A Foxo/Notch pathway controls myogenic differentiation and fiber type specification (original) (raw)
Interaction of Foxo1 and Notch signaling in C2C12 differentiation. To understand whether Notch and Foxo interact to control muscle development, we used a cellular differentiation model. C2C12 cells undergo myogenic conversion and myotube fusion upon growth factor withdrawal, a process associated with Foxo1 nuclear translocation (3). Accordingly, transduction of adenovirus encoding a constitutively active Foxo1 mutant (Foxo1-ADA; a mutant Foxo1 with the following amino acid substitutions: T24A, S253D, and S316A) (4) blocked the effect of serum withdrawal to induce C2C12 differentiation, as reflected by inhibition of myoblast fusion (Figure 1, A–C). Conversely, Foxo1 inhibition by siRNA did not affect these processes (Figure 1D). Similarly, constitutively active Notch1 (Notch1-IC) phenocopied Foxo1-ADA in blocking myoblast differentiation (Figure 1E). Virtually all cells became transduced with the adenoviruses (Supplemental Figure 1; supplemental material available online with this article; doi:10.1172/JCI32054DS1). Foxo1 siRNA effectively suppressed expression of both endogenous Foxo1 and transfected FLAG-Foxo1 (Supplemental Figure 2) in a dose-dependent manner, without affecting control proteins or other Foxo isoforms (Supplemental Figure 3). Neither Foxo1-ADA nor Notch1-IC affected C2C12 proliferation (Supplemental Figure 4).
Regulation of myoblast differentiation by Foxo and Notch. C2C12 cells were immunostained with anti-myosin antibody (green) and DAPI (blue). See text for panel description. Each experiment was repeated at least 6 times. Original magnification, ×10.
We asked whether we could preempt the effect of Foxo1-ADA by inhibition of endogenous Notch signaling. To this end, we used a truncated Notch1 receptor lacking the transmembrane anchor and intracellular domain, which acts as a decoy receptor by binding Notch ligands (32, 33) (our unpublished observations). The decoy did not affect C2C12’s ability to undergo differentiation in response to growth factor withdrawal (Figure 1F) but partly rescued Foxo1-ADA inhibition of myoblast differentiation (Figure 1G). As an alternative probe to block Notch signaling, the presenilin inhibitor (PSI) compound E (34) also rescued Foxo1-ADA inhibition of myoblast differentiation (Figure 1H).
To examine the effect of Foxo1 on Notch signaling, we cotransfected Foxo1 siRNA and Notch1-IC. Foxo1 siRNA rescued inhibition of myoblast differentiation and myosin expression by Notch1-IC (Figure 1I) while control siRNA had no effect (data not shown). To rule out nonspecific effects of Foxo1 siRNA on myoblast differentiation, we generated an siRNA-resistant Foxo1-ADA (Supplemental Figure 5). Foxo1 siRNA reversed the effects of Foxo1-ADA (Figure 1J) but failed to rescue inhibition of C2C12 differentiation caused by siRNA-resistant Foxo1-ADA (Figure 1K). We present a quantitative analysis of these data in Figure 2A, showing that Foxo1 and Notch1-IC decreased myosin levels by more than 80% while Notch decoy and Foxo1 siRNA restored them to approximately 70% of fully differentiated cells. We obtained similar data by performing a morphometric analysis of myosin-positive cells (Figure 2B). These data indicate that Foxo1 is required for the effect of Notch on myoblast differentiation.
Quantitative analysis of C2C12 differentiation. (A) Western blotting analysis of myosin expression in C2C12 cells. (B) Morphometric analysis of myosin-positive cells. Results from differentiation experiments were analyzed by scoring the number of myosin-immunostained cells as a percentage of all DAPI-positive cells. (C) DBD-Foxo1-ADA reporter gene assays. We carried out reporter gene assays using the canonical Foxo1-responsive Igfbp1 promoter (left panel) and the Hes1 promoter (right panel) in cells cotransfected with Foxo1-ADA or DBD-Foxo1-ADA. Western blot (inset) demonstrates that expression levels of the 2 proteins are similar. An asterisk indicates P < 0.01 by ANOVA.
We next determined whether Foxo1 affects differentiation via its transcriptional function. To this end, we generated a DNA-binding deficient (DBD) mutant in the backbone of the ADA mutant by replacement of N208A and H212R (DBD-Foxo1-ADA) (6, 35). We confirmed that this mutant is unable to bind DNA by measuring insulin-like growth factor–binding protein 1 (Igfbp1) promoter activity, a canonical Foxo1 target. Foxo1-ADA increased Igfbp1 promoter activity by 10-fold whereas DBD-Foxo1-ADA was unable to do so (Figure 2C). Surprisingly, this mutant was as effective as the DNA binding–competent Foxo1-ADA at inhibiting differentiation (Figure 1L). These data indicate that Foxo1 controls differentiation independently of its ability to bind DNA in a sequence-specific manner.
Foxo1 binds to Csl and is recruited to the Hes1 promoter. Notch1-IC binds to and coactivates Csl to promote Hes and Hey expression (11). Based on the results with the DBD-Foxo1-ADA mutant, we determined whether Foxo1 interacts with Csl in a Notch-dependent manner using coculture of C2C12 cells expressing Notch1 receptor with HEK293 cells expressing the Notch ligand Jagged1 or LacZ as a negative control. We provide several lines of evidence that Foxo1 and Csl interact in cultured cells. We detected endogenous Foxo1 in endogenous Csl immunoprecipitates, and the coimmunoprecipitation was significantly enhanced by activation of Notch signaling (Figure 3A). To confirm the specificity of the interaction, we expressed HA-tagged Foxo1 and FLAG-tagged Csl in C2C12 cells. Following immunoprecipitation with anti-HA (Foxo1) antiserum, we detected FLAG-Csl in immunoblots (Figure 3B). Conversely, following immunoprecipitation with anti-FLAG (Csl) antiserum, we detected HA-Foxo1 in immunoblots (Figure 3C). The ability to coimmunoprecipitate with Csl appears to be specific to Foxo1, as we failed to detect other Foxo isoforms in Csl immunoprecipitates (Supplemental Figure 6). A truncated Foxo1 mutant (Δ256, encoding aa 1–256) (36) retained the ability to interact with Csl. We detected FLAG-Csl in immunoprecipitates (Figure 3D) and HA-Δ256 in FLAG-Csl immunoprecipitates (Figure 3E), indicating that Csl interacts with the Foxo1 N terminal domain.
Foxo1 coimmunoprecipitates with Csl. (A) Coimmunoprecipitation of endogenous Foxo1 and Csl in C2C12 cells cocultured with LacZ-expressing (denoted by the minus sign) or Jagged1-expressing HEK293 cells (denoted by the plus sign). (B and C) Coimmunoprecipitation experiments in C2C12 cells cotransfected with FLAG-Csl and HA-Foxo1. (D and E) Coimmunoprecipitation experiments in C2C12 cells cotransfected with FLAG-Csl and the truncated mutant Myc- or HA-tagged Δ256 Foxo1. TCL, total cellular lysate.
To determine whether this is a direct protein-protein interaction and map the interaction domain(s), we first carried out pull-down assays with affinity-purified glutathione-S-transferase–Foxo1 (GST-Foxo1) produced in bacteria and FLAG-Csl expressed in HEK293 cells. We detected Csl association with full-length and N terminal Foxo1 (aa 1–300) but not with C terminal Foxo1 (aa 290–655) or GST (Figure 4A). We next mapped the Csl domain that interacts with Foxo1 using a cell-free system with GST-Foxo1 and GST-Flag-Csl purified from bacterial cultures. Again, we recovered full-length (aa 1–655) and N terminal (aa 1–300) but not C terminal (aa 290–655) Foxo1 in Csl immunoprecipitates. Conversely, N terminal Foxo1 interacts with N terminal Csl (Figure 4B).
Foxo1 binds directly to Csl. (A) GST pull-down assays of GST-Foxo1 fusion protein with Csl immunoprecipitated from HEK293 cells. (B and C) Binding of GST-Foxo1 and GST-FLAG-Csl in a cell-free system and mapping of the Csl interaction domain. Full-length and truncated fragments of GST-Foxo1 and GST-FLAG-Csl were purified from bacteria and coincubated. Thereafter, Csl was isolated using anti-FLAG antibody, and the immunoprecipitate was analyzed by immunoblotting with anti-Foxo1 or anti-FLAG antibodies. (D) Hes1 promoter ChIP assay spanning the Csl-binding site in C2C12 cells to detect endogenous Foxo1, Csl, and Notch1 (Endog) or following transduction with Foxo1-ADA during myoblast differentiation. Input represents DNA extracted from chromatin prior to immunoprecipitation. Hes1 (semiquantitative RT-PCR) and myosin (Western blot) expression corresponding to each time point are shown. Day 0 is defined as the time when cells were serum deprived to induce myoblast fusion.
We used Csl deletion mutants to map the Foxo1-binding domain in Csl. These studies indicate that Foxo1 binds to a domain encompassing aa 172–279 (Figure 4C), which is contained within the Csl NH2 terminal domain (NTD) domain (37) (Figure 4C). Interestingly, this domain is required for DNA and corepressor binding but does not contribute to Notch binding (38, 39).
Csl binds to a consensus sequence in the Hes1 promoter (40), which thus provides a useful readout assay of the Foxo/Csl interaction. If the latter were required to regulate C2C12 differentiation, 3 predicted conditions should be met: (a) Foxo1 should be detected in chromatin immunoprecipitation (ChIP) assays spanning the Csl element in the Hes1 promoter; (b) the interaction should be differentiation dependent; and (c) inhibition of differentiation by Foxo1-ADA should be accompanied by constitutive binding to the Csl element in the Hes1 promoter. Figure 4D demonstrates that all predictions are fulfilled. First, we performed ChIP assays using primers spanning the Csl-binding site of Hes1 in differentiating C2C12 cells. We detected endogenous Foxo1, Notch1, and Csl in immunoprecipitates from undifferentiated cells (Figure 4D). As the PCR-amplified sequence contains no forkhead binding sites, we concluded that Foxo1 binds to this DNA fragment via Csl. Moreover, binding of both Foxo1 and Notch1 decreased as cells became differentiated (days 1 and 2). When we transduced cells with constitutively nuclear Foxo1-ADA, differentiation was inhibited (Figure 1C) and the mutant Foxo1 was persistently bound to the Hes1 promoter, as were Csl and Notch1 (Figure 4D).
We next analyzed Hes1 expression. The prediction was that Hes1 levels should correlate with occupancy of the Hes1 promoter by Foxo1 and Notch1. Indeed, Hes1 mRNA expression declined as Foxo1 and Notch1 binding to Csl decreased while myosin protein levels increased (Figure 4D). To rule out a direct effect of Foxo1 on Csl transcription, we carried out reporter gene assays with the Csl promoter. Foxo1 failed to activate expression of a Csl reporter gene despite the presence of 10 repeats of a forkhead binding site in the Csl promoter (ref. 41 and data not shown). Moreover, Csl expression was unaffected in C2C12 cells expressing Foxo1-ADA (data not shown). These data indicate that Foxo1 regulates Notch-dependent differentiation via protein/protein interactions with Csl.
Foxo1 is required for Notch induction of Hes and Hey genes via Csl. We examined the ability of Foxo1-ADA to promote expression of endogenous Hes1, Hes5, and Hey1 in C2C12 cells. Both Foxo1-ADA and Notch1-IC increased the expression of the 3 genes while Foxo1 siRNA inhibited Hes1, Hes5, and Hey1 expression induced by Notch1-IC (Figure 5A). Foxo1 siRNA had no effect on Hes1, Hes5, and Hey1 expression in growth factor–deprived cells (Figure 5A).
Foxo1 regulates Notch-induced Hes1, Hes5, and Hey1 expression. (A) Hes1, Hes5, and Hey1 expression measured by semiquantitative RT-PCR in C2C12 cells transduced with Foxo1-ADA or Notch1-IC following transfection of GFP, Foxo1, or Csl siRNA as indicated. (B) Hes1 reporter gene assays in HEK293 cells transduced with Foxo1-ADA, Notch1-IC, Foxo1 siRNA, GFP siRNA, or control plasmid (empty). We measured luciferase activity and normalized it by β-galactosidase activity. The data represent arbitrary units relative to control empty vector.
We focused the next set of experiments on Hes1, as a prototypical Notch target gene. We tested Foxo1’s ability to regulate Hes1 transcription using reporter assays with the Hes1 promoter as well as measurements of Hes1 expression. Foxo1-ADA and Notch1-IC induced Hes1 promoter activity by 1.8- and 2.5-fold, respectively. Cotransfection of Foxo1-ADA with Notch1-IC caused a 2.5-fold increase (Figure 5B). Cotransfection of Foxo1 siRNA suppressed Notch-induced Hes1 activity in a dose-dependent manner while control siRNA had no effect (Figure 5B). We obtained similar results with a synthetic Hes1 reporter containing 4 tandem repeats of the Csl-binding motif (Supplemental Figure 7). Moreover, DBD-Foxo1-ADA was able to induce Hes1 reporter gene activity to an even greater extent than Foxo1-ADA, confirming that direct DNA binding is not required for Foxo1 activation of Hes1 (Figure 2C).
The failure of Notch1-IC to induce Hes1 expression in cells expressing Foxo1 siRNA suggests that Foxo1 is required for Csl/Notch interaction. Thus, we investigated the binding of Foxo1 and Notch1 to the Hes1 promoter in a coculture system. We cocultured C2C12 cells expressing Notch1 with HEK293 cells expressing the Notch ligand Jagged1 to induce activation of endogenous Notch signaling. Coculture in the presence of Jagged1-expressing cells increased endogenous Foxo1 (Figure 6A) and Notch1 binding to the Hes1 promoter in ChIP assays (Figure 6, A and B) (42). These data are consistent with the observation that Foxo1 coimmunoprecipitation with Csl increased upon coculture (Figure 3A). To determine whether Foxo1 binding to the Hes1 promoter is Csl dependent, we inhibited Csl expression with siRNA (Supplemental Figure 8). Transfection of Csl siRNA inhibited both Foxo1 and Notch1 binding to Hes1 promoter (Figure 6A), indicating that they are Csl dependent. Moreover, Foxo1-ADA failed to induce Hes1 expression in the presence of Csl siRNA (Figure 5A). The results of ChIP experiments were corroborated by Hes1 promoter assays. Expression of Jagged1 or Notch1 alone had no effect on Hes1 activity, but coculturing yielded a 3.7-fold increase in Hes1 reporter gene activity (Figure 6C). Foxo1 siRNA abolished Notch binding to the Hes1 promoter in ChIP assays (Figure 6B) and induction of Hes1 promoter activity (Figure 6C). These results suggest that Foxo1 is required for binding of Notch1 to the Hes1 promoter and provide a mechanism whereby inhibition of Foxo1 expression restores differentiation of myoblasts expressing Notch1-IC. The ability of Foxo1 siRNA to inhibit Notch induction of Hes1 in a coculture system rules out the possibility that the effects observed in differentiation experiments with Notch1-IC are due to nonphysiologic activation of Notch signaling by the truncated intracellular Notch1 mutant (15).
Foxo1 is required for Notch binding to the Hes1 promoter and activation of Hes1 target genes. (A) ChIP assays of endogenous Foxo1 and Notch1 in C2C12 cells cocultured with LacZ-expressing (denoted by a minus sign) or Jagged1-expressing HEK293 cells (denoted by a plus sign) in the absence (lanes 1 and 2) and presence (lanes 3 and 4) of Csl siRNA. (B) ChIP assays of endogenous Notch1 in coculture system in the absence (lanes 1 and 2) and presence (lanes 3 and 4) of Foxo1 siRNA. (C) Hes1 promoter assays following coculture in the absence and presence of Foxo1 or GFP siRNA. (D) ChIP assays of NcoR and Smrt and Maml1 binding to Hes1 in the coculture system in the absence (lanes 1 and 2) and presence (lanes 3 and 4) of Foxo1 siRNA. (E) Expression of MyoD, Myf5, and β_-actin_ in C2C12 cells by semiquantitative RT-PCR. (F) Model of Foxo1 and Notch regulation of Hes1 promoter.
Foxo1 promotes corepressor clearance and Maml1 binding to Csl. To clarify the molecular mechanism of Foxo1-dependent activation of Hes1 expression, we investigated corepressor/coactivator exchange at the Hes1 promoter. Activation of Notch cleared the corepressors nuclear corepressor (NcoR) and silencing mediator for retinoid and thyroid hormone receptor (Smrt) (43) and recruited the coactivator mastermind-like 1 (Maml1) (42) to the Hes1 promoter. Foxo1 siRNA prevented Notch-induced corepressor exchange (Figure 6D). These data are consistent with the observation that Foxo1 binds to the region 172–279 of Csl (Figure 4C), which has been shown to contain the NcoR/Smrt binding sites (38, 39).
To demonstrate that the observed changes in the transcriptional complex result in changes in Hes1 activity, we investigated expression of Hes1 target genes involved in myogenesis. Hes1 has been proposed to suppress myoblast differentiation by inhibiting the basic helix-loop-helix transcription factor MyoD without affecting Myf5 (16, 17). Expression analyses revealed that Notch1-IC or Foxo1-ADA suppressed MyoD, while Myf5 was unaffected. Notch decoy or Foxo1 siRNA partly restored MyoD expression (Figure 6E).
Altered fiber type composition in skeletal muscle lacking Foxo1. Based on the cellular data, we undertook to probe Foxo1 function in muscle differentiation in vivo using conditional gene inactivation. The predicted outcome of this experiment was accelerated differentiation of MyoD-containing but not Myf5-containing myoblasts. Because MyoD is the predominant myogenic factor in fast fibers while myogenin is the predominant factor in slow fibers (44), the removal of Foxo/Notch inhibition on MyoD expression should result in increased formation of fast fibers, potentially at the expense of slow fibers.
There are 3 Foxo isoforms in mice: Foxo1, Foxo3, and Foxo4 (8, 9). The latter is predominant in most muscle types (45) except soleus, where Foxo1 is the most abundant (Figure 7A). Coincidentally, soleus is also physiologically enriched in slow-twitch fibers and thus allowed us to readily test our hypothesis. We inactivated Foxo1 expression in skeletal muscle by crossing mice homozygous for a floxed Foxo1 allele with myogenin-cre transgenics. mRNA analysis indicated that the knockout occurred as planned (data not shown). Histological analyses revealed a reduction of type I (slow-twitch) fibers in soleus of myogenin-Foxo1 (Myog-Foxo) mice while type II fiber–enriched muscles were unaffected (Figure 7B). Consistent with the histological findings, expression of type I fiber markers decreased while type II fiber markers increased in Myog-Foxo1 mice (Figure 7C). We then analyzed expression of the myogenic transcription factors MyoD, Myf5, and myogenin. MyoD is the predominant factor in fast fibers and myogenin in slow fibers (44). Consistent with the histopathology, we found a 2-fold increase in MyoD expression and an approximately 80% decrease in myogenin while Myf5 expression was unchanged (Figure 7C). Moreover, expression of the Foxo1 coactivator Pgc1α, which regulates type I fiber determination (30) was unchanged, indicating that the phenotype of Myog-Foxo1 mice cannot be accounted for by decreased Foxo1-dependent Pgc1α transcription (Figure 7C) (46). As a functional correlate of the observed fiber type switch, we examined running performance on a treadmill. Indeed, Myog-Foxo1 mice displayed reduced running capacity, as predicted from the reduction in type I (endurance) fibers (Figure 7D).
Conditional ablation of Foxo1 in skeletal muscle. (A) Western blot analysis of Foxo1 and Foxo4 expression levels in various muscle types. Gastroc, gastrocnemius muscle; Vastus, vastus lateralis muscle. (B) Metachromatic and immunohistochemical analysis of soleus and plantaris muscle from Myog-Foxo1 mice and control (lox/lox) littermates. Original magnification, ×10. (C) Gene expression analysis of Myog-Foxo1 (black bars) and control mice (white bars); TropC, troponin-C; TropT, troponin-T; Mlc, myosin light chain; Myog, myogenin; Mck, muscle-type creatine kinase. Data are means ± SEM of 3 independent measurements (n = 6 for each genotype). An asterisk indicates P < 0.05 by ANOVA. (D) Treadmill performance test in 8-week-old Myog-Foxo1 mice and lox/lox littermates (n = 6 for each genotype). An asterisk indicates P < 0.05 by ANOVA.
Finally, to determine whether these changes reflected developmental alterations in fiber-type specification as opposed to adaptive or cell-nonautonomous factors, we determined MyoD expression in Foxo1 (24) and Notch1 knockout (25) embryos at E9.5. In Foxo1–/– embryos, MyoD levels increased 3.1 ± 1.1–fold, and in Notch1–/– embryos 7.3 ± 2.9–fold compared with controls (P < 0.05 in both mutants versus wild type, n = 4). The increase in MyoD expression observed in vivo is consistent with the physical and functional interactions between Foxo1 and Notch at this key signaling nexus in myoblast differentiation. Thus, we propose that the fiber-type switch in Myog-Foxo1 mice is the result of accelerated differentiation of MyoD-containing myoblasts during embryonic development.