Polyglutamylation of microtubules drives neuronal remodeling

cytoskeleton, axon, synapse, microtubule, posttranslational modifi cation, polyglutamylation, pruning, synapse elimination, remodeling, microtubule dynamics, developmental biology, motor neuron Figures: 8 Figures 3 Supplemental Figures 1 Supplemental Video (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 2

1 Developmental remodeling shapes neural circuits via activity-dependent pruning of synapses and 2 axons. The cytoskeleton is critical for this process, as microtubule loss via enzymatic severing is an 3 early step of pruning across many circuits and species. However, how microtubule-severing enzymes, 4 such as spastin, are activated in specific neuronal compartments remains unknown. Here, we reveal 5 that polyglutamylation, a posttranslational tubulin modification that is enriched in neurons, plays an 6 instructive role in developmental remodeling by tagging microtubules for severing. Motor neuron-7 specific gene deletion of enzymes that add or remove tubulin polyglutamylation—TTLL glutamylases 8 vs. CCP deglutamylases —accelerates or delays neuromuscular synapse remodeling in a 9 neurotransmission-dependent manner. This mechanism is not specific to peripheral synapses but also 10 operates in central circuits, e.g., the hippocampus. Thus, tubulin polyglutamylation acts as an activity-11 dependent rheostat of remodeling and shapes neuronal morphology and connectivity. 12

13 Microtubules are regulators of cellular shape, dynamics, and transport—processes that are especially 14 difficult to coordinate in the complex cells of the nervous system 1-3. As a consequence, microtubule 15 functions in neurons need to be locally regulated. In analogy to how posttranslational modifications 16 (PTMs) diversify histone function, the concept of the “tubulin code” posits that microtubules can be 17 endowed with unique functions through PTMs and a set of proteins that add, remove, or interpret 18 them. Such a system of “writer”, “eraser”, and “reader” microtubule-associated proteins (MAPs) and 19 enzymes can then locally modify microtubule stability, as well as the trafficking characteristics and 20 cargo selection of intracellular transport with high spatiotemporal precision 3-5. Thus, in principle, the 21 tubulin code can explain how a general system of filamentous proteins could locally regulate a 22 plethora of parallel functions to meet the compartmentalized needs of neural cells . Indeed, t he 23 microtubular cytoskeleton of neurons is very specialized. This specialization includes the expression 24 of neuron -specific tubulin isoforms (such as tubulin beta -3, Tubb3), pronounced and 25 compartmentalized PTMs, as well as a unique set of MAPs and severing enzymes that play important 26 roles in neurodegenerative and neurodevelopmental diseases 6-10. Nevertheless, while a PTM-based 27 implementation of the “tubulin code” is an appealing concept and clearly plays important roles in the 28 basic health and cell biology of neurons , e.g. via cell migration, axonal transport, growth cone 29 navigation, and cilia function 11, how the tubulin code could steer nervous system physiology in vivo 30 in general, and circuit development in particular, 12 is not well understood 5, 13-15. 31 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 3 Synaptic pruning—the wide-spread removal of exuberant synapses that is preserved across phyla, but 32 also between development and disease —involves a local regulation of cytoskeletal function and 33 stability 16-19. For instance, others and we have previously described the selective loss of microtubules 34 in destabilized presynaptic axon branches 16, 20, 21 . This loss can be mediated by the microtubule -35 severing enzyme, spastin 20, a protein commonly altered in hereditary spastic paraplegia 8. Indeed, 36 microtubule destabilization, together with glia-mediated engulfment of synaptic material, is thus far 37 one of very few shared mechanisms of synapse pruning across models and between development and 38 disease 22, 23. The local loss of microtubules in pruning axon branches is accompanied by changes in 39 PTMs 20, but whether these cytoskeletal changes are instructive or merely an epiphenomenon of the 40 breakdown, and which molecular signals locally regulate microtubule modifications, remains 41 unknown. Of particular significance in this context is polyglutamylation, a tubulin PTM that is highly 42 enriched in the brain 24. Tubulin polyglutamylation involves the enzymatic “seeding” of an initial 43 glutamate followed by addition of further glutamate residues, resulting in the formation of a 44 polyglutamate (PolyE) chain. This PTM enhances the negative charge of the microtubule lattice, thus 45 facilitating electrostatic interactions with neurodegeneration-associated MAPs, such as tau 25, 26, and 46 severases, including spastin 27. 47 Based on these prior observations, we undertook a systematic investigation into the role of 48 microtubule polyglutamylation during pruning. To this, we took advantage of the special features of 49 synaptic pruning at the neuromuscular junction , a synapse which transitions within a few days from 50 innervation by multiple to a single axon branch 28. Due to its accessibility and size, this synapse allows 51 unique dynamic investigations with subsynaptic resolution 20, 29 -31, which we used to reveal an 52 instructive role of tubulin polyglutamylation in synaptic remodeling . Specifically, we show that in 53 motor axon branches during pruning, polyglutamylation of tubulin alpha-4A (Tuba4a) results from the 54 balanced action of the “writer” enzyme, tubulin tyrosine ligase-like glutamylase 1 (TTLL1; but not the 55 functionally distinct TTLL7 13) and the “eraser” enzymes, cytosolic carboxy peptidases ( CCP) 1 and 6. 56 The local density of polyglutamyl side chains on microtubules, modified in cell type-specific knock-out 57 mice, in turn, determines the efficiency of spastin-mediated severing. We found that this interplay of 58 PTMs and microtubule severing paces the speed by which peripheral synapses remodel, but confirmed 59 a related mechanism for the timing of axon and spine pruning in the developing hippocampus. 60

61 Glutamylases and deglutamylases are translationally regulated during motor axon pruning 62 To establish the expression patterns of polyglutamylation “writer” and “eraser” enzymes during motor 63 axon pruning, we performed a translatome analysis of murine motor neurons using a ChAT-IRES-Cre 64 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 4 X Rpl22HA (called RiboTag hereafter) 32, 33 mouse cross -breeding. We immuno-isolated ribosome-65 bound mRNA from spinal cords , starting at postnatal day (P) 5 until 14 (P5, 7, 9, 11, 14 ; Fig. 1a ). 66 Specificity of this experiment for motor neuronal transcripts was verified by immunostaining for HA, 67 which illustrated clear co -localization with Choline Acetyl Transferase (ChAT) ( Fig. 1b), and 68 quantitative PCR, which demonstrated a 20x enrichment of ChAT and absence of Glial Fibrillary Acidic 69 Protein (GFAP) transcripts in the pull -downs compared to whole spinal cord mRNA (Extended data, 70 Fig. 1a and 1b). Principal component (PCA) analysis clearly separated the RiboTag pull -down vs. the 71 total spinal cord mRNA samples (Extended data, Fig. 1c). In total, we identified > 4,500 differentially 72 expressed transcripts at P9 (RiboTag vs whole spinal cord; Log2 |FC| ≥ 1.5; Padj ≤ 0.05; Extended data, 73 Table 1 ). Gene ontology (GO) analysis of this gene list pointed towards neuron- or axon -specific 74 biological processes (Extended data, Fig. 1d and 1e). Similarly, gene set enrichment analysis revealed 75 motor neuron markers (Isl1, Chat, Mnx1) augmented, while astrocyte markers (GFAP, Slc1a3, Aldh1l1, 76 Olig2, Cspg4) and oligodendrocyte-specific genes (Sox10, Olig2, Cspg4) were depleted in the RiboTag 77 pull-down samples ( Extended data, Fig. 1f). We then scrutinized the complete gene list (Extended 78 data, Table 2) obtained from our bona fide motor neuron translatome, for glutamylases (TTLL1, 3, 4, 79 5, 6, 7, 11 and 13) and deglutamylases (CCP1, 2, 3, 4, 5 and 6; Extended data, Table 2). 80 “Writer” TTLL glutamylases decorate the microtubule lattice with glutamate either in a “seeder” 81 fashion, by preferentially adding the first branch glutamate (E) onto the tubulin C -terminal tail 82 (monoE), or by elongating previously seeded E-chains 13, 34 (Fig. 1c). Among the known “elongators”, 83 TTLL1 showed the highest expression level s (e.g., at P14: approximately 69% of “elongator” TTLL 84 levels) that were stable across all analyzed time points. In contrast, TTLL11 wa s expressed at a 85 substantially lower level (e.g., at P14: 29 %), while TTLL6 and TTLL13 were virtually absent ( Fig. 1d). 86 Analysis of “seeders” yielded increasing TTLL7 mRNA reads between P5 and P14 (e.g., at P14: 87 % of 87 “seeder” TTLL levels), while TTLL5 and TTLL4 transcripts were much less abundant (at P14: 11 % and 1 88 % for TTLL5 and TTLL4, respectively; Fig. 1e). Amongst the deglutamylases, CCP1 transcripts were the 89 most abundant and increased during the pruning phases (e.g., increase from P5 to P14: 53 %; Fig. 1f). 90 Among all the other CCP-type deglutamylases (CCP2, 3, 4, 5, 6), CCP6 was the most abundant enzyme 91 which exhibited an increase of 28% between P5 and P14. The rest were present at very low transcript 92 rates, ranging e.g. from 0.03% to 1.1 % at P14 (Fig. 1f). These findings are consistent with previous 93 reports showing that CCP1 and CCP6 are the two most abundant deglutamylases in the central 94 nervous system 35. Thus, we focused our efforts on genetic disruption of TTLL1, TTLL7, CCP1 and CCP6 95 to probe the ir role as potential candidates that mediate polyE-instructed neuronal remodeling. We 96 first generated cholinergic neuron-conditional mouse mutants of the most expressed glutamylases, 97 i.e. TTLL1mnKO (ChAT-IRES-Cre X TTLL1flox/flox; see experimental procedures), which should mostly affect 98 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 5 polyglutamylation on tubulin alpha 13, and TTLL7 mnKO (ChAT-IRES-Cre X TTLL7flox/flox; see experimental 99 procedures), which would be expected to deplete monoglutamylation on tubulin beta 13. 100 Genetic ablation of TTLL1, but not of TTLL7, delays neuromuscular synapse elimination 101 We quantified microtubul e content and PTM patterns by immunohistochemistry and confocal 102 microscopy in the triangularis sterni muscles of these mice at P 9 20, 31, 36. To determine microtubular 103 content, we used an anti-tubulin beta-3 (Tubb3) antibody, the most abundant tubulin beta isoform in 104 motor neurons after P9 (Extended data, Fig. 1g ), while an anti-PolyE antibody, which recognizes C-105 terminal chains of  3 glutamate residues on tubulin tails, was used to assess axonal polyglutamylation 106 levels. As expected, PolyE normalized to Tubb3 was drastically reduced (to 0.09-fold), while 107 microtubule content was 2.8-fold increased ( normalized to neurofilament; Fig. 2 a-d). In contrast , 108 immunostaining using an anti-βmonoE antibody (normalized to Tubb3), which detects a single 109 glutamate seed on E435 of tubulin beta-2 tails, was largely unaffected (Fig. 2e and 2f). Finally, antibody 110 staining for acetylated tubulin (acetyl-K40) was also decreased (0.68-fold, normalized to Tubb3, Fig. 111 2g and 2h), in line with previous reports 37. Thus, microtubules in motor axons in TTLL1mnKO mice are 112 more abundant but may carry less PTMs. As revealed by time-lapse imaging of nerve-muscle-explants 113 of TTLL1mnKO that also carried a Thy1-EB3-YFP transgene 38, most parameters of plus-end 114 polymerization dynamics were unaffected (including comet density, orientation and velocity , 115 Extended data, Fig. 2a, 2c and 2d), but comet length distribution was significantly, albeit mildly altered 116 in TTLL1 mnKO compared to TTLL1 mnWT (Extended data, Fig. 2b). Specifically, EB3 comet elongation in 117 TTLL1mnKO increased aligning with recent in vitro findings that show augmented microtubule growth 118 rate upon deglutamylation 39. Concomitant to the increased microtubule mass and reduced PolyE 119 levels, a higher percentage of neuromuscular synapses remained innervated by two or more axons in 120 TTLL1mnKO vs. control muscles at all postnatal ages tested ( Fig. 2i and 2j). Collectively these findings 121 suggest that TTLL1 -mediated polyglutamylation instructs the local dismantling of microtubules , and 122 thus paces developmental axon pruning of motor neurons. 123 In contrast, the same set of analyses in TTLL7 mnKO mice revealed no c hanges in overall microtubule 124 mass and dynamism in terminal motor axons (Fig. 3a and 3b; Extended data, Fig. 2e-2h). In line with 125 the specificity of TTLL7 activity 13, the intensity of PolyE immunostaining was not altered (Fig. 3c and 126 3d), while quantification of βmonoE immunofluorescence revealed a decrease compared to control 127 axons as expected (0.55-fold, normalized to Tubb3, Fig. 3e and 3f). Consistently, the pruning speed of 128 motor axons was similar in TTLL7 mnKO vs. control muscles ( Fig. 3g). These data corroborate that only 129 the density of TTLL1-added polyglutamylation, but not TTLL7 -catalyzed βmonoE seeds, influences 130 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 6 pruning—suggesting a high level of specificity in th e regulation of pruning via microtubule PTMs, in 131 line with the tubulin code concept. 132 TTLL1KO leads to pruning defects in the CNS 133 We next explored whether reduced TTLL1-mediated polyglutamylation of microtubules would also 134

in defective pruning processes in the CNS. We chose granule neurons of the dentate gyrus, 135 which provide a well-established and easily quantified murine model for ‘stereotyped’ pruning. During 136 the first two postnatal months, granule cells retract their infrapyramidal bundle (IPB) of mossy fibers, 137 while the main suprapyramidal bundle (SPB), is maintained ( see schematic; Fig. 4a). We used a 138 constitutive TTLL1KO mouse model, which exhibits reduced PolyE levels in the brain 14. Quantification 139 of the IPB/ SPB length ratio in sections immunostained for Calbindin D -28K, which sel ectively stains 140 mossy fibers, revealed delayed pruning of the IPB in TTLL1 KO vs. control mice (Fig. 4b-d). We further 141 tested if elimination of spines in granule neuron dendrites —i.e., remodeling of input synapses further 142 upstream in hippocampal circuitry —was delayed as well. Gene-gun based DiI labelling 40 before (3 143 week-old mice) and after the pruning phase (8 week-old mice) showed that TTLL1KO mice developed 144 normal spine densities initially, but failed to prune (Fig. 4e-g). This suggests that TTLL1 -mediated 145 polyglutamylation might be a general pacemaker for the execution of remodeling processes, not only 146 during competitive elimination of peripheral synapses, but also during stereotyped pruning in the CNS. 147 Polyglutamylation of tubulin alpha-4A instructs remodeling 148 Exploration of tubulin isoforms in our motor neuron translatome data revealed an upregulation of 149 tubulin alpha-4A (Tuba4a) among the alpha isoforms, with a consistent increase across the synaptic 150 pruning phase (+63% from P5 to P14; Fig. 5a). In contrast, other tubulin alpha isoforms were either 151 stably expressed (e.g., Tuba1b and Tuba1c; Fig. 5a) or downregulated (Tuba1a; Fig. 5a). Interestingly, 152 Tuba4a is the isoform which seems to carry the longest glutamate side chains 41, 42. In light of th is 153 expression pattern, we investigated the effect of Tuba4a -specific polyglutamylation loss in neuronal 154 remodeling. For this, w e utilized Tuba4a knock-in mice (Tuba4aKI), which carry a mutated Tuba4a 155 allele, which prohibits glutamylation of this specific tubulin isoform 41. Using confocal analysis of 156 double immunostainings for PolyE and Tubb3 in terminal axons in Tuba4aKI triangularis sterni muscles, 157 we found a phenotype resembling to that of TTLL1 mnKO: Tubb3 intensity increased ( 1.2-fold; 158 normalized to neurofilament; Fig. 5b and 5c), while the intensity of the P olyE signal staining was 159 reduced (0.9-fold; normalized to Tubb3; Fig. 5d and 5e). As motor axon remodeling was significantly, 160 albeit transiently, delayed ( Fig. 5f), we conclude that polyglutamylation of tubulin alpha-4A is 161 necessary to initially drive the pruning of motor axons. 162 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 7 Genetic deletion of deglutamylases CCP1 and CCP6 in motor neurons accelerates pruning 163 We next wondered whether the described system operates in both direction s, meaning that hyper-164 polyglutamylation of microtubules would accelerate axon pruning. Deglutamylases (CCPs) catalyze the 165 removal of glutamate chains from tubulin C -terminal tails 43-45, and thus counteract the enzymatic 166 activity of elongating glutamylases. We focused our efforts on combined motor neuron -specific 167 deletions of CCP1 and CCP6 (CCP1&6mnKO), as—while CCP1 was the dominant CCP transcript in our 168 translatome data (Fig. 1f)—CCP6 is known to have compensatory potential. 169 Immunofluorescence analysis in terminal triangularis sterni axons at P9 revealed a reduction of Tubb3 170 levels in CCP1&6 mnKO mice compared to littermate controls (0.8-fold; normalized to neurofilament; 171 Fig. 6a and 6b). Unexpectedly, PolyE immunostaining intensity was significantly decreased (0.6-fold; 172 normalized to Tubb3; Fig. 6c and 6d), while βmonoE levels remained unaltered as predicted (Fig. 6e 173 and 6f). In converse to the TTLL1mnKO phenotype, acetylated microtubules were increased significantly 174 (Fig. 6g and 6h). Furthermore, we observed changes of EB3-YFP comet dynamics, with a 27 % drop in 175 density and a 12 % increase in mean velocity in CCP1&6mnKO compared to control mice (Extended data, 176 Fig. 2i and 2k). Microtubule plus-end growth (EB3 comet orientation) and length distribution remained 177 normal (Extended data, Fig. 2j and 2l). Correlative live imaging of EB3-YFP comet density normalized 178 to tubulin beta-3 immunostaining in fixed tissue on the single axon level was unchanged in the 179 CCP1&6mnKO compared to CCP1&6 mnWT controls ( Extended data, Fig. 3a-d). Of note , the synapse 180 elimination process in CCP1&6 mnKO mice was accelerated, in accordance with lower Tubb3 levels, as 181 fewer neuromuscular synapses were polyinnervated at several time points (Fig. 6 i). This finding 182 corroborates that the polyglutamylation “writer” and “eraser” enzymes bidirectionally or 183 “rheostatically” regulate the developmental remodeling of axons and synapses. 184 Spastin swiftly eliminates excessively polyglutamylated microtubules 185 The absence of excess polyglutamylation in young CCP1&6mnKO motor axons is surprising, as CCPs 186 remove glutamate residues from microtubules; consequently, their genetic deletion should result in 187 an increased PolyE immunostaining. We explored two possible explanations for this observation: (i ) 188 excessive polyglutamylation might result in highly branched, arborized polyglutamyl chains that could 189 evade PolyE-immunodetection due to steric hin drance 46 or (ii) instantaneous severing and 190 subsequent destabilization of excessively polyglutamyl-decorated microtubules by enzymes such as 191 spastin 27 might leave the polyglutamylated microtubule pool paradoxically depleted. 192 To test the former possibility , we performed anti-GT335 immunostainings, an antibody recognizing 193 the branching point of glutamate chains on tubulin tails 24, 47 independent of chain length . Confocal 194 quantification indicated a similar staining intensity in CCP1&6mnKO mice compared to controls (Fig. 6j 195 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 8 and 6k). This suggests that the absence of hyper-polyglutamylation in CCP1&6mnKO mice has a 196 biological, rather than technical , explanation. Next, we focused on a possible intersection of the 197 effects of absent CCP-type deglutamylase activity and spastin activation. To test this, we conducted 198 neonatal (P3) injections of an adeno associated virus ( AAV9-hSyn-Cre), which expressed Cre 199 recombinase in neurons only, in conditional CCP1 and spastin mice that also carried a reporter allele 200 (CCP1flox/flox X Spast flox/flox X TdTomato). The TdTomato reporter fluorescence was analyzed in motor 201 axons at P9 to validate successful Cre -mediated excision (Fig. 7a ). To circumvent background 202 ambiguities and the lack of reliable in situ detection tools for endogenous spastin or deglutamylases, 203 the tdTomato signal was indexed in quan tiles based on fluorescent intensity . We compared the 204 quantile with the highest intensity (tdTomatohigh; which we considered most likely for CCP1 and spastin 205 double-deletion) with the bottom quantile (tdTomatolow; likely non-recombined) for levels of PolyE 206 immunostaining. This indeed revealed a 2-fold increase of PolyE in the tdTomatohigh compared to the 207 tdTomatolow axons (Fig. 7b and 7d), which was accompanied by added microtubule content (1.3-fold; 208 Fig. 7b and 7c). Notably, spastin translation decreases from P5 onwards (which contrasts to other 209 severing enzymes, such as Katnal1; Extended data, Fig. 4a and Table 3 ). This suggests that 210 deglutamylases might be more relevant later in life —and, consistently, quantification of 211 immunostainings in adult CCP1&6 mnKO motor axons showed lower microtubule mass ( 0.4-fold; 212 normalized to neurofilament; Extended data, Fig. 4b and 4c) accompanied by a drastic increase in 213 PolyE intensity (4.1-fold; Extended data, Fig. 4d and 4e). In summary, microtubular content and PTM 214 composition in motor axons appears to be controlled by the interplay of tubulin glutamylases, 215 deglutamylases and specific microtubule severing enzymes. 216 Microtubular PTM levels are under control of neurotransmission 217 Neuronal activity is a main driver of sculpting neural circuits, including at the neuromuscular synapse 218 19, and it has been demonstrated to facilitate tubulin polygluta mylation 48. We previously 219 demonstrated that block of neurotransmission by injection of α-bungarotoxin (α-BTX), which blocks 220 postsynaptic acetylcholine rec eptors and stalls synapse elimination , leads to a reduction of tubulin 221 beta-3 immunofluorescence in the presynaptic motor axon 49. In order to test if this loss of tubulin 222 beta-3 is mediated by activity -driven modulation of P TMs, we analyzed polyglutamylation i n motor 223 axons at P9 following an injection of α-BTX into the thorax two days earlier (Fig. 8a). In axon branches 224 that terminated in α-BTX-blocked synapses (BTX+) polyglutamylation dropped significantly compared 225 to control axons (0.7-fold; normalized to Tubb3; Fig. 8b and 8c). This reduction following block of 226 neurotransmission is reminiscent of CCP1&6mnKO mice, where spastin appears to deplete 227 polyglutamylated microtubules faster than glutamyl side chains can accumulate. In order to test this 228 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 9 notion, we injected α -BTX into mice that globally lack spastin (Spast KO) and accumulate 229 polyglutamylation 20, enabling us to detect glutamylase activity u pon block of neurotransmission . 230 Indeed, PolyE immunostaining levels were increased (2.1-fold) in blocked axons ( α-BTX+/SpastKO) 231 compared to branches innervating unaffected synapses ( α-BTX-/SpastKO; Fig. 8b and 8c). Thus, our 232 data suggest that the enzymatic activity of the “elongator” TTLL glutamylase in the presynaptic axons 233 is governed by neurotransmission. In summary , polyglutamylation indeed appears to “en code” 234 activity-dependent signals that are subsequently “read” by spastin, which severs polyglutamylated 235 microtubules and thus facilitates axon branch and synapse removal. 236

237 Neurons remodel extensively during development, with remarkable temporal and spatial precision 19, 238 50. The fate of adjacent and otherwise indistinguishable neurites and synapses can starkly diverge—239 some are swiftly removed while others, often immediate neighbors, last for a lifetime 51. The 240 microtubule cytoskeleton appears to play a key role in such remodeling processes 16, 31. Albeit, how 241 cytoskeletal functions would be regulated in the required temporally- and spatially-controlled manner 242 has remained unclear. Here, we reveal a rheostatic system of enzymes 27 that determines the speed 243 of rem odeling and is governed by the “writer” polyglutamylase TTLL1 and opposing “eraser” CCP 244 deglutamylases. This system sets the glutamylation level, which in turn determines axon or synapse 245 fate via the microtubule destabilizing action of the “reader” severase spastin 20. 246 This concept is noteworthy from two vantage points: First, as a uniqu e in vivo manifestation of the 247 “tubulin code” concept, where PTMs convey specific local functionalit y to the cytoskeleton during 248 neurodevelopment, and second, as a possible general cell biological mechanism of branch -specific 249 axon or synapse dismantling. The “tubulin code” hypothesis suggests that PTMs modify microtubule 250 function, including stability, dynamics, and interactions 4, 52. Polyglutamylation is a modification that 251 is especially pronounced in neurons. By changing tubulin’s charge characteristics, glutamylation can 252 affect the recruitment of MAPs to microtubules 53, 41, 54. Accordingly, polyglutamylation has been linked 253 to transport deficits and consequent neurodegeneration 13, 14, 55, 56. However, physiological roles have 254 been harder to decipher for this PTM. Indeed, the large set of involved enzymes, which catalyze 255 addition and removal of glutamate residues, and the difficulty to probe them with sufficient spatial 256 and temporal precision in the developing nervous system make such investigations challenging. Our 257 translatome-guided and cell -type se pecific genetic approach, now reveal TTLL1-mediated 258 polyglutamylation of tubulin alpha-4A (Fig. 2 and 5) as an important step in neuronal pruning and 259 corroborate the high level of enzymatic specificity of the tubulin polyglutamylation system also in 260 neural development 13, in full concordance with the “tubulin code” concept. In contrast, the initiator 261 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 10 glutamylase TTLL7, which catalyzes the addition of the first glutamate onto tubulin beta isoforms 57, 262 does not seem to be involved in remodeling ( Fig. 3). Notably, tubulin alpha-4A—while not the most 263 abundantly expressed tubulin iso form—is the only tubulin alpha isoform that shows increasing 264 expression across the neuromuscular remodeling period ( Fig. 5a). This suggests that transcriptional 265 regulation of tubulin genes and perhaps enzymes might be a global factor that is permissive for 266 efficient polyglutamylation-driven remodeling. A further proposition that remains to be tested on the 267 basis of increasing expression ( Fig. 1e) is that TTLL5, a n alternative “seeder” glutamylase for tubulin 268 alpha 58, might act upstream of TTLL1 during remodeling. 269 Our data contain further examples of the complex interplay both on the transcriptional and the post-270 translational level between various players i n this system. For instance , the motor neuron -specific 271 deletion of deglutamylases (CCP1&6mnKO) had different outcomes regarding microtubular mass and 272 polyglutamylation levels in developmental remodeling vs. in adult motor axons (Extended data, Fig. 273 4b-e)—two time windows between which spastin mRNA levels dropped (Extended data, Fig. 4a). This 274 drop in polyglutamy lation could be reversed during remodeling by combined deletion of CCP1 and 275 spastin (Fig. 7), suggesting that microtubule stability and apparent PTM levels are the result of the 276 dynamic balance between “eraser” and “reader” activity. In this model, hyper-activated spastin 277 normally removes polyglutamyl-decorated microtubules faster than this population can be 278 replenished—leading to the observed paradoxical relative loss of polyglutamylated microtubules after 279 ablation of the enzymes that normally remove this PTM. 280 Polyglutamylation can profoundly vary even between adjacent axon branches that only differ in their 281 subsequent fate 20. Together with our data, this suggests that glutamate residues can be added onto 282 microtubules not only with temporal but also with high spatial precision. The se locally modified 283 microtubules can then, in turn, recruit and activate special sets of MAPs, including tau, kinesin motors, 284 and severing enzymes, such as spastin 59, 27, 41, 54. Our data also reveal that changing one PTM (e.g. via 285 genetic manipulation of TTLL1) can percolate within the cytoskeleton to affect other PTMs, such as 286 acetylation 37. Thus, the interplay between various PTMs results in a complex pattern of specialized 287 microtubule subsystems within individual axon branches that determines pruning via regulation of 288 microtubule stability and function. Overall, our work implies that polyglutamylation-based encoding 289 can tag neurites for removal —a new model of selective axon removal that can be tested in other 290 species and models. 291 Open questions that remain, are which events take place upstream and downstream of the spastin-292 mediated microtubular destabilization. For instance, how does microtubule breakdown result in the 293 characteristic but often d ivergent neuritic morphologies that accompany pruning? At the 294 neuromuscular junction, axo nal b reakdown involves the formation of axon bulbs, followed by 295 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 11 piecemeal degradation, resembling apoptotic body shedding 29, 50. In contrast , mossy fibers in the 296 infrapyramidal bundle and axons in remode ling neurons in the mushroom body seem to rather 297 undergo fragmentation 50, 60, while spines are typically absorbed without a morpholo gical trace 61. 298 Interestingly, many morphogenic processes in neurites are more commonly associated with actin 299 dynamics rather than with microtubule breakdown 62, 63. However, the switch between growth and 300 retraction of neu rites is known to involve an interplay of actin and microtubules 64. Similarly, the 301 balance between the longitudinally rigid co -axial microtubules vs. circumferentially-running actin 302 striations might normally prevent axons from beading 65, 66. Thus, the idea that polyglutamylation -303 triggered microtubule loss could trigger shape changes that are seen during pruning is supported by 304 existing eviden ce, while how the subsequent steps of axonal disintegration followed by glial 305 engulfment are triggered remains to be resolved. 306 Another key feature of some forms of neuronal remodeling is activity -dependence 50, 67. Also here it 307 remains overa ll unclear how altered activity patterns translate into morphogenic change. In our 308 experiments, chronic α-bungarotoxin blockage of neuromuscular synapses ( Fig. 8) changed the 309 polyglutamylation levels in presynaptic motor axons. However, it remains elusive, which factor in the 310 enzymatic equilibrium that governs polyglutamylation levels is affected and how the enzymes that 311 mediate or interpret polyglutamylation are regulated by the transcellular signaling that underlies 312 synapse elimination 64. For example, s pastin phosphorylation might regulate its binding to 313 microtubules or protein stability, although it is not clear how this would impact severing 68. For TTLLs 314 and CCPs, information about how their enzymatic activity is regulated is even sparser 37, 69. Indeed, 315 thus far no link between regulatory mechanisms that control polyglutamylation and the signaling 316 pathways that are presumed to execute neurite remodeling 50 has been made. Our prior work provides 317 a possible feedback mech anism, by which the relevant enzymes themselves might be provided by 318 microtubule-dependent transport, which we have shown to stall early in remodeling 20 and influence 319 other signaling events, such as myelination 49. Thus, changes in transport could alter the local 320 equilibrium between the addition and the removal of glutamate residues, with spastin actin g as a 321 local, branch-specific detector. 322 Indeed, our work strengthens the long -sought link between developmental neurite remodeling and 323 axon degeneration 50. Many of the enzymes of the remodeling mechanism that we describe here are 324 linked to neurodegenerative disease: Spastin to hereditary spastic paraplegia 8, CCP1 to infantile-onset 325 neurodegeneration 70, and Tuba4a to motor neuron disease 41. A unifying mechanism could involve 326 alterations on microtubule dynamics , followed by altered organelle and enzyme transport, which 327 again could feed back to cytoskeletal stability 71. Thus, developmental neurite remodeling could be 328 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 12 instructive for developing subtler strategies to target microtubules in settings of axon degeneration 329 by focusing drug development on polyglutamylation. 330 Acknowledgments 331 We t hank Emily Spaulding and Rob Burgess (The Jackson Laboratory, Bar Harbor, ME , USA) for 332 providing the RiboTag protocol ; Julien Gagneur (Technical University Munich, Computational 333 Molecular Medicine, Munich, Germany) for advice on RNA sequencing analysis; Leanne Godinho for 334 critically commenting on the manuscript and Adrian Marti Pastor for advice on software. For excellent 335 technical assistance, we thank Kristina Wulliman , Sabine Brummer and Yvonne Hufnagel; for animal 336 husbandry, we acknowledge Manuela Budak, Nebahat Budak and Matea Korica; we also thank Kristine 337 Kellermann for veterinarian consultancy and Sebastian Berger for IT-support. 338 Author contributions 339 Conceptualization AG, AZ, CJ, TM, MSB; data curation AG, AZ; methodology MB, TM, MSB; funding 340 acquisition TM, MSB; investigation AG, AZ, MW, MR; resources KAZ, MM, TJH, CJ, SE, MK, TM, MSB; 341 supervision CJ, SE, MK, MB, TM, MSB ; visualization AG, AZ; writing – original draft AG, TM, MSB; 342 writing – review & editing all authors. 343 Funding 344 TM and MSB were funded by the German Research Foundation (DFG) Excellence Cluster SyNergy (EXC 345 2145 – ID 390857198). MSB is the recipient of a DFG research grant (LE 4610/1-1 – ID 450131873) and 346 supported by the DGM foundation. TM was further supported by the European Research Council 347 under the European Union’s Seventh Framework Program (grant no. FP/2007 -2013; ERC Grant 348 Agreement no.: 616791), the German Center for Neurodegenerative Disease (DZNE), by DFG Mi 349 694/9-1 (FG Immunostroke 428663564) and by the DFG TRR 274/1 2020 (project s C02 and B03; ID 350 408885537). SE was supported by the DFG (ID 4 03584255–TRR267) and the Federal Ministry of 351 Education and Research (BMBF) in the framework of the Cluster4future program (CNATM - Cluster for 352 Nucleic Acid Therapeutics Munich). MK was supported by the DFG grants KN556/11-1 and KN556/11-353 2. MB was supported by Czech Science Foundation grant 21-24571S. MR was supported by the Grant 354 Agency of Charles University grant GAUK 275423. CJ is supported by the program “Investissements 355 d’Avenir” launched by the French Government and implemented by ANR with the references ANR-11-356 LBX-0038 and ANR-10-IDEX-0001-02 PSL, by the Institut Curie, the French National Research Agency 357 (ANR) award s ANR-17-CE13-0021 and ANR-20-CE13-0011, and the Fondation pour la Recherche 358 Medicale (FRM) grant s DEQ20170336756 and FRM MND2020030. MM obtained funding from the 359 Fondation Vaincre Alzheimer grant FR-16055p and the France Alzheimer grant 2023. 360 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 13 Experimental Procedures 361 Mouse lines, husbandry and genotyping 362 In all experiments, mice from both sexes were included. Animals were housed in i ndividually 363 ventilated cages with food and water ad libitum. All animal experiments conform to the regulations 364 by the local authorities (e.g., Government of Upper Bavaria). Experimental animals were kept together 365 with littermates. 366 Motor neuron knockout specific mice were generated by crossbreeding conditional knockout mice to 367 ChAT-IRES-Cre 32 mice, expressing Cre -recombinase under the ChAT -promotor (Jackson #6410). 368 ‘Ribotagging’ of motor neurons was conducted in ChAT-IRES-Cre crossbred to homozygous Rpl22HA 33 369 mice (RiboTagflox/flox; Jackson, #11029) and conditional Spast flox/flox (Fig. 1, Fig. 5 and Extended data, 370 Fig. 1 and 3). Glutamylases (TTLL1 or TTLL7) or deglutamylases (CCP1 and CCP6) were deleted in motor 371 neurons by crossbreeding conditional mutants of CCP1 72 &6 14, TTLL1 14 or TTLL7 13 mice (gift from Dr. 372 C. Janke, Institut Curie, Orsay) to ChAT-IRES-Cre 32 animals. All animals were CCP1&6, TTLL1 or TTLL7 373 homozygous; experimental animals – CCP1&6mnKO, TTLL1 mnKO or TTLL7 mnKO – were Ch AT-IRES-Cre-374 positive, whereas littermate controls were Ch AT-IRES-Cre-negative, named hereafter CCP1&6 mnWT, 375 TTLL1mnWT or TTLL7mnWT. Thy1-YFP 73 transgenic mice (cytoplasmic YFP in all motor neurons, Jacks on 376 #3709) were used to assess pruning speed in cross-breeding to CCP1&6mnKO, TTLL1 mnKO, TTLL7 mnKO 377 compared to their littermates CCP1&6 mnWT, TTLL1 mnWT, TTLL7 mnWT. Microtubule dynamics was 378 visualized and analyzed by Thy1-EB3-YFP 38 transgenic animals crossed to either CCP1&6 mnKO, 379 TTLL1mnKO or TTLL7 mnKO; CCP1&6 mnWT, TTLL1 mnWT or TTLL7 mnWT were littermate controls. Conditional 380 knockout of CCP1 72 and Spastin 20 in motor neurons, monitored by ROSA-CAG-TdTomato reporter 74 381 (Ai14; Jackson; #7914; CCP1flox/flox X Spastflox/flox X TdTomato) was generated by injection of viral vectors 382 encoding Cre-recombinase. Block of neurotransmission was conducted and analyzed in constitutive 383 spastin knockout 20 (SpastKO) mice (homozygous Spast KO vs controls SpastWT) and C57BL/6N (Charles 384 River, Strain Code 027) controls injected with α -BTX. Spine density measurements, hippocampal 385 pruning analysis was performed on TTLL1 constitutive knock-out 14 (named here TTLL1KO; gift from Dr. 386 C. Janke, Institut Curie, Orsay, France) backcrossed to CD-1 (Charles River, Strain Code 022) for three 387 generations. Cytoskeletal analysis and pruning speed were analyzed in homozygous tubulin alpha-4A 388 knock-in 41 (Tuba4aKI/KI) mice (kindly provided by Dr. M. Kneussel, ZMNH , Hamburg, Germany), and 389 negative, thus wildtype, littermates (Tuba4aWT/WT) served as controls. Genotyping of the Tuba4aKI was 390 performed as published 41. 391 Genotyping was performed as previously described 20, 49 (for the Tuba4aKI see above). Briefly, genomic 392 DNA was extracted from biopsies using a one -step lysis (lysis buffer in mM: 67 Tris, pH 8.8, 16.6 393 (NH4)2SO4, 6.5 MgCl2, 5 β-mercaptoethanol, 10 % Triton-X-100, and 50 μg/ml Proteinase K; incubation 394 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 14 at 55 °C for 5 h, followed by inactivation step 5 min at 95 °C). PCR was performed with GoTaq Green 395 Master Mix (Promega; #M7121) following a standard protocol, and then DNA was separated on a 1.5–396 2 % agarose gel. Primer sequences are available upon request. 397 Ribosomal pull down and RNA sequencing 398 Spinal cords were dissected from ChAT-IRES-Cre X Rpl22HA X Spastflox mice (see Mouse lines, husbandry 399 and genotyping above). The sampl es were kept at -80 °C until pull down of labeled ribosomes 400 described previously 33 (see also http://depts.washington.edu/mcklab/RiboTag.html). Spinal cords 401 were homogenized (buffer in RNase free water in mM: 50 TrisCl, pH7.5, 100 KCl (Sigma #P9541), 12 402 MgCl2 (Sigma #63068), 1 % NP-40 (Roche, #11332473001), 1 X Dithiothreitol (DTT, Sigma #646563), 1 403 X Protease Inhibitors (Sigma #11697498001), 1 mg/ml Heparin (Sigma #H3393), 1 00 µg/ml 404 cycloheximide (Sigma #C7698), RNAseOUT (Thermofisher #10777019). The mRNA-ribosome complex 405 was precipitated using a polyclonal HA -antibody (Sigma #H6908) and Dynabeads Protein G (Life 406 Technologies #10004D) as previously described. Ribosome -bound mRNA was isolated with RNeasy 407 Plus Micro Kit (Qiagen #74034) per manufacturer instructions (not precipitated fraction was used as 408 “input” spinal cord control). Prior sequencing, RNA quantity and integrity (RIN > 8.5) was controlled 409 with a Bioanalyzer (Agilent RNA 6000 Nano) and rRNA was depleted . Motor neuron mRNA was 410 sequenced in pair -end using an Illumina HiSeq4000 Kit, at a depth of ~40 million reads per sample 411 (Institute of Neurogenomics, Juliane Winkelmann, Helmholtz Munich, Germany). The raw sequencing 412 data (fastq files) were aligned to the mm9 mouse genome and read counts were extracted with HTSeq-413 count software with option “intersectionStrict”. Lowly expressed RNAs (< 10 re ad counts were 414 excluded and differential gene expression analysis was performed with DESeq2 software package 75. 415 A likelihood ratio test was performed on the timeline analysis (P5, 7, 9, 11, 14). Graphs were generated 416 using ggplot2 from custom ‘R’ statistical software scripts 76. 417 Reverse Transcription (RT)-qPCR 418 All following steps were conducted on ice. cDNA was obtained according to manufacturer’s 419 instruction. Briefly, 100 ng ‘ribotagged’ motor neurons mRNA, 100 µM random hexamer primers 420 (Roche # 11034731001) and 1 μl RNAsin Plus Inhibitor (Promega #N2615) were dissolved with water 421 in 14 μl total volume. The mixture was incubated for 5 min utes at 70 °C, and 10 min utes on ice. 5 μl 422 of M-MLV Reverse Transcriptase Buffer (Promega #M531A), 10 mM NTPs mixture and 200 U of MMLV 423 (Promega #M170A) were added and the mixture incubated for 1.5 h at 37 °C. cDNA was purified with 424 a QIAEX II Gel Extraction Kit (Qiagen #20021) as follows: The obtained cDNA was mixed with 2 μl QiaEX-425 II suspension (Silica-Matrix) and 80 μl QX-I buffer, incubated for 20 minutes at 25 °C at 1,000 rpm and 426 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 15 centrifuged at 13,000 x g for 2 min utes. The supernatant was discarded and the matrix was washed 427 twice with 90 μl PE buffer and centrifuged at 13,000 x g for 2 minutes. The supernatant was removed 428 and the pellet dried for 10 min utes with lid open at 300 rpm shaking at 25 °C. The cDNA was elut ed 429 with 20 μl elution buffer (1 mM Tris, pH 8.5), and incubated at 25 °C with 1 ,000 rpm for 430 homogenization. After centrifugation at 13,000 x g for 2 minutes the supernatant containing the clean 431 RNA was collected. 432 All qPCR reactions were performed with a LightCyler 1.3 Real-Time PCR system (Roche S/N: 140 6143). 433 The MgCl2 concentration (1-3 mM) and the annealing temperature were optimized for each primer 434 pair and confirmed through primer efficiency curves 40. Primer sequence were (5’ to 3’): Chat: 435 TTCTAGCTGTGAGGAGGTGC and CCCAAACCGCTTCACAATGG, GFAP: TCGCACTCAATACGAGGCAG and 436 TTGGCGGCGATAGTCGTTAG. For each reaction, a final concentration of 1 ng/μl cDNA was mixed with 437 2 μl FastStart DNA Master SYBR Green I (Roche #03 003 230 001), 0.5 μM forward and reverse primer, 438 1-3 mM MgCl2 and water to a total volume of 20 µl. Samples were run at least in triplicates. For each 439 primer pair, controls omitting either template or reverse transcriptase were included. 440 Immunofluorescence 441 The whole thorax was fixed in 4 % PFA for 1 h in 0.1 M phosphate buffer (PB) on ice and the triangularis 442 muscle was dissected and extracted as described previously 20, 31, 36, 49, 77. Prior to application of primary 443 antibodies, muscles were incubated in 5 % CHAPS (Carl Roth #75621-03-3) in 0.1 M PB for 1h at 37 °C. 444 Primary antibodies (see details below) were diluted in blocking solution (5 % BSA (Sigma #A7030, 0.5 445 % Triton-X in 0.1 M PB or 3 % BSA, 0.5 % Triton -X and 12 % NGS (Abcam #ab7481) in 0.1 M PB) and 446 incubated at 4°C overnight for postnatal muscles, 3 days when adult tissue or anti-acetylated tubulin 447 antibody was used. The following primary antibodies wer e used in this study: anti -tubulin beta-3 448 conjugated to Alexa Fluor 488 (BioLegend; AB_2562669; mouse IgG2a, 1:200), Alexa Fluor 555 (BD 449 PharMingen; #560339; mo use monoclonal, 1:200), or Alexa Fluor 647 (BioLegend; AB_2563609; 450 mouse IgG2a, 1:200), anti - GT335 (Adipogen, Mouse IgG1, 1: 200), anti -polyE (Adipogen, AG-25B-451 0030, Rabbit IgG, 1:1 ,000), anti -acetylated-tubulin (Abcam, 611B1, mouse IgG2b, 1:1 ,000), anti -452 neurofilament heavy polypetide (Abcam, ab4680, chicken, 1:500), anti-βmonoE (rabbit, IgG, gift from 453 Carsten Janke, 1:500). Muscles were washed in 0.1 M PB, incubated for 1 h at room temperature with 454 corresponding secondary antibodies coupled to Alexa Fluo r 488, Alexa Fluor 594, or Alexa Fluor 647 455 (Invitrogen; rabbit: #A-11070, #A-11072, #A-21246, #A-32790; mouse: #A-11005; chicken: #A-11042; 456 #A- 21449; goat: #A -11058) and washed again in 0.1 M PB. Muscles or sections were mounted in 457 Vectashield (Vector Laboratories) or Prolong-Glass (Thermo Fisher). Image stacks were recorded at a 458 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 16 confocal microscope (Olympus; FV1000 or FV3000) equipped with ×20/0.8 NA and ×60/1.42 NA oil -459 immersion objectives (Olympus). 460 For immunostaining of spinal cord sections ( Fig. 1), mice were transcardially perfused with PBS and 461 ice-cold 4 % PFA. Spinal cords were dissected and post -fixed overnight with 4 % PFA. 60 µm thick 462 vibratome sections were then immunostained with antibodies against Choline Acetyl -Transferase 463 (ChAT) (anti-ChAT, Novus Biologicals, goat, NBP1‐30052, 1:10) and HA (anti-HA, Sigma-Aldrich, H6908, 464 1:50), for 2 days over night. Spinal cord sections were washed in 1 x PBS, incubated for 1 h at room 465 temperature with corresponding secondary antibodies coupled to Alexa Fluor 488 and Alexa Fluor 594 466 (see above). The antibodies for spinal cord staining were diluted in blocking solution prepared with 5 467 % Normal Donkey Serum (Millipore #S30-M) and 0.5 % Triton-X in 1 x PBS. 468 Immunostaining of brain sections (Fig. 4) was carried out as described previously 40. Briefly, mice were 469 transcardially perfused with PBS and ice-cold 4 % PFA in PBS and brains were post-fixed in 4 % PFA in 470 PBS overnight, embedded in paraffin blocks and cut at 7 µm thick coronal sections. Prior staining, 471 antigen retrieval (Citrate-Based Antigen Unmasking Solution, Vector Laboratories) and blocking (1 % 472 BSA, 0.2 % Tween20) was performed. For IPB, visualization sections were incubated overnight at 4 °C 473 with anti-calbindin D28K antibody (Swant #CB-38a, 1:300) diluted in blocking solution. Subsequently, 474 samples were washed three times with PBS with 0.2 % Tween 20 and incubated for 2 hours at room 475 temperature with fluorescent secondary antibody coupled to Alexa Fluor 594 or Alexa Fluor 488 (Life 476 Technologies, #A-11037; A-11008) in blocking solution, washed three times with PBS with 0.2 % Tween 477 20 and embedded in Mowiol with 1 µg/ml Hoechst. Images were recording using a confocal 478 microscope (Leica SP8 and Leica Stellaris). 479 DiOlistics labeling of hippocampal granule cells and dendritic spine density analysis 480 For visualization of dendritic spines in TTLL1 KO brains a DiOlistic approach was used as described 481 previously 40. In brief, mice were perfused with 20 ml of 4 % PFA/PBS, and brains were isolated and 482 post-fixed in 4 % PFA/PBS for 30 minutes. Then, brains were washed in PBS for 30 minutes and 483 incubated in 15 % sucrose for 30 minutes , followed by 30 minutes in 30 % sucrose. 250 µm thick 484 coronal slices were cut in vibratome, washed in PBS and incubated for 5 minutes in 15 % sucrose and 485 subsequently in 30 % sucrose. The solution was removed and DiI was introduced into the brain slices 486 by DiI -labeled tungsten particles (prepared as previou sly described 40 with the use of a Gene Gun 487 helium‐powered system (Bio‐Rad) and 120 Psi pressure. After the labelling, slices were washed in PBS 488 to remove residual tungsten particles and kept for 30 minutes in PBS in the dark to let the dye diffuse. 489 Slices were mounted onto a glass slice in 0.5 % n‐propyl gallate/ 90 % glycerol/PBS (NPG) and the next 490 day imaged by a confocal microscope ( Leica Stellaris). Analysis of the dendritic spine density was 491 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 17 performed in 3 different TTLL1WT and TTLL1KO brains (at least 10 labelled gr anule cell dendrites were 492 analyzed per brain). 493 Virus production 494 AAV9-hSyn-iCre was produced as outlined previously 78. In brief, HEK293 -T cells were seede d in 10 -495 tray Cell Factories (Thermo Fisher Scientific) 24 hours before transfection, allowing the cells to reach 496 80-90 % confluence. Subsequently, 420 μg of the vector plasmid dsAAV -hSyn-iCre and 1.5 mg of the 497 helper plasmid pDP9rs, generously provided by Roger Hajjar (Phospholamban Foundation, 498 Amsterdam, The Netherlands), were introduced into the cells by polyethyleneimine -mediated 499 transfection (Polysciences, 24765 Warrington, PA). Cells were harvested after 72 hours, followed by 500 lysis and benzonase treatm ent. To purify the AAVs, ultracentrifugation on an iodixanol density 501 gradient (Progen, OptiPrep) was applied. Buffer exchange from iodixanol to Ringer's lactate buffer was 502 then carried out using Vivaspin 20 columns (Sartorius, VS2042, Göttingen, Germany). To obtain high 503 virus titers, the virus content of two 10 -tray Cell Factories was pooled and concentrated. Real -time 504 qPCR with SYBR Green Master Mix (Roche) was performed to evaluate AAV9 titers. 505 Neonatal AAV9 or α-BTX injections 506 Viral vectors were injected into neonatal pups according to previously published protocols 20, 49, 79. In 507 short, P3 pups were briefly anesthetize d with isofluorane (Abbott) and injected with 3 μl of AAV9 -508 hSyn-iCre (titer 1 x 10 13 to 2 x 10 14) into the right lateral ventricle using a nanoliter injector (World 509 Precision Instruments; Micro4 MicroSyringe Pump Controller connected with Nanoliter 2000) 510 attached to a fine glass pipette (Drummond; 3.5”, #3 -000-203-G/X) at a rate of 30 nl/s. The injection 511 was guided by ultrasound (Visualsonics, Vevo® 2100). 0.05 % (wt/vol) trypan blue was added to the 512 viral solution for visualizing the filling of the injected ventricles. Whole litters were injected, and pups 513 were allowed to recover on a heating mat before the litter was returned to their mother into the home 514 cage and sacrificed on P9 for experiments. 515 For analysis, Cre-mediated deletion in conditional CCP1 and Spast knockout (CCP1flox/flox 72 X Spastflox/flox 516 20) was verified by a robust expression of the TdTomato reporter allele 74 (homozygous) in motor 517 neurons of the triangularis sterni muscle. According to TdTomato fluorescence intensity in axons, 518 recombination was assumed, when fl uorescence was in the upper quan tile and negative when they 519 belonged to the lowest quantile. All other axons were excluded from analysis. 520 To block neurotransmission in the triangularis ste rni muscle of C57BL/6N (Charles River, Strain Code 521 027) or SpastKO 20 mice, 1 µl of 50 µg/µl α-BTX conjugated to Alexa Fluor 594 (Invitrogen; B13423) was 522 injected with a needle unilaterally in the thorax on P7, as previously described 49. In some cases, a 523 post-hoc stain of the triangularis sterni muscle with anti -Neurofilament (see above) or α -BTX Alexa 524 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 18 Fluor 488 or 594 further verified the degree of labeling with injected α -BTX conjugated Alexa Fluor 525 594 and absence of denervation (> 100 neuromuscular junctions ( NMJs) per mouse, n = 3 mice). 526 Immunostaining and quantification were conducted as described above with anti-tubulin beta-3 and 527 anti-polyE antibodies. 528 Live imaging of EB3 comet densities in nerve-muscle explants 529 EB3 comet imaging was performed on acute nerve -muscle explants as previously described 20, 38. In 530 brief, the thorax of euthanized mice was obtained by removing the skin over the rib cage, severing the 531 ribs close to the spinal column and dissecting the diaphragm. Further dissections were done under 532 oxygenated, ice-cold Ringer’s solution (in mM: 125 NaCl, 2.5 KCl, 1.25 NaH 2PO4, 26 NaHCO3, 2 CaCl2, 533 1 MgCl 2, and 20 glu cose, oxygenated with 95 % O 2/ 5 % CO 2) to remove thymus, pleura, lung and 534 pectoral muscles over the rib cage. The explant was fixed with insect pins (Fine Science Tools; 26001–535 25, 0.25 mm) on a Sylgard-coated 3.5 cm petri dish, with the inside of the thorax facing the objective. 536 Imaging was performed under continuous and steady perfusion with warmed oxygenated Ringer’s 537 solution and kept at physiological temperatures 33-36 °C by a heated stage connected to an automatic 538 temperature controller (Warner Instruments; TC-344C). Total imaging time on explants did not exceed 539 2 h ours. Live imaging was carried out with an epifluorescence microscope ( Olympus BX51WI ) 540 equipped with ×20/0.5 NA and ×100/1.0 NA water -immersion objectives, an automated filter wheel 541 (Sutter I nstruments; Lambda 10 –3), a charge -coupled device camera (Visitron Systems; CoolSnap 542 HQ2), controlled by μManager version 1.4 80. A total of 200 frames were acquired per movie, at a 543 frequency of 0.5 Hz and an expos ure time of 500 ms using a YFP filter s et (F36 -528; AHF 544 Analysentechnik). 545 Data analysis 546 ImageJ/FiJi 81 (http://fiji.sc) was used to determine pruning speed, by counting the number of 547 innervating terminal branches ending on each α -BTX-stained neuromuscular junction . For 548 quantification of immunostainings three background subtracted areas (ROIs; measured for mean gray 549 value) were averaged per single axon. Each ROI was obtained from a single optical section, as 550 previously described 20, 49. IPB length was quantified using the ratio of IPB length to the length of the 551 CA3 as described previously 82. To quantify EB3 comet parameters, out-of-focus frames were deleted 552 manually from movies a nd aligned with the TurboReg plugin with the parameter Rigid -body and 553 “Quality” set on “Accurate” and comet trajectories were manually analyzed using the MTrackJ-plugin 554 (developed by E. Meijering, Biomedical Imaging Group, Erasmus Medical Center, Rotterdam ). Only 555 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 19 EB3 comets that appeared for at least three consecutive frames were considered. Neurolucida 556 software (MBF Bioscience) was used to perform dendritic spine segmentation. 557 For image representation, maximum intensity projections were generated from confocal image stacks 558 with ImageJ/Fiji 81 and then further processed in Adobe Photoshop. All analyses were performed with 559 the experimenter blinded to the treatment or genotypes during imaging and scoring. 560 Statistics 561 Statistical analysis and visualization were performed using ‘R’ 76 or OriginLab software (origin.lab). All 562 experiments included at least 3 biological replicates. Normality was tested and outliers were removed 563 using the Grubbs test. A two -tailed Students t-test was performed if data passed the normality test, 564 otherwise Mann -Whitney test was used for non -parametric data. P < 0.05, indicated as *, was 565 considered significant. P < 0.01 is **, P < 0.001 is ***, and P < 0.0001 is ****. Bars – on the left – show 566 mean + SEM. Violin plots – on the right – depict data distribution, scaled based on constant maximum 567 width (graphs are cut at 90 % level of the dominant violin). 568

569 1. Burute, M. & Kapitein, L.C. Cellular Logistics: Unraveling the Interplay Between 570 Microtubule Organization and Intracellular Transport. Annu Rev Cell Dev Biol 35, 29-571 54 (2019). 572 2. Gudimchuk, N.B. & McIntosh, J.R. Regulation of microtubule dynamics, mechanics 573 and function through the growing tip. Nature reviews. Molecular cell biology 22, 777-574 795 (2021). 575 3. McKenna, E.D., Sarbanes, S.L., Cummings, S.W. & Roll-Mecak, A. The Tubulin Code, 576 from Molecules to Health and Disease. Annu Rev Cell Dev Biol 39, 331-361 (2023). 577 4. Janke, C. & Magiera, M.M. The tubulin code and its role in controlling microtubule 578 properties and functions. Nature reviews. Molecular cell biology 21, 307-326 (2020). 579 5. Moutin, M.J., Bosc, C., Peris, L. & Andrieux, A. Tubulin post-translational 580 modifications control neuronal development and functions. Dev Neurobiol 81, 253-581 272 (2021). 582 6. Matamoros, A.J. & Baas, P.W. Microtubules in health and degenerative disease of 583 the nervous system. Brain Res Bull 126, 217-225 (2016). 584 7. Magiera, M.M., Singh, P., Gadadhar, S. & Janke, C. Tubulin Posttranslational 585 Modifications and Emerging Links to Human Disease. Cell 173, 1323-1327 (2018). 586 8. Blackstone, C. Cellular pathways of hereditary spastic paraplegia. Annu Rev Neurosci 587 35, 25-47 (2012). 588 9. Sferra, A., Nicita, F. & Bertini, E. Microtubule Dysfunction: A Common Feature of 589 Neurodegenerative Diseases. Int J Mol Sci 21 (2020). 590 10. Lasser, M., Tiber, J. & Lowery, L.A. The Role of the Microtubule Cytoskeleton in 591 Neurodevelopmental Disorders. Front Cell Neurosci 12, 165 (2018). 592 11. Magiera, M.M., Singh, P. & Janke, C. SnapShot: Functions of Tubulin Posttranslational 593 Modifications. Cell 173, 1552-1552 e1551 (2018). 594 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 20 12. Al-Chalabi, A. et al. The genetics and neuropathology of amyotrophic lateral 595 sclerosis. Acta Neuropathol 124, 339-352 (2012). 596 13. Bodakuntla, S. et al. Distinct roles of alpha- and beta-tubulin polyglutamylation in 597 controlling axonal transport and in neurodegeneration. EMBO J 40, e108498 (2021). 598 14. Magiera, M.M. et al. Excessive tubulin polyglutamylation causes neurodegeneration 599 and perturbs neuronal transport. EMBO J 37 (2018). 600 15. Ziak, J. et al. Microtubule-binding protein MAP1B regulates interstitial axon 601 branching of cortical neurons via the tubulin tyrosination cycle. EMBO J (2024). 602 16. Atkins, M., Nicol, X. & Fassier, C. Microtubule remodelling as a driving force of axon 603 guidance and pruning. Semin Cell Dev Biol 140, 35-53 (2023). 604 17. Rumpf, S., Wolterhoff, N. & Herzmann, S. Functions of Microtubule Disassembly 605 during Neurite Pruning. Trends Cell Biol 29, 291-297 (2019). 606 18. Rolls, M.M., Thyagarajan, P. & Feng, C. Microtubule dynamics in healthy and injured 607 neurons. Dev Neurobiol 81, 321-332 (2021). 608 19. Riccomagno, M.M. & Kolodkin, A.L. Sculpting neural circuits by axon and dendrite 609 pruning. Annu Rev Cell Dev Biol 31, 779-805 (2015). 610 20. Brill, M.S. et al. Branch-Specific Microtubule Destabilization Mediates Axon Branch 611 Loss during Neuromuscular Synapse Elimination. Neuron 92, 845-856 (2016). 612 21. Luo, L. & O'Leary, D.D. Axon retraction and degeneration in development and 613 disease. Annu Rev Neurosci 28, 127-156 (2005). 614 22. Wilton, D.K., Dissing-Olesen, L. & Stevens, B. Neuron-Glia Signaling in Synapse 615 Elimination. Annu Rev Neurosci 42, 107-127 (2019). 616 23. Eroglu, C. & Barres, B.A. Regulation of synaptic connectivity by glia. Nature 468, 223-617 231 (2010). 618 24. Wolff, A. et al. Distribution of glutamylated alpha and beta-tubulin in mouse tissues 619 using a specific monoclonal antibody, GT335. Eur J Cell Biol 59, 425-432 (1992). 620 25. Boucher, D., Larcher, J.C., Gros, F. & Denoulet, P. Polyglutamylation of tubulin as a 621 progressive regulator of in vitro interactions between the microtubule-associated 622 protein Tau and tubulin. Biochemistry 33, 12471-12477 (1994). 623 26. Larcher, J.C., Boucher, D., Lazereg, S., Gros, F. & Denoulet, P. Interaction of kinesin 624 motor domains with alpha- and beta-tubulin subunits at a tau-independent binding 625 site. Regulation by polyglutamylation. J Biol Chem 271, 22117-22124 (1996). 626 27. Valenstein, M.L. & Roll-Mecak, A. Graded Control of Microtubule Severing by Tubulin 627 Glutamylation. Cell 164, 911-921 (2016). 628 28. Sanes, J.R. & Lichtman, J.W. Development of the vertebrate neuromuscular junction. 629 Annu Rev Neurosci 22, 389-442 (1999). 630 29. Bishop, D.L., Misgeld, T., Walsh, M.K., Gan, W.B. & Lichtman, J.W. Axon branch 631 removal at developing synapses by axosome shedding. Neuron 44, 651-661 (2004). 632 30. Walsh, M.K. & Lichtman, J.W. In vivo time-lapse imaging of synaptic takeover 633 associated with naturally occurring synapse elimination. Neuron 37, 67-73 (2003). 634 31. Brill, M.S., Lichtman, J.W., Thompson, W., Zuo, Y. & Misgeld, T. Spatial constraints 635 dictate glial territories at murine neuromuscular junctions. J Cell Biol 195, 293-305 636 (2011). 637 32. Rossi, J. et al. Melanocortin-4 receptors expressed by cholinergic neurons regulate 638 energy balance and glucose homeostasis. Cell Metab 13, 195-204 (2011). 639 33. Sanz, E. et al. Cell-type-specific isolation of ribosome-associated mRNA from complex 640 tissues. Proc Natl Acad Sci U S A 106, 13939-13944 (2009). 641 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 21 34. van Dijk, J. et al. A targeted multienzyme mechanism for selective microtubule 642 polyglutamylation. Mol Cell 26, 437-448 (2007). 643 35. Tort, O. et al. The cytosolic carboxypeptidases CCP2 and CCP3 catalyze 644 posttranslational removal of acidic amino acids. Molecular Biology of the Cell 25, 645 3017-3027 (2014). 646 36. Kerschensteiner, M., Reuter, M.S., Lichtman, J.W. & Misgeld, T. Ex vivo imaging of 647 motor axon dynamics in murine triangularis sterni explants. Nat Protoc 3, 1645-1653 648 (2008). 649 37. Bompard, G. et al. CSAP Acts as a Regulator of TTLL-Mediated Microtubule 650 Glutamylation. Cell Rep 25, 2866-2877 e2865 (2018). 651 38. Kleele, T. et al. An assay to image neuronal microtubule dynamics in mice. Nature 652 communications 5, 4827 (2014). 653 39. Chen, J. & Roll-Mecak, A. Glutamylation is a negative regulator of microtubule 654 growth. Mol Biol Cell 34, ar70 (2023). 655 40. Ziak, J. et al. CRMP2 mediates Sema3F-dependent axon pruning and dendritic spine 656 remodeling. EMBO Rep 21, e48512 (2020). 657 41. Hausrat, T.J. et al. Disruption of tubulin-alpha4a polyglutamylation prevents 658 aggregation of hyper-phosphorylated tau and microglia activation in mice. Nature 659 communications 13, 4192 (2022). 660 42. Redeker, V., Rossier, J. & Frankfurter, A. Posttranslational modifications of the C-661 terminus of alpha-tubulin in adult rat brain: alpha 4 is glutamylated at two residues. 662 Biochemistry 37, 14838-14844 (1998). 663 43. Garnham, C.P. & Roll-Mecak, A. The chemical complexity of cellular microtubules: 664 tubulin post-translational modification enzymes and their roles in tuning 665 microtubule functions. Cytoskeleton (Hoboken) 69, 442-463 (2012). 666 44. Ikegami, K. & Setou, M. Unique post-translational modifications in specialized 667 microtubule architecture. Cell Struct Funct 35, 15-22 (2010). 668 45. Rogowski, K. et al. A family of protein-deglutamylating enzymes associated with 669 neurodegeneration. Cell 143, 564-578 (2010). 670 46. Mahalingan, K.K. et al. Structural basis for polyglutamate chain initiation and 671 elongation by TTLL family enzymes. Nat Struct Mol Biol 27, 802-813 (2020). 672 47. Kann, M.L., Soues, S., Levilliers, N. & Fouquet, J.P. Glutamylated tubulin: diversity of 673 expression and distribution of isoforms. Cell Motil Cytoskeleton 55, 14-25 (2003). 674 48. Maas, C. et al. Synaptic activation modifies microtubules underlying transport of 675 postsynaptic cargo. Proc Natl Acad Sci U S A 106, 8731-8736 (2009). 676 49. Wang, M. et al. Completion of neuronal remodeling prompts myelination along 677 developing motor axon branches. J Cell Biol 220 (2021). 678 50. Schuldiner, O. & Yaron, A. Mechanisms of developmental neurite pruning. Cell Mol 679 Life Sci 72, 101-119 (2015). 680 51. Tapia, J.C. et al. Pervasive synaptic branch removal in the mammalian neuromuscular 681 system at birth. Neuron 74, 816-829 (2012). 682 52. Roll-Mecak, A. The Tubulin Code in Microtubule Dynamics and Information Encoding. 683 Dev Cell 54, 7-20 (2020). 684 53. Edde, B. et al. Posttranslational glutamylation of alpha-tubulin. Science 247, 83-85 685 (1990). 686 54. Genova, M. et al. Tubulin polyglutamylation differentially regulates microtubule-687 interacting proteins. EMBO J 42, e112101 (2023). 688 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 22 55. Bodakuntla, S. et al. Tubulin polyglutamylation is a general traffic-control mechanism 689 in hippocampal neurons. J Cell Sci 133 (2020). 690 56. Gilmore-Hall, S. et al. CCP1 promotes mitochondrial fusion and motility to prevent 691 Purkinje cell neuron loss in pcd mice. J Cell Biol 218, 206-219 (2019). 692 57. Ikegami, K. et al. TTLL7 is a mammalian beta-tubulin polyglutamylase required for 693 growth of MAP2-positive neurites. J Biol Chem 281, 30707-30716 (2006). 694 58. Devambez, I. et al. Identification of DmTTLL5 as a Major Tubulin Glutamylase in the 695 Drosophila Nervous System. Sci Rep 7, 16254 (2017). 696 59. Lacroix, B. et al. Tubulin polyglutamylation stimulates spastin-mediated microtubule 697 severing. J Cell Biol 189, 945-954 (2010). 698 60. Hoopfer, E.D. et al. Wlds protection distinguishes axon degeneration following injury 699 from naturally occurring developmental pruning. Neuron 50, 883-895 (2006). 700 61. Alvarez, V.A. & Sabatini, B.L. Anatomical and physiological plasticity of dendritic 701 spines. Annu Rev Neurosci 30, 79-97 (2007). 702 62. Hotulainen, P. & Hoogenraad, C.C. Actin in dendritic spines: connecting dynamics to 703 function. J Cell Biol 189, 619-629 (2010). 704 63. Luo, L. Actin cytoskeleton regulation in neuronal morphogenesis and structural 705 plasticity. Annu Rev Cell Dev Biol 18, 601-635 (2002). 706 64. Coles, C.H. & Bradke, F. Coordinating neuronal actin-microtubule dynamics. Curr Biol 707 25, R677-691 (2015). 708 65. Xu, K., Zhong, G. & Zhuang, X. Actin, spectrin, and associated proteins form a 709 periodic cytoskeletal structure in axons. Science 339, 452-456 (2013). 710 66. Datar, A. et al. The Roles of Microtubules and Membrane Tension in Axonal Beading, 711 Retraction, and Atrophy. Biophys J 117, 880-891 (2019). 712 67. Je, H.S. et al. ProBDNF and mature BDNF as punishment and reward signals for 713 synapse elimination at mouse neuromuscular junctions. J Neurosci 33, 9957-9962 714 (2013). 715 68. Li, S. et al. Spastin Interacts with CRMP2 to Regulate Neurite Outgrowth by 716 Controlling Microtubule Dynamics through Phosphorylation Modifications. CNS 717 Neurol Disord Drug Targets 20, 249-265 (2021). 718 69. Melo-Hanchuk, T.D. & Kobarg, J. Polyglutamylase activity of tubulin tyrosine ligase-719 like 4 is negatively regulated by the never in mitosis gene A family kinase never in 720 mitosis gene A -related kinase 5. World J Biol Chem 12, 38-51 (2021). 721 70. Shashi, V. et al. Loss of tubulin deglutamylase CCP1 causes infantile-onset 722 neurodegeneration. EMBO J 37 (2018). 723 71. Zempel, H. et al. Amyloid-beta oligomers induce synaptic damage via Tau-dependent 724 microtubule severing by TTLL6 and spastin. EMBO J 32, 2920-2937 (2013). 725 72. Munoz-Castaneda, R. et al. Cytoskeleton stability is essential for the integrity of the 726 cerebellum and its motor- and affective-related behaviors. Sci Rep 8, 3072 (2018). 727 73. Feng, G. et al. Imaging neuronal subsets in transgenic mice expressing multiple 728 spectral variants of GFP. Neuron 28, 41-51 (2000). 729 74. Madisen, L. et al. A robust and high-throughput Cre reporting and characterization 730 system for the whole mouse brain. Nat Neurosci 13, 133-140 (2010). 731 75. Love, M.I., Huber, W. & Anders, S. Moderated estimation of fold change and 732 dispersion for RNA-seq data with DESeq2. Genome Biol 15, 550 (2014). 733 76. Team, R.C. R: A language and environment for statistical computing. R Foundation 734 for Statistical Computing, Vienna, Austria. (2021). 735 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint 23 77. Brill, M.S., Marinkovic, P. & Misgeld, T. Sequential photo-bleaching to delineate 736 single Schwann cells at the neuromuscular junction. J Vis Exp, e4460 (2013). 737 78. Kaltenbacher, T. et al. CRISPR somatic genome engineering and cancer modeling in 738 the mouse pancreas and liver. Nat Protoc 17, 1142-1188 (2022). 739 79. Wang, M., Misgeld, T. & Brill, M.S. Neural labeling and manipulation by neonatal 740 intraventricular viral injection in mice. STAR Protoc 3, 101081 (2022). 741 80. Edelstein, A.D. et al. Advanced methods of microscope control using muManager 742 software. J Biol Methods 1 (2014). 743 81. Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. Nat 744

9, 676-682 (2012). 745 82. Bagri, A., Cheng, H.J., Yaron, A., Pleasure, S.J. & Tessier-Lavigne, M. Stereotyped 746 pruning of long hippocampal axon branches triggered by retraction inducers of the 747 semaphorin family. Cell 113, 285-299 (2003). 748 749 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted March 12, 2024. ; https://doi.org/10.1101/2024.03.11.584412doi: bioRxiv preprint