C3G (also known as RAPGEF1) plays a role in cell differentiation and is essential for early embryonic development in mice. In this study, we identify C3G as a centrosomal protein that colocalizes with cenexin (also known as ODF2) at the mother centriole in interphase cells. C3G interacts with cenexin through its catalytic domain, and the two proteins show interdependence for localization to the centrosome. C3G depletion causes a decrease in cellular cenexin levels. Centrosomal localization of C3G is lost as myocytes differentiate to form myotubes. Depletion of C3G by CRISPR/Cas9 results in the formation of supernumerary centrioles, whereas overexpression of C3G, or expression of a catalytically active C3G deletion construct, inhibits centrosome duplication. Cilium length is increased in C3G knockout cells, and this phenotype is reverted upon reintroduction of C3G or its catalytic domain alone. Association of C3G with the basal body is dynamic, decreasing upon serum starvation and increasing upon re-entry into the cell cycle. C3G inhibits cilium formation and length, and this inhibition is dependent on C3G catalytic activity. We conclude that C3G regulates centrosome duplication and maintains ciliary homeostasis, properties that could be important for its role in embryonic development.
Temporal regulation of cell fate in multicellular organisms during development and in adult tissues is dependent on signaling molecules that respond to distinct environmental cues, which are translated to altered gene expression and cytoskeletal remodeling. Guanine-nucleotide-exchange factors (GEFs) that activate small GTPases are important mediators in signaling pathways and act as hubs to receive signals and relay them to specific effector functions (Overbeck et al., 1995; Quilliam et al., 2002; Rossman et al., 2005). The ubiquitously expressed protein C3G (also known as RAPGEF1) regulates cell proliferation and differentiation and is highly conserved across vertebrate phyla (Radha et al., 2011). It is a 140 kDa protein with a catalytic domain at the C-terminus and a central CRK-binding region (CBR), which contains multiple polyproline tracts. C3G activates GTPases of the Ras and Rho families and interacts with several proteins, including CRK, SRC family kinases, c-Abl (ABL1), TCPTP (PTPN2) and β-catenin (CTNNB1) (Dayma et al., 2012; Gotoh et al., 1995; Knudsen et al., 1994; Mitra et al., 2011; Mochizuki et al., 2000; Radha et al., 2007; Shivakrupa et al., 2003). As such, C3G can participate in signaling as a GEF as well as an adaptor molecule. The GEF activity of C3G is regulated by protein interaction, membrane localization and phosphorylation (Ichiba et al., 1999, 1997; Sakkab et al., 2000). C3G phosphorylation at Y504 in human cells enhances its catalytic activity at distinct subcellular compartments (Mitra et al., 2011; Mitra and Radha, 2010; Radha et al., 2004).
C3G is involved in cytoskeletal remodeling and is required for neuronal and myogenic differentiation (Mitra et al., 2011; Radha et al., 2011; Sasi Kumar et al., 2015). Its loss causes embryonic lethality in mice indicating that early embryonic development is dependent on C3G functions (Ohba et al., 2001). Introduction of a hypomorphic allele prolongs survival, but the embryos show defects in the development of the brain and other organ systems (Ohba et al., 2001; Voss et al., 2003). We have previously shown that C3G interacts with, and regulates stability and activity of, β-catenin (Dayma et al., 2012). In addition, C3G expression is regulated by β-catenin activity. Endogenous and exogenously expressed C3G shows prominent cytoplasmic localization, and the protein undergoes regulated nuclear translocation in response to physiological stimuli (Hogan et al., 2004; Radha et al., 2004; Shakyawar et al., 2017). Nuclear functions, including chromatin remodeling and splicing, are regulated by C3G (Shakyawar et al., 2017, 2018). In a cell and tissue type-dependent manner, C3G functions to either increase or decrease cell proliferation, and altered C3G levels are associated with human tumors and other disorders (Che et al., 2015; Guerrero et al., 1998; Gutiérrez-Berzal et al., 2006; Hirata et al., 2004; Ishimaru et al., 1999; Okino et al., 2006; Radha et al., 2011; Sequera et al., 2018; Voss et al., 2006).
The centrosome, with its two centrioles, plays a pivotal role in cell division, and maintaining centrosome number is important for proper chromosome segregation (Bettencourt-Dias and Glover, 2007; Fu et al., 2015). The two centrioles are asymmetric, and the mother centriole has appendages that aid in the membrane anchoring and growth of primary cilia, a non-motile microtubule-based structure (Nigg and Stearns, 2011; Pelletier and Yamashita, 2012; Yadav et al., 2016). Cenexin (ODF2) is a scaffold protein in the sub-distal appendages (SDAs) of the mother centriole that is required for structural organization of the SDAs and for primary cilia formation (Ishikawa et al., 2005; Hehnly et al., 2013; Mazo et al., 2016; Tateishi et al., 2013; Huang et al., 2017). It maintains centrosome cohesion through regulation of β-catenin (Yang et al., 2018). Cells lacking cenexin show premature centrosome splitting, and cenexin knockout mouse embryos die before implantation (Salmon et al., 2006).
In cells that exit the cell cycle, the mother centriole matures into the basal body, from which the primary cilium protrudes out of the cell membrane. The primary cilium is retracted when cells re-enter the cell cycle, and is therefore dynamically regulated (Avidor-Reiss and Gopalakrishnan, 2013; Goetz and Anderson, 2010; Kim and Dynlacht, 2013; Pan et al., 2004; Sánchez and Dynlacht, 2016). The cilium serves as a signaling hub and plays an important role in cell fate decisions during development (Eggenschwiler and Anderson, 2007; Goetz and Anderson, 2010; Hilgendorf et al., 2016). Defects in the structure and functions of primary cilia are associated with developmental defects in humans called as ciliopathies (Gerdes et al., 2009; Habbig and Liebau, 2015; Nigg and Raff, 2009). A large number of proteins make up the centrosome, and mutations in centrosomal proteins, or their deregulation, cause a variety of human disorders. Although proteins in the centrosome–cilium interface have been mapped, how they function to regulate appropriate centrosomal duplication and the protrusion and resorption of primary cilia is not fully understood (Hsu et al., 2017; Gabriel et al., 2016; Gupta et al., 2015).
Because C3G has an essential role in mammalian early embryonic development and is a component of multiple signaling pathways important for cell fate decisions such as proliferation and differentiation, we examined whether the protein localizes to the centrosome and basal body. We show that C3G localizes to the SDAs of the mother centriole in interphase cells and functions to regulate centrosome duplication. C3G forms a complex with cenexin, and they show interdependence for their centrosomal localization. CRISPR/Cas9-mediated knockout (KO) of C3G caused an increase in the number of cells with cilia, as well as increased cilium length. Abnormal cilium length in KO cells could be rescued by reintroduction of C3G, suggesting that C3G is required for maintenance of cilium length.
C3G localizes to the centrosome
Cellular, as well as exogenously expressed, C3G predominantly localizes to the cytoplasm of exponentially growing cells (Hogan et al., 2004; Radha et al., 2004). In response to nerve growth factor (NGF) treatment of IMR-32 neuroblastoma cells, C3G is phosphorylated at Y504 and localizes to the core of the Golgi (Mitra et al., 2011; Radha et al., 2008). In interphase cells, the Golgi and centrosome are located adjacent to each other, close to the nucleus (Sütterlin and Colanzi, 2010). This prompted us to examine by confocal microscopy whether endogenous C3G localizes to the centrosome. We used a commercial antibody raised against sequences in the N-terminal region of C3G that specifically recognizes a 140 kDa polypeptide specific to C3G in a variety of cell types (Fig. S1A) (Begum et al., 2018; Guerrero et al., 2004; Radha et al., 2007; Shakyawar et al., 2017). Confocal stacks of methanol-fixed C2C12 cells showed C3G staining as a single sharp dot in a juxta-nuclear position in each cell (Fig. S1B). This staining was lost in cells when C3G expression was knocked down using shRNA treatment (Fig. S1C,D). An anti-γ-tubulin antibody marks the centrosome and is seen as two closely spaced spots in interphase cells (Vorobjev et al., 2000). The juxta-nuclear dot-like pattern of anti-C3G staining seemed consistent with the presence of C3G in the centrosome in C2C12 and ARPE-19 cells (Fig. 1A). To further confirm localization to the centriole, we examined staining in a clone of cells lacking C3G due to CRISPR/Cas9-mediated KO (Shakyawar et al., 2017). These cells did not show a signal colocalizing with γ-tubulin or centrin, a protein that marks the centrioles (D'Assoro et al., 2001; Lattao et al., 2017; Prosser and Fry, 2009; Conroy et al., 2012) (Fig. 1B). The reduction in C3G levels in KO cells was validated by western blotting (Fig. 1C) and quantification of fluorescence intensity at centrosomes (Fig. 1D). An alternative C3G antibody (rGRF2) also confirmed localization of C3G at the centrosome (Fig. S1E). In addition, co-staining of cells for C3G and PCM1, a protein that localizes to the pericentriolar material (PCM) of the centrosome and marks centriolar satellites (Ou et al., 2004), showed localization of C3G in an area central to the PCM positive signals (Fig. S1F). Overexpressed C3G–GFP also colocalized with γ-tubulin at the centrosome, in addition to its prominent presence throughout the cytoplasm (Fig. S1G). These results demonstrated that C3G specifically localizes to the centrosome.
Physiological cell fusion during skeletal muscle differentiation results in multinucleated cells with several centrioles (Bugnard et al., 2005). While undifferentiated myocytes have a single nucleus and a pair of centrioles, myotubes possess multiple centrioles that congregate close to the nuclei. C2C12 cells transfected with GFP–centrin as a centriole marker were either cultured in growth medium or induced to differentiate by culturing in differentiation medium, then immunostained to detect endogenous C3G. GFP–centrin was seen as a pair of closely spaced spots in interphase cells, but as multiple spots in fused myotubes. C3G was seen to distinctly colocalize with GFP–centrin in myocytes, but was not detected at GFP–centrin-marked centrioles in myotubes (Fig. 1E). Quantification of the colocalization coefficient between C3G and centrin showed a significant decrease in colocalization in myotubes (Fig. 1F). Similar loss of C3G at the centrioles in myotubes was seen by assessing colocalization with γ-tubulin (Fig. S2). As shown previously (Shakyawar et al., 2017), C3G localized prominently to the nuclei of myotubes. C3G localization to the centrosome may be dynamic and dependent on the physiological state of the cell.
Many proteins are known to show dynamic localization to the centrosome (Hames et al., 2005; Lui et al., 2016). Since C3G is primarily present in the cytoplasm, we examined whether its presence at the centrosome is dependent on intact cytoskeletal structures in the cell. C2C12 cells were treated with cytochalasin D (which disrupts actin microfilaments) or nocodazole (which disrupts microtubules) and examined for C3G localization to the centrosome. We observed that centrosomal localization of C3G was disrupted in cells treated with either of these drugs (Fig. S3A). Nocodazole-treated cells also showed weak staining for γ-tubulin, as described previously (Vorobjev et al., 2000). Experiments to assess the efficacy of the drugs used are shown in Fig. S3B,C. Nocodazole disrupted α-tubulin structures, and cytochalasin disrupted F-actin.
C3G and phosphorylated C3G (pC3G) localize to mitotic centrosomes
Localization of proteins to the centrosome may change as cells enter and exit mitosis (Hames et al., 2005; Lu et al., 2009; Lui et al., 2016). C2C12 cells expressing GFP–centrin and stained for C3G showed prominent staining for C3G at the centrosomes at various phases of mitosis (Fig. 2). Interestingly, interphase cells showed association of C3G with only one of the centrioles. In early prophase, where the newly divided centrosomes have not moved to the poles, C3G was associated with one of the centrosomes, but in all other mitotic phases C3G was present in both centrosomes. A similar pattern of C3G localization during mitosis was also observed when cells were co-stained for γ-tubulin (data not shown).
Phosphorylation of C3G (Y504 in human, Y514 in mouse) increases GEF activity of the protein towards Rap GTPase (Ichiba et al., 1999). Upon inhibition of cellular tyrosine phosphatases by pervanadate (PV), activation of SRC family kinases or activation of c-Abl, phosphorylated active C3G is seen in distinct subcellular regions (Mitra et al., 2011; Radha et al., 2008, 2004). In mitotic C2C12 cells, phosphorylated C3G (pY504) colocalized with γ-tubulin, whereas in interphase C2C12 cells the signal was barely detectable (Fig. 3A). These results indicate that catalytically active phosphorylated C3G (pC3G) is present particularly at the centrosome in cells undergoing mitosis, although C3G is seen at interphase and mitotic centrosomes. Colocalization of pC3G with γ-tubulin in IMR-32 cells was only observed after PV treatment, though the signal was diffuse and observed in the region surrounding the centrioles (Fig. 3B). The increase in pC3G upon PV treatment of IMR-32 cells was confirmed by western blotting (Fig. 3C). The pattern of pC3G staining we observed in PV-treated IMR-32 cells was similar to that seen previously in NGF- or forskolin-treated IMR-32 cells (Mitra et al., 2011; Radha et al., 2004). These results indicated that activation of C3G at the centrosome in interphase cells may be transient, occurring in response to distinct signals, and can be detected only upon inhibition of cellular tyrosine phosphatases.
C3G localizes to the mother centriole
The presence of C3G staining at one of the interphase centrioles suggested differential localization to the two centrioles. We examined C3G colocalization with GFP–centrin and γ-tubulin at higher magnification, and confirmed that C3G is associated with only one of the centrioles in C2C12 and ARPE-19 cells (Fig. 4A,B,C; Fig. S4A,B). Intensity scanning of a region of interest (ROI) across the centriole showed C3G staining corresponding to one side of the centrin and γ-tubulin intensity peaks in both cell types. Examination of the centrioles in the x-z plane also indicated that C3G staining did not fully coincide with centrin (Fig. S4C). Localization of C3G to the mother centriole was examined by transiently expressing GFP–cenexin, a marker for centriolar SDAs (Lange and Gull, 1995; Nakagawa et al., 2001). The pattern of cenexin localization in relation to γ-tubulin is shown in Fig. S4D. Endogenous C3G colocalized with GFP–Cenexin (Fig. 4D). Overexpressed GFP–cenexin is known to form aggregates (Steere et al., 2012), which were observed in some of the transfected cells. Interestingly, we found that endogenous C3G also colocalizes with these structures, appearing as rings in confocal sections. The intensity of C3G staining in these structures was far greater than at the centrosome, suggesting that a significant amount of cellular C3G colocalizes with cenexin when it forms aggregates due to overexpression. Intensity scanning of a ROI spanning the cenexin signal showed that the signal intensity of C3G matched that of cenexin (Fig. S4G). It was also observed that the diameter of the C3G ring was always greater than that of the cenexin signal (Fig. S4H). When alternative antibodies (rGRF2 and mGRF2) were used to detect C3G, we observed colocalization with GFP–cenexin and endogenous cenexin (Fig. S4E,F).
These results showed that C3G is localized to the SDAs of the mother centriole and suggested a possible interaction with cenexin. To examine this, we co-expressed C3G–Flag with GFP–cenexin or GFP alone, in HEK293T cells and carried out immunoprecipitation using GFP-trap beads. As shown in Fig. 4E, C3G was present in the complex pulled down only in GFP–cenexin-expressing cells. Immunoprecipitation of endogenous C3G from C2C12 cells expressing GFP–cenexin showed the presence of cenexin in C3G immunoprecipitates, but did not show immunoprecipitation of cenexin with control normal IgG (Fig. 4F). Interaction of C3G with centrin was also examined in lysates of cells co-expressing C3G–Flag and GFP–centrin. Unlike in the case of GFP–cenexin, no C3G was associated with GFP–centrin complexes under similar experimental conditions (Fig. S4I). These results indicated that C3G is specifically present in cellular complexes containing cenexin.
The domains in C3G responsible for interaction with cenexin were also examined. N- and C-terminal deletion constructs (schematic shown in Fig. 8A) that lack the regulatory and catalytic domains of C3G, respectively, were co-expressed with GFP–cenexin, and complexes were pulled down using GFP-trap beads. As shown in Fig. 4G, the deletion construct containing the C-terminal region was seen in the immunoprecipitated complex, whereas the construct lacking the C-terminal region was not present. The CBR domain alone, expressed as a recombinant GST fusion protein, did not show interaction with cenexin from cell lysates in an in vitro binding assay (Fig. S4J). Immunofluorescence assays also showed that the C-terminal region of C3G, but not the N-terminal region, could colocalize with cenexin (Fig. S5).
Reciprocal regulation of C3G and cenexin expression and localization
Cenexin has been shown to be required for ciliogenesis and regulation of centrosome duplication (Yang et al., 2018; Ishikawa et al., 2005), and its transcription is upregulated in quiescent cells (Pletz et al., 2013). Anchoring of proteins at the centriole appendages is dependent on molecular interactions with cenexin (Hehnly et al., 2013; Ishikawa and Marshall, 2011; Soung et al., 2009). Since C3G colocalized with cenexin, we examined cenexin levels and localization at the centrosome in cells lacking C3G. We observed weaker cenexin signals at centrosomes (marked by staining for centrin) in C3G KO cells compared to that seen in wild-type (WT) cells (Fig. 5A). Cenexin intensity was quantified in multiple cells and was significantly lower in the KO compared to the WT. Cellular cenexin protein levels were significantly reduced in C2C12 clones lacking C3G expression (Fig. 5B). This result was also confirmed in human cells using multiple clones of C3G KO MDA-MB-231 cells, generated using CRISPR/Cas9. Compared to WT cells, C3G KO clones showed reduced cenexin levels that correlated well with reduction in C3G levels (Fig. S6A). Knockdown of C3G by shRNA also resulted in reduction of total cellular cenexin (Fig. S6B). Overexpression of C3G caused a small increase in cellular cenexin levels in C2C12 and MDA-MB-231 cells (Fig. S6C), suggesting that C3G regulates cellular cenexin levels.
We also examined if C3G localization to the centrosome was dependent on cellular cenexin levels. siRNA was used to reduce cenexin protein expression, and C3G localization and protein levels were examined. C2C12 cells on coverslips were treated with control or cenexin-targeting siRNA. As shown in Fig. 5C, C3G localization at the centrosome was barely detectable in cells treated with cenexin siRNA compared to the localization in cells treated with control siRNA. Fluorescence intensity of C3G at the centrosome from a large number of cells confirmed the loss of C3G localization to the centrosome in cells depleted of cenexin. Efficacy of cenexin knockdown was validated by western blotting (Fig. 5D) and immunofluorescence (Fig. S6D). A small but consistent decrease in total cellular C3G level was seen upon cenexin knockdown. These results demonstrated a reciprocal dependence between C3G and cenexin for localization to the centrosome, as well as for maintenance of optimal cellular protein levels. Cenexin depletion has been shown to increase distance between γ-tubulin positive structures in NIH3T3 cells (Yang et al., 2018). A similar phenotype was seen in C2C12 cells. A greater proportion of cells depleted of cenexin (which also lacked C3G at the centrosome) showed γ-tubulin spots that were greater than 1.5 µm apart compared to the proportion observed in control cells (Fig. 5E).
C3G regulates centrosome duplication
Proteins that localize to the mother centriole play a role in centrosome duplication (Boutros and Ducommun, 2008), and optimal cenexin levels are required to regulate centrosome splitting (Yang et al., 2018). We therefore examined whether loss of C3G has an effect on centriole number. WT and C3G KO C2C12 cells were stained for centrin or γ-tubulin and examined for the number of centrioles in interphase cells. A large number C3G KO cells showed the presence of supernumerary centrioles (Figs 5A, 6A; Fig. S7A). Quantification showed that the percentage of interphase cells with more than two centrin or γ-tubulin spots was higher in KO cells when compared to WT cells (Fig. 6A; Fig. S7A). Weak cenexin staining was associated with only one of the many centrin positive structures in each cell (Fig. 5A), indicating that all the additional centrioles were newly divided and immature.
If loss of C3G causes abnormal centrosome duplication, we presumed that an increase in its levels might also alter centrosome duplication. Treatment with hydroxyurea (HU) inhibits DNA replication but not centrosome duplication, resulting in cells with multiple centrioles (Balczon et al., 1995). C2C12 cells transiently transfected with C3G–GFP were fixed and stained for γ-tubulin to visualize centrioles in the presence or absence of HU. The proportion of interphase cells with more than two γ-tubulin spots was quantified among C3G-expressing and non-expressing cells. C3G overexpression was inhibitory to centrosome duplication, both in normally growing cells and in HU-treated cells (Fig. 6B). Quantification indicated that fewer C3G-expressing cells showed supernumerary centrioles compared to non-expressing cells (Fig. 6C). A similar effect of C3G overexpression on centriole number was seen in IMR-32 cells (Fig. S7B). Therefore, optimal levels of C3G appear to be required for maintaining normal centriole number. We also found that expression of a phosphorylation-defective mutant of C3G, Y504F, and a deletion construct with constitutive catalytic activity could repress formation of supernumerary centrioles similar to expression of WT C3G (Fig. S7C).
C3G localizes to the basal body and regulates primary cilia
In cells that exit the cell cycle, the mother centriole forms the basal body, giving rise to the primary cilium, which can be detected by acetylated tubulin (Ac-tub) staining. C2C12 cells that put forth a primary cilium showed the presence of C3G colocalizing with GFP–cenexin at the base of the primary cilium (as detected by Ac-tub), confirming the presence of C3G at the basal body (Fig. 7A). Localization of C3G at the basal body was also seen in serum-starved ARPE-19 cells (data not shown). We observed that under conditions of serum starvation of C2C12 cells, when cells protrude and show prominent cilia, the intensity of C3G fluorescence at the basal body was weaker compared to that seen at the basal body in exponentially growing cells, where cells possess very short and few cilia. Under these conditions, cells exit the cell cycle but do not fuse to form myotubes. Quantification of C3G fluorescence intensity showed a significant reduction in intensity under conditions of serum starvation, which was relieved upon refeeding of the cells with serum for 24 h, during which time cells retracted their cilia (Fig. 7B). C3G KO C2C12 clones did not show any staining at the basal body, though primary cilia were present in these cells (Fig. 7C). Reduced C3G levels in the KO clone used was confirmed by western blotting (Fig. 7D). Staining for Ac-tub was intact in C3G KO cells, but we observed a difference in the morphology of primary cilia, which appeared longer (Fig. 7C, right panels). We quantitated the number of cells with primary cilia and measured primary cilium length in C3G KO clones and WT cells. This was carried out during exponential growth, where only a small fraction of C2C12 cells show cilia, and also under conditions of serum starvation for 24 h, when cells put forth cilia. Fig. 7C shows representative images of WT and KO cells. The C3G KO clone shows a significantly higher proportion of cells with primary cilium in exponentially growing as well as serum starved conditions, compared to the proportion in WT (Fig. 7E). Cilia were also significantly longer in KO cells than in WT (Fig. 7F).
Primary cilia are dynamic structures that are formed and retracted specifically in a cell-cycle-dependent manner. Loss of C3G leading to enhanced cilium formation and longer cilia could be because C3G functions as a negative regulator of ciliogenesis by restricting cilium length and aiding in resorption. Primary cilia can be induced to form by shifting cells to serum starvation, and subsequently induced to retract by re-feeding with serum. We stimulated WT and KO C2C12 cells to re-enter the cell cycle after serum starvation and quantitated the number of cells with cilia as well as their ciliary length. Both WT and C3G KO cells retracted their cilia when refed, but KO cells continued to show larger proportion of cells with cilia than WT cells, and cilia in KO cells were always longer than those in WT cells (Fig. 7E,F). These results demonstrate that, in addition to its role as an inhibitor of ciliogenesis, C3G functions to maintain cilium length.
Catalytic activity of C3G is required for normal cilium formation
To confirm that increase in cilium length was indeed due to loss of C3G, we transiently expressed full-length C3G or its catalytic domain alone (ΔN-C3G), which is known to be constitutively active, in the KO clone and examined cilium length in expressing and non-expressing cells. Schematic representation of the domains in C3G and its various deletion constructs is shown in Fig. 8A. As shown in Fig. 8B, cilium length was significantly reduced in C3G- and ΔN-C3G-expressing cells compared to that seen in non-expressing cells on the same coverslips. These results indicated that introduction of catalytically active C3G could rescue the phenotype of abnormally long cilia seen in KO cells.
Since C3G has functions dependent on its catalytic activity as well as protein interactions, we wished to examine whether the catalytic activity of C3G was essential for primary cilium maintenance. The presence of pC3G (the catalytically active form) at mitotic centrosomes suggested that the GEF function of C3G might be required at the centrosome. To address this, we used a dominant-negative approach of expressing deletion constructs with only the central CBR, or C3G lacking its catalytic domain (ΔC-C3G) (Fig. 8A). Expression of these constructs was verified by western blotting (Fig. 8D). These constructs have previously been shown to inhibit catalytic activity-dependent effector functions of cellular C3G (Guerrero et al., 1998; Martín-Encabo et al., 2007; Mitra and Radha, 2010; Radha et al., 2007, 2004; Sasi Kumar et al., 2015). CBR- and ΔC-C3G-expressing cells showed an increase in cilium formation when expressed in WT cells (Fig. 8C,E). CBR-expressing cells also showed longer cilia, mimicking the phenotype seen in cells lacking C3G (Fig. 8F). These results indicated that C3G regulates primary cilium length dependent on its catalytic activity.
Cells lacking C3G show altered cyclin levels and cell cycle profiles
Proteins that control centrosome duplication also regulate proliferation (Kim and Tsiokas, 2011). While generating C3G KO clones, we observed that clones with significantly reduced C3G expression grew slowly, and loss of C3G expression could not be sustained beyond 2–3 passages because the clones began showing C3G expression as they were passaged. We suspected that this may be due to the slow growth of cells lacking C3G, and their death during passaging. To examine if loss of C3G affects cell proliferation, we carried out assays to monitor growth of WT C2C12 and KO clones. Cell number was determined using MTT assays and proliferation was monitored over a period of 72 h. KO cells from the first passage proliferated very slowly compared to WT cells (Fig. S8A). The cell cycle profile of exponentially growing cells from the first passage, as determined by FACS analysis, showed that loss of C3G arrests cells in G1 phase (Fig. S8B). KO cells from the second passage showed increase in the size of the sub-G1 population, indicating that loss of C3G compromises cell survival (Fig. S8C). G1 arrest and delay in cell proliferation could be a consequence of changes in expression of genes that regulate the cell cycle. We therefore compared the expression of cyclins A, D1 and E in WT and C3G KO cells, and found that KO cells have significantly lower levels of these three cyclins (Fig. S8D). A reduction in cyclin D1 expression was also seen in a C3G KO clone of MDA-MB-231 cells (Fig. S8E).
In this study, we have identified C3G as a centrosomal protein and elucidated its role in regulating centrosomal duplication and primary cilium maintenance. The following evidence enabled us to conclude that C3G localizes to the centrosome and is required for maintaining centrosomal homeostasis as cells enter and exit the cell cycle. (1) C3G colocalized with γ-tubulin and centrin in interphase and mitotic cells, whereas centrosomal localization of C3G was lost upon differentiation of myocytes to myotubes. (2) C3G interacted with and colocalized with cenexin at the mother centriole, and C3G and cenexin showed interdependence for localization to the centrosome. (3) C3G was present in the basal body and its localization was dynamic, reducing upon serum starvation. (4) Cells lacking C3G due to CRISPR/Cas9 mediated KO had supernumerary centrioles and an increase in the number of cells with cilia. KO cells possessed longer cilia, and this phenotype was rescued by expressing C3G, or a deletion construct with catalytic activity.
A very large number of proteins are associated with the centrosome, and a few show asymmetric localization to one of the centrioles (Alves-Cruzeiro et al., 2014; Jakobsen et al., 2011). Proteins localized to the mother centriole regulate centrosome integrity and duplication, as well as primary cilia dynamics, but not all of them have been identified nor their functions well understood. Cellular C3G is seen predominantly in the cytoplasm and undergoes regulated nuclear entry in response to physiological stimuli (Shakyawar et al., 2017). The localization of C3G to the centrosome, and specifically to the mother centriole, is reported here for the first time. Perfect colocalization with cenexin at the centrosome, and formation of ring-like aggregates upon cenexin overexpression, agrees with properties of many SDA proteins (Huang et al., 2017; Mazo et al., 2016). Because overexpression of C3G inhibits centriolar duplication, and its loss results in formation of supernumerary centrioles, we concluded that C3G functions to maintain centrosome numbers.
Many centrosomal proteins show cell-cycle-dependent localization to the centrosome (Hames et al., 2005; Kim et al., 2009; Lui et al., 2016). We observed that C3G particularly localizes to the mother centriole, with no detectable signal seen at the daughter centriole in interphase cells. SDA proteins are known to associate with the centrosome in late prophase (Guarguaglini et al., 2005; Nigg and Stearns, 2011; Vorobjev and Chentsov Yu, 1982). During mitosis, C3G locates to the newly divided centrosome around the time that the mother centriole is marked by appendage proteins, providing additional evidence for its association with appendages.
Although C3G protein was associated with the centrosome in interphase cells as well as mitotic cells, pC3G (the active form of the protein) was seen at the centrosome only in mitotic cells. pC3G was seen at the centrosome of interphase cells only upon inhibition of tyrosine phosphatases, suggesting that C3G activation at the centrosome is very transient in interphase cells and more stable at mitotic centrosomes. It appears that C3G present at the centrosome may be specifically activated in a spatial and temporal manner at the onset of mitosis. Phosphorylation of proteins is important for expansion of the PCM at the onset of mitosis (Guarguaglini et al., 2005; Haren et al., 2009; Ramani et al., 2018; Runkle et al., 2011). Negative regulators of ciliogenesis are required to suppress untimely formation of basal bodies in dividing cells (Kobayashi et al., 2011). Inhibition of ciliogenesis in dividing cells is dependent on localization of proteins like Ndel1 to the SDAs of the mother centriole (Inaba et al., 2016). It is possible that the presence of active C3G at the centrosome prevents ciliogenesis during mitosis. The targets of activated C3G at the centrosome are yet to be identified.
The integrity of the centrosome, basal body and primary cilia are dependent on proper microtubule assembly, and C3G might be dynamically transported to the centrosome, aided by the cytoskeleton, in response to physiological stimuli. Localization of other proteins such as geminin, which regulates centrosome duplication and ciliogenesis, to the centrosome is also dependent on an intact actin cytoskeleton (Lu et al., 2009). In addition, cilium length is controlled by regulators of the cytoskeleton (Copeland et al., 2018; Zheng et al., 2016), indicating that microtubules and microfilaments can regulate cilia by localizing regulators to the centrosome, in addition to directly contributing to centrosome and cilia structure. Molecular control of ciliogenesis has been investigated extensively, but our understanding of ciliary maintenance is still poor. Few molecules that localize to the mother centriole, such as CP110, CEP97, Ndel1 and APC, are negative regulators of ciliogenesis (Inaba et al., 2016; Spektor et al., 2007; Tsang et al., 2008; Wang et al., 2014). The consequences of C3G depletion, rescue, and overexpression indicate C3G is a negative regulator of cilia length.
Cenexin regulates centrosomal cohesion by maintaining β-catenin levels, and restricts premature disjunction (Yang et al., 2018). C3G was previously shown by us to negatively regulate cellular β-catenin levels in epithelial cells (Dayma and Radha, 2011). Similarly, cenexin overexpression results in reduced β-catenin levels (Yang et al., 2018). In this study we have shown that C3G forms a complex with cenexin, and C3G localization to the centrosome was dependent on cenexin. Likewise, cenexin localization to the centrosome was also dependent on C3G, indicating reciprocal regulation. The C3G catalytic domain that is responsible for association with cenexin and localization to the centrosome was also sufficient to regulate cilia length. We have shown previously that the catalytic domain of C3G is required to induce differentiation of myocytes upon overexpression (Sasi Kumar et al., 2015). In addition to its catalytic activity, centrosomal localization of C3G may be important for myogenic differentiation.
Cellular cenexin levels were significantly reduced upon C3G depletion. Whether C3G regulates cenexin expression at the transcriptional or post-transcriptional level is to be examined. One possibility is that loss of C3G from the SDAs of the mother centriole may trigger degradation of cenexin protein. Presently, very little information is available on the regulation of cenexin, or other SDA protein levels. Many SDA proteins form complexes that are dependent on cenexin, through direct or indirect interaction. Cenexin depletion decreases levels of other SDA proteins (Huang et al., 2017). Because cenexin protein levels were observed to be very low in C3G KO cells, it would be expected that the effect of C3G depletion on centrosome duplication and primary cilia is due to reduction in cenexin levels. C3G and cenexin KO mice show preimplantation embryonic lethality (Salmon et al., 2006; Voss et al., 2003), and both proteins function as negative regulators of centrosome duplication (Yang et al., 2018; McKinley and Cheeseman, 2017). It is possible that defects seen upon C3G or cenexin depletion are both due to β-catenin accumulation. As part of future work, we wish to see if expression of cenexin, or depletion of β-catenin, in C3G KO cells can rescue abnormal centrosome duplication. Interestingly, we found differences in primary cilium phenotype when C3G or cenexin were depleted, indicating that some of their functions are distinct. C3G was not essential for primary cilium formation but was required for maintenance of cilium length. Cenexin is required for primary cilium formation and for their spatial configuration (Hehnly et al., 2013; Mazo et al., 2016).
Our findings show regulated movement of C3G in and out of the centrosome that corresponded with cellular proliferation (primary cilium withdrawn) and arrest (primary cilium protruded). Formation of supernumerary centrioles and longer cilia can directly affect cell proliferation. We observed reduced levels of various cyclins required for G1 and S phase in KO clones, which could be responsible for their arrest at G1. We have previously shown that C3G regulates chromatin modifications and gene expression (Shakyawar et al., 2017). C3G might, therefore, play a role in maintaining the balance between cell proliferation and quiescence by regulating gene expression as well as centrosomal duplication and primary cilia dynamics. Centrosomal localization of C3G is lost as myocytes fuse to form myotubes. During this process, primary cilia are retracted, and centrosomes do not divide (Bugnard et al., 2005; Fu et al., 2014; Srsen et al., 2009). Since C3G regulates centrosomal duplication as well as primary cilium formation, irreversibly arrested cells may not require C3G function at the centrosome.
The primary cilium is considered a signaling hub; enabling responses to chemical and mechano-sensory signals. C3G is involved in signaling from growth factors as well as mechanical cues and might therefore participate in transduction of signals received by the primary cilium (Marada et al., 2016; Radha et al., 2011; Takahashi et al., 2008; Tamada et al., 2004). Our results suggest that catalytic activity of C3G is required, and is also sufficient, for regulating centrosomal dynamics. Several proteins deregulated in human cancers have been shown to affect centrosomal duplication and ciliogenesis (Bettencourt-Dias and Glover, 2007; Nigg and Holland, 2018; Plotnikova et al., 2008; Wang and Dynlacht, 2018; Gönczy, 2015). C3G has been implicated in tumorigenesis, with altered C3G levels observed in different cancers (Che et al., 2015; Guerrero et al., 1998; Gutiérrez-Berzal et al., 2006; Hirata et al., 2004; Okino et al., 2006; Priego et al., 2016; Radha et al., 2011; Sequera et al., 2018). The function of C3G in regulating centrosome duplication might be responsible for its deregulation being associated with tumors.
Embryonic development is dependent on regulated division of each cell. Many molecules that are components of the centrosome or primary cilia are responsible for gross developmental disorders when mutated. Loss of C3G results in early embryonic lethality of mouse embryos, and this must be due to improper cell division and failure of timely differentiation. Regulation of centrosomal duplication and primary cilium dynamics are novel functions of C3G important for cell fate determination that might also be responsible for the early death seen in KO embryos. We also hypothesize that mutations in C3G might cause the specific developmental defects known as ciliopathies in humans.
MATERIALS AND METHODS
Cell culture, transfections and treatments
The mouse myoblast cell line C2C12, a kind gift from Professor Helen Blau (Stanford University, CA), was cultured in Dulbecco's Modified Eagle's Medium (DMEM; Gibco), containing 20% fetal bovine serum. ARPE-19 cells were grown in F12-DMEM (Gibco) with 10% serum. IMR-32, HEK-293T and MDA-MB-231 cells were grown in DMEM containing 10% FBS, at 37°C and 5% CO2. These cells were procured from ATCC (Manassas, VA) and validated by STR profiling. All cell lines used were checked for mycoplasma contamination intermittently. For myotube formation, C2C12 cells at 80% confluence were induced to differentiate by replacing FBS containing medium with medium containing 2% horse serum for 72–96 h (Kubo, 1991; Sasi Kumar et al., 2015). Lipofectamine Plus, 2000, 3000, LTX, and RNA Imax from Invitrogen were used for transfection, following the manufacturer's instructions. Treatments of cells were as follows: nocodazole (Calbiochem), 1 µg/ml for 4 h; cytochalasin D (Calbiochem), 1 µg/ml for 30 min and hydroxyurea (Calbiochem), 3 mM for 30 h. For serum starvation (SS), cells were grown in medium containing 0.5% serum for 24 h. For re-stimulation (STI) to enter the cell cycle, starved cells were re-fed with 20% serum containing medium for 24 h. C3G wild-type (WT) and knockout (KO) C2C12 and MDA-MB-231 clones were generated using mouse- and human-specific CRISPR/CAS9 constructs obtained from Santa Cruz, as described previously (Shakyawar et al., 2018). Cells from the first passage were used for all experiments unless otherwise mentioned. PV treatment was given as previously described (Radha et al., 2004).
Antibodies and plasmids
Antibodies were obtained from the following sources: rabbit anti-C3G (H300; sc-15359, RRID:AB_2177452), mouse anti-C3G (G4; sc-17840), rabbit anti-C3G (C19; sc-869, RRID:AB_2177454), rabbit anti-pC3G Tyr514 (sc3262, RRID: AB_1125719), rabbit anti-pC3G Tyr504 (sc-12926, RRID:AB_2177453), rabbit anti-PCM1 (H262; sc-67204, RRID:AB_2139591), rabbit anti-cyclin D1 (M-20; sc-718, RRID:AB_2070436), rabbit anti-cyclin A (H-432; sc-751, AB_631329), rabbit anti-cyclin E (M-20; sc-481, RRID:AB_2275345), mouse anti-α-tubulin (B-7; sc-5286, RRID:AB_628411) and mouse anti-GFP (B2; sc-9996, RRID:AB_627695) were from Santa Cruz Biotechnology. The C3G antibodies used have been validated and confirmed by us, as well as by others, using strategies suggested by the International Working Group for Antibody Validation, or as indicated in https://www.nature.com/articles/d42473-018-00082-4. For western blotting and immunostaining, antibodies were used at a dilution of 1:500 (to detect endogenous C3G) and 1:1000 (to detect overexpressed C3G). Mouse anti-actin (MAB1501; RRID:AB_2223041) and mouse anti-centrin (clone 20H5; 04-1624, RRID:AB_10563501) were obtained from Millipore. Mouse anti-γ-tubulin (T-6557, RRID:AB_477584) and mouse anti-acetylated-tubulin (T-7451, RRID:AB_609894) were from Sigma-Aldrich. Rabbit anti-cenexin (ab43840, RRID:AB_880577) was from Abcam. Rabbit anti-C3G-GRF2 (rGRF2; NBP1-88266, RRID:AB_11032624) and mouse anti-C3G-GRF2-2F5 (mGRF2; NBP2-45515) were from Novus Biologicals. A monoclonal antibody generated in-house (3F6) using the CBR domain of C3G, was used to detect all the deletion constructs used (Begum et al., 2018). Cells were stained for F-actin using Rhodamine–phalloidin (Molecular Probes). All antibodies were used at a dilution of 1:200-1:300 for immunofluorescence and 1:500-1:1000 for western blotting, with the exception of anti-γ-tubulin and anti-actin, which were used at 1:2500 for immunofluorescence and 1:5000 for western blotting.
C3G–GFP, C3G–Flag, GFP–CBR, ΔC-C3G, ΔN-C3G, Y504F, control shRNA (con-Sh) and C3G shRNA (ShC) plasmids have been described previously (Dayma and Radha, 2011; Radha et al., 2007), GFP–cenexin and GFP–centrin were kindly gifted by Kyung Lee (National Institutes of Health, Bethesda, MD) and Michel Bornens (Institute Curie, Paris, France), respectively. pEGFPC1, C3G CRISPR/dCas9 double nickase KO Plasmids (mouse sc-430750 and human sc-401616), and C3G HDR Plasmid (mouse sc-430750-HDR and human sc-401616-HDR), cenexin siRNA (a mix of three oligonucleotides that target different, non-overlapping regions of mouse cenexin; mouse sc-43411) and control siRNA-A (sc-37007) were from Santa Cruz Biotechnology. Fluorophore-conjugated secondary antibodies were from Millipore and Amersham GE. An adenoviral vector expressing human C3G was generated using the AdEasy System (Shakyawar et al., 2018).
Immunofluorescence and image analysis
Cells grown on cover slips were fixed with cold methanol at −20°C for 6 min, washed with PBS and incubated with 2% BSA in PBS for 1 h. Indirect immunofluorescence was carried out as described previously (Shakyawar et al., 2017) and coverslips were mounted in mounting medium containing 4,6-diamino-2-phenylindole (DAPI). For dual labeling, cells were incubated first with primary antibody, then the corresponding secondary antibody, followed by the second primary antibody and, finally, its corresponding secondary antibody. Unless otherwise mentioned, the H300 antibody was used to detect C3G in all experiments. The processed cells were scanned on a Leica TCS SP8 confocal microscope. Images were captured in sequential scan mode to detect non-overlapping emission ranges (thereby avoiding bleed-through) of 415–465 nm, 500–545 nm, 566–615 nm and 640–690, following excitation with 405 nm diode (for DAPI), Argon 488 nm (for GFP/488 dye), DPSS 561 nm (for Cy3) and 633 nm HeNe (for Cy5) lasers, respectively. Emission was detected using Photomultiplier tube detectors. Similar acquisition parameters were maintained during capture of images from samples belonging to the same experiment. Images were analyzed for colocalization by measuring fluorescence intensity in an ROI drawn across the centrosome. Proportion of cells showing a primary cilium was quantitated by observing acetylated tubulin staining. Averages were obtained by examining multiple fields from each coverslip of independent experiments carried out in duplicate. Cilia length was determined from scanned images using the LAS X (version 18.104.22.16851) software embedded in the system provided by Leica. Averages were obtained from a minimum of 50 cells on each coverslip, and statistical significance obtained from independent experiments carried out in duplicate. Cells expressing C3G, or its deletion constructs, were examined for the presence of supernumerary centrioles by staining for γ-tubulin, and for cilia by staining for acetylated tubulin. Proportion of interphase cells with three or more γ-tubulin spots were considered as having supernumerary centrioles. Quantitative data was obtained by examining a minimum of 50 expressing cells and 100 non-expressing cells from each of the duplicate cover slips from independent experiments. All positively stained cells were considered as expressing, irrespective of the intensity of signal.
Immunoprecipitation and western blotting
Immunoprecipitation of over-expressed GFP-tagged protein was carried out using GFP-trap (Chromotek) as indicated by the manufacturer’s instructions. Immunoprecipitation of constructs lacking a GFP tag was carried out using the corresponding primary antibody and Protein A/G plus agarose as described previously (Mitra and Radha, 2010). Western blotting was carried out using standard protocols (Radha et al., 2007, 2004). ImageJ software was used for quantitation of band intensity, and intensities were adjusted using a loading control to estimate relative differences. Band intensities were averaged from three independent experiments to show quantitative differences.
An equal number of WT C2C12 and C3G KO C2C12 cells were seeded in a 96-well plate in triplicates. Cell viability was assessed at 24-h intervals for 72 h. A 20 μl volume of MTT was added and incubated for 5 h. Medium was removed and the purple formazon crystals trapped in cells were dissolved by adding 100 μl of DMSO for 30 min before absorbance at 540 nm was measured using a plate reader.
Cell cycle analysis
Cells were harvested as a single-cell suspension in PBS and added slowly to a Falcon tube containing 70% ethanol and incubated for 45 min. Cells were centrifuged and resuspended in Propidium Iodide staining buffer (Ormerod, 1994), then processed for cell cycle analysis using a Beckman coulter Gallios flow cytometer. Cell cycle profiles were analyzed using FlowJo X software.
Quantitative data was obtained as an average of multiple experiments and represented as mean±s.d. Two-tailed Student's t-tests were used to calculate significance of differences between two averages, and a one-tailed test was used to determine significance of the relative difference in a test sample compared to normalized controls.
We are grateful to Dr Kyung Lee, NIH and Dr Michel Bornens, Institute Curie, for the gift of plasmids. We thank Dr Jyotsna Dhawan, CCMB, for critical comments on the manuscript and Ms Divya Sriram, CCMB, for help with preparation of the manuscript and figures. We thank Mr B.V.V. Parthasaradhi, CCMB, for help with the MTT assay.
Conceptualization: V.R.; Methodology: S.C.N., V.R.; Validation: S.C.N.; Formal analysis: S.C.N., V.R.; Investigation: S.C.N.; Data curation: S.C.N., V.R.; Writing - original draft: V.R.; Writing - review & editing: V.R.; Visualization: S.C.N., V.R.; Supervision: V.R.; Project administration: V.R.; Funding acquisition: V.R.
We acknowledge funds received from the Council of Scientific and Industrial Research, India (BSC0108) and the Department of Biotechnology, Ministry of Science and Technology, India (BT/PR11759/BRB/10/1301/2-14) for carrying out this work.
The authors declare no competing or financial interests.