Spatiotemporal manipulation of ciliary glutamylation reveals its roles in intraciliary trafficking and Hedgehog signaling

Tubulin post-translational modifications (PTMs) occur spatiotemporally throughout cells and are suggested to be involved in a wide range of cellular activities. However, the complexity and dynamic distribution of tubulin PTMs within cells have hindered the understanding of their physiological roles in specific subcellular compartments. Here, we develop a method to rapidly deplete tubulin glutamylation inside the primary cilia, a microtubule-based sensory organelle protruding on the cell surface, by targeting an engineered deglutamylase to the cilia in minutes. This rapid deglutamylation quickly leads to altered ciliary functions such as kinesin-2-mediated anterograde intraflagellar transport and Hedgehog signaling, along with no apparent crosstalk to other PTMs such as acetylation and detyrosination. Our study offers a feasible approach to spatiotemporally manipulate tubulin PTMs in living cells. Future expansion of the repertoire of actuators that regulate PTMs may facilitate a comprehensive understanding of how diverse tubulin PTMs encode ciliary as well as cellular functions.


Introduction 1
The primary cilium is a microtubule-based sensory organelle protruding from the apical 2 surface of resting cells; it is crucial in phototransduction, olfaction, hearing, embryonic 3 development, and several cellular signaling pathways, such as Hedgehog (Hh) signaling 1,2 .

4
Defects in primary cilia lead to a number of human diseases 3 . Structurally, the cilium is 5 composed of nine microtubule doublets called the axoneme, which offer mechanical support 6 to the cilium, and also provide tracks for motor protein-dependent trafficking, known as 7 intraflagellar transport (IFT) 4 . Polyglutamylation generates glutamate chains of varying lengths 8 at the C-terminal tails of axonemal tubulin 5,6 . This post-translational modification (PTM) occurs 9 on the surface of microtubules and provides interacting sites for cellular components, such as 10 microtubule-associated proteins (MAPs) and molecular motors 6 . However, the detailed 11 mechanisms of how axonemal polyglutamylation regulates the stability and functionality of 12 cilia remain to be understood.

7
The effects of tubulin polyglutamylation on the structure and functions of microtubules 8 have been studied mainly through the following approaches: 1) biochemical characterization 9 of glutamylated microtubules, 2) cell biology assays for hyper-or hypo-glutamylation induced 10 by genetically controlling the expression level of corresponding PTM enzymes, and 3) cell 11 biological analysis of genetically mutated tubulins. As a result, it has been shown that chemical 12 conjugation of glutamate side chains on purified microtubules increases the processivity and 13 velocity of Kinesin-2 motors 15 . Tubulin hyperglutamylation leads to microtubule disassembly 14 owing to the binding of a severing enzyme, namely spastin, to hyperglutamylated 15 microtubules 16,17 . Mice lacking a subunit of the polyglutamylase complex display 16 hypoglutamylation in neuronal cells, which is accompanied by a decreased binding affinity of kinesin-3 motors to microtubules 18 . Moreover, the genetic or mopholino-mediated perturbation 1 of polyglutamylases or deglutamylases across different model organisms results in 2 morphological and/or functional defects in cilia and flagella [19][20][21][22][23][24][25][26][27][28][29][30][31][32][33] . Collectively, these studies 3 strongly suggest the importance of tubulin polyglutamylation in the structural integrity and 4 functionality of microtubules in cilia as well as other subcellular compartments. However, these 5 approaches also revealed technical limitations. First, the distribution pattern of 6 polyglutamylated tubulin is spatiotemporally dynamic; i.e., polyglutamylation is abundant in 7 axoneme, centrioles, and neuronal axons in quiescent cells, which converges to the mitotic 8 spindle and midbody during cytokinesis 6 . This dynamic feature cannot be directly addressed 9 by conventional genetic manipulations or pharmacological inhibitors. Second, constitutive 10 genetic perturbation often allows for compensation where cells adapt to their new genetic 11 environment, likely leading to a missed detection of immediate consequences of loss-of-12 function such as an effect on transient interactions between tubulins with specific PTMs and 13 their molecular partners 34 . Third, besides tubulin, many nucleocytoplasmic shuttling proteins 14 such as nucleosome assembly proteins also have been identified as substrates of 15 glutamylases 10,35 . Therefore, global manipulation of genes encoding enzymes that modulate 16 glutamylation is insufficient to specifically perturb axonemal glutamylation.

6
To circumvent these limitations, we developed a method named STRIP for 1 SpatioTemporal Rewriting of Intraciliary PTMs. The method is based on chemically inducible 2 dimerization (CID) to spatiotemporally recruit a catalytic domain of deglutamylase onto ciliary 3 axoneme, where the axonemal polyglutamylation can be rapidly stripped in an inducible 4 manner. By implementing STRIP with simultaneous live-cell, time-lapse fluorescence imaging, 5 we report cell biological analysis on the immediate consequences of deglutamylation that is 6 rapidly induced inside primary cilia. We begin by describing a new approach, STRIP, for the spatiotemporal perturbation of 3 axonemal polyglutamylation to study its immediate effect on the interplay among the 4 microtubules, molecular motors, and MAPs (Fig. 1a). The basis of our technology is chemically 5 inducible dimerization 36,37 , in which a chemical such as rapamycin induces the dimerization of 6 FK506 binding protein (FKBP) and FKBP-rapamycin binding domain (FRB) (Fig. 1a). To 7 anchor a fusion protein of Cerulean3 and FRB onto ciliary axoneme, we employed 8 microtubule-associated protein 4 (MAP4) that accumulates at the axoneme 38 . In particular, we was confirmed to exhibit no significant effect on cilium length, axonemal acetylation, or 1 axonemal glutamylation ( Supplementary Fig. 1c). Next, we tested whether Cerulean3-FRB-2 MAP4m would affect IFT by monitoring Neon-IFT88, which labels IFT particles, with live-cell,  Fig. 1d,e). Collectively, these results confirmed that 6 Cerulean3-FRB-MAP4m is a valid fusion construct to be anchored at the ciliary axoneme.

7
Our previous study showing that most cytosolic proteins have access to the ciliary lumen 40 8 suggests that cytosolic FKBP proteins can be trapped at the axoneme upon rapamycin 9 addition in cells expressing Cerulean3-FRB-MAP4m. To test this possibility, we transfected 10 NIH3T3 cells with Cerulean3-FRB-MAP4m (axoneme), 5HT6-mCherry (cilia membrane), and 11 YFP-tagged FKBP (YFP-FKBP; cytosol). Subsequent exposure to rapamycin increased the 12 YFP fluorescence signal at the ciliary axoneme over 5 min  13 Movie 1). The time required for half-maximal accumulation of YFP-FKBP in the axoneme (t1/2) 14 was 98.1 ± 24.0 s (mean ± s.e.m.; Fig. 1d). To generalize the method, we next tested the 15 MAP4m approach with a gibberellin-based CID that works orthogonally to the rapamycin-16 based CID 41 . Here, we constructed two fusion proteins: a codon-optimized Gibberellin 9 insensitivity DWARF1 (mGID1) fused to Neon (Neon-mGID1), and MAP4m fused to an N-1 terminal 92 amino acids of Gibberellin insensitive (GAIs-CFP-MAP4m). The addition of a 2 gibberellin analog (GA3-AM) to cells co-expressing these fusion proteins led to accumulation 3 of the Neon fluorescence signal at the ciliary axoneme ( Supplementary Fig. 2 and   4 Supplementary Movie 2), albeit with slower kinetics (127.3 ± 25.9 s) compared with the 5 rapamycin-based CID ( Supplementary Fig. 2c). Taken together, these results confirmed that 6 the CID systems utilizing MAP4m can relocate cytosolic proteins onto the ciliary axoneme of 7 living cells within minutes.

10
To change the polyglutamylation status of tubulins in cilia, we implemented STRIP to recruit a 11 deglutamylase to the axoneme. Among the tubulin deglutamylase family, CCP5 is 12 demonstrated to efficiently remove glutamate residues at a branch point as well as those in a 13 linear tandem both in vitro and in vivo 12,14 , suggesting CCP5 as a valid candidate to actuate 14 deglutamylation upon the STRIP execution. We thus constructed Neon-FKBP-tagged full-15 length CCP5 (CCP5FL-Neon-FKBP), intending to relocalizing this fusion protein from cytosol 16 to axoneme upon rapamycin addition. However, CCP5FL-Neon-FKBP localized inside cilia even prior to rapamycin addition, thereby constitutively depleting axonemal glutamylation 1 ( Supplementary Fig. 3a-c), clearly hampering the use of the full-length CCP5 for the STRIP 2 operation. Therefore, we assessed subcellular localization of truncation mutants of CCP5, 3 namely N-terminus (CCP5N, residues 1-160) and catalytic domain (CCP5CD, residues 161-4 531). When expressed in cells as a fusion protein of Neon-FKBP, these two CCP5 truncations 5 both localize in cytosol ( Supplementary Fig. 3b). For further characterization of these 6 truncation mutants, we introduced two point mutations (H252S and E255Q) to CCP5CD 7 (CCP5CDDM) to impair the deglutamylation activity 9 . A fusion construct of this mutant 8 (CCP5CDDM-Neon-FKBP) was also confirmed to be cytosolic ( Supplementary Fig. 3b). In 9 contrast to full-length CCP5, none of the truncated mutants impacted axonemal glutamylation 10 ( Supplementary Fig. 3b,c). To directly assess enzymatic activities of these CCP5 truncations,

11
we forcefully anchored each of them to axoneme via fusion with MAP4m ( Supplementary Fig.   12 4a). The immunofluorescence assay showed that CCP5CD, but not CCP5N or CCP5CDDM, 13 significantly reduces axonemal glutamylation, indicating that CCP5CD retains viable enzyme 14 activity ( Supplementary Fig. 4b,c). Importantly, axonemal deglutamylation induced by 15 CCP5FL-Neon-FKBP or CCP5CD-Cerulean3-MAP4m did not affect axonemal acetylation or 16 cilium length (Supplementary Figs. 3c and 4c). Taken together, our results identified CCP5CD as an ideal candidate that becomes functional only at the axoneme without affecting other 1 aspects of primary cilia such as cilia length and tubulin acetylation status.

2
To trigger deglutamylation of axonemal tubulin, CCP5CD-Neon-FKBP was recruited to reduced the glutamylation within 30 min (Fig. 1e,f). Interestingly, deglutamylation did not alter 10 the localization of MAP4m, suggesting that MAP4m binding to microtubules is independent of 11 tubulin glutamylation (Fig. 1e). Importantly, the recruitment of Neon-FKBP alone or catalytically 12 inactive CCP5CDDM onto the axoneme was insufficient to deplete axonemal glutamylation, 13 confirming that CCP5CD-induced axonemal deglutamylation is dependent on the enzymatic 14 activity of CCP5 (Fig. 1e,f). It is noteworthy that the CCP5-or STRIP-induced deglutamylation 15 left residual glutamylation at the proximal end of primary cilia ( Fig. 1e and Supplementary Fig.   16 3b). The residual glutamylation localized at the inversin zone instead of the transition zone ( Supplementary Fig. 6). The detailed mechanism of how axonemal tubulins in the inversin 1 zone resist deglutamylation treatment is unclear.

11
To determine the spatial specificity of STRIP-triggered axonemal deglutamylation, we 12 measured the level of tubulin glutamylation in other subcellular compartments. Glutamylated 13 tubulins in the cytosol and basal body can be visualized after "cold treatment" before fixation 14 (see the method, Supplementary Fig. 8a) 10 . Consistent with previous reports 8,12 , 15 deglutamylation by CCP5FL is prevalent in the cytosol, basal body, and axoneme 16 ( Supplementary Fig. 8a,b). In contrast, CCP5CD does not lead to noticeable change in tubulin glutamylation compared with a control condition with Neon alone. Strikingly, the STRIP-1 triggered deglutamylation was most prominent at the axonemes, and non-significant 2 elsewhere in the same single cells, validating the high spatial precision (Supplementary Fig.   3 8a,b). It is somewhat interesting that expression of soluble CCP5CD in cells had only marginal 4 effect on glutamylation in the cytosol, despite a nature of free diffusion. We speculate that the 5 expression level of cytosolic CCP5CD may be low enough to take an effect, and/or that 6 endogenous polyglutamylases in the cytosol may counteract CCP5CD-mediated 7 deglutamylation in the cytosol. Nevertheless, a design principle of the STRIP enables drastic 8 concentration of CCP5CD on the axoneme at the expense of a fraction of the CCP5CD pool 9 in the cytosol, notably due to >10,000-fold volume difference between cilia and cytosol.

11
The roles of glutamylation in ciliary structure

12
To reveal the effect of axonemal deglutamylation on the ciliary structure, we first measured 13 the cilia length based on Arl13B staining. The cilium length measured before and after STRIP 14 treatment at least for 2 h was comparable among the control (no CCP5), wild-type (CCP5CD),

11
CCP5CD onto the growing axoneme significantly slows the extension of cilium length ( Fig. 3 12 b,c). Taken together, these results indicate that axonemal glutamylation is important for cilia 13 elongation during ciliogenesis, but not for maintenance of steady-state cilia.

16
Besides mechanical support, the axoneme also serves as railways for IFT 45,46 . We thus evaluated whether axonemal deglutamylation impacts the rate of IFT. To test this, the IFT 1 activity was quantified before and after STRIP-mediated axonemal deglutamylation by 2 monitoring Neon-IFT88 with live-cell fluorescence microscopy (Fig. 4). A subsequent linescan 3 analysis of the Neon-IFT88 signal in individual cilia at two time points (before and 30 min after 4 rapamycin treatment) showed that Neon-IFT88 initially spread across the entire cilia in the 5 form of puncta converges to the base of the cilium after deglutamylation (Fig. 4a,b). To assess 6 IFT motility, a kymograph was then generated based on the time-lapse images of Neon-IFT88 7 before and after deglutamylation (Fig. 4c). The velocity of IFT in both directions (i.e., 14 Besides rapid deglutamylation, we also evaluated the effect of long-term deglutamylation   . Taken together, these results confirmed that axonemal deglutamylation slows anterograde IFT without blocking the tethering of the IFT machinery to the basal body.

1
Anterograde IFT is mainly powered by kinesin motors 45 , whose distribution and motility 2 are reportedly modulated by tubulin glutamylation 15,18 . We therefore hypothesized that 3 axonemal deglutamylation inhibits anterograde IFT by impairing kinesin motility. To address 4 this, we examined the effects of axonemal deglutamylation on the distribution of Neon-Kif3B, Neon-Kif3B particles was measured to be comparable to that of the Neon-IFT88 trains 8 ( Supplementary Fig. 12b,c), suggesting that Neon-Kif3B can be used to assess the motility of 9 kinesin motors. We then found that Neon-Kif3B accumulates at the proximal end of 10 deglutamylated cilia (Fig. 6a,b), just like Neon-IFT88 (Fig. 4a,b). To reinforce this finding, we 11 performed quantification of Neon-Kif3B distribution in the cilia divided into two compartments, 12 base and axoneme, which exhibited more Neon-Kif3B at the base of deglutamylated cilia 13 compared with control cilia (70.7  2.2% vs. 59.6  3.9%) (Fig. 6c). In summary, these results 14 suggest that glutamylation controls anterograde IFT through Kif3B of the kinesin-2 complex.

15
The Lechtreck group once showed that the demand of anterograde IFT in newly growing 16 cilia is much higher than that in mature cilia 47 . Therefore, the rapid deglutamylation-induced defects in anterograde IFT may cause more severe inhibition of cilia elongation than cilia 1 maintenance, a prediction consistent with our finding that rapid deglutamylation impaired 2 elongation of immature, but not mature cilia (Figs. 2, 3).

13
The kinesin-2 complex interacts with Hh components to regulate Hh signaling 48-52 , making 14 us hypothesize that deglutamylation-induced defects in kinesin-2-mediated anterograde IFT 15 would collaterally affect Hh signaling. We thus examined the subcellular location of Hh 16 components such as Smoothened (Smo) and Gli3 in control and deglutamylated cilia after stimulation with the SAG (Fig. 7a). The SAG-induced accumulation of Smo (Fig. 7b,c) and 1 Gli3 (Fig. 7d,e) in the cilia was significantly impaired by axonemal deglutamylation. The 2 distribution and motility of GFP-Gli3 in deglutamylated cilia upon treatment with SAG was also 3 evaluated by FRAP ( Supplementary Fig. 14). Deglutamylation induced by CCP5CD-mCherry-  We also tested whether axonemal deglutamylation inhibits Gli activation using 8xGBS-

11
Defective ciliogenesis was previously observed after complete genetic depletion or 12 mutation of kinesin-2 in ciliated organisms [54][55][56][57] , which seems at first glance inconsistent with 13 our present finding that deglutamylation-induced defects in kinesin-2 does not affect cilia 14 maintenance. There are at least two explanations for this apparent discrepancy. In addition to 15 axoneme, kinesin-2 also localizes at the ciliary base to regulate its organization 58 (Fig. 6a and   16 Supplementary Fig. 12a). Therefore, genetic manipulation of kinesin-2 may have affected functions of kinesin-2 not only at the axoneme but also the ciliary base, possibly resulting in 1 defective basal bodies that could devastate ciliogenesis. Another explanation is that residual 2 anterograde IFT in the STRIP cells may have been sufficient for cilia growth and maintenance, 3 unlike the cases for near complete loss-of-function of kinesin-2 proteins 54-57 .

4
Our work showed that axonemal deglutamylation preferentially hampers anterograde IFT 5 but not retrograde IFT (Fig. 4c.d and Supplementary Fig. 11). In several ciliated model

Statistical analysis
Statistical analysis was performed with an unpaired two-tailed Student's t test and whether 1 variances were equal or not was determined by F test. P values were calculated, when P value 2 ≥0.05 represent no significant difference, P value < 0.05 represent significant difference, and 3 P value < 0.01 represent highly significant difference. The data that support the findings of this study are available from the corresponding author 7 upon reasonable request.