Tubulin detyrosination promotes monolayer formation and apical trafficking in epithelial cells

Polarization of epithelial cells involves the separation of the plasma membrane into two membrane domains and a complete reorganization of the cytoskeletal architecture. Microtubules, as central elements of the cytoskeleton, play critical roles in the generation of an apico-basolateral polarity and specific transport pathways to the apical or the basolateral cell pole (Müsch, 2004). Alterations in microtubule composition thus affect the integrity of epithelial cell sheets (Yap et al., 1995). Tubulin is subject to distinct post-translational modifications, which include acetylation, tyrosination, detyrosination, Δ2 modification, polyglutamylation, palmitoylation and phosphorylation (Hammond et al., 2008). Tubulin tyrosination was the first tubulin-specific modification reported (Barra et al., 1973). Pools of tyrosinated (tyr-tubulin) and detyrosinated (detyr-tubulin) microtubules are generated by a cycle of removal and subsequent re-ligation of tyrosine to the C-terminus of α-tubulin. Whereas dynamic microtubules are tyrosinated, detyrosination leads to stabilisation and inhibition of microtubule disassembly (Peris et al., 2009). This process can be modulated by the depolymerizing motor mitotic centromere-associated kinesin (MCAK). Tubulin tyrosination is dependent on the synthesis of tubulin tyrosine ligase (TTL), which acts on αβ-tubulin heterodimers and restores tyr-tubulin (Raybin and Flavin, 1975). A knockout of TTL is lethal in mice within 1 hour after birth due to neuronal disorganization, suggesting that this enzyme is essential (Erck et al., 2005). The distribution of tyr- and detyr-tubulin has been studied intensively in different cell types and during different phases of cell division. Dynamic tyrosinated microtubules are commonly found in the interphase network and the metaphase spindle, whereas stable detyrosinated microtubules have only been detected in a limited subset of interphase microtubules from TC-7 cells (Gundersen et al., 1984). In neuroblastoma cells detyr-tubulin was restricted to microtubules of elongated cell processes, with a concentration of tyr-tubulin in the cell bodies (Wehland and Weber, 1987). In general, the ratio of detyr- to tyr-tubulin is higher in cells that have established cell-cell contacts than in isolated cells 1 day after seeding (Bré et al., 1991). In polarized neuronal cells tyrosinated tubulin acts as a directional cue to navigate motor heads of kinesin-1 to axons (Konishi and Setou, 2009). Tubulin modification thus has the potential to determine cellular polarity by a precise regulation of kinesin-1 transport. This prompted us to examine the dynamics of tubulin detyrosination to tyrosination during the formation of polarized epithelial cell sheets. We found an increase followed by a decrease in the number of detyrosinated microtubules during polarization of MDCK cells. Constant depletion of detyr-tubulin by overexpression of TTL resulted in prolonged time periods for monolayer formation, and perturbed the surface delivery of secreted and membrane-bound cargo molecules. In particular apical transport pathways were affected, which indicates that detyr-tubulin-containing microtubules are involved in at least two processes of MDCK cells; the correct assembly of cells into monolayers and guiding apical protein trafficking in fully polarized cells.

The arrangement of detyrosinated, acetylated and tyrosinated α-tubulin in polarized epithelial cells was first studied by immunofluorescence. Line scan analysis of MDCK cell cysts grown in Matrigel showed an accumulation of all three tubulin variants close to the apical membrane, followed by a gradual decline towards the basal cell pole. Detyr- and tyr-tubulin were equally distributed along the apical–basolateral axis. Steeper gradients were occasionally observed for acetylated tubulin (Fig. 1A; see also supplementary material Fig. S1), which supports data from Jaulin and Kreitzer who showed an enrichment of acetylated tubulin at the apical cell pole (Jaulin and Kreitzer, 2010). In a further study, the quantity of detyr- and tyr-tubulin at the two cell poles was studied by TIRF microscopy, which detects fluorescent dyes in close proximity to the apical or basal plasma membrane of polarized MDCK cells (Schneider et al., 2010). Most of the membrane-proximal microtubules were enriched in tyr-tubulin, while a smaller proportion was double stained by detyr- and tyr-tubulin (supplementary material Fig. S2A,B). Overall, similar ratios of detyr- and tyr-tubulin were detected at the apical and basal domain (Fig. 1B), which indicates that neither tubulin variant preferentially concentrates at the poles of polarised cells and confirms the line scan data from MDCK cell cysts.

The distribution of detyr- and tyr-tubulin on double-positive microtubules was then analysed by high-resolution ground state depletion with individual molecular return (GSDIM) microscopy of MDCK cells. The cells were grown for 1 day after seeding on coverslips and fixed with methanol, then immunostained with anti-tyr- or anti-detyr-tubulin antibodies. As depicted in Fig. 1, short detyr-tubulin sections of up to 1 µm long were disrupted by longer segments of tyr-tubulin, thus leading to a string-of-pearls-like appearance of the two tubulin variants with a preferential concentration of tyr-tubulin at the tubulus ends. Similar arrangements of interspersed detyr- and tyr-tubulin stretches had been seen for interphase and metaphase microtubules using immuno-EM (Geuens et al., 1986). However, antibodies directed against α-tubulin as a control were more or less evenly distributed along microtubules (supplementary material Fig. S2C).

Next we addressed whether the balance between the two tubulin variants varies during the course of epithelial polarisation. Therefore, MDCK cells were incubated on filters and analysed after 1–5 days. Immunofluorescence data suggested that detyr-tubulin decreased during the process of epithelial cell polarisation (supplementary material Fig. S2D). This was confirmed by immunoblot analysis of MDCK cell lysates harvested after different incubation intervals with antibodies directed against α-tubulin and tyr-, detyr- or acetylated tubulin (supplementary material Fig. S2E). Densitometric quantification of the immunoblots shows a slow increase in tyr-tubulin as well as acetylated tubulin during cell polarisation (Fig. 1D), which is in agreement with the overall level of cellular α-tubulin. The level of detyr-tubulin, however, increased to a maximum on the second day, followed by a decrease to less than 10% at day 5. This decrease is statistically significant and corroborates our immunofluorescence observations, as well as previously published data by Quinones et al. (Quinones et al., 2011). Thus, the pattern of tubulin modifications changes during the course of epithelial polarisation with an initial increase followed by a decrease in the amount of detyr-tubulin, resulting in a high prevalence of tyr- and a minor pool of detyr-tubulin in fully polarised cells.

Tubulin tyrosine ligase (TTL), the enzyme that converts detyr- into tyr-tubulin, is expressed in the cytosol of polarized MDCK cells as assessed by immunofluorescence staining with TTL-specific antibodies (Fig. 2A). To monitor changes in TTL expression during epithelial polarization, MDCK cells were harvested on different days after seeding. Equal amounts of protein from cell lysates were separated by SDS gel electrophoresis and further processed by western blot analysis with anti-TTL antibodies. A housekeeping protein, GAPDH was used as a loading control. As depicted in Fig. 2B,C, TTL expression changed with cellular differentiation. Just following seeding TTL expression was low, but dramatically increased during cell polarization from day 2 to 4. Once the monolayer had been formed at day 5, TTL quantities once again declined. Hence, the synthesis of TTL is at a maximum level when MDCK cells are fully polarized. This is in accordance with the observed decline of detyr-tubulin during cell polarization (Fig. 1D), since an increase of TTL is expected to enhance the tubulin tyrosination.

We next artificially decreased detyr-tubulin in MDCK cells by continuous expression of GFP-tagged TTL (TTL–GFP). TTL–GFP-expressing MDCK cells (MDCK_TTL–GFP) were selected after transfection in the presence of neomycin and positive clones stably expressing TTL–GFP were verified by western blot with anti-TTL and anti-GFP antibodies (supplementary material Fig. S3A). The amount of detyr-tubulin was dramatically decreased in MDCK_TTL–GFP (Fig. 3A,B). Interestingly, polyglutamylated tubulin was also depleted to about 50% by TTL overexpression. The concentration of tyr- or acetylated tubulin was not significantly altered. The reduction in detyr-tubulin in MDCK_TTL–GFP cells was verified by immunohistochemistry with antibodies against tyr- and detyr-tubulin in polarized epithelial cells. A mixture of MDCK and MDCK_TTL–GFP cells was analysed by immunofluorescence microscopy using antibodies directed against tyr- or detyr-tubulin 1–4 days after seeding (Fig. 3C; supplementary material Fig. S3B). In direct comparison, detyrosinated microtubules of non-polarized TTL-overexpressing cells were less frequent compared to control cells. Furthermore, detyr-tubulin was barely detectable in polarized MDCK_TTL–GFP cells, whereas several detyr-tubulin-enriched tubules could be seen in control MDCK cells. For a quantitative assessment of the effect of TTL overexpression on detyr-tubulin, we used a nocodazole depolymerisation assay (Khawaja et al., 1988). Nocodazole was added for various times at a concentration (10 mM) that slowly depolymerizes stable detyr-tubulin microtubules (supplementary material Fig. S3C). Following a 1 hour treatment with nocodazole, about 40% of polarized MDCK cells were positive for detyr-tubulin. In contrast, the number of detyr-tubulin-positive polar MDCK_TTL–GFP cells was negligible. After nocodazole treatment for longer periods of time, only a few detyr-tubulin-positive MDCK cells remained. Thus, polarized MDCK cells harbour significant amounts of stable detyr-tubulin-enriched microtubules, which slowly depolymerize in the presence of nocodazole. Our data indicate that this pool of stable microtubules was almost absent in MDCK_TTL–GFP cells at the start of the experiment.

We also addressed whether overexpression of TTL in MDCK cells affects the ratio of tubulin polymer to soluble dimer. It had previously been shown that TTL binds preferentially to non-assembled tubulin (Arce et al., 1978), so changes in the level of this enzyme might also influence the total amounts of assembled microtubules. We therefore, separated assembled microtubules from soluble tubulin dimers by centrifugation. Antibodies directed against cytosolic GAPDH and the microtubule-associated motor protein KIF5B where used as a control to discriminate between microtubules in the pellet fractions and soluble material in the supernatants. Surprisingly, the amount of assembled versus soluble tubulin was almost equal or significantly elevated in MDCK_TTL–GFP cells up until 3 days after seeding (supplementary material Fig. S4). This changed following cell polarization, with a decrease in the amount of polymerized tubulin. This change in the balance can be explained by constantly elevated TTL levels in MDCK_TTL–GFP cells; the enzyme would normally decline following cell polarization (Fig. 2B,C). Elevated TTL levels may thus sequester tubulin subunits and prevent them from being incorporated into microtubules, as previously described (Szyk et al., 2011).

Based on our working hypothesis that changes in the level of detyr-tubulin during the process of epithelial polarization are required for the correct formation of epithelial cell sheets, we then studied the formation of polarized MDCK_TTL–GFP cell monolayers in greater detail. Already by 1 day after plating, MDCK_TTL–GFP cells showed a much higher tendency to join into islands of 10–30 cells than MDCK cells (Fig. 4A). As indicated by immunostaining of claudin-1 in the upper part of laterally aligned plasma membranes, tight junctions had already been assembled 12 hours after plating in MDCK_TTL–GFP cells (Fig. 4B, upper panel). In contrast, in MDCK_GFP control cells the majority of claudin-1 was still in intracellular vesicles. Moreover, ZO1 and E-cadherin were aligned earlier along lateral membranes in MDCK_TTL–GFP than in MDCK_GFP cells (supplementary material Fig. S5A). If cells polarized after 5 days were immunostained, no differences in the distribution of these polarity markers could be detected (Fig. 4B, lower panel; supplementary material Fig. S5B). Comparison of the ground areas of the two cell lines revealed that 12 hours after plating, each MDCK_TTL–GFP cell was significantly smaller than an average MDCK_GFP cell (supplementary material Fig. S5A,C), thus suggesting that the cells become taller during island formation as indicated by x/z scans (Fig. 4A). TTL overexpression also correlated with a higher tendency to assemble into islands of prematurely polarized cells, as demonstrated by time-lapse analysis (Fig. 4C; supplementary material Movies 1, 2). MDCK_GFP cells formed an almost confluent cell layer 32 hours after seeding whereas MDCK_TTL–GFP cells seeded at the same density still exhibited extensive lacunae between cell islands. Another parameter that can be used to determine monolayer formation is the transepithelial resistance (TER). This value increased during monolayer formation of MDCK_GFP cells until it reached a plateau when the monolayer became tight (supplementary material Fig. S5D). TER values for MDCK_TTL–GFP cells increased slightly more slowly compared to controls, indicating a delay in monolayer formation. In summary, these data suggest that TTL overexpression in MDCK cells induces changes in the course of monolayer formation.

Next, we wished to study whether enhanced levels of TTL also affect the directed transport of cargo molecules in polarized epithelial cells. To this end, surface molecules of the apical or basolateral membrane of MDCK_GFP and MDCK_TTL–GFP cells were biotinylated and harvested with streptavidin–Sepharose beads. Fig. 5A depicts the pattern of proteins isolated from the two membrane domains in comparison to the corresponding cell lysates separated by SDS-PAGE analysis and silver stained. Prominent proteins from the apical and, to a lesser extent, from the basolateral surface of MDCK_GFP cells were dramatically depleted in MDCK_TTL–GFP cells. This can be explained either by a defect in forward trafficking to the cell surface, or an accelerated uptake into the cell in detyr-tubulin depleted cells. To elucidate this, we studied the kinetics of surface delivery of soluble gp80, the major secretory protein of MDCK cells (Wada et al., 1994). Secretion of this predominantly apically sorted glycoprotein into the apical or basolateral medium from MDCK_GFP and MDCK_TTL–GFP cells was monitored after different periods of incubation (Fig. 5B,C). Quantification of the data revealed a substantial decrease in the transport kinetics of apical gp80 secretion in TTL-overexpressing cells with a significance below 0.01 according to Student’s t-test. The effect on basolateral trafficking was less pronounced and membrane trafficking of the two basolateral proteins CD29 and E-cadherin was not significantly affected in TTL-overexpressing cells (supplementary material Fig. S6A,B). Confocal microscopy revealed strong intracellular immunostaining of gp80 in MDCK_TTL–GFP cells, whereas only a few tubular gp80-positive cisternae-like structures were detected in MDCK_GFP control cells (Fig. 5D), which were possibly endosomal organelles, as previously described (Apodaca et al., 1994). Together with the surface biotinylation data, this would suggest that predominantly apical transport pathways are affected in MDCK_TTL–GFP cells and that a reduced level of detyr-tubulin leads to a dramatic reduction in apical protein secretion. The opposite effect would be predicted upon TTL depletion, leading to an increase in levels of detyr-tubulin.

Therefore, a specific siRNA-mediated knockdown of TTL was performed with two TTL siRNAs in transiently transfected MDCK cells and luciferase siRNA transfection as a mock control. Following siRNA transfection, at 48 hours TTL was depleted to less than 35% compared to controls (supplementary material Fig. S7A,B). Fluorescence microscopic analysis of the polarity markers E-cadherin, claudin and ZO-1 revealed that MDCK cells form polarized cell monolayers after TTL depletion (supplementary material Fig. S7C,D). Concurrently, transport kinetics of gp80 to the apical medium of TTL-depleted cells increased by 25% within 4 hours (Fig. 6A,B). Basolateral delivery of CD29 or E-cadherin was not affected by TTL knockdown (supplementary material Fig. S6C–E). In combination with the results from TTL-overexpressing MDCK_TTL–GFP cells, these data indicate that changes in TTL expression substantially alter the apical transport kinetics of soluble gp80.

To further assess whether apical trafficking of membrane-associated glycoproteins is affected in TTL-depleted cells, we studied the delivery of two polypeptides that use distinct apical transport platforms and pathways. Lipid-raft-associated sucrase isomaltase (SI) and non-raft p75–GFP (Delacour et al., 2007) were biosynthetically labelled and immunoprecipitated from the two membrane domains of polarized MDCK_SI or MDCK_p75–GFP cells after TTL depletion or luciferase siRNA transfection. Fig. 6C indicates that the proportion of p75–GFP that had reached the apical plasma membrane was 30% higher in cells with reduced amounts of TTL than in control cells. Lipid-raft-dependent apical trafficking, however, was not affected by TTL depletion, as no significant changes in the quantities of apical SI could be observed (Fig. 6D). This was unexpected, but could be due to the involvement of actin microfilaments as a rate-limiting step in this transport pathway (Jacob et al., 2003). To clarify this possibility, we analysed surface delivery of endogenously expressed raft-associated gp135/podocalyxin in MDCK_TTL–GFP cells, which contain radically depleted levels of detyrosinated microtubules. As assessed by biotinylation, the amount of surface-resident gp135 was significantly reduced by TTL overexpression (Fig. 6E). A partial compensation for this effect by simultaneous knockdown of endogenous TTL and TTL_–GFP (Fig. 6F) suggests that, also for a raft-associated glycoprotein, apical trafficking can be modulated by lowering the level of overexpressed TTL. Moreover, in conjunction with the data observed for SI, apical delivery of gp135 was not significantly improved if TTL was knocked down in MDCK wild-type cells. Thus, we conclude that the kinetics of raft-mediated apical trafficking cannot be further optimized by TTL knockdown if the level of TTL is not artificially increased.

Finally, the influence of TTL knockdown on the tubulin composition of MDCK_SI or MDCK_p75–GFP cells was verified by immunoblots and immunofluorescence (Fig. 7). At first, antibodies directed against detyr-, tyr- or acetylated tubulin were used for immunoblots of control and TTL-depleted cells. Knockdown of TTL resulted in a substantial increase in detyr-tubulin, whereas the levels of tyr- or acetylated tubulin were not significantly altered (Fig. 7A–C). Four days after seeding, levels of detyr- and tyr-tubulin were monitored in control and TTL-depleted MDCK cells by immunofluorescence. As seen in Fig. 7D detyr-tubulin staining is much stronger than in cells expressing normal levels of the ligase (see also supplementary material Fig. S7E,F). Altogether, these data clearly demonstrate that TTL depletion raised the level of detyr-tubulin in microtubules in MDCK cells.

In summary, elevated detyr-tubulin increases the transport rate of soluble gp80 or non-raft-p75–GFP to the apical membrane. In contrast, apical transport of gp80 or lipid-raft-associated gp135 is disrupted by detyr-tubulin depletion, which can be rescued by TTL knockdown. This suggests that detyrosinated microtubules are involved in lipid-raft-dependent and -independent apical protein trafficking in MDCK cells.

The role of post-translational modifications in the generation and maintenance of polarized epithelial cells is only poorly understood.

In this study, we analysed differences in the quantities of tyr- and detyr-tubulin and monitored the consequences of an imbalance in these two tubulin variants on the cell architecture and the transport of glycoproteins to both membrane domains of MDCK cells. A polarized distribution of tyr- or detyr-tubulin was not found in these cells, as equal amounts of each population were detected at the apical and the basolateral cell poles in cell cysts or epithelial monolayers. In contrast, previous studies on the distribution of post-translationally modified tubulin variants in nerve cells have described a predominant location of tyr-tubulin in the distal region of the axon and in cell bodies; whereas highly detyrosinated domains of microtubules were found at their minus ends (Baas and Black, 1990; Brown et al., 1992). In MDCK cells, both the tyrosinated and detyrosinated isoforms were present along most of the length of double positive microtubules with alternating sections of the two isoforms. Segmental labelling of microtubules has been described previously (Geuens et al., 1986; Webster et al., 1987). Detyr-tubulin stretches can be generated by an individual turnover from tyr- to detyr-tubulin, which is facilitated on pre-existing microtubules (Arce and Barra, 1985). Long stretches of detyr-tubulin have been found in the Chlamydomonas flagellum along the outer wall of the B tubules (Johnson, 1998). The authors speculate that these post-translational modifications of the tubule surface may enhance dynein motor interactions.

As shown in this study and by Quinones et al., the number of detyrosinated microtubules increases as the cells become polarized (Quinones et al., 2011). In agreement with observations on neuronal cells (Bisig et al., 2009), these detyr-tubulin-enriched microtubules are frequently curled. Comparatively rapid accumulation of detyr-tubulin has been reported for myoblasts cultured in differentiation medium (Gundersen et al., 1984). This morphological differentiation was blocked by inhibition of microtubule detyrosination (Chang et al., 2002), which suggests a detyr-tubulin-dependent mechanism for morphological transformations. In agreement with our data, the development of extensive cell to cell interactions between MDCK or Vero cells was found to be accompanied by an increase in microtubule detyrosination (Bré et al., 1991) in confluent, subconfluent or individual cells within a time interval of 1 day, which is reflected by the initial rise in detyr-tubulin up to 2 days after seeding detected in this study.

Interestingly, our data suggest that TTL overexpression affects the ratio of soluble to polymerized tubulin in MDCK cells. Since it has been shown that TTL binds efficiently to dimeric tubulin (Arce et al., 1978) and forms an elongated complex with the tubulin dimer it may prevent its incorporation into microtubules (Szyk et al., 2011). This scenario seems to prevail in polarized MDCK_TTL–GFP cells, which maintain constantly high levels of the enzyme. However, converse ratios between assembled and soluble tubulin measured 3 days after seeding cannot be explained by the sequestration of tubulin dimers. These ratios are in accordance with the situation in human osteosarcoma cells, where TTL overexpression reduced the average growth rates of microtubules (Szyk et al., 2011). In addition, microtubule dynamics can be modulated by the presence or absence of cell-specific factors such as CLIP-170 (Peris et al., 2006), which indicates the complexity of this process. Do variations in the level of assembled tubulin within the detected range affect membrane trafficking in polarized MDCK_TTL–GFP cells? Since microtubules appear to be involved in the transport of proteins to both plasma membranes of MDCK cells (Grindstaff et al., 1998), and since the delivery of two basolateral cargo proteins was unaffected in MDCK_TTL–GFP cells, a general effect on trafficking does not seem plausible. Moreover, it is unlikely that the dramatic reduction in apical p75 transport or gp80 secretion results exclusively from the observed 20–30% reduction in assembled tubulin. However, we cannot exclude the participation of this decrease on apical trafficking in MDCK_TTL–GFP cells.

Another query is how TTL overexpression can result in the depletion of detyrosinated microtubules. TTL acts preferentially on soluble tubulin, and the detyrosination reaction preferentially takes place on microtubules (Arce et al., 1978; Arce and Barra, 1985; Deanin et al., 1980). One could argue that a shift in the equilibrium towards tyr-tubulin elevates the assembly of tyrosinated microtubules and pushes the balance towards a relative decrease in detyrosinated microtubules. Indeed, the opposite scenario, of low TTL protein expression being paralleled by increased detyr-tubulin is a common feature of several types of cancer cells (Kato et al., 2004; Mialhe et al., 2001; Soucek et al., 2006). Increasing the amount of TTL in HEK293T cells accordingly results in a decrease in detyr-tubulin (Kato et al., 2004). It seems that the story is not simple, as we did not detect elevated levels of tyr-tubulin in MDCK_TTL–GFP cells, possibly suggesting that changes in the equilibrium are being compensated by additional cellular mechanisms. Alternatively, it might well be that alterations in the amount of soluble tyr-tubulin are covered by the considerably large pool of assembled tyr-tubulin (Gundersen et al., 1984).

Nevertheless, variations in the equilibrium of the two tubulin modifications have dramatic effects on cellular morphology and function. An increase in tubulin detyrosination during epithelial-to-mesenchymal transition (EMT) in immortalized human mammary epithelial cells promotes morphological alterations, which results in enhanced reattachment to endothelial cells (Whipple et al., 2010). However, excess tubulin detyrosination in mammalian TTL-null fibroblasts results in defects in both spindle positioning during mitosis and cell polarization during interphase (Peris et al., 2006). In conclusion, a sudden increase in the level of detyr-tubulin to a specific degree is required for morphological changes during cellular differentiation in a variety of cell types. Reciprocally, artificial induction of tubulin-tyrosination by TTL overexpression perturbs the differentiation into epithelial monolayers. MDCK_TTL–GFP cells are less mobile, assemble into cell islands and acquire a prematurely polarized morphology. The lack of mobility could be explained by a depletion of detyr-tubulin-enriched stable microtubules in MDCK_TTL–GFP cells, which are needed for the developing asymmetry of a cell as it begins to migrate (Gundersen and Bulinski, 1988). It is thus likely, that these immobile cells settle down, form cell islands and develop cell-to-cell contacts with neighbouring partners in their vicinity.

In polarized three-dimensional monolayers of MDCK cells, the level of detyr-tubulin decreases from day 2 to day 5, whereas the concentration of tyr-tubulin remains constantly high and even increases slightly in this time interval. A 100% increase in TTL levels in the same period can be regarded as a reason for this observation, since knockdown of TTL reverses the effect and results in high levels of detyr-tubulin. Similarly, EMT induction in breast cancer cells is characterized by a decrease in the synthesis of TTL and an increase in detyr-tubulin (Whipple et al., 2010). This indicates that constant amounts of TTL in epithelial cell layers maintain the concentration of detyr-tubulin at a basic level. In addition, expression of TTL plays a vital role in mice and is essentially required for the polarization of neuronal cells, since TTL-null neurons display morphogenetic anomalies (Erck et al., 2005).

The question arises about the function of a remaining pool of detyr-tubulin-positive microtubules in fully polarized epithelial cells. Evidence for detyrosinated microtubules as cytoskeletal tracks for intracellular trafficking came from CHO cells. In these cells, stable detyrosinated microtubules play a role in the organization of endocytic recycling compartments and facilitate vesicular transport to the cell surface (Lin et al., 2002). A collapse of the recycling compartment was observed in anti-kinesin antibody-injected cells, which suggests that kinesin motors are involved in endosome export. Particularly the kinesin-1 isoform KIF5b is essential for apical delivery of membrane proteins from the trans-Golgi network to the cell surface (Jaulin et al., 2007). We have previously shown that the kinesin motor Kif5c is also involved in apical trafficking events following trans-Golgi network exit (Astanina and Jacob, 2010). This motor preferentially binds to and traffics along detyrosinated microtubules in live cells (Dunn et al., 2008). In this present report we demonstrate that apical trafficking of soluble, raft-associated or non-associated cargo proteins can be modulated by TTL expression and alterations in intracellular detyr-tubulin, which suggests that stable detyrosinated microtubules serve as tracks for kinesin-driven post Golgi transport to the apical membrane domain of MDCK cells, especially considering that kinesin motors do not move preferentially along dynamic microtubules (Cai et al., 2009). Recent data on the function and stability of sensory cilia in C. elegans support the idea of post-translationally modified microtubules that guide molecular motors to specific subcellular destinations (O’Hagan et al., 2011). The authors propose that these tubulin modifications regulate the velocity of kinesin-2 motors. This ‘tubulin code’ would thus provide signposts that regulate the localization or activity of motors (Janke and Kneussel, 2010), which even contributes to the establishment and maintenance of neuronal polarity in mammalian cells (Konishi and Setou, 2009). Autoinhibition is an alternative instrument for the regulation of kinesin motors involved in intracellular trafficking (Hammond et al., 2010). However, apart from a general subapical concentration of microtubules, we did not observe that detyrosinated or tyrosinated microtubules were specifically focused to one of the two cell poles. A predominantly subapical accumulation could be detected for acetylated tubulin as previously shown (Jaulin and Kreitzer, 2010). Therefore, we propose that additional factors assist in intracellular guidance to the specific membrane domains of these cells.

Further evidence for the role of post-translational modifications of microtubules in intracellular transport comes from cultured neuronal cells, in which acetylation promotes kinesin-1 binding and transport of cargo to a specific subset of neurites (Reed et al., 2006). Yet another kinesin motor, kinesin-3 UncA, from Aspergillus nidulans, preferentially binds to detyrosinated microtubules and is required for fast hyphal extension (Zekert and Fischer, 2009). Kinesin-binding to detyrosinated microtubules is not merely involved in intracellular transport events because it can, in addition, cross-bridge these microtubules with specific intermediate filaments (Liao and Gundersen, 1998). Cytoskeletal tracks can further be characterized by the alignment with small GTP-binding septins. Spiliotis et al. have shown that septin-2 coupling of polyglutamylated microtubules to post-Golgi vesicle transport is required for the morphogenesis of polarized MDCK cells (Spiliotis et al., 2008).

Irrespective of the fundamental insight in epithelial biology, a greater understanding of the role of alterations in tubulin modification in epithelial cells is of wider interest because pathomechanisms that occur in EMT change the balance of tubulin tyrosination and thus promote cell metastasis. We show here, for the first time, that changes in the pattern of post-translational tubulin modifications affect monolayer formation of, and membrane trafficking in MDCK cells. Further studies on all components involved and their interactions should provide an integrated picture of the function of tubulin modifications in the establishment and maintenance of epithelial integrity.

MDCK cells were cultured at 37°C under 5% CO₂ in minimum essential medium (PAA) containing 5% foetal calf serum or in Dulbecco’s modified Eagle’s medium (4.5 g/l glucose) containing 10% FCS supplemented with antibiotics and glutamine. For generation of a cell line stably expressing TTL–GFP the transfected cells were selected in medium containing neomycin for at least 90% fluorescence. For the generation of polarized monolayers of MDCK or MDCK_TTL–GFP cells ∼3×10⁷ cells per 10 cm culture dish or equivalent quantities per six-well filter were seeded. Cells were counted in a Neubauer chamber. Three-dimensional MDCK cell cysts were obtained by using BD Matrigel™ Basement Membrane Matrix Phenol-Red-Free (BD Biosciences). In short, cells were diluted to ∼2×10⁴ cells/ml PBS and mixed with the same volume of Matrigel in a cold environment. 30 µl of this mixture were spread out on cold 12 mm coverslips lying in cell culture dishes, covered with medium after gelling and incubated for 7–10 days at 37°C, in 5% CO₂ with regular medium change. Plasmid and small interfering RNA (siRNA) transfection of MDCK cells was performed with Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions. Biosynthetic labelling of MDCK_p75 or MDCK_SI cells, accumulation of newly synthesized material in the trans-Golgi network at 20°C and surface immunoprecipitation were performed essentially as described before (Cramm-Behrens et al., 2008). Images generated on a phosphoimager were analysed with the QUANTITY ONE software (Bio-Rad).

For construction of a TTL-containing plasmid the total RNA was isolated from MDCK cells using the RNeasy Mini Kit (Qiagen) followed by generation of TTL cDNA according to the manufacturer’s specifications (Fermentas RevertAid™ M-MuLV reverse transcriptase) using a reverse TTL primer (5′-ggatcccccagcttgatgaacgcagc-3′). The cDNA was used as template in a PCR (forward TTL primer: 5′-gaattcatgtacaccttcgtggtg-3′ and reverse TTL primer) and the PCR product was then cloned into pEGFP-N1 vector using the EcoRI and BamHI restriction sites.

For RNA-mediated interference (RNAi) experiments, specific depletion of TTL was performed with the following two duplexes – (1) 5′-UGUUUGGAUUGCAAAGUCAUU-3′/3′-UUACAAACCUAACGUUUCAGU-5′, (2) 5′-CAUUACCUUGGAAAGUAGUUU-3′/3′-UUGUAAUGGAACCUUUCAUCA-5′.

As a control, luciferase siRNA was used (5′-CGUACGCGGAAUACUUCGATT-3′/5′-UCGAAGUAUUCCGCGUAC+GTT-3′).

The following tubulin antibodies were used: monoclonal anti-α-tubulin (Clone DM 1A) and anti-acetylated α-tubulin (Sigma-Aldrich), monoclonal anti-tyrosinated α-tubulin (YL1/2, Santa Cruz), polyglutamylated tubulin (GT335, AdipoGen) polyclonal anti-detyrosinated α-tubulin (Millipore) and monoclonal anti-detyrosinated α-tubulin antibody (Synaptic Systems). The polyclonal anti-detyrosinated tubulin antibody (Millipore) not recognizing the polyglutamylated form of tubulin was used for immunoblots and GSDIM imaging. The following polyclonal antibodies were used: anti-TTL (Proteintech Group), anti-Claudin-1 (Invitrogen), anti-clusterin-α (gp80; Santa Cruz), anti-β-catenin (Sigma), anti-Kif5B (Abcam) and anti-ZO1 (Zymed). The monoclonal antibodies anti-E-cadherin, anti-CD29 and anti-GAPDH (6C5) were purchased from BD Biosciences (E-cadherin, CD29) and HyTest Ltd (GAPDH). The monoclonal anti-SI, monoclonal anti-p75 (ME20.4) and anti-gp135 antibodies were generously provided by H. P. Hauri (Biozentrum Basel, Switzerland), A. Le Bivic (Faculté des Sciences de Luminy, Marseille, France) and G. Ojakian (State University of New York Health Science Center, New York, USA).

For preparation of cell lysates the cells were collected in lysis buffer (25 mM Tris-HCl, 50 mM NaCl, 0.5% sodium desoxycholate, 0.5% Triton X-100, pH 8.0) after the indicated times, incubated at 4°C on a rotating platform, centrifuged for 5 minutes at 12,000 g and the supernatants were separated by SDS-PAGE using the Hoefer-Mini-VE system (Amersham Pharmacia Biotech) and transferred to nitrocellulose membranes. The membranes were blocked in 5% skimmed milk powder in PBS for 1 hours and incubated overnight at 4°C with specific antibodies. Detection was performed with horseradish-peroxidase-conjugated secondary antibody visualized by ECL (Thermo Fischer Scientific) and an Intas gel imager CCD camera. The results were quantified using LabImage 1D software.

Microtubule assembly was studied essentially as previously published (Ligon et al., 2003). Cells were harvested 1 (non polar), 3 or 5 days (polar) after seeding, centrifuged and resuspended in pre-warmed medium (37°C) containing 30 mM HEPES, pH 6.1. The cells were washed once in pre-warmed medium, resuspended and left to recover at 37°C for 15 minutes. After two additional washing steps in pre-warmed PBS, PHEM/2 M glycerol/2% Triton X-100 (50 mM PIPES, 50 mM HEPES, 1 mM EDTA, 2 mM MgCl₂, pH 6.9) was added, mixed and incubated for 10 minutes at 37°C. Following a centrifugation step (2 minutes, 8000 g) the soluble tubulin fraction was separated from the polymerized tubulin pellet. The polymer to dimer ratio was calculated using the LabImage 1D software. Tubulin signals were normalized to GAPDH from identical blots. Values of polymerized tubulin (P) were divided by total tubulin (sum of P and S) for ratio presentations.

For surface biotinylation cells were incubated with the crosslinker Sulfo-NHS-Biotin (Thermo Scientific) to biotinylate surface proteins. After several washing steps with PBS plus 1 mM CaCl₂ and 1 mM MgCl₂ (PBS++)/0.1 M glycin and PBS++ alone, the cells were frozen at −20°C and lysed in 20 mM Tris, 150 mM NaCl, 5 mM EDTA, 1% Triton X-100. Together with BSA the cell lysates were incubated with neutravidin beads (Thermo Scientific) to select the biotinylated surface proteins. After renewed washing steps the proteins were eluted with SDS/dithiothreitol-containing buffer, separated by SDS-PAGE and stained using a silver staining procedure. The secretion of gp80 was analysed by replacing the apical and basolateral medium of the relevant cells, then after the indicated time intervals removing and concentrating this medium. The gp80 content was tested by immunoblotting. For quantification, cell lysates were separated by SDS-PAGE and silver stained. The samples were normalized to prominent lysate bands.

For immunofluorescence analysis of tubulin modifications, Claudin-1, E-cadherin and ZO-1, the cells were fixed with ice-cold methanol for 5 minutes and blocked in 1% milk powder/PBS++ for 1 hour. For TTL analysis the fixation was performed in 4% paraformaldehyde for 10–15 minutes followed by a permeabilization step in acetone (5 minutes) and 1 hour blocking in 5% FCS/PBS++. The incubation with primary antibody was performed in blocking reagents/PBS++ for 2 hours and with secondary antibody labelled with the indicated AlexaFluor dye in PBS++ for 1 hour. Nuclei were stained with Hoechst 33342. Three-dimensional cysts were stained in a similar manner but with longer incubation times: methanol fixation 20 minutes, blocking 2 hours, primary antibody 24 hours, secondary antibody 14 hours, Hoechst 33342 30 minutes, washing steps 20 minutes each.

MDCK_GFP or MDCK_TTL–GFP cells were seeded on glass-bottomed dishes (WillCo Dish®, Willco Wells) and incubated until ∼40–50% confluency (2 hours). After replacement of medium with 100 µM HEPES-containing medium the cells were used for microscopic time-lapse experiments.

Confocal images were acquired on a Leica TCS SP2 AOBS microscope using a 40× oil planapochromat objective (Leica Microsystems). The live-cell time-lapse experiments were recorded on a Leica DMI6000 B using a 20× dry fluotar objective. The total internal reflection fluorescence microscopy (TIRFM) was performed as described before (Schneider et al., 2010) using a Leica DMI6000 B microscope with an HCX PL APO 100× objective. The high resolution images were acquired using a Leica SR GSD super-resolution system (Leica) according to the manufacturer’s specifications. Data evaluation was performed with the Leica software in combination with the Volocity imaging software package (Improvision).

For the determination of the transepithelial resistance, MDCK_GFP and MDCK_TTL–GFP cells were seeded onto filters, incubated at 37°C, and starting 8 hours after seeding, the medium was changed and TER measurement was performed every 24 hours using the Millicell ERS-2 Voltohmmeter (Millipore) in duplicate.

We are grateful to M. Dienst and W. Ackermann for technical assistance. We thank P. Holz and L.A. Ligon for advice and for providing the tubulin ratio protocol, J.M. Przyborski for careful reading of the manuscript and B. Lamp for help with the Millicell ERS-2 Voltohmmeter (Millipore).