Primary cilia are elongated and shortened in Tsc1
−/− and Tsc2
−/− MEFs, respectively
At low nutrient levels, the TSC1-TSC2 complex downregulates mTORC1 activity so that the loss of either protein is associated with constitutive activation of mTOR signaling [47, 48]. To confirm this, Tsc1−/− and Tsc2−/− MEFs were starved (0.5% fetal bovine serum, FBS) for 48 h and analyzed by SDS-PAGE and western blotting (WB) for the expression of Tsc1 and Tsc2, and the phosphorylation status of S6 (pS6), which is a marker for mTORC1 activity (Fig. 1a). As expected, TSC1 and TSC2 were absent in Tsc1−/− and Tsc2−/− MEFs, respectively. It has previously been demonstrated that TSC1 stabilizes TSC2 by protecting TSC2 from ubiquitination and proteosomal degradation [49]. This concurs with the lower level of TSC2 observed in Tsc1−/− MEFs (Fig. 1a). Furthermore, phosphorylation of S6 (pS6) was detectable in both Tsc1−/− and Tsc2−/− MEFs upon serum depletion, demonstrating constitutive mTORC1 activity. This activity was hampered by treatment with the mTORC1 inhibitor, rapamycin, verifying that phosphorylation of S6 is mTORC1-dependent (Fig. 1a).
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 re-addition 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 TSC1-encoding 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 net flux, and the amount of LC3B-I and LC3B-II in the Tsc1−/− cells to levels comparable to that of WT cells. No effect of rapamycin on LC3B was observed for the WT or the Tsc2−/− cells (Fig. 3a–d). This indicates that the long cilia phenotype in Tsc1−/− cells may be a result of increased net flux. Compared to Tsc2−/− and WT cells, Tsc1−/− cells seem to contain larger amounts of LC3B protein. Recalculating the efflux relative to the amount of LC3B-I eliminated the difference between the cell lines and the effect of rapamycin on the Tsc1−/− cells (legend Fig. 3d). Thus, the increased amount of LC3B-II in Tsc1−/− cells correlates with the increased LC3B-I level.
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), purmorphamine-induced 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 purmorphamine-induced 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.
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 clathrin-dependent 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 Tsc1−/− MEFs after 1 h of stimulation 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, siRNA-mediated 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).