bp of the 5′ and 3′ untranslated regions of the target gene were amplified and introduced into the modified vector using the enzymes HindIII/XbaI (insertion upstream of the resistance gene) and XhoI/BamHI (insertion downstream of the resistance gene). Before transfection, plasmids were linearised with HindIII and BamHI. Transfection was done by electroporation in an Amaxa Nucleofector II (Lonza, Germany) (Burkard et al., 2007). Cell clones were screened for correct integration of resistance cassettes by PCR using T. brucei genomic DNA. The double knockout was successful if both wild-type alleles were replaced by the respective resistance cassettes. The primer pairs were selected with the reverse primer downstream and outside of the target DNA sequence to validate the accurate integration of the resistance cassettes. The resulting gene deletion strain Δttl1::BLA/Δttll1::PURO is referred to as ttll1^−/− throughout this publication. The oligonucleotides used in this study are listed in Table S1.

The open reading frame of TTLL1 was inserted either into the tubulin locus using the promoterless vector pTag8-tub (Wickstead et al., 2003) to generate a construct with TTLL1 expression levels similar to those of wild-type cells or into the vector pHD1800, a derivative of pHD1700^2xmyc (Colasante et al., 2006), containing a doxycycline-inducible procyclin promoter for the overexpression of TTLL1.

The N-terminal endogenous tagging of T. brucei XMAP215 (TriTrypDB identifier Tb927.6.3090) with mNeonGreen using the template plasmid pPOTv7-phleomycin-3xTy-mNG-3xTy was performed as described (Dean et al., 2015). The primer sequences were obtained from the TrypTag database (Billington et al., 2023). Ty is an epitope tag recognised by the monoclonal mouse antibody BB2.

The transcript levels of TTLL1 in the wild-type and genetically modified cells were analysed by RT-qPCR. RNA was isolated with the RNeasy Plus Mini Kit (Qiagen, Germany) and transcribed into cDNA with the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). qPCR was performed in a StepOne real-time PCR system (Thermo Fisher Scientific) with Maxima SYBR Green/ROX qPCR Master Mix (Thermo Fisher Scientific) and 5 ng cDNA template. The primers and the protocol for the PCR reaction can be found in Table S1. The constitutively expressed gene PFR-A was used as an endogenous control. Gene expression levels were calculated using the ΔΔCT method.

Microscope slides with cytoskeletal preparations of trypanosomes were prepared as described previously (Schock et al., 2021). Microscopy was performed using a Zeiss Axio Imager M2 microscope equipped with a SPOT Pursuit CCD camera (Diagnostic Instruments) and recorded with VisiView software (Visitron, Germany). Ultrastructure expansion microscopy was performed as described previously (Kalichava and Ochsenreiter, 2021). Image overlays and measurements were done with ImageJ (version 1.53). To measure fluorescence intensity across the flagellum, small rectangles of identical sizes covering selected regions of the flagellum were drawn and intensity profiles were measured using the ImageJ plugin ‘Plot profile’. Measurements of the posterior cell aperture were done using the ImageJ line measurement tool as shown in Fig. S7B. For all morphometric measurements, only cells with one non-dividing nucleus and one or two kinetoplasts (G1, S and G2 phases of the trypanosome cell cycle) were considered and cells in mitosis were excluded. Images were edited with ImageJ. Images were deconvoluted with the plugin ‘DeconvolutionLab2’ and the algorithm ‘Richardson-Lucy’ (RL). The point-spread function (PSF) image was generated using the ImageJ plugin ‘Diffraction PSF 3D’. Statistical analysis and data presentation was performed using the Past v3.24 software package (University of Oslo; https://www.nhm.uio.no/english/research/resources/past/). To examine the significance of differences between samples, a Kruskal–Wallis test was used. As the H-test only provides information about the difference of all samples considered, we used an additional Mann–Whitney-U-test for pairwise comparisons. All figures were assembled in Microsoft PowerPoint. Information on image acquisition for determining trypanosome motility is provided below. All microscope images shown are representative of at least three independent experiments.

Cytoskeletal protein samples for SDS-PAGE were prepared as described previously (Schock et al., 2021). For fractionation, the samples were separated into soluble and insoluble components after treatment with the detergent by centrifugation at 15,000 g for 5 min. Pelleted cytoskeletons were resuspended in hot SDS sample buffer [125 mM Tris-HCl pH 6.8, 5% (v/v) glycerol, 4% (v/v) SDS, 5% β-mercaptoethanol, some crystals of Bromophenol Blue] and incubated at 100°C for 10 min. Flagella were isolated via sodium chloride extraction (Robinson et al., 1991). An equivalent of 2.5×10⁴ cells per lane were loaded onto 10% SDS polyacrylamide gels. High-resolution separation of α- and β-tubulin on polyacrylamide gel was done as described previously (Banerjee et al., 2010). Proteins were transferred to nitrocellulose membranes and probed with primary and HRP-conjugated secondary antibodies. Blots were developed by chemiluminescence with Lumigen substrate (Takara, Japan) and visualized on an ImageQuant LAS-4000 detection system (GE Healthcare).

The following primary antibodies were used: mouse monoclonal anti-monoglutamylated α/β-tubulin [GT335; cat. no. AG-20B-0020-C100, Biomol, Germany; 1:10,000 for western blotting (WB), 1:6000 for immunofluorescence (IF)]; rabbit polyclonal anti-polyglutamylated α/β-tubulin (PolyE; cat. no. AG-25B-0030-C050, Biomol; 1:10,000 for WB, 1:6000 for IF); rabbit polyclonal anti-monoglutamylated β-tubulin (β-monoE; gift from Carsten Janke, Institute Curie, University Paris-Saclay, France; 1:7000 for WB) [the precise epitope specificities of the glutamylation antibodies are described in Bodakuntla et al. (2022)]; mouse monoclonal IgG anti-tubulin (TAT; gift from Keith Gull, University of Oxford, UK; 1 µg/ml for WB, 5 µg/ml for IF); mouse monoclonal IgG anti-PFR (L8C4, gift from Keith Gull; 1:10 for IF) and anti-Ty epitope (BB2, gift from Keith Gull; 1:1000 for IF) (Bastin et al., 1996); mouse monoclonal IgM anti-EB1 [our laboratory (Jentzsch et al., 2020); 1:5 for IF]; rat monoclonal IgG anti-tyrosinated anti-tubulin (YL1/2, cat. no. MAB1864, Sigma-Aldrich, Germany; 1:250 for WB, 1:10 for IF); mouse monoclonal IgG anti-BIP [our laboratory (Winter et al., 2017); 1:3000 for WB]; and mouse monoclonal IgG anti-CAP5.5 (this laboratory). CAP5.5 (TriTrypDB identifier Tb927.4.3950) is a procyclic-specific cytoskeleton-associated protein and often used as a marker for successful preparations of cytoskeletons from T. brucei (Hertz-Fowler et al., 2001). The mouse monoclonal anti-CAP5.5 was generated as follows: a 1482 bp fragment of the N-terminal domain and a 768 bp fragment of the C-terminal domain of CAP5.5 were cloned in the vector pTrcHisC (Invitrogen) and expressed in Escherichia coli Rosetta (Merck-Millipore, Germany). Purification was done under denaturing conditions on Protino Ni-NTA resin (Macherey–Nagel, Germany) and a mixture of both proteins at 50 µg for the primary injection and 25 µg for each booster injection, with TiterMax Gold adjuvant (Sigma-Aldrich) was used for the immunization of female BALB/c mice (Janvier Labs, France). After a routine immunisation scheme, spleen cells were isolated and fused to P3X63-Ag8.653 myeloma cells. Fusion was done with polyethylene glycol 4000 and cells were plated into 24-well plates containing OptiMEM, 10% fetal calf serum, hypoxanthine, aminopterin and thymidine (Thermo Fisher-Gibco, Germany). To condition the medium, peritoneal mouse macrophages were seeded into the plates 3 days prior to plating. Supernatants were screened by immunofluorescence on procyclic T. brucei cytoskeletons. Positive hybridoma cells were subcloned by limiting dilution. The antibody isotype was determined using the Isostrip mouse mAb typing kit (Roche, Germany) and is of the IgG1 isotype. Production of monoclonal antibodies was done in accordance with local regulations and licensed by the Government of Lower Franconia.

The motility of T. brucei cells was analysed via brightfield imaging in a similar fashion as described before (Jentzsch et al., 2020). To support long-term monitoring of cell movement in two dimensions, we designed and fabricated specialized chambers via soft lithography techniques (Whitesides et al., 2001) that restricted cell motion to a 20 µm thin slab (see Fig. S6A). In brief, a uniform layer of photoresist (SU8-2015, MicroChem, USA) was applied to a silicon wafer via spin coating at 2000 rpm, creating a consistent thickness of 20 µm. After soft baking the wafer for 3 min on a 95°C hot plate, the chamber design (generated with AutoCAD) was precisely written onto the photoresist with a Microwriter M3 Baby Plus (Durham Magneto Optics). The exposed/activated photoresist was hardened in a post-exposure bake (5 min at 95°C). Subsequently, the remaining areas were washed off with developer (mr-Dev 600, Micro Resist Technology, Germany) and isopropanol. The resulting negative master of the chamber pattern was employed as a mold for PDMS casting. To this end, the PDMS precursor (Sylgard 184, Dow Europe, Germany) and its curing agent were mixed in a ratio of 10:1, degassed in an exicator and then poured onto the wafer. The PDMS device was cured at 75°C for 3 h. After punching two inlet holes for subsequent filling, chambers were covered with a glass coverslip (Menzel, 24 mm×60 mm, #1.5). To this end, cover slips were cleaned in an ultrasonic isopropanol bath, while PDMS surfaces were cleaned using magic tape. Then both components were activated in a plasma oven at 1.12×10² Pa for 20 s. Upon assembling in a sandwich-like manner, the exposed PDMS and glass surfaces were covalently bound and further sealed by heating to 75°C for 40 min. In this step, the chambers were also hydrophobically coated by filling them with 1% trichloro(perfluorooctyl)silane (Sigma-Aldrich, Germany) in HFE 7500 oil (IoLiTec, Germany) after 20 min. Drying of the coating occurred during the second half of the heating cycle. Prior to motility experiments, chambers were soaked in Milli-Q water to prevent swelling of the PDMS during the measurement. After loading the chamber with cells (typical cell density of 3–4 million cells per ml), both inlets were sealed to prevent evaporation-driven flows.

Motility measurements typically consisted of ten 3-min videos with 10 frames per second, each yielding ∼50 trajectories of individual cells. For our analyses, we considered approximately 500 trajectories per condition. Imaging of swimming trypanosomes was performed in brightfield mode on a Leica DMI600B inverted microscope equipped with a Leica DFC360FX camera and a 20×/0.7 NA air immersion objective. Individual time series of TIF-files were then subjected to further analysis using ilastik (Berg et al., 2019), ImageJ and customized MATLAB codes (Rehfeldt and Weiss, 2023). Details on the image analysis and the subsequent extraction and analysis of the trajectories of individual cells are described in detail in the section ‘Analysis of the motility assay’ in the supplementary information.

We thank Keith Gull (University of Oxford, UK) for the TAT, L8C4 and BB2 antibodies, Carsten Janke (Institute Curie, University Paris-Saclay, France) for the β-monoE antibody, André Schneider (University of Bern, Switzerland) for the pALC14 plasmid, Bill Wickstead (University of Nottingham, UK) and Samuel Dean (University of Warwick, UK) for the pTAG-tub, pNAT pEnG0P and pPOTv7-mNeonGreen plasmids, and Frank Voncken (University of Hull, UK) for the pHD1800 plasmid. We thank Rita Grotjahn for assistance with electron microscopy and Markus Retsch (University of Bayreuth, Physical Chemistry I) for access to the clean room and the microwriter.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261740.reviewer-comments.pdf