TSC1 and TSC2 regulate cilia length and canonical Hedgehog signaling via different mechanisms

Primary cilia are sensory organelles that coordinate multiple cellular signaling pathways, including Hedgehog (HH), Wingless/Int (WNT) and Transforming Growth Factor-β (TGF-β) signaling. Similarly, primary cilia have been implicated in regulation of mTOR signaling, in which Tuberous Sclerosis Complex proteins 1 and 2 (TSC1/2) negatively regulate protein synthesis by inactivating the mTOR complex 1 (mTORC1) at energy limiting states. Here we report that TSC1 and TSC2 regulate Smoothened (SMO)-dependent HH signaling in mouse embryonic fibroblasts (MEFs). Reduced SMO-dependent expression of Gli1 was demonstrated in both Tsc1−/− and Tsc2−/− cells, and we found that Tsc1 is required for TGF-β induced phosphorylation of SMAD2/3 and subsequent expression of the HH signaling effector and transcription factor GLI2. Hedgehog signaling was restored in Tsc1−/− cells after exogenous expression of Gli2, whereas rapamycin restored HH signaling in Tsc2−/− cells. Furthermore, we observed that Tsc1−/− MEFs display significantly elongated cilia, whereas cilia in Tsc2−/− MEFs were shorter than normal. The elongated cilium phenotype of Tsc1−/− MEFs is likely due to increased mTORC1-dependent autophagic flux observed in these cells, as both the autophagic flux and the cilia length phenotype was restored by rapamycin. In addition, ciliary length control in Tsc1−/− MEFs was also influenced by reduced expression of Gli2, which compromised expression of Wnt5a that normally promotes cilia disassembly. In summary, our results support distinct functions of Tsc1 and Tsc2 in cellular signaling as the two genes affect ciliary length control and HH signaling via different mechanisms. Electronic supplementary material The online version of this article (10.1007/s00018-018-2761-8) contains supplementary material, which is available to authorized users.


Introduction
Mutations in the tumor suppressor genes encoding TSC1 (Hamartin) and TSC2 (Tuberin) cause a multisystemic tumor syndrome termed tuberous sclerosis complex (TSC). TSC is an autosomal dominant genetic disorder of high penetrance with a prevalence of 1:10,000 [1] and is characterized by widespread dysplastic and neoplastic lesions [2]. TSC1 and TSC2 were originally found to form a heterodimeric complex that acts as a switch for turning off mTOR signaling by inactivating the mTOR complex 1 (mTORC1). Active mTORC1 phosphorylates the eukaryotic initiation factor 4E-binding protein-1 (4E-BP1) and 40S ribosomal protein S6 kinase 1 (S6K1) to promote protein synthesis [3,4]. The phosphorylated form of 40S ribosome protein S6 is thus a marker for mTORC1 activity. The heterodimeric TSC complex negatively regulates mTORC1 activity via the GTPase activity of TSC2, towards the small G-protein RHEB (Ras homologue enriched in brain) [5]. The mTORC1 signaling cascade receives inputs from various upstream cues related to the metabolic and nutritional status of the cell and translates these into cell growth [4,6]. In the presence of nutrients, mTORC1 is activated and promotes cell growth, including protein synthesis and energy storage. Conversely, during starvation mTORC1 is inhibited and autophagy is induced by AMP-activated protein kinase (AMPK), leading to generation of intracellular nutrients and energy during degradation of non-functional or non-essential organelles or protein aggregates [4,6], in turn contributing to cell survival. During starvation AMPK promotes autophagy by activating ULK1 via phosphorylation at Ser317, Ser555, and Ser777, whereas in the presence of nutrients mTORC1 suppresses autophagy by hampering interaction between AMPK and ULK1 through phosphorylation of ULK1 at Ser757 [7]. In addition, the TSC complex activates mTOR complex 2 (mTORC2) [8], which regulates the cytoskeleton through its stimulation of F-actin stress fibers and Rho GTPases [9]. TSC1 and TSC2 also play a role in regulating pathways other than mTORC1/2. For example, TSC inactivation has been shown to cooperate with non-canonical Smoothened (SMO)-independent Hedgehog (HH) signaling to drive tumor growth in murine cerebellar granule neuron precursors [10]. Furthermore, TSC1 was demonstrated to be required for the proper activation of the transforming growth factor-β (TGF-β)-SMAD2/3 pathway in HeLa cells [11]. Despite these advances, the roles of TSC1 and TSC2 in regulating signaling pathways independently of mTORC1/2 remain poorly understood.
The vertebrate HH signaling pathway is one of the best studied ciliary signaling pathways and is essential for embryonic development and the maintenance of multiple organs [36,37]. The 12-pass transmembrane receptor patched homolog 1 (PTCH1) is localized to the ciliary membrane in the absence of HH and prevents ciliary localization of SMO, a class F, G-protein coupled receptor that is required for the activation of GLI transcription factors. In this "OFF" stage, GLI2 and GLI3 are proteolytically cleaved to form repressor forms (GLI-R), which block transcriptional activation of target genes in HH signaling. The binding of HH to PTCH1 induces ciliary exit of PTCH1 causing concomitant ciliary enrichment of SMO, which results in decreased ciliary cAMP production, and in stabilization of full-length forms of GLI2 and GLI3 (GLI2-FL and GLI3-FL) in the cilium followed by translocation of these proteins to the nucleus in their activated forms (GLI2-A and GLI3-A) to induce transcriptional activation of target genes, including Gli1 and Ptch1 [38][39][40][41][42].
Compared to HH signaling, less is known about the role of primary cilia in mTOR signaling. TSC1 was shown to localize to the base of the cilium [43], and its loss in zebrafish was associated with left-right asymmetry defects in addition to ciliary elongation [33]. An aberrant cilia phenotype was also reported in murine cells where loss-offunction mutations in either Tsc1 or Tsc2 induced elongation of primary cilia [43,44]. In addition, primary cilia were shown to regulate mTORC1 activity through the LKB1-AMPK pathway in murine kidney cells [32]. While the heterodimeric TSC1-TSC2 complex plays a crucial regulatory role in mTORC1 activity, the need to evaluate the separate functions of TSC1 and TSC2 is evident. Although evidence supports the interdependence of TSC1 and TSC2 in the regulation of mTORC1 [45], more than 50 proteins have been reported to interact with TSC1 and/or TSC2 [46], suggesting that both proteins function in other cellular contexts independent of each other.
Here we report that both Tsc1 −/− and Tsc2 −/− mouse embryonic fibroblasts (MEFs) affect ciliary length control and cellular signaling pathways, but via different mechanisms. First, we found that Tsc1 −/− cells exhibited elongated cilia, which we show is due to dysregulated autophagy and mTORC1 activity as well as reduced GLI2-mediated expression of Wnt5a. In contrast, we observed significantly shortened cilia in Tsc2 −/− cells. Second, while both Tsc1 −/− and Tsc2 −/− MEFs display impaired SMO-dependent HH signaling, TSC2 affects HH signaling via mTORC1, whereas TSC1 regulates HH signaling via the TGF-β-SMAD2/3 pathway, which in turn regulates expression of the GLI2 transcription factor required for HH target gene expression. Collectively, our results indicate that TSC1 and TSC2 affect ciliary length control and canonical HH signaling via different mechanisms.
Next, we evaluated the consequence of TSC1 and TSC2 depletion on the frequency and length of primary cilia during serum starvation. To this end, Tsc1 −/− and Tsc2 −/− MEFs were serum starved (0.5% FBS) for 48 h to induce growth arrest and formation of primary cilia, and subsequently analyzed by immunofluorescence microscopy (IFM) with antibodies against γ-tubulin that marks the ciliary base, acetylated α-tubulin that marks the cilium, and IFT88, which is component of the intraflagellar transport system and marks both the ciliary base and the cilium (Fig. 1b). While the frequency of ciliated cells appeared to be unaffected by depletion of either TSC1 or TSC2 (wild type (WT): 53.6 ± 10.4; Tsc1 −/− : 56.6 ± 2.3; Tsc2 −/− : 60.7 ± 5.9), we found that cilia in Tsc1 −/− MEFs were significantly longer with an average length of 2.98 μm as compared to WT MEFs that displayed an average ciliary length of 2.20 μm (Fig. 1b, c), which is in accordance with previous observations [43]. In contrast, we observed that the length of primary cilia in Tsc2 −/− cells was significantly decreased with a mean length of 0.74 μm (Fig. 1c). The observed ciliary length phenotypes of the mutant cells were not an indirect effect of increased cell size [50], since microscopic analysis showed that the sizes of mutant and wild type cells were largely similar (Fig. 1d). Furthermore, the ciliary length phenotype of the Tsc1 −/− MEFs did not seem to be secondary to cell cycle defects, as judged by staining with antibody against acetylated tubulin and KI67, a potent proliferation marker; primary cilia were only induced in quiescent cells and readdition of serum lead to cell cycle re-entrance as expected (Fig. 2a).
We then asked if exogenous expression of TSC1 and TSC2 could rescue the ciliary length phenotype of Tsc1 −/− and Tsc2 −/− cells. While transfection with TSC1encoding plasmid (pTSC1) partly restored the ciliary length in Tsc1 −/− cells, only minor effect was observed by transfection of Tsc2 −/− cells with TSC2-encoding plasmid (pTSC2) (Supplementary Figure 1a, b). The latter result could be due to a toxic effect of the transfection and/or aberrant production of a smaller protein product (Supplementary Figure 1c, d). The transfection efficiency, measured by transfection with pGFP plasmid, is 14-15% (Supplementary Figure 1e). As an alternative approach to confirm the ciliary length phenotypes of the Tsc1 −/− and Tsc2 −/− cells, we used siRNA to deplete TSC1 or TSC2 from WT MEFs (Fig. 2b). Using this approach, we found that the siRNA-depleted WT cells displayed similar ciliary lengths as observed in Tsc1 −/− and Tsc2 −/− cells, respectively (Fig. 2c, d), confirming the mutant cell phenotypes. To test the dominancy of the two genes, we furthermore subjected Tsc1 −/− cells to siRNA-mediated depletion of TSC2, and vice versa ( Fig. 2e; Supplementary  Figure 2). In both cases, a reduction in the ciliary length was observed, indicating that the absence of Tsc2 is dominant for the resulting phenotype (Fig. 2f).
Finally, we induced cilia formation in starvation medium (0.5% FBS) in combination with rapamycin and found that the ciliary length phenotype of Tsc1 −/− MEFs was rescued by this treatment, while rapamycin had no observable effect on the ciliary length in Tsc2 −/− MEFs (Fig. 2g). These results support the conclusion that the increased ciliary length in Tsc1 −/− MEFs is associated with aberrant activity of mTORC1. Consistent with Hartman et al. [43], the ciliation frequency in WT, Tsc1 −/− and, Tsc2 −/− cells did not appear to be affected by rapamycin.

Tsc1 −/− MEFs exhibit increased mTORC1-dependent autophagy
Several studies have shown that autophagy regulates ciliary length [34,35,[51][52][53][54][55]. To test for differences in autophagic activity between Tsc1 −/− and Tsc2 −/− MEFs, we performed an autophagic flux assay, measuring the net flux of LC3B-II in starvation medium (0.5% FBS), stimulating both autophagy and formation of primary cilia (Fig. 3a). LC3B is expressed as a pro-LC3B protein, which is subsequently cleaved by removing the C terminus to produce the cytosolic form, LC3B-I. The cytosolic form is conjugated with phosphatidylethanolamine to form the membrane-bound LC3B-II, which is attached to the autophagosome membrane reflecting maturation of the autophagosomes. After fusion of the autophagosomes with lysosomes, LC3B-II is degraded or released as LC3B-I through delipidation. Turnover of LC3-II thus reflects the autophagic activity [35]. The net flux and the amount of LC3B-I and LC3B-II were significantly increased in the Tsc1 −/− cells compared to the WT and Tsc2 −/− cells ( Fig. 3b-d). As the cilia length in Tsc1 −/− cells was reduced in the presence of rapamycin (Fig. 2g), we investigated the effect of rapamycin on autophagy. Rapamycin reduced the antibodies and the ciliary base/centrosomes (asterisks) were labeled with anti-γ-tubulin (γ-TUB) antibody. Nuclei were visualized with differential interference contrast (DIC) microscopy or DAPI staining. Lower panels show merged and shifted overlays of Ac-TUB-and IFT88-labeled cilia. Scale bar: 2 µM. c Quantification of cilia lengths for experiment shown in b. Three hundred cilia (Ac-TUB) from three independent experiments were used for quantification. d IFM analysis of WT, Tsc1 −/− and Tsc2 −/− MEFs labeled with DAPI, Ac-TUB and IFT88 for cell size comparison. Nuclei were visualized with DAPI staining. Scale bar: 5 µM WT +PI ( To test the effect of the autophagic activity on the ciliary length more directly, we investigated the ciliary length in the Tsc1 −/− cells after treatment with lysosomal protease inhibitors for 4 h to inhibit autophagy, and found a profound reduction (Fig. 3e, f). However, a significant reduction in the ciliary length in both the Tsc2 −/− and the WT cells was also observed (Fig. 3f). In contrast, treatment with the autophagy inducer ABT-737 [56] strongly induced ciliary elongation in the Tsc1 −/− cells, whereas no effect was observed in the WT cells (Fig. 3g, h). In summary, these results indicate that autophagy initiated by serum starvation is important for the growth of the primary cilium in general, whereas increased autophagy specifically causes ciliary lengthening in the Tsc1 −/− cells.

Tsc1 −/− and Tsc2 −/− MEFs display impaired SMO-dependent HH signaling
Some evidence points to a possible link between the HH and mTOR signaling, since a combination therapy inhibiting both pathways results in an enhanced therapeutic effect compared to inhibiting either pathway alone [57]. Therefore, we speculated whether HH signaling is altered in Tsc1 −/− and Tsc2 −/− MEFs. To monitor HH signaling, we initially performed quantitative RT-PCR (qPCR) analysis of the expression of the HH target genes, Gli1 and Ptch1 in cells starved (0.5% FBS) for 48 h to induce cilia formation. To activate canonical HH signaling, the SMO agonist purmorphamine [58] was added for the last 24 h of cultivation. Furthermore, cells were cultured in the presence and in the absence of rapamycin to evaluate the role of mTORC1 in purmorphamine-induced expression of the target genes. As shown (Fig. 4a, b), purmorphamineinduced expressions of Gli1 and Ptch1 were significantly reduced in both Tsc1 −/− and Tsc2 −/− MEFs as compared to WT cells. However, rapamycin treatment restored the mRNA levels to WT levels only in Tsc2 −/− MEFs, indicating that defects in HH signaling are associated with increased mTORC1 activity in Tsc2 −/− MEFs. Since rapamycin treatment abrogated the elongated cilia phenotype observed in Tsc1 −/− MEFs (Fig. 2g), we further conclude that the long cilia phenotype observed in these cells (Fig. 1) is not directly caused by aberrant HH signaling or vice versa.
To further investigate the mechanisms by which HH signaling is impaired in Tsc1 −/− and Tsc2 −/− MEFs, we then performed qPCR analysis of the mRNA levels of Gli2 and Gli3, which are generally constitutively expressed in normal cells [59,60]. In these experiments, we observed that the expression of Gli2, but not Gli3, is greatly reduced in Tsc1 −/− cells, both in the presence and absence of purmorphamine and rapamycin (Fig. 4c, d). Cellular fractionation and WB analysis confirmed the absence of detectable GLI2 in the nucleus of Tsc1 −/− cells (Supplementary Figure 3). This indicates that reduced expression of Gli1 and Ptch1 may be caused by the lack of a sufficient level of activator forms of GLI2 independent of mTORC1 activity. In contrast, Tsc2 −/− MEFs displayed normal expression levels of both Gli2 mRNA and Gli3 mRNA compared to WT cells (Fig. 4c, d). To determine whether decreased HH signaling in Tsc1 −/− MEFs can be explained by the loss of GLI2, we examined HH signaling in Tsc1 −/− MEFs in which a plasmid expressing GLI2 was introduced. In agreement with our hypothesis, exogenous expression of Gli2 caused significantly elevated expression of Gli1 in response to purmorphamine treatment in these cells ( Fig. 4e; Supplementary Figure 4). We also evaluated the level of Gli2 mRNA in Tsc1 −/− MEFs subjected to exogenous expression of TSC1, by transfecting the cells with the pTSC1 plasmid. In this case, we observed a ~ tenfold increase in the level of Gli2 transcript, establishing a direct link between TSC1 and Gli2 expression ( Fig. 4f; Supplementary Figure 5). Finally, inhibition of mTORC1 activity by rapamycin did not affect the purmorphamineinduced increase in Gli1 expression in Tsc1 −/− MEFs even after introducing exogenous Gli2 (Fig. 4e). This indicates that TSC1 regulates HH signaling through the expression of Gli2 and this regulation is independent of mTORC1 activity. Fig. 3 Loss of Tsc1 leads to increased mTORC1-dependent autophagic activity. a SDS-PAGE and WB analyses on the level of LC3B and phosphoS6 (pS6) in WT, Tsc1 −/− and Tsc2 −/− MEFs starved (0.5% FBS) for 48 h in the absence or presence of rapamycin (Rapa) (24 h) followed by treatment with lysosomal protease inhibitors (PI) as indicated. α-tubulin (α-TUB) was used as loading control. b Normalized LC3B-II flux for experiments shown in a as measured by the LC3B-II protein levels after 2 h PI treatment subtracted the protein levels at 0 h. c Normalized LC3B-I protein levels for experiments shown in a before PI treatment. d Normalized LC3B-II protein levels for experiments shown in a before PI treatment. b-d Quantifications of LC3B-I/II protein levels were performed by densitometric analysis and normalized to α-TUB, and to a control sample loaded on all gels to correct for blotting efficiency.

TSC1 regulates HH signaling via a TGF-β-SMAD2/3-dependent pathway
Previous studies have shown that primary cilia in fibroblasts coordinate canonical TGF-β signaling through clathrindependent endocytosis of activated receptors at the ciliary pocket followed by phosphorylation and activation of SMAD2/3 transcription factors in early endosomes around the ciliary base region [28]. Activated SMAD2/3 then forms a heterotrimeric complex with SMAD4, which translocates to the nucleus for expression of target genes. Interestingly, several reports have indicated that GLI2 is a target gene for SMAD2/3 signaling in various human cell lines including fibroblasts [61,62], and it has recently been shown that TSC1, independently of TSC2, is required for the association between TGF-β receptors and SMAD2/3, so that the knockdown of TSC1 causes an impaired phosphorylation and nuclear translocation of SMAD2/3 in HeLa cells [11]. We therefore speculated whether canonical TGF-β signaling is impaired in Tsc1 −/− MEFs, leading to a reduced expression of Gli2. To address this, we investigated SMAD2/3 phosphorylation in response to TGF-β1 stimulation. In contrast to WT and Tsc2 −/− MEFs, SMAD2/3 phosphorylation was significantly reduced in -  - with TGF-β1 (Fig. 5a), showing that TSC1, but not TSC2, is required for activation of canonical TGF-β signaling, as previously observed in HeLa cells [11].
To further examine whether the loss of TSC1 merely results in a delayed response to TGF-β1 stimulation, we evaluated the level of SMAD2/3 phosphorylation in Tsc1 −/− MEFs after 1, 6, 12, and 24 h of ligand treatment. These experiments showed that SMAD2/3 phosphorylation is reduced at all investigated time points (Fig. 5b, c), thereby substantiates the conclusion that TSC1 plays a critical role in the activation of SMAD2/3. No SMAD2/3 phosphorylation could be observed after 24 h (Fig. 5b) in agreement with a previous study showing that SMAD2/3 phosphorylation peaks after 1 h of ligand exposure, and declines over the next 8 h, to finally cease [63].
To investigate a potential relationship between impaired Gli2 expression and SMAD2/3 phosphorylation in Tsc1 −/− MEFs, we carried out qPCR analysis on Gli2 expression in response to TGF-β1 stimulation in these cells. Interestingly, an increase in Gli2 expression was evident in WT MEFs for at least 6 h while no response in Tsc1 −/− MEFs was observed during this time interval (Fig. 5d). As a control for ligand specificity in receptor-mediated SMAD2/3 phosphorylation and requirement for TGF-β signaling in Gli2 expression, SDS-PAGE, WB and qPCR analyses showed that administration of the TGF-β receptor antagonist, SB431542 [64], abolished SMAD2/3 phosphorylation (Fig. 5e) and greatly reduced the expression of Gli2 in WT MEFs (Fig. 5f).

Reduced Gli2 expression affects Wnt5a expression resulting in elongated cilia in Tsc1 −/− MEFs
The prominent differences in the lengths of primary cilia in Tsc1 −/− and Tsc2 −/− MEFs (Fig. 1b, c) suggest that signaling defects in these cells may also have an impact on ciliary length control. Both TGF-β and HH signaling have been shown to regulate the expression of the gene encoding the non-canonical WNT ligand, WNT5A [65][66][67], which is known to control ciliary length by inducing disassembly of the primary cilium [65,66,68,69]. We therefore hypothesized that Wnt5a expression in Tsc1 −/− MEFs is compromised as a result of low Gli2 expression. To investigate this, growth-arrested ciliated cells were subjected to re-addition of serum for different times to monitor the expression of Wnt5a by qPCR analysis during ciliary disassembly and cell cycle re-entry. As shown in Fig. 6a, the level of Wnt5a mRNA was clearly detected in WT and Tsc2 −/− MEFs, but almost undetectable in Tsc1 −/− MEFs (Fig. 6a), in agreement with our hypothesis. As Wnt5a has been shown also to be regulated by GLI3 [65], we furthermore investigated the GLI3-FL/GLI3-R ratio in WT and Tsc1 −/− MEFs before and after purmorphamine stimulation. Whereas a significant increase in the GLI3-FL/GLI3-R ratio could be observed in the WT MEFs as an effect of the purmorphamine stimulation, indicating higher amount of active GLI3, no effect could be observed in the and Tsc1 −/− cells (Supplementary Figure 3). Therefore, we cannot exclude that reduced formation GLI3-FL in conjunction with low levels of GLI2 is involved in the reduced amount of Wnt5a mRNA observed in the Tsc1 −/− MEFs.
To further investigate the relationship between the expression of Gli2 and Wnt5a, we transfected Tsc1 −/− MEFs with the GLI2-encoding plasmid and found an increased expression of Wnt5a (Fig. 6b, Supplementary Figure 4). Indeed, stimulation with recombinant WNT5a restored the length of cilia in Tsc1 −/− MEFs to a level comparable to that of WT cells (Fig. 6c, d), supporting the conclusion that reduced expression of Wnt5a contribute to the elongated cilia phenotype of Tsc1 −/− MEFs. To confirm this observation, we used siRNA to knock down Wnt5a in WT MEFs and subsequently measured ciliary length after culturing the cells for 24 h in starvation medium (0.5% FBS). An increased ciliary length was observed, compared to cells transfected with an siRNA construct containing a scramble sequence (Fig. 6e,   f; Supplementary Figure 6), confirming that the expression of Wnt5a promotes ciliary shortening. Moreover, siRNAmediated knock down of Gli2 in WT MEFs ( Fig. 6g; Supplementary Figure 7) lead to ciliary lengthening (Fig. 6h), supporting that Wnt5a expression depends on GLI2.
It is possible that other factors than GLI2, including GLI3, may be involved in regulation of Wnt5a expression as elevated Wnt5a expression was observed in Tsc2 −/− MEFs (Fig. 6a), which is consistent with their short cilia phenotype, although the level of Gli2 mRNA in Tsc2 −/− MEFs was comparable to the level in WT cells (Fig. 4c).

Discussion
The heterodimeric TSC complex that consists of TSC1 (Hamartin) and TSC2 (Tuberin) plays a critical role in coordinating extracellular signals with intracellular energy status to control cell proliferation and growth [48]. However, the independent functions of the two TSC proteins are less well known. In this study, we investigated the cilia phenotypes in MEFs lacking either TSC1 or TSC2. While our observation of elongated cilia in Tsc1 −/− MEFs is in accordance with previous results [33,43,44], our observations of reduced ciliary length in Tsc2 −/− are not [43]. In a previous study performed by Hartman et al. [43], it was found that the cilia in both Tsc1 −/− and Tsc2 −/− MEFs were 17-27% longer than WT cilia resulting in the average length of 1.54 µm in Tsc1 −/− cells and 1.40 µm in Tsc2 −/− cells, respectively [43]; in both cases, shorter than the average cilia length of 2.7 µm measured in our WT cells. The reason for this discrepancy is unknown. Hartman et al. cultured the cells for 48 h completely without serum [43], but in our system, we were not able to obtain viable cells under these conditions. We confirmed our results by demonstrating similar cilia phenotypes in cells subjected to siRNA-mediated knock down of Tsc1 and Tsc2, respectively. In addition, whereas Hartman et al. [43] found that the length of the cilia was insensitive to rapamycin treatment, we found that inhibition of mTORC1 activity abrogated the long cilia phenotype in Tsc1 −/− MEFs. That loss of TSC1 causes elongated cilia in a rapamycin-sensitive manner has previously been reported in Tsc1 conditional knockout distal convoluted tubules in mice [44], and in Danio rerio (zebrafish) [33,51].
Increased mTORC1 activity was observed in both the Tsc2 −/− and Tsc1 −/− cells, as expected, due to lack of inhibition of mTORCI. It is also generally accepted that increased mTORC1 activity is associated with reduced autophagy [47,48], but surprisingly we observed that loss of Tsc1 led to increased autophagic flux, and both increased mTORC1 activity and increased autophagy was inhibited by rapamycin treatment. The increased autophagic flux in the which may indicate increased autophagy as it has been shown that starvation leads to fast conversion of LC3B-I to LC3B-II (after 15-30 min) and to increased expression of LC3B after long time starvation [70]. In line with a recent study in mouse bone marrow macrophages [71], we also found that loss of Tsc1 in MEFs leads to increased amount of LC3B protein, which can be normalized by rapamycin. In addition, Tsc1 −/− macrophages dispalyed increased number of autophagosomes, activation of AMPK, phosphorylation of ULK1 (site Ser555), and increased amount of Beclin1 protein, all observations supporting increased autophagic activity in the Tsc1 −/− cells [71]. Several studies have shown that autophagy regulates cilia length [34,35,[51][52][53][54][55], but with disagreement about whether increased autophagy leads to longer or shorter cilia. Three different studies [52,53,55] demonstrated that cigarette smoke, flavonoid silibinin and the anti-malaria medicine mefloquine lead to increased autophagy. However, whereas cigarette smoke and silibinin treatments was associated with shortening of cilia in mouse lung tissue and embryonic preadipocytes [52,55], cilia were elongated as an effect of mefloquine treatment in RPE cells [53]. Furthermore, the effect of Atg5, a gene essential for autophagy, was investigated in three other studies, with different outcomes [34,35,54]. Whereas longer cilia were observed in mutant MEFs in one study [35], shorter and reduced number of cilia were observed in two other studies in Atg5 mutant MEFs and mice kidney proximal tubular cells, respectively [34,54], indicating that autophagy induces cilia growth, in agreement with our results.
Further, our data show that treatment with lysosomal protease inhibitors reduces the ciliary length in both WT and mutant MEFs, and treatment with the autophagy inducer ABT-737 leads to cilia elongation in Tsc1 −/− cells. These observations indicate that nutrient deprivation-induced ciliogenesis depends on proteins delivered by autophagy. Furthermore, the rapamycin-sensitive long cilia phenotype observed in Tsc1 −/− cells indicates that active mTORC1 in these cells is a prerequisite for increased cilium elongation. This result is in agreement with previous observations of cilia in Kupffer's vesicle (ciliated epithelial organ) of Tsc1 −/− zebrafish, where rapamycin treatment was shown to cause cilia shortening and overexpression of S6K1 led to ciliary lengthening [51].
As a result of low energy stress due to the constitutive mTORC1 activation in Tsc1 −/− and Tsc2 −/− cells, it is possible that AMPK is activated. AMPK induces autophagy, directly through phosphorylation of ULK1 (Ser555) and indirectly through phosphorylation of TSC2, which then inhibits mTORC1 [7]. The increased autophagic activity in Tsc1 −/− relative to Tsc2 −/− cells might be due to the lack of TSC2 in Tsc2 −/− cells. The reduced autophagic activity in WT cells compared to Tsc1 −/− cells might be due to an intact TSC1/TSC2 complex in WT cells capable of keeping the mTOR activity under control, preventing overuse of resources. Thus, the increased level of autophagy in Tsc1 −/− cells compared to WT cells might be a result of constitutive mTORC1 activation, which in turn may promote ciliary extension, in agreement with the observed ciliary lengthening in the Tsc1 −/− cells, but not in WT cells, as a result of ABT-737-induced autophagy. Alternatively, the observed increased autophagy in Tsc1 −/− cells could be due to a yet unknown TSC1-dependent mechanism on autophagy (Fig. 7).
While few studies have focused on cross-talk between HH and mTORC1 signaling, the progression of epithelial to mesenchymal transition, which marks the potential invasiveness and metastatic potential of tumors, was shown to be diminished by rapamycin in a xenograft model of rhabdomyosarcoma in a manner that not only inhibited mTORC1 activity but also HH signaling [72]. Furthermore, a cross-talk between the two pathways has been described in a model of esophageal cancer where mTORC1/S6K1 promotes the oncogenic function of GLI1, through S6K1mediated phosphorylation of GLI1, in a SMO-independent manner [57]. We here show that TSC1 and TSC2 via different mechanisms play critical roles in activation of canonical, SMO-dependent HH signaling adding further complexity to the role of mTOR signaling components in coordinating diverse HH pathways. In both Tsc1 −/− and Tsc2 −/− MEFs, the levels of Gli1 and Ptch1 expression in purmorphaminestimulated cells were greatly reduced as compared to WT cells, but the ability to respond to purmorphamine was to a large degree restored by rapamycin only in Tsc2 −/− MEFs. These results suggest that the activation of canonical HH signaling relies on the inhibition of mTORC1 activity in a TSC2-dependent manner, while TSC1 largely operates through mTORC1-independent pathway(s) to ensure robust HH signaling in response to activation of SMO. Indeed, we demonstrate that impaired HH signaling in Tsc1 −/− MEFs, is due to low expression of Gli2, likely as a result of impaired TGF-β-SMAD2/3 signaling (Fig. 5a). This is in line with previous work showing that SMAD2/3 activation is impaired in Tsc1 −/− but not in Tsc2 −/− cells [11].
It is important to address the question of how and where TSC1 contributes to proper TGF-β signaling. We did not investigate expression levels and the cellular localization of TGF-β receptors, but previous studies showed that receptors and downstream signaling components in TGF-β signaling are enriched in primary cilia and at the ciliary base region [28]. In fibroblasts, TGF-β1 stimulation leads to a translocation of TGF-β receptors from the cilium to the ciliary pocket, for clathrin-mediated endocytosis of the activated TGF-β receptors, and phosphorylation of SMAD2/3 in early endosomes at the ciliary base region [28] (Fig. 7). Since TSC1 contains a putative transmembrane domain [73] and localizes to the ciliary base in human RPE cells [43] and early endosomes in HeLa cells [11], TSC1 may function at the primary cilium to regulate TGF-β receptor-mediated activation of SMAD2/3, although speculative at this point.
TGF-β-SMAD2/3 signaling has previously been associated with TSC as the expression of TGF-β ligand is upregulated in angiofibromas obtained from TSC patients [74]. However, the genotype of the TSC patients examined in this study was not specified. Furthermore, TGF-β activates mTORC1 in murine fibroblasts but not in epithelial cells via a PI3 K/AKT/TSC2-dependent pathway [75]. It remains to be investigated whether this effect on mTOR activity is compromised only in Tsc2 −/− MEFs.
Since TGF-β and HH signaling converge at the level of GLI2, we further elaborated on the link between low Gli2 expression and the observed elongated ciliary length phenotype in Tsc1 −/− MEFs. TGF-β1-driven Gli2 expression contributes to canonical HH signaling, but Gli2 may also regulate target gene expression downstream of TGF-β signaling, including expression of genes encoding members of the WNT protein family [68]. In the canonical WNT pathway, binding of WNT ligands to Frizzled receptors induces β-catenin stabilization and entry into the nucleus to regulate target gene transcription. However, the so-called non-canonical WNT proteins, such as WNT5A [76], are involved in alternate signaling cascades including a role in WNT5A-PLK1-dependent stimulation of Aurora A kinase, leading to the disassembly of the primary cilium [69]. We here demonstrate that a low Gli2 mRNA level correlated with reduced Wnt5a expression, which in turn contributes to an elongated ciliary length phenotype. The cilia length in the Tsc1 −/− cells was reduced by both recombinant WNT5A  Fig. 7 Model illustrating the effect of TSC1 on HH signaling, ciliogenesis and cilia disassembly. 1: TGF-β receptors (TGF-β-RI and TGF-β-RII) localize to the ciliary tip. 2: TGF-β1 stimulation leads to heterotetramerization of the receptors and activation by phosphorylation of SMAD2/3 at the ciliary base. 3: TSC1 is required for this step. 4: phosphorylated SMAD2/3 (pSMAD2/3) acts (in a trimeric complex with SMAD4) as transcription factor leading to expression of Gli2. 5: GLI2 is translocated to the cilium where it is inhibited by SUFU (negative regulator of HH). 6: inactive SMO is activated by its agonist purmorphamine. 7: activation of SMO leads to removal of SUFU followed by expression of Gli2. 8: active GLI2 enters the nucleus as a transcription factor leading to expression of Gli1 and Wnt5a. 9: WNT5A mediates primary cilia disassembly. 10: Tsc2-deficient cells display shortened primary cilia. 11: the TSC1/TSC2 complex downregulates mTORC1, leading to increased autophagy. 12: during starvation AMPK promotes autophagy directly and indirectly through phosphorylation of TSC2, which then inhibit mTORC1. 13: Tsc1-deficient cells display increased mTORC1 activity but at the same time mTORC1-dependent increased autophagic flux. 14: autophagic activity stimulates ciliogenesis and by rapamycin treatment. Whereas reduced expression of Wnt5a could be explained by aberrant TGF-β-SMAD2/3 signaling, the effect of rapamycin could be explained by mTORC1-dependent increased autophagic flux in the Tsc1 −/− MEFs. However, we cannot exclude the possibility that rapamycin treatment also leads to even shorter cilia in Tsc2 −/− cells, as the cilia are already very short and difficult to measure by our microscopy setup.
In summary, we have demonstrated that loss of TSC1 causes both increased autophagic flux and impaired SMOdependent HH signaling due to reduced expression of Gli2. The long cilia phenotype in Tsc1 −/− MEFs could be explained by a combination of increased cilia growth due to increased autophagic flux and reduced rate of disassembly due to reduced expression of the transcription factor Gli2 that regulates the expression of Wnt5a (Fig. 7). Although substantial progress has been made in understanding the role of the heterodimeric complex formed by the proteins encoded by TSC1 and TSC2 in TSC pathology, many questions remain regarding the separate functions of these two genes. Addressing these questions in the future will contribute to a better understanding of the molecular etiology of TSC and potential new opportunities for therapy.

Immunofluorescence microscopy (IFM) and imaging analysis
Approximately 0.5 × 10 6 cells per well were seeded in a 6-well plate containing glass coverslips and allowed to incubate in complete medium (DMEM containing 10% FBS, 1% penicillin-streptomycin) for 24 h prior to cilia induction by culturing cells in starvation medium (0.5% FBS) for 48 h followed by three times washing in ice-cold phosphatebuffered saline (PBS). They were then fixed in 4% paraformaldehyde for 15 min, washed three times in PBS, permeabilized with 1% Triton-X-100 in PBS for 15 min, and incubated in blocking solution [PBS containing 3% bovine serum albumin (BSA)] for 30 min. The cells were incubated with primary antibodies overnight at 4 °C and washed three times in blocking solution. Secondary antibodies were added for 45 min at room temperature followed by washing for 5 min in blocking solution. They were then incubated with 0.5 μg ml −1 4′, 6-diamidino-2-phenylindole for 30 s to stain DNA, washed three times in PBS and mounted with an anti-fading mounting gel containing N-propyl gallate (Sigma-Aldrich #P3130). All procedures were performed at room temperature. Confocal microscopy was performed using an Olympus Fluoview 1000 FV. Adjustments to brightness and contrast were minimal and applied to the whole image. Z-stacked images were Z-projected and Fiji software (http://fiji.sc) was used for measuring cilia length.

SDS-PAGE and WB analyses
The cells were lysed with ice-cold lysis buffer (50 mM Tris-HCl pH 6.8, 10% glycerol, 2.5% SDS, 10 mM DTE, 10 mM β-glycerophosphate, 10 mM NaF, 0.1 mM Naorthovanadate, 1 mM phenylmethylsulfonyl fluoride and 1× complete protease inhibitor cocktail tablets (Roche #11697498001), vortexed for 10 s and left on ice for 30 min followed by 3 × 10 s. sonication, and centrifuged for 5 min at 13,000 rpm at 4 °C. Supernatant fractions were used for protein analysis. Total protein (20 μg) was separated by SDS-PAGE and transferred to nitrocellulose membranes. After applying blocking solution (100 mM Tris pH 7.6, 0.1% Tween 20, 5% dry milk) for 45 min at room temperature, the membranes were incubated with primary antibodies and left over night at 4 °C. They were then washed five times in TBS-T (100 mM Tris pH 7.6, 0.1% Tween 20), followed by incubation with HRP-conjugated secondary antibodies, which were detected with SuperSignal™ West Dura (Ther-moFisher Scientific # 34076) and digitally developed by chemiluminescence using G:Box Chemi XX6 (Syngene). Where necessary, the membranes were stripped by two rinses in stripping solution (1% Tween 20, 0.1% SDS, 1.5% glycine, pH 2.2) followed by two rinses in PBS and TBS-T. The membranes were then ready to be blocked again. The optical density of the bands was analyzed with the gel analyzer of the Fiji software.

Plasmid transfection and siRNA-mediated gene knock down
The following plasmids were used for transient plasmid transfection: GFP-tagged Gli2 (pGli2, mouse) (Addgene #37672), Myc-tagged human WT-TSC1, pTSC1 and human TSC2-Flag, pTSC2 (both a gift from Mark Nellist, Erasmus MC, Rotterdam). JetPRIME (Polyplus transfection, Illkirch, France) was used as the transfection reagent following the protocol provided by the manufacturer. RNA silencing was achieved using siRNA targeting mouse Tsc1 (#1: TTG GTT GAT TAT TAC CTG GAA #2: ATC GAG AAA GAT AAG GAA GAA), Tsc2 (#1: CAG CAT TAA TAT CTT ATC ACA #2: CTG GAC ATC ATT GAA CGA CTA), Wnt5a (AAG ATC TTA AAT ATA GAT ATA) and Gli2 (CAC CAA CCC TTC AGA CTA TTA) or a scramble sequence for control at a final concentration of 25 nM. All siRNAs were from Qiagen and delivered to cells using DharmaFECT1 transfection reagent (Dharmacon), following the protocol provided by the manufacturers.

Quantitative RT-PCR analysis (qPCR)
For cDNA preparation total RNA was extracted from the cell cultures using the RNeasy Mini kit (Qiagen) following the instructions provided by the manufacturer. To validate the transfection efficiency with plasmids (p.TSC1, p.TSC2, p.Gli2), an isolated RNA was treated with DNase I (11284932001 Roche) before cDNA preparation according to the manufacturer's protocol. An empty p.cDNA3 vector was used for Mock transfection. The RNA concentration was determined using a NanoDrop Spectrophotometer (Thermo Fisher Scientific). For cDNA preparation, reverse transcription was carried out using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). Standard qPCR was carried using predesigned TaqMan probes (Thermo Fisher Scientific) targeting the gene of interest. The measurements are performed in triplicates. The expression levels were normalized to the level of TATA binding protein (TBP) in the same samples after amplification with the TaqMan Gene Expression Master Mix (Thermo Fisher Scientific) on a 7500 Real-Time PCR system (Applied Biosystems) and the − ΔΔC t method was used to evaluate differential expression [77]. The C t value is the cycle number at which the fluorescence generated within a reaction crosses the threshold line. C t levels are inversely proportional to the amount of target in the sample, the lower the C t level the greater the amount of target. ΔC t is the difference in C t values for our target gene and our housekeeping gene (Tbp) for a given sample (normalized values) and ΔΔC t is the difference in ΔC t values between treated and untreated samples [77]. The negative value − ΔΔC t , is used as the exponent of 2 in the equation 2 −ΔΔC t and represents the difference in normalized number of C t values. The exponent 2 is used based on the assumption that each cycle doubles the amount of product. If the control sample ΔC t is 4 and the treated sample ΔC t is 2, then 2 −ΔΔC t (2 −(2-4) ) yields 4, meaning that the gene of interest in the treated sample is increased four times in comparison to the control sample.

Cellular fractionation
Cellular fractionation was performed as described previously [78].