FBW7 couples structural integrity with functional output of primary cilia

Structural defects in primary cilia have robust effects in diverse tissues and systems. However, how disorders of ciliary length lead to functional outcomes are unknown. We examined the functional role of a ciliary length control mechanism of FBW7-mediated destruction of NDE1, in mesenchymal stem cell (MSC) differentiation. We show that FBW7 functions as a master regulator of both negative (NDE1) and positive (TALPID3) regulators of ciliogenesis, with an overall positive net effect on primary cilia formation, MSC differentiation to osteoblasts, and bone architecture. Deletion of Fbxw7 suppresses ciliation, Hedgehog activity, and differentiation, which are partially rescued in Fbxw7/Nde1-null cells. We also show that NDE1, despite suppressing ciliogenesis, promotes MSC differentiation by increasing the activity of the Hedgehog pathway by direct binding and enhancing GLI2 activity in a cilia-independent manner. We propose that FBW7 controls a protein-protein interaction network coupling ciliary structure and function, which is essential for stem cell differentiation.


Introduction
The primary cilium is a solitary, antenna-like, sensory organelle protruding into the extracellular space 1 . It is present in virtually all cell types of the human body functioning as a signaling center for receptor tyrosine kinases, G protein coupled receptors, the Hedgehog, Notch, and Wnt pathways [2][3][4][5][6] . A unique feature of the primary cilium is its ability to oscillate out of phase with the cell cycle, as primary cilia are formed when cells exit the cell cycle, whereas they start disassembling upon entry into the cell cycle [7][8][9][10][11][12] . This oscillatory pattern of cilia enables cells to coordinate multiple signaling pathways with the cell cycle. It has been suggested that changes in ciliary length may influence the output of cilia signaling pathways 13,14 , but the mechanisms underlying this coordination of cilia structure and function are complex and not well understood.
One of the signaling pathways that requires intact primary cilia for maximal activity is the Hedgehog pathway 15,16 . In the absence of the Hedgehog ligand, GLI2 and GLI3 transcription factors are proteolytically processed at the base of the cilium to generate transcriptional repressors, GLI2R and GLI3R. In the presence of Hedgehog ligand, proteolytic cleavage is inhibited allowing the accumulation of full length GLI2 and GLI3 (GLI2A or GLI3A) that initially translocate to the tip of the cilium and subsequently in the nucleus, where they activate transcription of target genes 5,17 . Some target genes are activated by simple removal of GLI2R and/or GLI3R from their promoters, while others require transactivation mediated by GLI2A and/or GLI3A. Although all details of pathway activation have not been completely defined, it is well-established that the primary cilium provides the structural framework for maximal Hedgehog activity in vertebrates. This property makes it an ideal pathway to study mechanisms of coupling of ciliary structure and signaling output. Because of fundamental roles of the Hedgehog pathway in multiple systems, including the skeletal system and especially mesenchymal stem cell differentiation to osteoblasts and/or chondrocytes 18 , functionality of this pathway can be further used in the context of stem cell differentiation in response to changes in ciliary structure.
Bone marrow derived MSCs are a progenitor cell source that has the ability to differentiate into multiple cell lineages, including osteoblasts, adipocytes, and chondrocytes [19][20][21] . In addition, MSCs contribute to the formation of the niche for hematopoietic stem cells in the bone marrow, which give rise to all blood cell types. Reciprocal interactions between MSCs and HSCs govern the differentiation process of numerous cell types and defects in MSC differentiation can have widespread effects ranging from osteopenia and chondrodysplasias to obesity and cancer metastasis. MSCs possess a primary cilium and previous work has shown that it is essential for the differentiation of these cells to osteoblasts 22 . However, exact effects of cilia per se, ciliary length, and contribution of cilia-based signalling and their underlying mechanisms in MSC differentiation are not completely understood.
We have identified an endogenous program regulating ciliary length in terminally differentiated cells 13 . This program involves FBW7, the recognition receptor of the SCF FBW7 E3 ubiquitin ligase. FBW7 has established roles in maintenance, self-renewal and differentiation of several adult stem cell types including cancer initiating cells 23,24 , but the contribution of cilia to these effects is unknown. We previously showed that FBW7 mediates its effect on ciliary length by the timely destruction of NDE1, a negative regulator of ciliogenesis 25 , upon cell cycle exit. NDE1 is phosphorylated by CDK5 at a specific site, which is in turn recognized by FBW7 ligase and targeted for degradation through the Ubiquitin-proteasomal system (UPS). The gradual destruction of NDE1 allows ciliogenesis to proceed normally and cilia to reach their appropriate length 13 . In the present study, we wished to understand how FBW7 and its target, NDE1, can affect cilia function and whether FBW7 mediates its effects on stem cell differentiation via its effects on cilia. We find that deletion of Fbxw7 in MSCs suppressed osteoblast differentiation.
Consistently postnatal deletion of Fbxw7 in mice resulted in reduced bone formation and mild osteopenia in 3-month-old mice. Mechanistic experiments expand the role of FBW7 to the proteasomal degradation of a positive regulator of ciliogenesis (TALPID3) in addition to NDE1.
TALPID3 is present at the mother centriole, positively regulates the Hedgehog pathway, and is required for ciliogenesis [26][27][28][29] . NDE1 is also present at the mother centriole, its effect on the Hedgehog pathway had been unknown, and is a negative regulator of ciliogenesis. We show that NDE1 physically interacts with GLI2 and increases the transcriptional activity of the Hedgehog pathway. These data suggest that mechanisms controlling ciliary length and signalling are tightly intertwined, maintaining both structural and functional integrity of primary cilia that is essential in osteoblastogenesis. Because FBW7 is one of the most commonly mutated tumour suppressors, TALPID3, a gene mutated in patients with Joubert Syndrome, and NDE1 a microcephaly gene, our data have implications in understanding the pathophysiology of all these diseases which although, they seem unrelated, they may share a common root in defective ciliary structure-function coupling.

Results
Postnatal deletion of Fbxw7 leads to changes in bone architecture The role of Fbxw7 in osteoblast differentiation in vivo has not been determined. Because of embryonic lethality of Fbxw7-null mice 30 , we used the tamoxifen-inducible UbcCre ERT2 driver to delete Fbxw7 later, in postnatal mice and examined bone architecture of the distal femur in 12-week-old male mice using microcomputed tomography (µCT). Bone volume fraction (BV/TV), connectivity density and trabecular number (Tb.N), were significantly decreased ( Fig. 1a-d), whereas trabecular separation (Tb.Sp), a parameter that reflects overall space between trabeculae, was increased ( Fig. 1e). No difference between wild type and mutant mice was detected in trabecular thickness (Tb.Th) (Fig 1f). Serum bone-specific alkaline phosphatase levels (BALP) (Fig. 1g), but not Tartrate resistant acid phosphatase (TRAcP) (Fig. 1h) were reduced in mutant mice indicating that reduced bone mass may have been primarily caused by reduced osteoblast differentiation and/or function.

Deletion of Fbxw7 reduces ciliation in MSCs or MSC-like cell types Ex vivo cultures of
MSCs derived from the bone marrow of UbcCre ERT2 ;Fbxw7 f/f mice were used to test whether deletion of Fbxw7 can affect ciliation and differentiation of MSCs to osteoblasts. In terms of ciliation, we specifically examined CD106+ MSCs, called skeletal stem cells (SSCs), which give rise to osteoblasts, adipocytes, and chondrocytes 31, 32 33, 34 . Ex vivo deletion of Fbxw7 via 4hyrdroxytamoxifen (4-OHT) treatment in serum starved primary cultures of mouse SSCs (CD106 + MSCs) resulted in the reduction of the percentage of ciliated cells compared to untreated cells (Fig. 2a,b). Because MSC cultures can be highly heterogeneous, we also examined the effect of FBW7 depletion using RNAi in transiently transfected C3H10T1/2 cells, an MSC-like stem cell line 35,36 . Depletion of FBW7 led not only to a reduction in the percentage of ciliated cells, but also a reduction in ciliary length in remaining ciliated cells ( Fig. 2e-g).
Reduced levels of ciliation were also observed in mouse embryonic fibroblasts (MEFS) transiently transfected with a Fbxw7-specific sgRNA construct (Fig. 2c,d).These data were consistent with our previous results in hTERT-RPE1 cells 13 , that FBW7 has an essential role in ciliary length control.

Fbxw7 deletion suppresses differentiation of MSCs to osteoblasts Primary cultures of
UbcCre ERT2 ; Fbxw7 fl/fl MSCs allowed us to delete Fbxw7 at either a stem cell stage by inducing deletion before treatment with differentiation media (Fig. 3a), or at a stage where cells had already been committed to differentiate, by inducing deletion after addition of differentiation media ( Supplementary Fig. 1). Deletion of Fbxw7 in MSCs before induction of differentiation resulted in significant suppression of osteoblast differentiation (Fig. 3b), as demonstrated by Alizarin red S staining. In addition, mRNA levels of four out of five osteogenic markers examined were significantly reduced at all time-points after osteogenic induction (Fig. 3c-e). On the other hand, deletion of Fbxw7 after osteogenic induction resulted in a milder suppression of differentiation ( Supplementary Fig. 1). Tamoxifen treatment did not have an effect on cells in which Fbxw7 was not floxed ( Supplementary Fig. 2). In sum, these results suggested that FBW7 promoted osteoblast differentiation by acting predominantly at the stage of osteoblast lineage commitment rather than by supporting maintenance of already committed preosteoblasts.
In contrast to what was seen in osteoblasts, deletion of Fbxw7 before adipogenic treatment promoted adipogenic differentiation of MSCs isolated from UbcCre ERT2 ; Fbxw7 fl/fl mice ( Supplementary Fig. 3). However, deletion of Fbxw7 after commitment to adipogenesis showed a trend to increase differentiation without reaching statistical significance ( Supplementary Fig.  3). These results indicated that deletion of Fbxw7 did not generally impair the ability of MSCs to differentiate, but it rather skewed differentiation towards the adipogenic lineage at the expense of osteoblastic differentiation. This type of effect suggested that FBW7 contributes to stem cell lineage determination rather than affecting committed cell types.
FBW7 regulates ciliary length through NDE1 in the C3H10T1/2 stem cell line Since we established a cellular role of FBW7 in osteoblast differentiation, we next determined whether this role was mediated via an effect on cilia and NDE1. Thus, first we tested whether deletion of both Nde1 and Fbxw7 would rescue reduced ciliation and ciliary length induced by loss of Fbxw7 alone, as previously shown in hTERT-RPE1 cells 13 . This hypothesis was tested in the cell line C3H10T1/2 for several reasons. First, these cells are multipotent and have been extensively used as a surrogate cell culture system to model osteoblast, adipocyte, or chondrocyte differentiation in vitro 35,36 . Second, they express FBW7 and NDE1, but not NDEL1 ( Supplementary Fig. 4). NDEL1 is a homolog of NDE1 that functions redundantly to NDE1 in ciliogenesis 37,38 . This effect has been described in vivo and is believed to account for the much milder phenotype of Nde1-null compared to Ndel1-null mice restricted only to neuronal cell types 39 . Most relevant to our study, NDEL1 is not a substrate of FBW7 and therefore it would evade FBW7-mediated degradation. Finally, the Hedgehog pathway drives osteoblast differentiation of C3H10T1/2 cells and therefore, cilia-based signaling is functionally relevant in these cells 40,41 .
We generated three stable lines, lacking Fbxw7, Nde1, or both using CRISPR/Cas9 gene editing. Inactivation of these genes was confirmed by both Sanger sequencing of the edited locus and protein analysis ( Supplementary Fig. 4). Because current batches of the NDE1 antibody recognize an additional band in mouse lysates, we used MEFs from Nde1 -/mice to identify the band corresponded to NDE1. The lower band detected by the NDE1 antibody corresponded to NDE1 ( Supplementary Fig. 4). Cells lacking Fbxw7 showed a reduction in the percentage of ciliated cells and ciliary length in remaining ciliated cells (Fig. 4), as shown in SSCs (Fig. 2a,b). Deletion of Nde1 did not increase the percentage of ciliated cells 24 h after serum starvation, but rather ciliary length, as we had previously shown in NIH3T3 and hTERT-RPE1 cells 25 . In double mutant cells, the percentage of ciliated cells was similar to wild type levels, while ciliary length was significantly higher than wild type cells.

FBW7 regulates functional integrity of cilia in C3H10T1/2 cells While experiments above
showed that the FBW7/NDE1 module is essential for cilia structure, they could not inform us about cilia function. Thus, we tested whether cilia per se are essential for osteoblast differentiation. We inhibited cilia formation by two means, CRISPR/Cas9-mediated inactivation of the Ift88 gene or treatment of cells with Ciliobrevin A, an inhibitor of cytoplasmic dynein, Hedgehog pathway, and cilia formation 42 . Cells lacking cilia failed to differentiate to osteoblastlike cells, as determined by expression of alkaline phosphatase (ALP), a well-established marker of osteoblast differentiation in these and other multipotent cells ( Supplementary Fig. 5).
Osteoblast differentiation was severely reduced in single Fbxw7 and Nde1-deleted cells compared to wild type cells, but partially restored in double mutant cells (Fig. 5a-d). In addition, we used MLN4924, an inhibitor of Cullin-dependent ligases, including FBW7 43 . Initiation of treatment of wild type C3H10T1/2 cells with MLN4924 before osteogenic induction resulted in suppression of differentiation. However, initiation of MLN4924 treatment only after osteogenic induction had no effect on differentiation ( Supplementary Fig. 6). These data were consistent with data using 4-OHT-inducible deletion of Fbxw7 in MSCs ( Fig. 3 & Supplementary Fig. 1), suggesting that SCF FBW7 activity at the pre-commitment stage, but not after commitment, is more relevant to osteoblast differentiation. Further, because effects of single knockouts were partially corrected in double mutant cells, these results supported the hypothesis that the FBW7mediated degradation of NDE1 is essential for osteoblast differentiation. To test whether cilia were functional in double mutant cells, cells were treated with Ciliobrevin A (Fig. 5f-h) or transiently transfected with an Ift88-specific sgRNA (Fig. 5e). Indeed, ALP levels were dramatically suppressed in these cells, suggesting that deletion of Fbxw7 and Nde1 led to the formation of functional cilia (Fig. 5). While the reduced level of differentiation in FBW7 depleted cells could be consistent with the reduced percentage of ciliated cells and ciliary length, the reduced level of differentiation in Nde1-null cells was unexpected based on effects on cilia alone and suggested that knocking out NDE1 uncoupled structure from function. Similar results were obtained in another clone of C3H10T1/2 cells lacking Nde1, ruling out clonal effects. Based on these data, we reasoned that while deletion of Nde1 led to abnormally long cilia, these cilia were not functional, yet they acquire some level of functionality in double mutant cells.
Considering the central role of the Hedgehog pathway in osteoblast differentiation and the dependence of this pathway on structural integrity of primary cilia 5, 17, 18 , we transfected all four cell lines with a Gli-reporter construct and determined basal level activity of the Hedgehog pathway. As shown in Fig. 6a, pathway activity correlated with differentiation levels. We reasoned that low activity of Hedgehog in Fbxw7-null cells was due to decreased levels of ciliation and reduced ciliary length. However, since Nde1-null cells have normal ciliation and even longer primary cilia compared to wild type, we hypothesized that functionality of ciliarelated Hedgehog effectors might have been compromised. Thus, we tested whether overexpression of various forms of GL12 or GLI3 could rescue differentiation in Nde1-null cells (Fig. 6b,c and Supplementary Fig. 5). Full length or constitutively active GLI2 increased differentiation in these cells, suggesting the possible defects in GLI2-mediated signaling may account for the impaired cilia functionality in Nde1-null cells 17,44 . Interestingly, overexpression of a constitutively active full-length GLI3 construct (GLI3 P1-P6) had no effect in Nde1-null cells 17 , while acting as a dominant negative allele in Fbxw7-null cells ( Supplementary Fig. 5), highlighting the importance and specificity of GLI2 in cilia-mediated osteoblastogenesis in this system. Here, it is interesting to note that it was not GLI3R that could account for the dominant negative effect of GLI3 P1-P6 on Fbxw7-null cells, because this GLI3 form is resistant to proteolytic cleavage.

NDE1 physically interacts with and increases the transcriptional activity of GLI2
Hedgehog activity was decreased in Nde1-null cells despite the presence of cilia. This prompted us to test whether NDE1 increased the activity of GLI2 and physically interacted with GLI2.
Transient transfection in C3H10T1/2 and HEK293T cells indicated that GLI2 activity was increased in the presence of NDE1 (Fig. 6d,e) and NDE1 co-immunoprecipitated with GLI2 (Fig.   6f). These data could help explain the reduced GLI activity in Nde1-null cells and further indicate that while NDE1 functions as a negative regulator of ciliogenesis, it functions as a positive regulator of cilia based signaling such as Hedgehog signaling, coupling ciliary structure and function.

Identification of TALPID3 as a possible target of FBW7 Because osteoblast differentiation
was partially rescued in double mutant cells, while cilia appeared grossly structurally similar to cilia of Nde1-null cells, we reasoned that positive regulator(s) of the Hedgehog pathway might be additional potential targets of FBW7 that could accumulate in double mutant cells providing some degree of rescue. Screening all known positive regulators of the Hedgehog pathway for the presence of optimal FBW7 phosphodegrons, we identified Fused and TALIPD3 as potential targets. We proceeded with TALPID3, because it had 4 optimal FBW7 phosphodegrons ( Fig.   7d). In addition, TALPID3 is required for early stages of cilia formation and organization of transition fibers to assemble the ciliary gate at the base of the cilium 27 , which would be consistent with the small but significant increase in ciliary length in double mutant cells. As predicted, all FBW7 isoforms co-immunoprecipitated with TALPID3, albeit at different levels ( Fig. 7a,b). Ubiquitinylation experiments in HEK293T cells suggested that co-expression of TALPID3 with wild type, but not ubiquitin ligase dead FBW7 mutants, resulted in increased ubiquitinylation of TALPID3 (Fig. 7c). Interestingly, an N-terminally truncated form of FBW7 that has been widely used for functional studies was stabilized when co-transfected with TALPID3, suggesting that binding of TALPID3 to FBW7 may have blocked its auto-ubiquitylation 45 supporting the idea that TALPID3 is a genuine target of FBW7 (Fig. 7a) Fig. 4). These data suggest FBW7 controls the abundance of TALPID3 in addition to NDE1 and the coordinated effect of these, and possibly other yet unidentified proteins can contribute to the cilia structure-function coupling in cells lacking Fbxw7.

Discussion
While it is known that reduction in ciliary length can result in severe phenotypes in diverse tissues and systems, it is not clear how changes in ciliary length can influence signaling output resulting in these robust phenotypes. We have shown earlier that the FBW7-mediated proteasomal degradation of NDE1 functions as a rheostat for ciliary length 13 . Results from our present study suggest that ciliary signaling output is tuned to ciliary length by direct physical interactions of proteins that mediate structural roles with proteins mediating functional roles (Fig.   9). This conclusion is based on the following lines of evidence. First, FBW7 controls the abundance of both positive and negative regulators of ciliogenesis, TALPID3 and NDE1, respectively. Second, NDE1 physically interacts with GLI2 increasing its transcriptional activity.
Third, the FBW7/NDE1/TALPID3/GLI2 network of protein-protein interactions and activities is essential for both cilia formation and differentiation of MSCs to osteoblasts and postnatal deletion of Fbxw7 leads to reduced bone mass and osteoblast activity (Fig. 8). While the differentiation of MSCs to osteoblast was utilized as a model system to decipher molecular and cellular mechanisms of cilia structure and function, our results have implications in the pathophysiology of not only bona-fide ciliopathies, such as Joubert syndrome (TALPID3) 46 , but also of emerging ciliopathies such as cancer (FBW7) 47,48 and microcephaly (NDE1) 38 .
Our approach to mechanistically determine structure-function relationships of primary cilia entailed the identification of a functional assay for the primary cilium in cell culture that would also bear physiological relevance in the whole organism. Several lines of evidence have suggested that primary cilia are essential for the differentiation of MSCs to osteoblasts [49][50][51][52][53] .
Therefore, we determined whether the FBW7-mediated degradation of NDE1 can also affect ciliary function in this context. Our data showed that deletion of Fbxw7 in primary MSCs and C3H10T1/2 suppressed both ciliation and differentiation. Consistently, postnatal deletion of In Nde1-null cells, while ciliation levels are normal, if not enhanced, Hedgehog activity and specifically, GLI2 activity is reduced because of a direct effect of NDE1 on GLI2 protein and activity. Overexpression of GLI2, but not GLI3 in Nde1-null cells rescued differentiation. In double mutant cells, the partial rescue in Hedgehog activity could be attributed to an activation of GLI2 triggered by a possible stabilization of TALPID3 levels, due to the loss of FBW7. This increase in TALPID3 could further promote ciliogenesis in the absence of NDE1 and increase Hedgehog activity directly, as has been shown previously by others 28 or indirectly. This model is consistent with our data unraveling an interplay between cilia structure and cilia function.

Knowledge obtained from our study could have important implications in conditions where
Hedgehog activity is upregulated such as in tumors.
FBW7 is one most commonly mutated tumor suppressors. Tumorigenicity induced by the inactivation or loss-of-function mutations in FBW7 depends on cell context and relevance of target protein(s) 54 . Based on our current and previous data showing that depletion of FBW7 affects structure and function of primary cilia 13 , it is tempting to speculate that part of the tumorigenic program induced upon the loss of functional FBW7 could be attributed to cilia and cilia-based signaling. We envision that loss of functional FBW7 could promote tumorigenicity in GLI2-dependent tumors or tumors that arise independently of Hedgehog signaling and they do not require an intact cilium for initiation and/or progression. However, loss or inactivation of NDE1 is mutated in patients with primary microcephaly. We had proposed earlier that loss of NDE1 in radial glial progenitors may have caused a delay in G1 to S transition due to abnormally long cilia 25 . This hypothesis was supported by in vivo experiments in rats, in which NDE1 was depleted in utero 38 . Our current data raise the possibility that reduced Hedgehog activity could also contribute to the delay in cell cycle progression in this context, leading to premature cell cycle exit and depletion or reduction of the radial glial progenitor pool size.
Interestingly, bi-allelic loss-of-function variations in SMO or Hedgehog acyl-transferase (HHAT) genes in humans led to many developmental anomalies including microcephaly 56,57 . While it is unknown whether microcephaly in humans caused by inactivating mutations in NDE1 is due to abnormally long cilia and/or reduced Hedgehog activity (GLI2), our data lend support to this idea. We attributed the delay in G1-to-S transition in NIH 3T3 and hTERT-RPE1 cells to abnormally long cilia and the extra time needed to disassemble these cilia before progressing into S. The present data lead us to refine this hypothesis to include the possibility that the reduced rate of serum-induced G1/S entry could be attributed to attenuated GLI2 activity in Nde1-null cells. Whether this is a specific effect of NDE1 on GLI2 or a general property of all or a subset of negative regulators of ciliogenesis is a subject of future studies.
Joubert syndrome represents a genetically and phenotypically heterogeneous group of disorders characterized by hypoplasia of the cerebellar vermis and other neurologic symptoms.
Additional features include skeletal abnormalities, retinal dystrophy, and renal anomalies.
TALPID3 is mutated in patients with Joubert syndrome 23, which specifically show hydrocephaly, short-rib thoracic dysplasia 14 with polydactyly, and other skeletal malformations 46 . Molecular studies in model cell culture systems have shown that TALPID3 is a mother centriolar protein, required for ciliogenesis, and positively regulates Hedgehog signaling activity.
Several studies have shown that TALPID3 is required for basal body docking at early stages of ciliogenesis 26,27 . However, the specific cell type(s) mediating these in vivo effects have not been identified. Our data show that TALPID3 could be a possible target of FBW7 along with NDE1. However, the net effect on cilia in cells lacking Fbxw7 is a reduction in the number of ciliated cells and reduced ciliary length in remaining ciliated cells, suggesting that NDE1 has a dominant effect on ciliogenesis. From a cilia-centric point of view, it appears that Joubert syndrome 23 and microcephaly may be mediated by diametrically opposite mechanisms, where Joubert Syndrome 23 is caused by loss of cilia, while microcephaly may be caused by abnormally long cilia. However, our data indicate that they are both caused by the failure to couple cilia structure with function, as both diseases seem to share a common defect to maintain a certain level of Hedgehog activity despite cilia being structurally very different. Osteoblast, but not osteoclast, metabolic activity is reduced in these mice. These data lead us to suggest that FBW7 can have an essential role in MSC differentiation to osteoblasts, at least in the animal model used in our studies. Previous studies in MSCs showed that deletion of Fbxw7 led to an upregulation of NOTCH2, which led to the increased production of the macrophage CCL2 cytokine that promoted metastasis 59 . No effects on osteoblast differentiation using ex vivo cultures or bone architecture were determined in these mice. It is conceivable that FBW7 can have multiple effects on MSCs affecting both differentiation and cytokine production and additional indirect effects on other cell types such as hematopoietic stem cells cannot be excluded.
Overall, our study identified a protein-protein interaction network with an essential role in coupling cilia structure with function. Although the studies were performed in a model system of stem cell differentiation, which is highly amenable to in vitro, ex vivo, and in vivo approaches of cell differentiation, and in which primary cilia have an established role, new mechanistic information was generated that could help understand cilia function not only in normal conditions, such as stem cell differentiation but also in diverse diseases such as cancer, microcephaly, and Joubert syndrome.
The antigen-antibody complexes were then incubated with Protein G Sepharose beads for 3 hours in 4ºC and were analyzed for the presence of FBW7 by Western blot. The same immunoprecipitation conditions were applied for the interactions between GLI2-NDE1, or TALPID3 and FBW7. In the GLI2-NDE1 interactions cell lysates were incubated with the FLAG antibody, and in the TALPID3-FBW7 interaction cell lysates were incubated either with a-TALPID3 or with a-HA.

Indirect immunofluorescence
MSCs, Mouse embryonic fibroblasts and C3H10T1/2 cells were grown on glass coverslips and fixed in 4% paraformaldehyde, permeabilized in 0.1% Triton X-100 in PBS, blocked in 3% heatactivated goat serum (or donkey serum for ALP) /0.1% Triton X-100 in PBS (blocking buffer), and incubated overnight with primary antibodies diluted in blocking buffer at 4°C. Primary antibodies were used against mouse acetylated a-tubulin at 1:1,000 (Sigma Aldrich, 1:1000), CD106 (Abcam, 1:200) or goat ALP (Thermofisher, 1:50). Cells were washed three times with PBS and incubated for 2 h at 4°C with appropriate combinations of AlexaFluor-conjugated secondary antibodies (Invitrogen,1:2000) for 2 hours at 4C protected from light. Excess of secondary antibodies were removed by four washes in PBS Samples were mounted with Diamond DAPI (Thermofisher) to counterstain the nuclei. Images were obtained with an Olympus FV1000 confocal microscope and processed with ImageJ software for ciliary length measurements.

Quantitative polymerase chain reaction (qPCR)
RNA was extracted and purified from Wild type mesenchymal stem cells or Fbxw7 null-MSCs at different time points of osteogenic treatment or at 5days of adipogenic treatment using Trizol reagent (Invitrogen). RNA was reverse-transcribed to cDNA and samples were amplified by qPCR. mRNA levels of the genes of interest were normalized to wild type via the ΔΔCt method.

Osteogenic differentiation
Osteogenic differentiation induction was performed as follows: Mesenchymal stem cells or C3H10T1/2 cells were trypsinized and seeded in a 96-well or 24-well plate respectively. When 100 % confluency was reached, the medium was replaced with osteogenic induction medium (STEMCELL TECHNOLOGIES, #05504). The osteogenic differentiation for mesenchymal stem cells was evaluated by Alizarin Red S staining at days 14 and 21 and 28 days of differentiation and quantitative real-time PCR of various osteogenic marker genes at these days. Alkaline phosphatase (ALP) staining was used as an indicator of osteoblast differentiation in C3H10T1/2 at 5 days of osteogenic treatment. Triplicate tests were conducted in each experiment.

Alizarin Red S Staining
Mesenchymal stem cells were differentiated to osteoblasts with the osteogenic induction medium. Then, cells were washed with PBS, fixed with 4% paraformaldehyde and incubate at room temperature for 15minutes. The fixative was then removed and cells were washed twice with PBS. Then samples were incubated with Alizarin Red Solution at room temperature for 20minutes. After 20 minutes of incubation, the dye was removed and the cells were gently washed three times with ddH20. After the last wash plates were let to dry and were ready for image acquisition at the dissection microscope.

Alkaline Phosphatase Staining
Undifferentiated C3H10T1/2 cells show weak alkaline phosphatase (ALP) activity whereas differentiated osteoblasts show very high activity of ALP. ALP activity is therefore a good indicator of osteoblast differentiation. One BCIP/NBT tablet (SigmaFast TM BCIP-NBT, Sigma Aldrich) was dissolved in 10ml distilled ddH20 to prepare the substrate solution which detects ALP by staining the cells blue-violet when ALP is present. It should be stored in the dark and used within 2hours. The washing buffer was made by adding 0.05% Tween 20 to PBS without Ca ++/ Mg ++ (Corning, #21-040-CM). Cells were washed one time with PBS and fixed with 4% paraformaldehyde for 60 seconds. After the fixation, cells were washed with washing buffer and then treated with BCIP/NBT substrate solution at room temperature in the dark for 10 minutes.
The staining process is being observed every 2-3 minutes. Cells are then washed with 1x with washing buffer and 1x with PBS respectively. After the last wash, the staining results were analyzed at the dissection microscope.

Adipogenic differentiation
Adipogenic differentiation induction was performed as follows: Mesenchymal stem cells were trypsinized and seeded in a 96-well plate respectively. When 100 % confluency was reached, the medium was replaced with Mesencult TM Adipogenic Differentiation Kit (STEMCELL TECHNOLOGIES, #05507). The adipogenic differentiation for mesenchymal stem cells was evaluated by Oil Red O staining and quantitative real-time PCR of adipogenic marker genes at 5 days of differentiation.

Oil O Red Staining
Oil Red O was purchased from Sigma (#O-0625). The stock solution was made by adding 0.35g Oil Red O in 100ml of isopropanol. The Oil Red O working solution was made by mixing 6ml of Oil Red O stock solution with 4ml of ddH20, let it at room temperature for 20minutes followed by filtering. Cells were washed 1X with PBS and fixed with 4% of paraformaldehyde for 15 minutes at room temperature. Cells were washed twice with ddH20 and then 1X with 60% isopropanol for 5 minutes at RT. After the aspiration of isopropanol cells should be dried completely and then incubated with Oil Red O working solution for 10minutes at room temperature. The next step is the washing of cells 4X with ddH20 and the acquisition of the images acquisition of the images at the dissection microscope.

Hedgehog Activity
Hedgehog Activity was analyzed in C3H10T1/2 or HEK293T cells by using a GLIBS reporter plasmid, expressing firefly luciferase in a manner dependent on the activity of Hedgehog pathway. Full growth medium was replaced 24h after transfection and cells were lysed in lysis buffer (Promega) 48h after transfection. Cell lysates were incubated with Stop & Glo substrates (Promega) and luciferase activity was measured in a Synergy Neo2 reader (Biotek).

Micro-Computed Tomography (µCT) Analysis
All procedures were conducted based on the requirements of the IACUC. In this study, UbcCre ERT2 ; Fbxw7 f/f male mice were crossed with Fbxw7 f/f female mice (all mice have C57Bl/6background). For the deletion of Fbxw7, nursing dams were intraperitoneally injected with 4-Hydroxy-Tamoxifen from postnatal days 2 to 6 (P2-P6) as previously described 61 . Then mice at 12 weeks old were sacrificed, soft tissue was removed and femur was analyzed at the UCSF CCMBM Skeletal Biology and Biomechanics Core for changes in 3-dimension structural parameters.       c,d) Expression levels of ALP in indicated cells (c) and summary data (d). Data are presented as means ± SEM. One-way ANOVA with Dunnett's multiple comparisons test, ****p < 0.0001. e,f) ALP staining after transfection of Fbxw7Nde1KO C3H10T1/2 cells with pcDNA3 or an Ift88specific sgRNA (e) or after treatment with DMSO or Ciliobrevin A (f). 2X indicates magnification of corresponding 1X images. g,h) Expression levels of ALP in Fbxw7Nde1KO C3H10T1/2 cells after treatment with DMSO or Ciliobrevin A (g). Summary data of (g) are shown in (h). Data are presented as means ± SEM. Student's t-test, **p < 0.01  Student's t-test, **p < 0.01, ****p < 0.0001.

Figure 8. Working model.
Working model depicting the role of the FBW7-NDE1-TALPID3 pathway in primary cilia structure, Hedgehog pathway activity, and osteoblast differentiation.