|Home | About | Journals | Submit | Contact Us | Français|
SAG/RBX/ROC protein is an essential RING component of SCF E3 ubiquitin ligase. The role of SAG during embryogenesis remains unknown. We report here a critical role for SAG in controlling vascular and neural development by modulating RAS activity via promoting degradation of neurofibromatosis type 1 (NF1). Mice mutant for Sag died at embryonic day 11.5-12.5 with severe abnormalities in vascular and nervous system. Sag inactivation caused Nf1 accumulation and Ras inhibition, which blocks embryonic stem (ES) cells from undergoing endothelial differentiation and inhibits angiogenesis and proliferation in teratomas. Simultaneous Nf1 deletion fully rescues the differentiation defects in Sag−/− ES cells, and partially rescues vascular and neural defects in Sag−/− embryos, suggesting that the effects of Sag deletion may not be solely explained by Nf1 misregulation. Collectively, our study identifies NF1 as a physiological substrate of SAG-CUL1-FBXW7 E3 ligase and establishes a ubiquitin-dependent regulatory mechanism for the NF1-RAS pathway during embryogenesis.
The SCF E3 ubiquitin ligase is the largest family of E3 ligases, consisting of SKP1, Cullins, F-box proteins, and a RING protein, RBX (RING Box protein-1), also known as ROC (Regulators of Cullins). The substrate specificity of SCF complex is determined by the F box proteins that bind to SKP1 and Cullins through its F-box domain and to substrates through its WD40 or LRR domains (Jin et al., 2004; Zheng et al., 2002), whereas the core SCF E3 ubiquitin ligase is a complex of Cullins-RBX, in which RBX binds to E2 and facilitates ubiquitin transfer from E2 to the substrates (Wu et al., 2000). By promoting the ubiquitination of various regulatory proteins for degradation by 26S proteasome, SCF ligases regulate many biological processes, including apoptosis, cell cycle progression, signal transduction and DNA replication (Deshaies and Joazeiro, 2009; Nakayama and Nakayama, 2006). Although a large number of F-box proteins were found in the human genome (Jin et al., 2004) that selectively target various protein substrates, there are only two family members of RING proteins in human or mouse, RBX1 and RBX2, also known as SAG (Sensitive to Apoptosis Gene) (hereafter referred to as SAG) (Duan et al., 1999; Kamura et al., 1999; Ohta et al., 1999; Tan et al., 1999; Wei and Sun, 2010). Both family members, having a functional RING domain at the carboxyl terminus, are evolutionally conserved with a similar tissue expression pattern (Sun et al., 2001). While either RBX1 or SAG is capable of binding to six members of cullin family (CUL 1-3, CUL4A, B and CUL-5) and has in vitro E3 ubiquitin ligase activity when complexed with cullin-1 (Furukawa et al., 2002; Swaroop et al., 2000), RBX1 is constitutively expressed and prefers to bind with CUL2/VHL, whereas SAG is stress-inducible and preferably binds to CUL5/SOCS (Gu et al., 2007; Kamura et al., 2004).
The role of SAG during development has been previously studied in several model organisms. The Hrt1, the only yeast homologue of RBX1/SAG, is a growth-essential gene whose targeted disruption causes yeast death, which can be fully rescued by either human RBX1 (Ohta et al., 1999; Seol et al., 1999) or SAG (Swaroop et al., 2000). In C. elegans, siRNA knockdown of Rbx1 (ZK287.5) induced the death during embryogenesis as well as in adult animals (Jia et al., 2011; Sasagawa et al., 2003), whereas knockdown of Rbx2/Sag (R10A10.2) did not cause any significant phenotypic changes (Moore and Boyd, 2004). In Drosophila, disruption of Roc1a or Roc1b caused lethality or male sterility, respectively, whereas disruption of Roc2/Sag caused no overt developmental phenotype (Donaldson et al., 2004; Reynolds et al., 2008). These studies suggest that RBX2 homologues in C. elegans and Drosophila are functionally redundant and RBX2 loss can be compensated by its family member, RBX1 or ROC1a/1b during embryogenesis. Our recent knockout study revealed that the in vivo functions of Rbx1 and Sag are non-redundant in mice. In Sag wild-type background, the Rbx1 inactivation caused early embryonic lethality at embryonic day 7.5 (E7.5) due to proliferation defects, which is partially caused by p27 accumulation, as simultaneous deletion of p27 extended the life span of Rbx1 deficient embryos from E6.5 to E9.5. Thus, one in vivo physiological function of Rbx1 is to ensure cell proliferation by preventing p27 accumulation during the early stage of embryogenesis (Tan et al., 2009). However, the role of Sag during mouse embryogenesis is unknown.
The NF1 tumor suppressor gene is frequently mutated in many types of sporadic human cancers including glioblastoma multiforme, malignant peripheral nerve sheath tumor and epithelial cancers in the ovary and lung (Ding et al., 2008; Parsons et al., 2008; Sangha et al., 2008; TCGA Research Network, 2008; Woodruff, 1999). Individuals with NF1 mutations are predisposed to developing a variety of benign and malignant tumors, many of which affect the peripheral and central nervous system (Cichowski and Jacks, 2001). NF1 contains a functional domain that shows homology to the members of the RAS GTPase Activating Protein (GAP) family. As a GAP, NF1 negatively regulates RAS proto-oncogene by accelerating the conversion of active RAS-GTP to inactive RAS-GDP (Le and Parada, 2007). Upon growth factor stimulation, NF1 is rapidly degraded by the ubiquitin-proteasome pathway (Cichowski et al., 2003; McGillicuddy et al., 2009) to ensure a proper activation of RAS signals for proliferation. Previous studies on NF1 ubiquitination and degradation only focused on its GAP-related domain (GRD). One study found that a fragment containing NF1 GRD domain and its adjacent 80 amino acids (codons 1096-1534) was degradation-sensitive, but did not identify corresponding E3 ubiquitin ligase (Cichowski et al., 2003). A recent study reported that ETEA, a UBA and UBX domains-containing protein, bound to NF1 GRD domain, and negatively regulated NF1 levels (Phan et al., 2010). However, ETEA is not a bona fide E3 ubiquitin ligase. Thus, the E3 ubiquitin ligase responsible for targeted degradation of NF1 remains unknown.
In this study, we used a knockout approach and revealed the critical roles of Sag in embryonic vascular and neural development, and endothelial differentiation of ES cells. We found that Sag disruption caused Nf1 accumulation to inhibit Ras-Mapk signals. Simultaneous Nf1 deletion rescues the defects fully in endothelial differentiation of Sag−/− ES cells and partially in developing vascular and nervous system of Sag−/− embryos, suggesting Nf1 misregulation contributes partly to the defects derived from Sag deletion . We further showed that NF1 is a physiological substrate of SAG-CUL1-FBXW7 E3 ubiquitin ligase. Upon recognized and bound by FBXW7, NF1 is targeted for ubiquitination and degradation by SAG E3. Thus, our study reveals the mechanisms that regulate the stability of NF1 proteins, which might provide insights on designing novel therapy for NF1-related diseases.
To understand the in vivo physiological function(s) of SAG, we inactivated Sag in mice via a gene-trap approach. Mouse Sag (NM_011279) was mapped onto chromosome 9 with three exons and two introns (Fig. S1A). One ES clone (XE423) with a gene-trap vector inserted in the second intron of the Sag gene was identified (Nord et al., 2006) and characterized by PCR genome walking (Fig. S1A). The insertion disrupts the Sag transcript resulting in a truncated fusion mRNA that encodes partial Sag N-terminal sequence without the RING domain and Sag function (Sun, 1999). The confirmed ES clone was injected into C57BL/6 blastocysts to obtain chimeras, which were backcrossed with C57BL/6 mice to obtain germline transmission. Offspring carrying the Sag genetrap allele were identified by PCR genotyping (Fig. S1B), and further confirmed by genomic Southern blot analysis (Fig. S1C&D). Expression of Sag protein in Sag+/+, but not in Sag−/− embryos was detected at E10.5 (Fig S1E).
Intercrossing of heterozygous mice, which survived to term without any observable abnormalities, failed to generate Sag-disrupted homozygous mice among a total of 624 genotyped offsprings. The ratio of wild-type mice to heterozygous mice was about 1:2, exactly as expected for a condition with homozygous lethal phenotype (Fig. S1F). To define at which stage of development the Sag−/− embryos die, we dissected uteri of pregnant mice between E9.5 and E15.5. No viable Sag−/− embryos were detected after E13.5 days, whereas a few viable Sag−/− embryos were detected at E11.5-12.5 with reduced frequency. At E9.5-10.5, a Mendelian ratio of 1:2:1 (+/+:+/−:−/−) was observed (Fig. S1F). Thus, homozygous deletion of Sag causes embryonic lethality at E11.5-12.5.
Dissection of Sag−/− embryos at E10.5 revealed that Sag−/− embryos were growth-retarded with frequent pericardial edema (Fig. 1A, arrow). Furthermore, the yolk sacs of Sag−/− embryos have dramatic reduction of blood vessels, which was confirmed by CD31/PECAM whole-mount staining for endothelial cells (Fig. 1B). Compared to Sag+/+ embryos with extensive vascular network of both large and small vessels in the head region (Fig. 1C), Sag−/− embryos exhibited disrupted vasculature with missing major blood vessels and disorganized secondary branches (Fig. 1D). Similar vascular defects were also found in Sag−/− E9.5 embryos (data not shown). H&E stained sagittal sections revealed a substantial reduction in blood vessels and the number of nucleated red blood cells within blood islands in the yolk-sac of Sag−/− embryos (Fig. 1E, top). Vascular defects in the head region were also evident with a reduction in the number of blood vessels surrounding and within the neuroepithelium (Fig. 1E&1F).
To dissect the underlying mechanisms responsible for the observed vascular defects in Sag−/− embryos, we turned into an in vitro cell culture system to circumvent early lethality caused by Sag deficiency. We generated Sag−/− embryonic stem (ES) cells from blastocysts, obtained from Sag+/− inter-crosses. Sag+/+ ES cells underwent endothelial differentiation to form cystic embryoid bodies (cEBs) (Fig. 2A) with blood island structures, lined with endothelial cells (Fig 2B, arrows). In contrast, Sag−/− ES cells completely failed to differentiate into cEBs, though they did form EBs (Fig. 2A&B). These results suggest that Sag is required for EBs to undergo endothelial differentiation to form cEBs under in vitro cultured conditions. We next determined the growth rate and incidence of teratoma derived from ES cells upon injected into in vivo nude mice. We used two independent pairs of Sag−/− vs. Sag+/+ ES cells and found that teratomas derived from Sag−/− ES cells grew much slower and formed significantly smaller tumors with a lower tumor incidence (AB1 teratomas developed in 2 out of 7 mice) (Fig. 2C, Fig. S2A). Teratomas derived from Sag−/− ES cells also had a significantly reduced blood vessel density (Fig. 2D&E). Immuno-costaining experiment revealed that these blood vessels were derived from transplanted ES cells, not from host progenitor cells, since the endothelial cells co-expressed both CD31 and β-gal (targeting vector contains β-gal cDNA) in teratomas derived from Sag−/−, but not from Sag+/+ ES cells (Fig. S2B). Furthermore, we observed a reduced proliferation rate in teratomas from Sag−/− ES cells, compared to those from Sag+/+ ES cells (Fig. 2F). Thus, Sag disruption remarkably inhibits angiogenesis and growth/proliferation of teratomas in vivo.
Because RAS/MAP kinase pathways are known to regulate proliferation, angiogenesis and endothelial cell differentiation (Kawasaki et al., 2008; Kranenburg et al., 2004), we sought to determine whether the Ras signaling pathway is inactivated during Sag−/− ES cell differentiation. As shown in Figure 3A, phosphorylation of p44/Erk1 was significantly induced during differentiation of Sag+/+ EBs, but not in Sag−/− EBs. No difference was seen in the levels of total or phosphorylated forms of Akt and p38. A Ras activity assay further showed that Ras was active in ES cells, regardless of Sag status, but was completely inactive in EBs derived from Sag−/− ES cells (Fig. 3B). Similar results were obtained from another independent set of ES clones (not shown). Thus, Ras activation and Erk1 phosphorylation were maintained during cystic EB formation, whereas they were inactivated in the absence of Sag. We further found that FTI277, an inhibitor of Ras farnesyltransferase (Kawasaki et al., 2008) that inhibited p44/Erk1 phosphorylation (Fig. 3C), could significantly reduce cystic EB formation from Sag+/+ ES cells (Fig. 3D). A similar result was obtained with PD098059, a Mek inhibitor (Fig S3). Consistently, a reduced Erk phosphorylation, an indicator of Mek inactivation, was found in Sag−/− teratoma tissues, as compared to Sag+/+ teratomas (Fig 3E). Taken together, the findings demonstrated that the inability to maintain active Ras/Erk signals, upon Sag disruption, contributes significantly to the failure of endothelial differentiation in vitro, angiogenesis, and possibly reduced proliferation in teratomas in vivo.
RAS signaling pathways are inhibited by naturally occurring inhibitory proteins or inactivating phosphatases (Karnoub and Weinberg, 2008; Kolch, 2005). These proteins, upon Sag disruption, would accumulate to inactivate the RAS pathway during endothelial differentiation if they were the direct substrates of SAG-SCF E3 ubiquitin ligases. We therefore examined the levels of several Ras signaling inhibitory proteins in ES cells and in EBs formed during differentiation. As shown in Figure 3F, the basal level of neurofibromin (encoded by Nf1) was slightly higher in Sag−/− ES cells than that in Sag+/+ ES cells. The Nf1 level gradually increased, reaching the peak at 6 day of Sag−/− EB differentiation, but remained low in Sag+/+ EBs (panel 2). No significant difference was observed between the two groups in the levels of p120RasGap, a family member of Nf1 with a similar Ras inhibiting activity (Henkemeyer et al., 1995), nor of other Ras/Raf inhibitors or phosphatases of the Ras pathway tested (Dhillon et al., 2007b; Kolch, 2005), including Spred2, Mkp-1, and PP-2A (subunits B and C), except Rkip, a Raf inhibitor (Hagan et al., 2006), which showed higher levels in Sag+/+ ES cells and EBs, but was not regulated by ES cell differentiation (Fig. 3F). RT-PCR analysis revealed that Nf1 mRNA remained consistent during ES cell differentiation, regardless of Sag status (Fig. 3G), indicating that Nf1 accumulation upon Sag disruption likely results from a reduced degradation. We further measured the Nf1 levels in mouse embryonic fibroblasts (MEF) derived from E10.5 embryos with three Sag genotypes and found an elevated Nf1 level in Sag−/− MEFs, as compared to that in Sag+/+ or Sag+/− MEFs (Fig. 3H). Taken together, our results suggested that Nf1 could be a substrate of Sag-SCF E3 ubiquitin ligase, which accumulates upon Sag disruption, to inhibit Ras/Erk signals.
The crystal structure of SCF complex (Zheng et al., 2002) reveals that the substrate is recognized and recruited by an F-box protein to the ligase complex for ubiquitination. Thus, Sag-mediated Nf1 degradation must involve an F-box protein that recognizes Nf1 for binding. Examination of the NF1 sequence for the consensus binding motifs of several F-box proteins revealed an evolutionarily conserved perfect binding site for FBXW7 (also known as CDC4) (L-I/L/P-pT-P-XXXX) at the C-terminus (Fig. S4A), which is consistent with the binding motif found in other known substrates of FBXW7 (Fig. S4B). No FBXW7 binding site was identified in p120RasGAP, a family protein not subject to proteasome-dependent proteolysis (Cichowski et al., 2003). We therefore determined the potential binding of FBXW7 and NF1. Using an immunoprecipitation assay, we found that transiently transfected FBXW7 pulled down endogenous NF1 (Fig. 4A). In a reciprocal experiment, NF1C, a C-terminal portion of NF1 (codons 2180-2840) containing the FBXW7 binding site, pulled down both FBXW7 and SAG when co-transfected, indicating in vivo binding of SAG-FBXW7-NF1 (Fig. 4B). Furthermore, SAG, when co-transfected with FBXW7, pulled down FBXW7, as well as endogenous NF1 and CUL1 (Fig. 4C), indicating an in vivo formation of SAG-CUL1-FBXW7 E3 ligase. More significantly, SAG cotransfection promoted FBXW7-mediated ubiquitination of endogenous NF1, as demonstrated by slower migrating bands (Fig. 4C, top panel), which was not observed when FBXW7 was transfected alone (Fig. 4A, top). In order to detect endogenous FBXW7-NF1 binding, we examined several commerically available antibody against FBXW7 and identified one from Sigma that was capable of detecting overexpressed FBXW7 (data not shown). Using this antibody, we were able to detect endogenous FBXW7-NF1 binding, as evidenced by detection of FBXW7 (likely being enriched when complexed with NF1) in NF1 immunoprecipitates (Fig 4D, top panels). Reciprocally, NF1 can also be detected in FBXW7 immunoprecipitates (Fig 4D, bottom panels).
We next determined if FBXW7-NF1 binding and NF1 ubiquitination were dependent on the FBXW7 consensus-binding motif on NF1. We generated two NF1 mutants on the FBXW7 binding motif (Fig. S4C) and found that these mutations had a significant reduction in binding to FBXW7 (Fig. 4E). Consequently, FBXW7-induced ubiquitination of exogenously expressed NF1-C was remarkably inhibited in these mutants, without affecting ubiquitination of endogenous NF1 (Fig. 4F). We further determined NF1 ubiquitination using an in vitro purified system and found that FBXW7-CUL1 promoted NF1 ubiqutination in a manner dependent on E2 ubiquitin conjugating enzyme (Fig. 4G). The involvement of CUL1 in promoting NF1 ubiquitination was further demonstrated as follows: (1) endogenous NF1 could be pulled down by exogenously expressed CUL1 (Fig S4D), (2) siRNA knockdown of CUL1 significantly inhibited NF1 ubiquitination (Fig. S4E), and (3) endogenous SKP1 were present in SAG immunoprecipitates (Fig. S4F). Thus, SAG-SKP1-CUL1-FBXW7 could form an active E3 ubiquitin ligase in vivo that regulates cellular levels of NF1 by promoting its ubiquitination and degradation. To a lesser extent, CUL-5 appears to be also involved in NF1 degradation (Fig 4C&H, Fig S4D&E), although it generally belongs to an E3 complex with SOCS-Sag (Kamura et al., 2004).
We next determined whether FBXW7 shortens protein half-life of NF1, and whether SAG could further accelerate it. The protein half-life of transfected NF1C was about 2 hrs upon TPA stimulation, which was reduced to ~1 hr when co-transfected with wt FBXW7 (Fig. 5A), but extended to ~4 hrs when co-transfected with an F-box deleted dominant negative FBXW7 mutant (Fig. 5B). The protein half-life of two NF1 mutants was much extended under the same conditions (Fig. S5A&B). Moreover, cotransfection of SAG with FBXW7 further shortened the protein half-life of NF1C from 3 hr (FBXW7 alone) to ~1.5 hrs (when combined) in the absence of TPA stimulation (Fig. 5C). We further determined the effect of FBXW7 on protein half-life of endogenous NF1. Under normal growth conditions, endogenous NF1 is relatively stable with a protein half-life much longer than 3 hrs, regardless of FBXW7 status (Fig. S5C). NF1 was, however, subject to mitogen-induced reduction in an FBXW7 dependent manner. In serum-starved FBXW7−/− DLD1 cells (Rajagopalan et al., 2004), serum addition had no effect on NF1 levels up to 120 min. In contrast, serum addition to starved FBXW7+/+ DLD-1 cells induced a rapid elimination of NF1 within 15 min, which lasted up to 120 min before it started to recover (Fig. 5D, top). Likewise, the protein half-life of endogenous NF1 after exposure to TPA was about 30 min, which was extensively extended in FBXW7−/− cells (Fig. 5D, bottom). No change at the level of p120RasGAP was observed, regardless of FBXW7 status (Fig 5D). We further confirmed that FBXW7-dependent NF1 reduction was via the ubiquitin-proteasome pathway, since concurrent MG-132 treatment blocked NF1 degradation, along with accumulation of ubiquitinated NF1 in FBXW7+/+ cells, but not in FBXW7−/− cells (Fig. 5E). Finally, we found that both SAG and CUL1/5 were involved in controling the NF1 turn-over. The Nf1 half-life was much extended in Sag−/− MEF cells, as compared to Sag+/+ cells under the conditions of serum starvation, followed by serum addtion to trigger Nf1 degradation (Fig. 5F). Similarly, TPA-induced NF1 degradation was much delayed upon siRNA knockdown of either CUL1, or to lesser extent, CUL5 (Fig. 5G). Thus, NF1 is a physiological substrate of SAG-CUL1-FBXW7 (and possibly a SAG-CUL5 based E3 ligase) for targeted degradation.
We next attempted to define the physiological importance of Nf1 accumulation upon Sag disruption in endothelial differentiation of ES cells and vascular development of embryos. We reasoned that if Nf1 accumulation inhibits endothelial differentiation of EBs by inhibiting Ras/Erk activation, then simultaneous deletion of Nf1 would restore the formation of cystic EBs in the Sag−/− background. We generated Sag/Nf1 double null ES clones from blastocysts obtained by the mating of Sag+/− and Nf1+/− mice. Among 64 ES lines established, we identified a single clone (ISG8) with the Sag−/−;Nf1−/− genotype. However, this clone did not grow well and was readily attached onto the Petri dish during suspension culture for differentiation, leading to much fewer numbers of embryoid bodies, although we did observe the formation of cEBs at a rate comparable to that of wild type (Fig. S6A&C) with full restoration of Ras activity (Fig. S6B). We then focused our attention to three independent Sag−/−;Nf1+/− ES cell lines and found that all three were able to undergo endothelial differentiation to form cystic EBs, although the frequency was relatively lower than that of wild type (Fig. 6A, two clones, A1 and A10 were shown). The blood island structures with endothelial lining were apparent in these rescued cEBs (Fig. 6B), indicating that heterozygous loss of Nf1 is sufficient to rescue the defect in endothelial differentiation caused by Sag inactivation. We further confirmed that the Nf1 status on the Sag+/+ background did not affect endothelial differentiation, since both Nf1+/− and Nf1−/− ES cells are capable of forming cEBs (Fig S6C). We next determined if Ras activity was also rescued upon Nf1 inactivation and found that while Sag−/−;Nf1+/+ EBs showed a high level of Nf1 with a undetectable Ras activity (Fig 6D, lane 1), Sag−/−;Nf1+/− EBs had a significant lower level of Nf1, but a high level of activated Ras, similar to that seen in wild type EBs (Fig. 6D, lanes 2 vs. 3). These results suggested that loss of one Nf1 allele is sufficient to reduce abnormally high levels of Nf1 (caused by Sag disruption) to a level that permits cystic EB formation during ES cell differentiation. Taken together, our results demonstrated that Ras inactivation, as a result of Nf1 accumulation upon Sag disruption, is the major cause for the failure of Sag−/− ES cells to undergo endothelial differentiation.
We then determined whether simultaneous deletion of Nf1 would rescue vascular defects in Sag−/− embryos (Fig. 1D) by crossing of Sag+/− mice with Nf1+/− mice (Zhu et al., 2005). Although homozygous deletion of Nf1 caused embryonic lethality by E13.5 due to cardiac defects, mutant embryos appeared normal through E10.5 (Brannan et al., 1994; Jacks et al., 1994). We examined the blood vessel networks in head areas of E10.5 embryos with 4 genotypes, using CD31 whole-mount staining (Fig 6E). While Sag−/−;Nf1+/+ brains showed reduced primary vascular networks with missing major trunk vessels, Sag+/+;Nf1−/− brains showed relatively normal vasculature, similar to that of Sag+/+;Nf1+/+. Significantly, the vasculature in the head areas of Sag−/−;Nf1−/− embryos showed a partial restoration with reappearance of some trunk blood vessels (Fig. 6E), even though these Sag−/−;Nf1−/− embryos remained growth-retarded (Fig. 6E&F). Furthermore, CD31-positive endothelial cells in control embryos formed a chain-like structure that covered the entire neuroepithelium, whereas only residual CD31-positive cells were found in the Sag−/− embryos (arrows, Fig. 6F). Importantly, the chain-like structure formed by CD31-positive endothelial cells was partially restored by both Nf1 heterozygous and homozygous mutations (Fig. 6F&H).
We further observed that Sag disruption caused a widespread apoptosis in embryos evidenced by cleaved caspase 3 staining (Fig. 6F). To rule out the possibility that excessive apoptosis observed in Sag-deficient embryos could potentially exert non-specific effects on normal development of endothelial cells, we further analyzed control and Sag-deficient embryos at earlier stages. At E9.5, compared to controls, Sag-deficient embryos exhibited relatively normal morphology with only a minor increase in apoptosis within developing nervous system (Fig. 6G). However, the number of CD31-positive endothelial cells was greatly reduced in Sag-deficient embryos (arrows, Fig. 6G). Particularly, compared to control embryos, Sag-deficient embryos only contained less than 10% of CD31-positive endothelial cells in the head areas (Fig. 6I). Strikingly, this defect was significantly rescued by loss of one or two alleles of Nf1 (arrows, Fig. 6G, 6I). When comparing the relative number of CD31-positive endothelial cells at E9.5 and E10.5 among control, Sag single, and Sag/Nf1 double mutant embryos, we found that CD31-positive endothelial cells in all the mutant embryos, albeit at a much lower level, expanded proportionally from E9.5 to E10.5, as compared to those in controls (Fig. 6J). Taken together, our studies, using both in vitro ES cells and in vivo embryos, suggest that Sag-mediated Nf1 degradation is critical for the genesis of endothelial cells in the vascular system. This notion was further supported by the fact that CD31-positive endothelial cells in Sag-deficient embryos at both E9.5 and E10.5 stages did not colocalize with caspase-3 positive cells (Fig. 6F&G), excluding the involvement of apoptosis.
To further determine genetic interaction between Sag and Nf1 in vivo, we studied Nf1-dependent Sag function in the developing nervous system, as 40% to 80% of NF1 children exhibit learning deficits (Hyman et al., 2005). During development, the onset of neurogenesis in the neocortex occurs at the basal surface of the neuroepithelium at E10.5 (Farkas and Huttner, 2008; Gotz and Huttner, 2005). We found that neuronal differentiation, revealed by Tuj1 expression, was not significantly different between control and Sag-deficient embryos with or without additional Nf1 mutations (Fig. 7A, B). These results indicate that Sag−/− neural precursor cells in the neocortex undergo neuronal differentiation according to a normal developmental schedule, albeit in overall growth-retarded embryos. We next examined whether there are proliferation defects in Sag−/− neural precursor cells. During cell cycle, neural precursor cells undergo DNA replication (S phase) in the basal part of the neuroepithelium (Farkas and Huttner, 2008), which were labeled by a single BrdU pulse (Fig. 7C, upper left panel). The nuclei of neural precursor cells migrate apically through G2 phase to undergo M phase in apical position along the ventricle, which were revealed by an M phase marker, phospho-histone H3 (PH3) (Fig. 7C, bottom left panel). Strikingly, little or no BrdU- or PH3-positive neural precursor cells were identified in the Sag−/− neocortex (Fig. 7C). More importantly, none of the defects in cell cycle progression was rescued by Nf1 inactivation (Fig. 7C). Thus, Sag-deficient neural precursor cells in the developing neocortex exhibited severe defects in entering S and M phases of the cell cycle in an Nf1-independent manner. Consequently, compared to control neocortex, the number of neural precursor cells labeled by Ki67 expression was significantly reduced in the Sag−/− neuroepithelium, consistent with the fact that these cells had proliferation defects (Fig. 7D&E). Given that Nf1 deficiency does not rescue cell-cycle progression defects in Sag−/− neural precursor cells, we unexpectedly observed a significant level of rescue of Ki67-positive neural precursor cells upon Nf1 deletion (Fig. 7D&E). We, therefore, determined whether apoptosis additionally contributed to Nf1-dependent reduction of neural precursor cells in Sag−/− neocortex. While control neocortex had little or no caspase3-positive apoptotic cells, nearly 40% of Sag−/− cells in E10.5 neocortex were undergoing apoptosis, which was significantly rescued by Nf1 inactivation in a dose-dependent manner (Fig. 7D&F). In addition to the developing brain, Nf1 inactivation also partially rescued apoptosis in the Sag-deficient spinal cord (arrows, Fig.7G), though apoptosis in surrounding mesoderm-derived tissues was not affected (Fig. 7G). Moreover, neuronal apoptosis of Sag−/− E9.5 embryos could also be rescued by Nf1 deletion (Fig. (Fig.6G6G&S6D). Taken together, our observations demonstrate that during neocortical development, Sag plays critical roles in regulating apoptosis and cell cycle progression in an Nf1-dependent and Nf1-independent manner, respectively. Furthermore, Nf1-dependent Sag function in suppressing apoptosis is specifically required for developing nervous system, but not for surrounding mesenchymal tissues.
Despite Sag being dispensable during development in Drosophila and C. elegan (Moore and Boyd, 2004; Reynolds et al., 2008), we demonstrate that Sag is absolutely required for mouse development. Sag disruption causes embryonic lethality at E11.5-12.5, which is associated with severe developmental defects including those in the vascular and nervous system. Since SCF E3 ligases are essential for many developmental processes (Petroski and Deshaies, 2005), it is not surprising that Sag, one of the two Rbx/Roc family members essential for E3 ligase activity, plays critical roles in many aspects of embryonic development. Although accumulation of many critical Sag/SCF-targeted substrates could contribute to severe developmental defects observed in Sag−/− embryos, our biochemical and genetic experiments indicate a causal involvement of Nf1, which, upon accumulation as a result of Sag disruption, inactivates Ras-Mapk signals to block endothelial differentiation and induce apoptosis in the nervous system.
The RAS-MAPK signaling pathway is actively involved in cell proliferation, angiogenesis, and tumorigenesis (Dhillon et al., 2007a; Kranenburg et al., 2004) as well as endothelial specification of VEGFR2-expressing vascular progenitor cells (Kawasaki et al., 2008). Here we showed that active Ras/Erk signals are also required for endothelial differentiation of mouse ES cells. Blockade of this pathway by inhibitors of RAS (FTI-277) or ERK (PD098059) repressed ES cell differentiation into endothelial cells to form blood islands. We further showed in vivo that reduced Ras/Erk activation was associated with reduced angiogenesis and decreased proliferation in teratomas derived from Sag−/− ES cells. Our mechanistic study linked Ras inactivation, upon Sag disruption, to accumulation of Nf1, a naturally occurring Ras inhibitor (Cichowski and Jacks, 2001).
Our study also defined SAG-SKP1-CUL1-FBXW7 as a physiological E3 ligase for targeted ubiquitination and degradation of NF1. First, NF1 contains an evolutionarily conserved FBXW7 binding motif at the C-terminus (codons 2755-2758), which is required for FBXW7-NF1 binding and NF1 ubiquitination. Second, SAG-CUL1-FBXW7-NF1 forms a complex in vivo that promotes NF1 ubiquitination. Third, FBXW7 shortens NF1 protein half-life in a binding motif dependent manner, which is further promoted by SAG. Fourth, ubiquitination of exogenously expressed NF1-C is promoted by FBXW7-CUL1, but inhibited by siRNA knockdown of CUL1 or genetic deletion of FBXW7. Finally and most importantly, endogenous FBXW7 binds to endogenous NF1 under physiological conditions and SAG-CUL1-FBXW7 is required for ubiquitination and degradation of endogenous NF1 upon mitogen stimulation. Our study also demonstrates that SAG could form an active E3 complex with SKP1, FBXW7 and CUL1, in addition to well-known CUL5 (Kamura et al., 2004), for targeted degradation of NF1.
FBXW7 is a p53-dependent haploinsufficient tumor suppressor (Kimura et al., 2003; Mao et al., 2004). Similar to Sag disruption, Fbxw7 gene ablation also causes embryonic lethality at E10.5-E11.5 with remarkable abnormalities in vascular development in the brain and yolk sac (Tetzlaff et al., 2004; Tsunematsu et al., 2004). These phenotypic similarities suggested that Sag and Fbxw7 might have a similar function, such as preventing Nf1 accumulation, during vascular development. However, unlike Fbxw7 whose deletion caused accumulation of Notch-1 and Notch-4 (Masuda et al., 2010; Tetzlaff et al., 2004; Tsunematsu et al., 2004), we found that neither Notch-1 nor Notch-4 was accumulated in Sag−/− ES, EB, or embryos (data not shown), suggesting that accumulation of Notch family proteins do not contribute to observed defects in Sag−/− ES cells and embryos. Instead, we showed a causal involvement of Nf1 in some aspects of vascular and neural development, regulated by Sag. First, Sag is required for endothelial cell differentiation, which is Nf1 dependent. Second, Sag plays an essential role in suppressing apoptosis during embryogenesis, with a partial involvement of Nf1 in neural precursor cells in developing brains and spinal cords, but not in mesoderm-derived tissues. Third, despite the presence of a significant number of neural precursor cells in Sag−/− neocortex, these Ki67-positive precursor cells completely lack the ability to undergo DNA replication and mitosis, which is completely independent of Nf1. It will be interesting to determine the specific stage(s) in which Sag−/− neural precursor cells are arrested during cell cycle. These observations suggest that Nf1- independent apoptosis and cell cycle failure, potentially as a result of accumulation of other critical Sag substrates, could be the major cause for growth retarded and lethal phenotypes observed in Sag−/−;Nf1−/− embryos with an embryonic life not extended beyond E11.5-12.5. Nevertheless, our genetic rescue data, by demonstrating that Nf1 inactivation rescued vascular defects, particularly at earlier developmental stages when Sag deficiency causes no overt morphological defects, strongly argue that genetic interaction between Nf1 and Sag is specific and primary to vascular development and not secondary to overall developmental retardation. Furthermore, it is worth noting that despite the presence of excessive and widespread apoptosis at E10.5, CD31-positive endothelial cells expanded proportionally from E9.5 to E10.5 in Sag-deficient embryos compared to controls. This observation further supports the notion that Sag/Nf1 regulatory network controls the generation of CD31-positive endothelial cells, but not subsequent proliferation or apoptosis.
In summary, our study demonstrates an in vivo physiological function of Sag in regulation of vascular and neural development by modulating Ras-Mapk signaling pathway through targeted degradation of Nf1. Upon induction by different stimuli (Duan et al., 1999; Gu et al., 2007; Tan et al., 2008), SAG recruits FBXW7 and CUL1 to promote ubiquitination and degradation of NF1, thus activating RAS-ERK signals to induce differentiation and proliferation, and to suppress apoptosis during vascular and neural development, respectively (see graphical abstract). Thus, under physiological conditions, SAG regulates NF1 levels to maintain active RAS/ERK signals for effective vascular and neural development. However, under pathological condition, such as in cancer, SAG, which is overexpressed with a poor patient prognosis (Jia et al., 2010; Sasaki et al., 2001), would inhibit apoptosis, promote proliferation and angiogenesis (Duan et al., 1999; Gu, 2007; Sun et al., 2001). SAG, therefore, could be a valid anti-cancer and anti-angiogenesis target (Jia et al., 2010; Nalepa et al., 2006; Sun, 2006; Wei and Sun, 2010). Moreover, 40 to 80% of individuals with NF1 have cognitive deficits, which are caused by loss of one allele of the NF1 gene - haploinsufficiency in the brain (Costa et al., 2002; Cui et al., 2008; Hyman et al., 2006). Thus, targeting the SAG/FBXW7 E3 may provide an attractive therapeutic strategy to increase NF1 expression for treating NF1-associated diseases caused by haploinsufficiency.
Blastocysts were isolated from inter-crossing of Sag+/− mice or Sag+/−;Nf1+/− mice, and placed in culture on irradiated mouse embryonic feeder cells in high glucose DMEM (Invitrogen, Carlsbad, CA) supplemented with 15% FBS (Harlan, Indianopolis, IN), 0.1 mM β-mercaptoethanol (Sigma, St. Louis, MO), 103U/ml LIF (ESGRO, Millipore),25 μM PD098059 (Sigma) and penicillin/streptomycin. Inner cell mass outgrowths were trypsinized and passaged until ES cell lines were established in 35 mm cell culture dishes (Hughes et al., 2007).
Whole-mount CD31 immunostaining was performed on E9.5 and E10.5 yolk sacs and embryos (Shen et al., 2005). Briefly, embryos or yolk sacs were fixed in 4% PFA/PBS, dehydrated by methanol, quenched by H2O2, and blocked in 4% BSA. The samples were stained by incubating with anti-CD31 (rat monoclonal MEC13.3, BD Biosciences) at 4°C overnight, followed by peroxidase-conjugated secondary antibodies. The embryos were developed in 0.25% DAB with H2O2 in PBS, and photographed on a dissecting microscope (model S6D; Leica) with a progressive 3CCD camera (Sony).
Single suspended mouseES cells were plated at density of 1 × 106 cells/10 cm Petri dish containing 15% DMEM. Cells were cultured for up to 13 days with media change every other day. Embryoid bodies were first fixed in the 4% PFA, blocked with 3% milk and 0.1% Triton X-100 in PBS, and incubated with CD31 antibody overnight. After washing, the samples were incubated with secondary antibody overnight. Color was developed using DAB. The samples were then dehydrated, embedded and sectioned. The sections were counter-stained with Eosin-Y.
Paraffin sections were deparaffinized, rehydrated and analyzed by immunofluorescence (Wang et al., 2009; Zhu et al., 2001). Briefly, sections were incubated with primary antibodies in blocking solution overnight. The antibodies used are as follows: Ki 67 (1:500, mouse, BD Biosciences), Tuj-1 (1:500, Rabbit, Covance), Cleaved Caspase3 (1:500, Rabbit, Cell signaling), Phospho-histone H3 (1:500, Rabbit, Abcam), BrdU (1:500, Rat, Abcam) and CD31 (1:200, Mouse, Dako). The secondary antibodies were conjugates of Alexa Fluor 488 or Alexa Fluor 555 (1:500, Invitorgen). DAPI (1:1000, Invitrogen) was used as nuclear counter staining. Sections were examined under a fluorescence microscope (Olympus).
The paired Student t test was used for statistical analysis, using SAS software for two paired samples. A mixed model was fit to the tumor volumes (excluding mice with no tumors). Presence or absence of Sag knockout was included as a fixed effect. ES cell line was included as a random effect in the model to account for correlation between observations within a cell line. An F-test from the mixed model fit was used to test whether the mean value for tumors from cell lines with Sag knockout was the same as that for the tumors with wild type Sag.
We thank Dr. B. Vogelstein for providing us DLD-1 and HCT-116 cells with FBXW7 gene deleted, and Drs. K. Nakayama and B. E. Clurman for plasmids expressing FBXW7. We also thank Dr. Matthew Schipper for his help in statistical analysis of the teratoma data. This work is supported by NCI grants (CA111554, CA118762 and CA156744) to YS and (NS053900 and W81XWH-11-1-0251) to YZ.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.