Acute inhibition of centriolar satellite function and positioning reveals their functions at the primary cilium

Centriolar satellites are dynamic, membraneless granules composed of over 200 proteins. They store, modify, and traffic centrosome and primary cilium proteins, and help to regulate both the biogenesis and some functions of centrosomes and cilium. In most cell types, satellites cluster around the perinuclear centrosome, but their integrity and cellular distribution are dynamically remodeled in response to different stimuli, such as cell cycle cues. Dissecting the specific and temporal functions and mechanisms of satellites and how these are influenced by their cellular positioning and dynamics has been challenging using genetic approaches, particularly in ciliated and proliferating cells. To address this, we developed a chemical-based trafficking assay to rapidly and efficiently redistribute satellites to either the cell periphery or center, and fuse them into stable clusters in a temporally controlled way. Induced satellite clustering at either the periphery or center resulted in antagonistic changes in the pericentrosomal levels of a subset of proteins, revealing a direct and selective role for their positioning in protein targeting and sequestration. Systematic analysis of the interactome of peripheral satellite clusters revealed enrichment of proteins implicated in cilium biogenesis and mitosis. Importantly, induction of peripheral satellite targeting in ciliated cells revealed a function for satellites not just for efficient cilium assembly but also in the maintenance of steady-state cilia and in cilia disassembly by regulating the structural integrity of the ciliary axoneme. Finally, perturbing satellite distribution and dynamics inhibited their mitotic dissolution, and mitotic progression was perturbed only in cells with centrosomal satellite clustering. Collectively, our results for the first time showed a direct link between satellite functions and their pericentrosomal clustering, suggested new mechanisms underlying satellite functions during cilium assembly, and provided a new tool for probing temporal satellite functions in different contexts


Introduction
The mammalian centrosome/cilium complex consists of the centrosome, cilium, and centriolar satellites, which together regulate polarity, signaling, proliferation and motility in cells and thereby development and homeostasis in organisms. Centriolar satellites (hereafter satellites) are 70-100 nm membrane-less electron-dense granules that localize and move around centrosomes and cilia [1]. Satellite assembly is scaffolded by the large coiled-coil protein Pericentriolar Material 1 (PCM1), which physically interacts with many known and putative centrosome proteins [2,3]. Satellite resident proteins function in a wide range of cellular processes including cilium assembly, ciliary transport, centriole duplication, mitotic regulation, and microtubule dynamics and organization, and include proteins mutated in developmental and neuronal disorders [4]. Accordingly, acute or constitutive loss of satellites through depletion or deletion of PCM1 resulted in defects in cilium assembly, ciliary signaling, epithelial cell organization and autophagy [4][5][6][7].
Satellites mediate their functions in part by regulating the cellular abundance or centrosomal and ciliary targeting of specific proteins [5][6][7][8]. These results, together with their microtubule-mediated active transport, led to the current model, which defines satellites as trafficking modules and/or storage sites for their protein residents.
Although satellites are ubiquitous structures in vertebrate cells, their number and distribution changes in a context-dependent way. For example, in most cell types, satellites are predominently clustered around the centrosome, and to a lesser extent scattered throughout the cytoplasm [9]. However, in specialized cell types, their distribution varies from clustering at the nuclear envelope in myotubes and at the apical side of polarized epithelial cells, to being scattered throughout the cell body in neurons 7 at the C-terminus for inducible dimerization with the motor domains (Fig. 1B).
Analogous to GFP-PCM1 and endogenous PCM1, GFP-PCM1-FKBP localized to satellites in transfected human cervical carcinoma (HeLa) cells without altering their distribution (Fig. S1A). To target satellites to the microtubule plus ends at the cell periphery, we induced their recruitment to the FRB fusion of constitutively active kinesin-1 HA-Kif5b (1-269 a.a.) motor domain (hereafter HA-Kif5b-FRB or Kif5b), which lacks the tail region required for interaction with endogenous cargoes and thus can only bind to satellites via the FKBP-FRB dimerization (Fig. 1B) [22]. HA-Kif5b-FRB localized to the cytosol and did not affect the distribution of satellites (Fig. S1B). To target satellites to the microtubule minus ends at the cell center, we induced their recruitment to the FRB fusion of the constitutively active N terminus of dynein/dynactin cargo adaptor bicaudal D homolog 2 BICD2 (1-198 a.a.) (hereafter HA-BICD2-FRB or BICD2), which interacts with dynein and dynactin but lacks the cargo-binding domain (Fig. 1B) [22]. HA-BICD2-FRB localized diffusely throughout the cytosol in low-expressing cells and formed granules in high-expressing cells (Fig. S1B). In contrast to HA-Kif5b, we noticed that HA-BICD2 expression by itself resulted in satellite dispersal throughout the cytosol (Fig. S1B), which is consistent with its previous characterization as a dominantnegative mutant that impairs dynein-dynactin function [24][25][26]. Therefore, we did not use BICD2-based centrosomal satellite clustering as a way to assay satellite functions in ciliated and proliferating cells.
To test the feasibility of the inducible satellite trafficking assay, we first determined the localization of satellites in transfected HeLa cells before and after rapamycin addition by live imaging and quantitative immunofluorescence. In cells co-expressing GFP-PCM1-FKBP and HA-Kif5b-FRB, satellites had their typical pericentrosomal clustering pattern in the absence of rapamycin (Fig. 1C, S1C). Satellite localization ranged from clustering around the centrosomes to scattering throughout the cytosol in cells co-expressing GFP-PCM1-FKBP and HA-BICD2 depending on ectopic BICD2 expression levels (Fig. 1D, S1D). Upon rapamycin addition to cells expressing HA-Kif5b-FRB and GFP-PCM1-FKBP, satellites were targeted to the cell periphery where the majority of microtubule plus ends localize (Fig. 1C, 1G, S1C, MovieS1). In contrast, in cells expressing HA-BICD2-FRB and GFP-PCM1-FKBP, rapamycin addition resulted in clustering of satellites at the centrosome (Fig. 1D, 1G, S1D, MovieS2).
Notably, we observed partial distribution of both GFP-PCM1-FKBP and PCM1 to the cell periphery or center upon rapamycin addition in a subset of cells, likely due to insufficient expression of the active motors relative to the satellites (Fig. S1E). The integrity and organization of the microtubule network remained unaltered in both cases upon rapamycin-induced redistribution of satellites (Fig. S1F).
In Kif5b-expressing cells, satellites formed large clusters that were heterogeneously distributed at the peripheral cell protrusions, whereas in BICD2expressing cells, they formed a single cluster at the centrosome (Fig. 1C, 1D). A similar heterogeneous distribution pattern at the periphery was also observed when GFP-PCM1-FKBP was co-expressed with the motor domain of another kinesin Kif17 (Fig.   S1H). Staining of BICD2-expressing cells for satellites and the mother centriole protein Cep164 revealed tight clustering of satellites around the mother centriole in a ring-like pattern (Fig. 1H). Given that a subset of centrosomal microtubules are anchored at the subdistal appendages of the mother centrioles, this localization pattern suggests preferential trafficking of satellites along these microtubules [27]. Because satellite targeting to the cell periphery or center was dependent on microtubules, we next examined whether these satellite cluster(s), once they formed, were maintained independently of microtubules or not. The clusters remained mostly intact and did not split into smaller granules upon depolymerization of microtubules by nocodazole treatment (Fig. 1I, Movie S3, S4). Together, these results show that rapamycin-induced dimerization of satellites with molecular motors compromises satellite distribution and size irreversibly, and thus this assay provides a tool to investigate satellite functions in a temporally-controlled manner.

Satellites exert variable effects on pericentrosomal abundance of their residents
Previous work has shown that the cellular loss of satellites, by acute or chronic depletion of PCM1, alters the centrosomal levels and dynamics of a subset of proteins [4,5,14,28]. Given that satellites interact with and store a wide range of centrosome proteins, their pericentrosomal localization is likely required for efficient protein targeting to the centrosomes. To test this, we co-transfected HeLa cells with the indicated constructs and used quantitative immunofluorescence to measure the pericentrosomal levels of various known satellite proteins. We chose satellite proteins that also localize to different parts of the centrosome (e.g., pericentriolar material, distal appendages) and are implicated in different centrosome/cilium-associated cellular functions ( Fig. 2A, 2B).
As controls, we quantified the transfected cells in which GFP-PCM1-FKBP localized like endogenous PCM1 and excluded the ones that were impaired for pericentrosomal satellite clustering due to overexpression of molecular motors. For quantifying centrosomal levels in cells with peripheral or centrosomal satellite clustering after rapamycin treatment, we only accounted for the cells that exhibited complete redistribution to the cell center or periphery.
The pericentrosomal levels of the ciliogenesis factors, including the transition zone component Cep290, the E3 ubiquitin ligase Mib1, Cep131 and Cep72, increased significantly in centrosomally-targeted BICD2-expressing cells and decreased significantly in peripherally-targeted Kif5b-expressing cells at 6 h and 24 h after rapamycin treatment (Fig. 2C, S2A). We also note that these proteins were recruited to the PCM1-positive clusters at the cell periphery in Kif5b-expressing cells (Fig. 2B, 2C).
Pericentrosomal accumulation of other satellite proteins implicated in cilium assembly and centriole duplication including KIAA0753, CCDC14 and OFD1 were affected similarly upon satellite mispositioning (Fig. S2B). In contrast, there was no significant change in the pericentrosomal levels of the centriole duplication factors Cep63 and Cep152, the distal appendage protein Cep164, the centriole component centrin and the centrosomal linker protein C-Nap1 in cells with peripheral satellite clustering (Fig. 2B,   S2A). Additionally, except for centrin, these proteins were not recruited to the peripheral aggregates in Kif5b-expressing cells.
We next examined the potential role of satellite distribution in regulating the daughter centriole composition by using the rapamycin-treated BICD2-expressing cells, which leads to satellite accumulation around the mother centriole (Fig. 1I). We chose the daughter centriole protein Cep120 for further study as it was identified in the satellite proteome and 60% of its centrosomal pool was shown to be mobile [3,29,30]. We stained HeLa cells co-expressing GFP-PCM1-FKBP and HA-BICD2 before and after rapamycin addition with Cep120, the centrosome marker gamma-tubulin, and the mother centriole marker Cep164 (Fig. S3A). Despite its enrichment in the daughter centriole in control cells, Cep120 was strongly enriched in the mother centriole in BICD2-expressing cells (Fig. 3A). This indicates that satellite positioning regulates the association of Cep120 with the daughter centriole.
Satellite interactome is highly enriched in microtubule-associated proteins including the key regulators of microtubule nucleation and dynamics such as gamma-tubulin and the components of the HAUS/augmin complex [2,3]. While the centrosomal gamma-tubulin levels did not change in cells with peripheral satellite clustering, gammatubulin concentrated at the peripheral satellite clusters (Fig. S3B). To test whether recruitment of gamma-tubulin to these clusters results in microtubule nucleation, we To determine whether, and if so, how the composition of satellites upon their peripheral targeting, we took a systematic approach. To this end, we applied the BioID proximity labeling approach and identified the PCM1 proximity interactome before and after rapamycin treatment in Kif5b-expressing cells. PCM1-FKBP was fused to myc-BirA* at the N terminus and co-expressed with HA-Kif5b in HEK293T cells. Myc-BirA*-PCM1-FKBP localized to and induced biotinylation at the satellites, as assessed by staining for myc, streptavidin and gamma-tubulin (Fig. 3A). As expected, localized biotinylation was observed at the pericentrosomal granules in control cells and at the peripheral satellite clusters in rapamycin-treated cells (Fig. 3A). Given that satellites are biotinylated at the periphery away from the centrosomes, their proteome will likely exclude the contaminating interactions of PCM1 with the centrosome due to their close proximity.  (Table S1).
SAINT analysis identified 541 proximity interactions for PCM1 before rapamycin treatment and 601 interactions after rapamycin treatment (Bayesian false discovery rate (BFDR)<0.01) ( Table S2). The control and peripheral PCM1 interactomes shared 476 components, while 65 proteins were specific to peripheral satellites and 125 proteins were specific to control cells (Fig. S4A). The proteins enriched (>1.25 fold relative to control) or depleted (<0.75 fold relative to control) at the peripheral satellites with their associated GO biological processes are shown in Table 2 and Fig. S4. The 71% overlap before and after rapamycin treatment shows that the composition of the PCM1 proximity interactome remains mostly unaltered upon peripheral targeting of satellites. The peripheral PCM1 interactome is significantly enriched in previously identified MTOC, centrosome, satellite and microtubule cytoskeleton components based on Gene Ontology (GO) analysis for "cellular compartment" (Fig. 3D). Of note, we did not identify enrichment for components of microtubule cargoes such as endosomes and lysosomes, which shows that the effects of the mispositioning assay are specific to satellites (Fig.   3D).
GO analysis of the proteins identified in the PCM1 interactome of peripheral satellites for biological processes revealed significant enrichment for processes related to cilium assembly and mitotic progression (Fig. 3C). Given that satellite-less epithelial cells were mainly defective in cilium assembly, we aimed to use this dataset to gain into the specific ciliary processes regulated by satellites [6,7]. To this end, we determined the PCM1 proximity interactors previously associated with the primary cilium and generated an interaction network by classifying these proteins them based on their functions in the ciliogenesis program (Fig. 3D). Among the highly represented categories are the microtubule-associated proteins previously localized to cilia such as CCDC66, CSPP1 and MAP4 and the centrosome/satellite proteins that regulate cilium biogenesis and function. Additionally, ciliary trafficking complexes IFT-B, kinesin 2 and dynein; proteins implicated in ciliary vesicle formation, septins and transition zone components were also identified in the proteome of peripheral satellite clusters.
Collectively, systematic analysis of the composition peripheral satellites suggest ciliumand mitosis-related cellular processes as potential satellite functions.

Peripheral centriolar satellite clustering results in defective cilium assembly, maintenance and disassembly
Given our results thus far showed that satellites and associated centrosome proteins can be targeted to the periphery rapidly and efficiently in an inducible way, we employed this assay to investigate temporal satellite functions in ciliated and proliferating cells.
Satellite-less PCM1 -/kidney and retina epithelial cells had significant defects in their ability to assemble primary cilia [6,7,31]. We therefore first examined whether pericentrosomal clustering of satellites is also required for cilium assembly. To this end, we generated inner medullary collecting duct (IMCD3) and IMCD3::PCM1 -/cells that The inducible nature of the satellite trafficking assay enabled us to study the temporal function of satellites in ciliated cells during primary cilium maintenance and disassembly, which could not have been tested in satellite-less PCM1 -/cells [1,2]. To study the function of satellites in regulation of steady-state cilia stability, cells were serum starved for 48 h, treated with rapamycin for 1 h, and the percentage of cilia was quantified in a time-course manner after rapamycin washout (Fig. 4E). As expected, the percentage of ciliation did not change over 24 h in control cells (Fig. 4E, Fig. S5D).
However, induction of peripheral satellite clustering resulted in a 22% decrease in the percentage of ciliated cells at 6 h after rapamycin treatment (control: 69% ± 0.5%, Rapa:47.6% ± 3.3, p<0.0001). The fraction of ciliated cells continued to decrease over time until only 38% ± 1% of cells retained primary cilia after 24 h of rapamycin treatment, compared to the stable 69.5% ± 0.5% of control cells (p< 0.0001) (Fig. 4E,   Fig. S5D). Rapamycin treatment had no effect on maintenance of control cells (Fig.   S5C). Next, we examined whether satellite mispositioning can affect cilium disassembly after serum re-stimulation. To this end, cells were serum starved for 48 h, treated with rapamycin for 1 h, and the percentage of cilia was quantified over a 6h serumstimulation time course (Fig. 4F, Fig. S5E). In control cells, the percentage of ciliation decreased from 68.5% ± 0.5 to 56% ± 1 at 2 h, 46% ± 0 at 4 h and to 38% ± 4.5 at 6 h ( Fig. 4F, Fig. S5E). In rapamycin-treated cells, there was a rapid and significantly greater decrease in the fraction of ciliated cells at 2 h, from 68.5% ± 0.5 to 45% ± 0, which corresponds to the first wave of cilium disassembly [32, 33] (Fig. 4F, Fig. S5E).
The enhanced deciliation and disassembly phenotypes upon peripheral satellite clustering identified satellites as regulators of cilium maintenance and disassembly in addition to their reported functions in assembly.
The maintenance and disassembly of cilium requires modifications of the axonemal tubulins [34]. A major event that promotes cilium disassembly is the phosphorylation of histone deacetylase 6 (HDAC6) by Aurora A kinase and subsequent deacetylation of the modified tubulins of the ciliary axoneme and cortactin [35,36]. To test whether satellites regulate cilium maintenance and disassembly in a HDAC6dependent manner, we performed cilium maintenance and disassembly experiments in control and rapamycin-treated cells using tubacin as a potent and selective inhibitor of HDAC6 deacetylase activity [48]. Analogous to control cells, tubacin treatment rescued the enhanced cilium disassembly phenotype of rapamycin-treated cells at 4 and 6 h (4 h: Rapa: 38% ± 4.5%, tubacin+Rapa: 49.5 ± 1.5% (p<0.01); 6h: Rapa: 30.5% ± 2.5%, tubacin+rapa: 45% ± 1% , p<0.01), but did not rescue the cilium maintenance defects at 6,12 or 24 h.(p>0.05) (Fig. 4D, 4E). These results show differential requirements for regulation of cilium maintenance and disassembly, and suggest that satellites function in cilium disassembly by regulating the structural integrity of the ciliary axoneme upstream of HDAC6.

Satellites are not required for proper mitotic progression and cytokinesis
Satellite integrity and localization is dynamically modulated during cell division and the functional significance of these changes remains poorly understood [5,37]. The enrichment of key regulators of microtubule nucleation and dynamics, and mitosis such as NUMA, Plk1 and CDK1 in the satellite proteome suggests functions for satellites during mitosis [2,3]. Unexpectedly, satellite-less IMCD3 and RPE1 PCM1 KO cells did not exhibit defects in cell proliferation, cell cycle progression and mitotic times [1]. The lack of mitotic functions in these cells could be due to the activation of compensatory mechanisms that rescues defects upon constitutive loss of satellites. To investigate whether, and if so, how satellites contribute to mitosis, we assayed the mitotic behavior of IMCD3 PCM1 KO peripheral cells after rapamycin induction of peripheral satellite clustering. IMCD3 PCM1 KO peripheral cells were treated with rapamycin for 1 h followed by time-lapse imaging over 16 h. In control cells, as expected, satellites dissolved during mitosis, which was accompanied by an increase in the soluble cytoplasmic pool of PCM1 (Fig. 5A). In rapamycin-treated cells, the satellite clusters at the periphery lost their peripheral association. Notably, the clusters did not dissolve during mitosis and preferentially accumulated at the cytokinetic bridge (Fig. 5A). Similar mitotic behavior for centriolar satellites were observed in HeLa cells co-expressing GFP-PCM1-FKBP and HA-Kif5b after rapamycin treatment (Fig. 5B).
To examine the phenotypic consequences of satellite mispositioning and inhibition of their mitotic dissolution, we analyzed the movies before and after rapamycin treatment to quantify the percentage of mitotic cells that 1) completed mitosis, 2) underwent apoptosis, 3) arrested in mitosis for more than 5 h. In cells with peripheral satellite clustering, there was not significant change in the fraction of mitotic cells that arrested in mitosis or underwent apoptosis, as assessed by membrane blebbing and DNA fragmentation (3 independent experiments, p>0.05) (Fig. 5C). Next, we quantified the mitotic and cytokinesis time for the group of control and rapamycin-treated cells that completed mitosis during the time of imaging. Mitotic time was determined as the time from nuclear envelope breakdown to anaphase onset by using mCherry-H2B dynamics or brightfield images as references. While the average mitotic time was 31.4±2.9 min for IMCD3 peripheral cells that are not treated with rapamycin (n=26), it was 32.8±2.4 for IMCD3 peripheral cells after rapamycin treatment (n=27) (2 independent experiments, p>0.05) (Fig. 5D). Cytokinesis time was quantified as the time frame between anaphase onset and cleavage and it was similar between control and rapamycin-treated cells (rapamycin: 32,95 ± 6,7 min (n=20), + rapamycin: 48,56 ± 12,94 (n=27)) (Fig. 5D).
Rapamycin treatment by itself did not affect mitotic and cytokinesis times (Fig. S5F).
Collectively, these results provide further support that satellites are not required for proper mitotic progression and cytokinesis.

Discussion
We developed a chemically-inducible satellite trafficking assay that allowed us to rapidly and specifically target satellites to the cell periphery and we used this new tool to study satellite functions in a temporally-controlled way in ciliated and proliferating cells.
Application of this assay in epithelial cells for the first time identified functions for satellites not just for cilium assembly, but also for cilium maintenance and disassembly.
Targeted and systematic quantification of the satellite proteome at the cell periphery suggested that defects in the pericentrosomal abundance of proteins implicated in these processes as the mechanism underlying these defects. In addition to probing temporal satellite functions, our results for the first time also revealed a direct link between satellite functions and cellular positioning. Importantly, we note that the satellitetrafficking assay provides a powerful tool for future studies aimed at investigating celltype specific functions and mechanisms of satellites. Its use in specialized cell types, in particular the ones with non-centrosomal MTOCs such as differentiated muscle cells and polarized epithelial cells, also has the potential to provide new insight into why and how satellite distribution varies in different contexts.
Depletion or deletion of various centrosome proteins perturbs cellular positioning of satellites either by inducing their dispersal throughout the cytoplasm or tighter concentration at the centrosome [29,[38][39][40]. The satellite distribution defects have been proposed to underlie the phenotypes associated with loss of these proteins such as defective ciliogenesis. However, these loss-of-function studies remained insufficient in directly testing this proposed link. The results of our study provide direct evidence for the functional significance of proper satellite positioning during primary ciliumassociated processes, and thus strengthens the conclusions of the previous work on how perturbed satellite localization might contribute to the function of their residents.
Satellites are proposed to mediate their functions in the context of centrosomes and cilia by trafficking proteins to or away from the centrosome and by sequestering centrosome proteins to limit their recruitment at the centrosome [38]. Because our results demonstrated the functional importance of pericentrosomal satellite clustering, it is compelling to propose a complementary function for microtubules and motors in regulating satellite functions. In addition to active transport of centrosome cargoes along microtubules, we propose that microtubules and motors cooperate to maintain satellite proximity around the centrosome and that this is required for timely and efficient exchange of proteins between centrosomes and satellites. Future studies are required to address several key questions that pertains to these models: What is the precise mechanism that establishes and maintains pericentrosomal satellite clustering? Do satellites exchange material with the centrosome through active transport or simple diffusion? What are the upstream signaling pathways that induce protein release from satellites?
The direct link between cellular satellite positioning and their functions revealed by this study corroborates the notion that functions associated with different cellular compartments are regulated by their distinct spatial distribution within cells [49]. This relationship has been reported for multiple organelles and functions. One example is the nutrient-induced peripheral and starvation-induced perinuclear positioning of lysosomes, which was proposed to coordinate mTOR activity and autophagosome biogenesis during cellular anabolic and catabolic responses [41]. Another example is the requirement for proper mitochondrial positioning during T cell activation to allow sustained local Ca+2 influx and during neuronal differentiation to allow terminal axon branching [42][43][44].
In summary, our findings provide key insight into the satellite functions as key regulators of primary cilium as well as functional relevance of satellite distribution, dynamics, and size. In addition to addressing these mechanisms through in vitro and in vivo studies, it is essential in future to identify the molecular players that dictate satellite size, distribution, and material properties in different cell types and in distinct cell states.
Development of the recently-developed reversible optogenetic and streptavidin-based tools for spatiotemporal manipulation of organelle positioning in the context of satellites will be required to address some of these questions [49,50].

Cell culture, transfection and transduction
Human cervical cancer HeLa and human embryonic kidney HEK293T cells were

Immunofluorescence and antibodies
Cells were grown on coverslips, washed twice with PBS and fixed in either ice cold methanol at -20°C for 10 minutes or 4% PFA in PBS for indirect immunofluorescence.  For functional assays, quantification and analysis were performed only in cells that exhibited complete redistribution of satellites to the cell periphery or center.

Microscopy and image analysis
Complete redistribution to the cell periphery was defined by lack of centrosomal GFP-PCM1-FKBP signal in the pericentrosomal area within 3 μm 2 area centered on the centrosome. Complete redistribution to the cell center was defined by the lack of GFP-PCM1-FKBP signal in the cytoplasm beyond the 3 μm 2 pericentrosomal area. As controls, cells that were not treated with rapamycin in which satellites display their typical distribution pattern were quantified.

Mass spectrometry and data analysis
After on-based tryptic digest of biotinylated proteins, peptides were analyzed by online C18 nanoflow reversed-phase nano liquid chromatography (Dionex Ultimate 3000 RS LC, Thermo Scientific) combined with orbitrap mass spectrometer (Q Exactive Orbitrap, Thermo Scientific). Samples were separated in an in-house packed 75 μm i.d. Carbamidomethylation of cysteine was used as fixed modification and acetylation (protein N-termini) and oxidation of methionine were used as variable modifications.
Maximal two missed cleavages were allowed for the tryptic peptides. The precursor mass tolerance was set to 10 ppm and both peptide and protein false discovery rates (FDRs) were set to 0.01. The database search was performed against the human Uniprot database (release 2016).
Mass spectrometry data for each sample were derived from two biological and two technical replicates. Spectral counts of identified proteins were used to calculate BFDR and probability of possible protein-protein interaction by SAINTexpress v.3.6.3 with -L option [47]. Proteins were filtered out according to SAINTscore (>0.5) and BFDR (<0.05). Interaction maps were drawn using Cytoscape. GO-enrichment analysis was done using EnrichR. The significantly altered GO categories (biological process and cellular compartment) were selected with a Bonferroni-adjusted cutoff p-value of 0.05.

Statistical analysis
Statistical results, average and standard deviation values were computed and plotted by using Prism (GraphPad, La Jolla, CA). Two-tailed t-tests and ANOVA were applied to compare the statistical significance of the measurements. Error bars reflect SD. Following key is followed for asterisk placeholders for p-values in the figures: *P < 0.05, **P < 0.01, ***P < 0.001, **** P < 0.0001

Acknowledgements
We acknowledge all members of Firat-Karalar laboratory and Jennifer Wang from the Stearns laboratory at Stanford University for insightful discussions regarding this work.
FKBP and FRB constructs were a kind gift from Lukas Kapitein (Utrecht University). We acknowledge the Koç University Proteomics Facility for mass spectrometry analysis and Altug Kamacioglu for the SAINT analysis of the mass spectrometry data. We acknowledge Life Science Editors for editing assistance. This work was supported by ERC Grant 679140 to ENF, Royal Society Newton Advanced Fellowship to ENF and EMBO Installation Grant to ENF.

Competing interests
No competing interests declared.    The proximity PCM1 interactome of satellites were identified using the BioID approach.              Assembly γ-Tubulin