LUZP1, a novel regulator of primary cilia and the actin cytoskeleton, is altered in Townes-Brocks Syndrome

Primary cilia are sensory organelles that are crucial for cell signaling during development and organ homeostasis. Cilia arise from the centrosome and their formation is governed by numerous regulatory factors. We show that the leucine-zipper protein LUZP1 localizes to the pericentriolar material and actin cytoskeleton. Using TurboID proximity labeling and pulldowns, LUZP1 associates with factors linked to centrosome and actin filaments. Loss of LUZP1 reduces F-actin levels, facilitating ciliogenesis and altering Sonic Hedgehog signaling, pointing to a key role in the cytoskeleton-cilia interdependency. Moreover, we show that LUZP1 interacts with a truncated form of the transcription factor SALL1 that causes Townes-Brocks Syndrome. TBS is characterized by digit, heart and kidney malformations and is linked in part to defective cilia. Truncated SALL1 increases the ubiquitin proteasome-mediated degradation of LUZP1. Alteration of LUZP1 levels may be a contributing factor to TBS, suggesting possible therapies using modulators of cilia and cytoskeletal function.

To gain some clues into the function of LUZP1, we sought to identify its 142 proximal interactome using the TurboID approach (33). We used hTERT-RPE1 cells 143 stably expressing low levels of FLAG-TurboID-LUZP1, and after a brief biotin-144 labeling, biotinylated proteins were captured for analysis by liquid chromatography 145 tandem mass spectrometry (LC-MS/MS). 311 high-confidence proximity LUZP1 146 interactors were identified in at least two replicates (Table S1). With the purpose of 147 obtaining a functional overview of the main pathways associated to LUZP1, a 148 comparative Gene Ontology (GO) analysis was performed with all the hits ( Figure 1C-149 E and Table S1). In the Cellular Component domain, "cytoplasm", "actin cytoskeleton", 150 "centrosome", "midbody", "cell junction" and "vesicle" terms were highlighted ( Figure  151 1C and Table S1). In the category of Biological Process, LUZP1 proteome shows 152 enrichment in the "cytoskeleton organization", vesicle-mediated transport and cell 153 adhesion categories among others ( Figure 1D and Table S1). With respect to Molecular 154 Function, LUZP1 also showed enrichment in cytoskeleton-related proteins ("structural 155 component of cytoskeleton" and "actin binding" terms; Figure 1E and Table S1). 64 or 156 138 of the verified or potential, respectively, centrosome/cilia gene products previously 157  Figure 2C). Next, to compare LUZP1 with additional centriolar markers, we 171 labelled U2OS cells expressing YFP-LUZP1 with the distal centriolar markers CCP110 172 and ODF2 (outer dense fiber of sperm tails 2). We did not observe colocalization of 173 LUZP1 with these markers, indicating that LUZP1 is likely found at the proximal end 174 of both centrioles ( Figure 2D). We further imaged LUZP1 along with PCM1 and 175 centrobin, markers of PCM and of the MC, respectively. Interestingly, we observed 176 LUZP1 being surrounded by PCM1 ( Figure 2E), while LUZP1 surrounded centrobin 177 at the MC ( Figure 2G). The profile histograms confirm that LUZP1 localizes between 178 PCM1 and centrobin ( Figure 2F and 2H, respectively), suggesting that LUZP1 might 179 be a novel PCM associated-protein, forming a basket around the proximal end of both 180 centrioles. We also examined LUZP1 localization in the centrosome in synchronized 181 human RPE1-cells. LUZP1 was reduced at the centrosome during G2/M and G0 phases 182   inhibition during G0 led to increased LUZP1, suggesting that active degradation occurs 218 in G0 arrested-RPE1 cells. Intriguingly, when LUZP1 levels were examined in TBS 275 219 cells, a reduction in both actin-associated LUZP1 and phalloidin-labelled stress fibers 220 was observed when compared to control cells ( Figures 4A-C). These results indicate 221 that actin cytoskeleton might be altered in TBS cells. By pulldown assays, we 222 confirmed that LUZP1-YFP interacts with both actin and FLNA ( Figure 4D and Figure  223 4-figure supplement 1). Of note, actin, FLNA and other stress fibers-associated proteins 224 were also found to be associated with LUZP1 by proximity labeling and mass 225 spectrometry (Table S1) These results reflect that the integrity of the actin cytoskeleton may influence the 232 solubility but not the stability of LUZP1. 233

LUZP1 plays a role in primary cilia formation and F-actin stabilization 235
Based on the LUZP1 localization at the centrosome, its interaction with 236 centrosomal proteins and the defects in ciliogenesis previously observed in TBS cells 237 (18), we hypothesised that LUZP1 might have a role in cilia formation. To examine 238 this, we analyze ciliogenesis in Shh-LIGHT2 cells, a cell line derived from 239 immortalized mouse NIH3T3 fibroblasts that display primary cilia and carry a Shh 240 luciferase reporter (herein considered as WT fibroblasts) (37). Additionally, using 241 CRISPR/Cas9 gene editing directed to exon 1 of murine Luzp1, we generated Shh-242 LIGHT2 mouse embryonic fibroblasts null for Luzp1 (Luzp1 -/cells), and for genetic 243 rescue experiments, LUZP1 was restored to these cells by the expression of human 244 LUZP1-YFP fusion (+LUZP1 cells). We examined LUZP1 localization associated with cells were plated at equal densities and induced either to ciliate for 48 hours by serum 249 withdrawal (starved), or to reabsorb their cilia by serum replenishment for 4 hours 250 (refed) ( Figure 5A). We quantified ciliation rates and primary cilia length at the 251 mentioned timepoints. Luzp1 -/fibroblasts displayed higher ciliation rate (60%) than 252 WT (10.5%) and +LUZP1 (22.2%) when the cells were not subjected to starvation 253 ( Figure 5B). However, Luzp1 -/cells were not significantly more ciliated than WT or 254 +LUZP1 fibroblasts upon 48 hours of starvation or 4 hours after inducing cilia 255 disassembly ( Figure 5B). In addition, primary cilia in Luzp1 -/cells were significantly 256 longer than in non-starved WT cycling cells ( Figure 5A Based on the LUZP1 localization to the actin cytoskeleton and that a reduction 278 in LUZP1 was accompanied by a diminishment in F-actin levels in TBS 275 cells, we 279 hypothesised that LUZP1 might affect F-actin levels. First, we observed a reduction in 280 F-actin (labelled by phalloidin) in the Luzp1 -/cells compared to WT, which was 281 recovered in +LUZP1 cells ( Figure 5F). Furthermore, LUZP1 levels and actin filaments 282 were diminished in WT fibroblasts upon starvation ( Figure 5G  Upon induction by purmorphamine for 24 hours, WT cells increased significantly the 296 expression of both targets, while Luzp1 -/cells did not, indicating that Luzp1 -/cells fail 297 to induce Shh signaling. To further study the role of LUZP1 in Shh signaling, we 298 analyzed GLI3 processing by Western blot using total lysates extracted from WT vs 299 Luzp1 -/cells. Without purmorphamine induction, we found a significantly higher ratio 300 of GLI3 activating form vs GLI3 repressive form (GLI3-A:GLI3-R) in Luzp1 -/cells 301 compared to WT (2.9 fold increase in Luzp1 -/cells vs WT) ( Figure 6C and Figure 6-302 figure supplement 1). After induction, the values were similar for Luzp1 -/and WT cells. 303 Since the Luzp1 -/parental line was Shh-LIGHT2, we also examined the effects of 304 lacking Luzp1 on Shh signaling by measuring the activity of a GLI-responsive Firefly 305 luciferase reporter ( Figure 6D). Prior to purmorphamine treatment, Luzp1 -/cells 306 showed higher Shh activity compared to control or +LUZP1 cells, as observed in TBS-307 Our results suggest that LUZP1 could be a mediator of TBS cilia phenotype and 352 that this could be caused, at least in part by the increased degradation of LUZP1 353 triggered by truncated SALL1. Therefore, increasing LUZP1 levels in TBS cells might 354 affect the cilia and actin cytoskeleton phenotypes. To check whether LUZP1 355 overexpression is sufficient to repress ciliogenesis in primary human fibroblasts, 356 Control and TBS 275 cells were transduced with YFP or LUZP1-YFP ( Figure 8A). to the mouse brain (28, 29). We tested two different commercial antibodies against 386 LUZP1 and, while nuclear localization was weakly detected by immunofluorescence, 387 we observed a more prominent localization of LUZP1 to the actin cytoskeleton and 388 centrosome, both in human and mouse cells. This localization is consistent with our 389 TurboID analysis that showed an enrichment of factors associated with the actin 390 cytoskeleton and/or centrosomes among the potential interactors of LUZP1. The 391 localization of LUZP1 to the actin cytoskeleton, as well as being expressed in tissues 392 beyond the brain, is consistent with independent validation in cell lines by the Human 393 Protein Atlas (HPA; proteinatlas.org) and other expression databases (e.g. EMBL EBI 394 Expression Atlas ebi.ac.uk/gxa). Moreover, two independent proximity labeling studies 395 identified LUZP1 as a proximal interactor of centriole (35) and centriolar satellite-396 related proteins (45). Here, we report that LUZP1 forms a basket-like 3D structure 397 surrounding the proximal end of both centrioles. Like LUZP1, a large number of 398 centrosomal scaffold proteins (as for instance Cep120, Cep57, Cep63, Cep152, CPAP, 399 Cdk5Rap2, PCNT, among others) contain coiled-coil regions, and the proteins are 400 concentrically localized around a centriole in a highly organized fashion (5-7). 401 Furthermore, here we show that LUZP1 interacts with centrosome and actin-related 402 proteins (Figure 3 and Figure 4). LUZP1 has also been identified as an interactor of 403 ACTR2 (ARP2 actin related protein 2 homologue) and FLNA (19,20), and it has been 404 recently described as an actin cross-linking protein (19). Additionally, we found that 405 LUZP1 localizes not only to centrioles and actin cytoskeleton, but also to the midbody 406 in dividing cells, which was recently reported to influence ciliogenesis in polarised 407 epithelial cells (46). 408 Discrepancies with the previously reported LUZP1 localization and distribution 409 might be due to technical differences, or perhaps the epitope specificity for the 410 previously reported antiserum. Our data suggest that the association of LUZP1 to 411 centrosomes and actin filaments in many tissues may contribute to its overall roles. 412

LUZP1 is altered in TBS-derived cells 414
TBS is caused by mutations in SALL1 gene, which lead to the formation of a 415 truncated protein that interferes with the normal function of the cell. Here we found that 416 LUZP1interacts with truncated SALL1 and with SALL1 FL , which suggests that 417 interaction occurs through an N-terminal domain shared by both. We believe that, in 418 control cells, LUZP1 and SALL1 FL might have minimal interaction due to their 419 respective localizations to the cytoplasm and nucleus. However, truncated SALL1, 420 alone or together with SALL1 FL that is retained in the cytoplasm, likely interacts with 421 cytoplasmic LUZP1, promoting its degradation and functional inhibition. Importantly, 422 we detected an increase in LUZP1 levels upon treatment with the proteasome inhibitor 423 MG132 (Figure 7), suggesting that LUZP1 degradation is proteasome-mediated. Next, 424 we demonstrated that LUZP1 is ubiquitinated, and that truncated SALL1 both increases 425 LUZP1 ubiquitination and decreases its stability. LUZP1 ubiquitination was detected 426 in several proteomic screens for ubiquitinated proteins (47-51). The mechanism by 427 which truncated SALL1 can influence LUZP1 ubiquitination is yet to be revealed, but 428 one possibility could be de novo complexes involving specific Ub E3 ligases or de-429 ubiquitinases which could influence LUZP1 stability. In fact, various E3s/de-430 ubiquitinases were found as proximal interactors of truncated SALL1 and LUZP1, as 431 well as other components of the UPS. Furthermore, regulation by the UPS system has 432 been reported for centrosomal factors, necessary for the process of ciliogenesis such as 433

LUZP1 as an integrator of actin and primary-cilium dynamics 445
Actin dynamics coordinate several processes that are crucial for ciliogenesis. 446 For example, placing the MC to the appropriate area at the cell cortex is an actin-447 dependent process (55, 56). A reduction in cortical actin might potentially promote 448 ciliogenesis, as there would be no physical restriction to prevent cilium growth. 449 Supporting this hypothesis, several studies have found that changes in the actin network 450 architecture, induced either chemically or genetically, promote ciliogenesis or affect 451 cilia length (57-62). How actin regulates cilium length is not clear. One hypothesis is 452 that actin is involved in ectocytosis and cilium tip scission, preventing the axoneme 453 from growing too long (63, 64). Moreover, the removal of the CCP110/CEP97 complex 454 from the centrosome is thought to be an essential event at the beginning of cilia 455 formation. Many proteins are known to interact with the CCP110/CEP97 complex to 456 regulate ciliogenesis (65). We found that LUZP1 is associated with CCP110 and CEP97 457 and that CCP110 was displaced in Luzp1 -/cells. However, our TurboID analysis did 458 not detect CCP110 and CEP97 in the vicinity of LUZP1. This divergence might result 459 from the limitation of TurboID to detect proteins that are separated further than 10 nm 460 from each other. In fact, we found LUZP1 and CCP110 localizing to de proximal and 461 distal end of centrioles, respectively ( Figure 2D).

Statistical analysis 734
Statistical analysis was performed using GraphPad 6.0 software. Data were 735 analyzed by Shapiro-Wilk normality test and Levene´s test of variance. We used two-736 tailed unpaired Student´s t-test or Mann Whitney-U tests for comparing two groups, 737 One-way ANOVA or Kruskall-Wallis and the corresponding post-hoc tests for more 738 than two groups and two-way ANOVA to compare more than one variable in more than 739 two groups. P values were represented by asterisks as follows: (*) P-value < 0.05; (**) 740 P-value < 0.01; (***) P-value < 0.001; (****) P-value < 0.0001. Differences were 741 considered significant when P < 0.05. 742

DATA AVAILABILITY 744
The data that support the findings of this study are available from the 745 corresponding author upon reasonable request.    Bonferroni post-hoc test. 1130 The following figure supplement is available for Figure 5: 1131