Identification of Tubulin Deglutamylase among Caenorhabditis elegans and Mammalian Cytosolic Carboxypeptidases (CCPs)

Tubulin polyglutamylation is a reversible post-translational modification, serving important roles in microtubule (MT)-re-lated processes. Polyglutamylases of the tubulin tyrosine ligase-like (TTLL) family add glutamate moieties to specific tubulin glutamate residues, whereas as yet unknown deglutamylases shorten polyglutamate chains. First we investigated regulatory machinery of tubulin glutamylation in MT-based sensory cilia of the roundworm Caenorhabditis elegans . We found that ciliary MTs were polyglutamylated by a process requiring ttll-4 . Conversely, loss of ccpp-6 gene function, which encodes one of two cytosolic carboxypeptidases (CCPs), resulted in elevated levels of ciliary MT polyglutamylation. Consistent with a deglutamylase function for ccpp-6 , overexpression of this gene in

The microtubule (MT) 2 cytoskeleton plays critical roles in multiple cellular processes such as chromosome segregation, intracellular transport, cell morphogenesis, and polarity. Tubulin, which is the major protein subunit of MT fibers, exhibits polymorphic protein variation and undergoes a variety of unique post-translational modifications at its C-terminal tail, such as detyrosination/tyrosination (1,2), polyglycylation (3), and polyglutamylation (4). These modifications regulate tubulin and MT function. Polyglutamylation enzymes add multiple glutamate moieties to specific glutamate residues within substrate proteins. Physiological roles of tubulin polyglutamylation for MT-related processes were reported recently (5,6) and reviewed by Ikegami and Setou (7).
Polyglutamylase enzymes have recently been identified as belonging to the tubulin tyrosine ligase-like (TTLL) protein family (8 -10) (supplemental Fig. 1). In contrast, the enzyme(s) underlying deglutamylation have not been discovered yet, although such proteins are suggested to be present in mouse brain neurons and in cultured cells (11,12).
In this study, we employed the genetically tractable nematode, Caenorhabditis elegans, to identify genes that regulate tubulin polyglutamylation/deglutamylation. Specifically, we investigated tubulin post-translational modification in C. elegans sensory cilia, which are MT-based organelles that extend from the distal dendrite tips of 60 sensory neurons found in sensory organs named amphid (head) and phasmid (tail) sensilla (see Fig. 1A). In this system, we identified a cytosolic carboxypeptidase (ccpp) gene, ccpp-6, as a candidate tubulin deglutamylase gene, whose functional properties are opposite to those of the tubulin polyglutamylase gene, ttll-4. Furthermore, we identified mammalian CCP5 as a functional homolog of nematode CCPP-6 and confirmed its deglutamylase activity in cultured cells and in vitro. To our knowledge, this is the first report revealing the molecular nature of the tubulin deglutamylase enzyme.

EXPERIMENTAL PROCEDURES
Worm Culture and Genetics-Nematodes were grown and maintained at 20°C under standard conditions (13). All mutant and transgenic strains used in this study are listed in the supplemental Methods.
Immunohistochemical Analyses-Immunohistochemical analyses were performed as described in previous studies (14) with a slight modification. Full details are described in the supplemental Methods.
C. elegans Microscopy and Image Analysis-Immunostained worm samples were mounted on a slide glass with VECTASH-IELD (Vector Laboratories) and analyzed under an FV1000 confocal microscope (Olympus). At least 20 adult worms for each strain were analyzed, and all images were collected using the same optical condition. Images were analyzed using FV10-ASW software (Olympus). Immunostained polyglutamylated (GT335) and ␣-tubulin signal intensities were measured in sensory cilia from the same optical slice. Following subtraction of background signals in non-ciliary regions, a GT335/␣-tubulin signal intensity ratio was calculated for each image. GT335/␣tubulin ratios were averaged for each strain. The relative intensity of each strain was obtained from the average of GT335/ ␣-tubulin normalized by that of wild type and statistically analyzed by StatPlus (AnalystSoft).
Cell Culture and RNAi-The conventional methods of cell culture and gene knockdown by RNAi are described in the supplemental Methods.

RESULTS AND DISCUSSION
C. elegans Sensory Ciliary MTs Are Strongly Polyglutamylated, Requiring ttll-4-We determined the distribution of polyglutamylation signals in C. elegans sensory cilia using antibodies that specifically recognize polyglutamylation. Using the monoclonal antibody (mAb) GT335, which specifically recognizes both monoglutamylation and polyglutamylation (10), robust signals were detected in the ciliary regions of the nematode head and tail (Fig. 1, A and B, left; supplemental Fig. 2A; and supplemental Movie 1). To ensure that the observed signals are indeed those of ciliary MTs, we employed strains expressing ciliated cell-specific transcriptional GFP transgenes (derived from promoter-GFP constructs), which illuminate the entire sensory neuron. In such strains, we found that the cilia of most amphid head neurons (ASE, ADF, ASH, ASK, AWA, AWC, and AFD) showed GT335 signals (Fig. 1B middle and right; supplemental Fig. 3). Strong signals were also seen in the ciliary structures of the inner (Fig. 1B, left panel, white arrows) and outer labial sensilla (Fig. 1B, left panel, yellow arrows; supplemental Movie 1). In contrast, no GT335 signals were detected in other ciliary neurons such as the neurons of the cephalic sensilla (supplemental Fig. 3, bottom row, arrow). In the tail, the ciliary structures of the phasmid neurons PHA, PHB, and PQR were also labeled with GT335 (supplemental Fig. 2B). Another mAb, B3, which specifically recognizes polyglutamylated tubulins (15), yielded similar signal staining patterns (supplemental Fig. 2C). Together, these data show that the MTs of C. elegans sensory cilia are highly polyglutamylated.
Tubulin polyglutamylation is mediated by enzymes of the TTLL family (8 -10). The C. elegans genome contains six TTLL genes (ttll-4, -5, -9, -11, -12, and -15), and each gene belongs to a different clade (supplemental Fig. 1). Next, we studied the expression pattern of C. elegans ttll genes using transgenic worms carrying transcriptional GFP reporters (GFP directed by ttll promoter) or translational GFP reporters (GFP was fused to the entire open reading frame of the gene) under the control of the endogenous promoter of the gene (supplemental Figs. 4 and 5; summarized in supplemental Table 1). We found that three ttll genes, ttll-4, -9, and -11, are expressed in many ciliated sensory neurons in amphid (supplemental Fig. 5, D and F-H). C. elegans does not have any polyglycylase orthologs (TTLL3, -8, and -10) (16 -18), which are found in other animals having motile cilia (18,19), where polyglycylation exclusively occurs. Consistent with an absence of motile cilia and polyglycylase orthologs in C. elegans, immunostaining for polyglycylation in worms revealed no signals. 3 To identify which ttll gene contributes to the polyglutamylation observed in the sensory cilia, we studied worm strains with mutations in the ttll genes (summarized in supplemental Table 1). The GT335 signals completely disappeared in ttll-4 worms (Fig. 1C, second row, asterisks, and D).
To clearly evaluate the level of tubulin polyglutamylation, we quantified the signals of GT335 and of counterstained ␣-tubulin (Fig. 1C, middle column). The complete loss of GT335 signal was greater than a partial decrease of the ␣-tubulin signal in ttll-4 mutants (Fig. 1D). In contrast, only a slight reduction in GT335 signals was observed ( Fig. 1, C, third row, and D, p Ͻ 0.01, Student's t test). We observed no overt decrease of GT335 immunoreactivities in a hypomorphic ttll-11(gk482) mutant (data not shown). The GT335 immunoreactivities were not altered in ttll-5(tm3360) and ttll-15(tm3871) mutants (data not shown), presumably because these genes are not expressed in ciliated sensory neurons (supplemental Fig. 5, E and J). The prominent loss of polyglutamylation observed in the ttll-4 mutant was reversed by overexpression of ttll-4::mCherry in sensory neurons ( Fig. 1, C bottom row, and D). These findings demonstrate that TTLL-4 is a key enzyme that autonomously mediates tubulin polyglutamylation in sensory cilia.
CCPP-6 Is a Tubulin Deglutamylase in C. elegans-Previous reports suggest the presence of dynamic reversible regulation of tubulin polyglutamylation by deglutamylase activity (11,12). Having identified TTLL-4 as a critical protein for tubulin polyglutamylation in C. elegans sensory cilia, we next turned our attention to identifying candidate deglutamylase genes in C. elegans. Cytosolic carboxypeptidases (CCPs) were recently identified as a subfamily of M14 metallocarboxypeptidases (20,21). CCP1/Nna1, the first identified member of the CCP family, is mutated in Purkinje cell degeneration (pcd) mice (22); in these mice, the mitral cells in olfactory bulb show decreased signals of detyrosinated tubulins (20). C. elegans has only two CCP genes in its genome, F56H1.5 and EEED8.6, which we name here ccpp-1 and -6, respectively by their amino acid similarity and protein structure in comparison with six mammalian CCP genes (CCP1-6) ( Fig. 2A; supplemental Fig. 6). CCPP-1 belongs to the 3 Y. Kimura and M. Setou, unpublished data. JULY 23, 2010 • VOLUME 285 • NUMBER 30
To determine whether ccpp-1 and -6 are involved in the regulation of tubulin polyglutamylation, we first examined their expression in sensory neurons. A transcriptional GFP reporter of ccpp-1 was expressed in amphid neurons (supplemental Fig. 7C). On the other hand, that of ccpp-6 was expressed only in the labial neurons but not in amphid neurons (data not shown).
A translational reporter mCherry fused to CCPP-1 protein showed the weak dotted signals around the pharynx (supplemental Fig. 7D, arrowheads). Meanwhile, a translational GFP reporter for ccpp-6 showed weak signals in the cell body of putative amphid neurons (supplemental Fig. 7, A and   B, arrowheads), suggesting that expression in amphid cells was regulated by an enhancer element in a ccpp-6 intron(s). The weak signals observed for GFP/mCherry-tagged CCPP-1 and CCPP-6 transgenes indicate that both CCPP proteins are maintained at very low levels, possibly because they are rapidly degraded. Nevertheless, these results suggest that both CCPP-1 and CCPP-6 are expressed and function cell autonomously in amphid neurons.

Identification of Tubulin Deglutamylase
the amphids of these animals showed increased GT335 signals along with the increase in the tubulin signals (Fig. 2, C, lower  row, and D, lower panel). Hence, CCPP-1 might be involved in regulating tubulin function in the sensory cilia but does not likely act as a deglutamylase.
Mammalian CCP5 Functions as a Tubulin Deglutamylase-We next examined whether mammalian CCPs ( Fig. 2A) harbor tubulin deglutamylase(s). To this end, we exogenously overexpressed Myctagged CCP5 and -6, which are orthologs of C. elegans CCPP-6, and CCP2 as a control along with tubulin polyglutamylases, TTLL5 and -6, in HEK293T cells. Remarkably, GT335-detected tubulin bands completely disappeared when CCP5 was expressed (Fig. 3A). We could not detect such an effect for CCP2 and -6 at least in our system (Fig. 3A, data not shown). Strikingly, CCP5 did show no effect on the level of tyrosinated tubulin (Fig. 3B), indicating that CCP5 does not have detyrosinase activity. This deglutamylation activity was also detected for other proteins (supplemental Fig. 8, arrowheads). We further analyzed the effect of CCP5 in the endogenous polyglutamylation by the overexpression in mouse cortical neurons. The CCP5-expressing cells showed marked reduction in the GT335 signals (Fig.  3C, middle column).
To further examine whether mammalian CCP5 effectively counteracts tubulin polyglutamylation, we performed RNAi treatment of CCP5 in NIH3T3 cells. Two independent small interfering RNAs against CCP5 clearly increased intensities of GT335-detected bands at 50 kDa (Fig. 3D), which was overlapped with the ␤-tubulin band ( Fig.  3D; green versus red). The two-dimensional electrophoresis revealed that the CCP5 knockdown caused an increase in the intensity of GT335-detected ␤-tubulin spots (Fig. 3E, top row), and more acidic ␤-tubulin spots were detected (Fig. 3E, bottom row). These results indicate that CCP5, one of the sequence orthologs of nematode CCPP-6 ( Fig. 2A), is a tubulin deglutamylase in mammalian cells. The reason why ␤-tubulin polyglutamylation is effectively increased in CCP5 knockdown cells could be explained by the presence of endogenous glutamylation activity preferential to ␤-tubulin in non-neuronal cells (12).
We next carried out an in vitro enzyme assay of recombinant CCP5. We first tried to use nematode CCPP-6 and murine CCP5 expressed in bacteria. However, these recombinants did not show the enzyme activity or were hardly solubilized (see the supplemental Methods). This is probably because CCPP-6 and CCP5 fails to fold into the correct three-dimensional  . Murine CCP5 has tubulin deglutamylase activity. A, murine CCP2 or -5 was expressed in HEK293T cells along with TTLL5 and -6. PolyE, polyglutamylation signal recognized by GT335; ␣-Tub, ␣-tubulin; EGFP, enhanced GFP; GAPDH, glyceraldehyde-3-phosphate dehydrogenase. B, murine CCP5 alone was expressed in HEK293T cells. C, cortical neurons (6 days in vitro) in which a vector for GFP together with the expression plasmid for CCP5 or its control vector was introduced were subjected to the immunocytochemistry by using GT335 and ␤-tubulin antibody. D, two independent stealth small interfering RNAs were introduced into NIH-3T3 cells. CCP5 knockdown samples were subjected to the long range SDS-PAGE and reverse transcription-PCR. The arrowhead points to the newly emerging band, which is overlapped with the ␤-tubulin band. E, two-dimensional electrophoresis of CCP5 knockdown cell lysates. The numbers of predicted glutamates added to ␤-tubulin represent are numerically indicated. F, in vitro enzyme assay was performed by using pig tubulin as a substrate and CCP5-expressing HEK293T lysate as the enzyme source. Reaction mixture was analyzed by Western blotting with GT335. G, comparison between CCP5 and -6 enzyme activity in vitro. In vitro enzyme assay was performed in the same optimized condition as in F. CCP5-expressing lysate showed significant reduction (*, p Ͻ 0.05), whereas CCP6-expressing lysate did not show activity in this condition. The reaction with a mixture of both lysates did not show any enhancement. Bars indicate mean Ϯ S.E. H, CCP5-expressing and control HEK293T cell lysates were immunoprecipitated with anti-Myc polyclonal antibody. The immunoprecipitants were electrophoresed using 5-20% gradient gel and stained with SYPRO Ruby protein stain. Protein bands indicated by an asterisk were the putative degradation products of Myc-CCP5 (supplemental Fig. 10). The arrowheads indicate subunits of immunoglobulin. I, after immunoprecipitation, immunoprecipitants were analyzed by in vitro enzyme assay using pig tubulin as a substrate. Measurement results were indicated in the bottom chart. The lanes with (Ϫ) in E, G, and I indicate the substrate pig tubulin without lysate treatment. structure when expressed in bacteria. Thus, we decided to use Myc-tagged CCP5 expressed in HEK293T cells and successfully detected deglutamylase activity in vitro (Fig. 3, F and G; supplemental Fig. 9). The salt concentration drastically affected the CCP5 activity in the cell lysates; the enzyme was more active with 100 mM NaCl than with 0 or 50 mM NaCl (supplemental Fig. 9). CCP5 activity appeared to be saturated at 15 min as the decrease in GT335 signals of tubulin reached plateau (Fig. 3F). The faint decrease of GT335 signals in the mock transfection sample could be caused by the presence of endogenous deglutamylase in the cell lysates as reported previously (11,12). The increase of ␣-tubulin signals in CCP5-containing samples could occur by the preferential binding of anti-␣-tubulin antibody to less glutamylated tubulins (6,23).
Despite the highest similarity between nematode CCPP-6 and mammalian CCP6, we were not able to detect deglutamylase activity of CCP6 in cultured cells. We were not able to detect the reduction of GT335 signal in the in vitro assay either (Fig. 3G). We would not, however, exclude possibilities that CCP6 functions as a deglutamylase. CCP6 might function under other experimental conditions or act on specific substrates, e.g. axonemal tubulins in cilia, as CCPP-6 works in the sensory cilia of C. elegans.
Finally, we challenged the enzyme activity assay by using Myc-CCP5 immobilized on antibody-protein G beads. In the immunoprecipitant of anti-Myc antibody, the antibody-purified CCP5 was detected in a major band (Fig. 3H). The broad bands with the faster migration than Myc-CCP5 (Fig.  3H, asterisk) were likely to be degraded Myc-CCP5 as the anti-Myc antibody detected some proteins at the same position (supplemental Fig. 10). Other minor bands seemed to be nonspecifically precipitated proteins because they were also detected in the bead fraction of the empty vector-transfected cell lysates. Thus, Myc-CCP5 was the major and specific component of the immunoprecipitant. Incubation of substrate pig brain tubulin with this immobilized Myc-CCP5 showed clear reduction in the GT335 signal (Fig. 3I). The background activity detected in cell lysates (Fig. 3F) was completely eliminated in this assay (Fig. 3I). Taken together, these results clearly demonstrate that CCP5 has a tubulin deglutamylase activity.
Conclusion-By the genetic approach with C. elegans, we isolated CCPP-6 as a negative regulator of tubulin polyglutamylation in vivo. We also demonstrated that CCP5, a mammalian sequence ortholog of CCPP-6, had clear enzyme activity of tubulin deglutamylase by in vitro assay. In conclusion, we have identified C. elegans CCPP-6 and mammalian CCP5 as a tubulin deglutamylase by both in vitro and in vivo assay results.