The locus coeruleus (LC) is the main source of noradrenaline in the brain and is implicated in a broad spectrum of physiological and behavioral processes. However, genetic pathways controlling the development of noradrenergic neurons in the mammalian brain are largely unknown. We report here that Rbpj, a key nuclear effector in the Notch signaling pathway, plays an essential role in LC neuron development in the mouse. Conditional inactivation of Rbpj in the dorsal rhombomere (r) 1, where LC neurons are born, resulted in a dramatic increase in the number of Phox2a- and Phox2b-expressing early-differentiating LC neurons, and dopamine-β-hydroxylase- and tyrosine-hydroxylase-expressing late-differentiating LC neurons. In contrast, other neuronal populations derived from the dorsal r1 were either reduced or unchanged. In addition, a drastic upregulation of Ascl1, an essential factor for noradrenergic neurogenesis, was observed in dorsal r1 of conditional knockout mice. Through genomic sequence analysis and EMSA and ChIP assays, a conserved Rbpj-binding motif was identified within the Ascl1 promoter. A luciferase reporter assay revealed that Rbpj per se could induce Ascl1 transactivation but this effect was counteracted by its downstream-targeted gene Hes1. Moreover, our in vitro gene transfection and in ovo electroporation assays showed that Rbpj upregulated Ascl1 expression when Hes1 expression was knocked down, although it also exerted a repressive effect on Ascl1 expression in the presence of Hes1. Thus, our results provide the first evidence that Rbpj functions as a key modulator of LC neuron development by regulating Ascl1 expression directly, and indirectly through its target gene Hes1.
The locus coeruleus (LC) is a dense cluster of neurons located in the dorsolateral tegmentum of the pons. This nucleus is the chief source of noradrenaline in the mammalian brain and projects to most regions of the central nervous system (CNS) (Aston-Jones et al., 1995; Foote et al., 1983). Noradrenaline transmission is associated with a wide range of physiological and behavioral processes, such as anxiety, arousal, learning and memory, nociception, sleep/wake cycles and stress (Berridge and Waterhouse, 2003; Singewald and Philippu, 1998). Dysfunctions of the noradrenergic system lead to a variety of neurological disorders, including attention deficit hyperactivity disorders, circadian rhythm sleep disorders, depression, epilepsy and Parkinson's disease (Aston-Jones et al., 2001; Aston-Jones et al., 2004; Berridge and Waterhouse, 2003; Warnecke et al., 2005).
Noradrenergic neurons in the LC are derived from progenitors in the dorsal rhombomere (r) 1 and are among the earliest born neurons in the mouse brain (Aroca et al., 2006; Pattyn et al., 2000; Steindler and Trosko, 1989). Previous studies have identified several genes involved in the development of LC neurons (Goridis and Rohrer, 2002). In zebrafish and chick embryos, bone morphogenetic proteins (BMPs) secreted from the roof plate regulate the early stages of LC specification by inducing the expression of the proneural bHLH gene Ascl1 (Guo et al., 1999; Hatakeyama et al., 2004; Vogel-Höpker and Rohrer, 2002). A recent study revealed that an orphan nuclear receptor Nr2f6, functioning downstream of Ascl1, is also required for LC neuron development (Warnecke et al., 2005). In the absence of Bmps (Guo et al., 1999; Hynes and Rosenthal, 1999; Vogel-Höpker and Rohrer, 2002), Ascl1 (Hirsch et al., 1998) or Nr2f6 (Warnecke et al., 2005), the LC is either completely absent or reduced in size. It has also been shown that Ascl1 activates Phox2a and Phox2b, two paired-like homeobox genes, which are critical for the early differentiation of LC neurons in the mouse, chick and zebrafish (Guo et al., 1999; Hirsch et al., 1998; Morin et al., 1997; Pattyn et al., 1997; Vogel-Hopker and Röhrer, 2002; Yang et al., 1998). Phox2a and Phox2b in turn initiate the expression of dopamine-β-hydroxylase (Dbh) and tyrosine hydroxylase (TH), two genes expressed in late-differentiating and mature LC neurons. Additionally, the homeobox transcription factor Tlx3 (Hornbruch et al., 2005; Qian et al., 2001), the tyrosine kinase receptor TrkB and its ligands BDNF and NT-4 (Holm et al., 2003) were also reported to play essential roles in the maintenance and survival of LC neurons. Although a number of factors have been identified to be critical in the LC development, genetic pathways controlling the LC neurogenesis in the mammalian brain are still elusive.
Notch signaling plays important roles in the neurogenesis of developing mammalian brain (Furukawa et al., 2000; Louvi and Artavanis-Tsakonas, 2006; Scheer et al., 2001). In mice, established components of the canonical Notch pathway include four Notch receptors (Notch1-4), five Notch ligands (Delta1, Delta3, Delta4, Jagged1 and Jagged2) and a DNA-binding transcription factor recombination signal-binding protein Jκ (Rbpj; also known as CBF1). When Notch signaling is inactive, Rbpj is bound to DNA as part of a repressor complex; upon ligand binding, the intracellular domain of Notch (NICD) is released and translocates to the nucleus, where it binds to Rbpj, thereby leading to Rbpj-dependent transcriptional activation. All four mammalian Notch receptors ultimately bind to Rbpj, thus Rbpj serves as a key integrator of canonical Notch signaling (Bray, 2006; Han et al., 2002). Target genes of the Rbpj–NICD activator complex include Hes1 and Hes5 (Kageyama and Ohtsuka, 1999), which repress the expression of proneural genes, such as Ascl1 (Bertrand et al., 2002), Ngn1 and Ngn2 (Imayoshi et al., 2008). A group of loss-of-function studies have revealed that Notch signals inhibit neuronal differentiation and help maintain neuronal progenitors in a pluripotent state (Louvi and Artavanis-Tsakonas, 2006; Yoon and Gaiano, 2005).
In the present study, we showed that conditional ablation of Rbpj gene in the dorsal r1, where LC neurons are born, selectively enhanced the generation of LC neurons, but not of other r1 derivatives. Meanwhile, Rbpj deletion resulted in a significant downregulation of Hes1 expression and upregulation of Ascl1 expression. Electrophoretic mobility shift assay (EMSA) and chromatin immunoprecipitation (ChIP) assay identified an Rbpj-binding site within Ascl1 promoter where Hes1-binding sites are present. Luciferase reporter assay revealed that Rbpj promoted Ascl1 transactivation, and this effect was repressed by Hes1. Further in vitro and in vivo gene expression assays showed that Rbpj upregulated Ascl1 expression when Hes1 expression was knocked down by RNA interference. These findings provide the first evidence that Rbpj function as a novel regulator of noradrenergic neurogenesis in the mouse LC.
Rbpj is expressed in the LC progenitors
To explore the effects of Rbpj on LC development, we first examine whether Rbpj is expressed in the developing dorsal r1, where the LC is born. Using whole mount in situ hybridization, we found that Rbpj transcripts were detected as early as E8.5 with strong expression in the neural tube of prosencephalon, mesencephalon and metencephalon including dorsal r1 (Fig. 1A, arrowhead). At E9.5, Rbpj expression became ubiquitous through the neural tube and relatively weak in the dorsal r1 (Fig. 1B). Afterwards, the expression of Rbpj was still observed through the total neural tube but at a low level (data not shown). Sense RNA probe revealed no in situ hybridization signals (Fig. 1C). In addition, on the transverse sections through E9.5 dorsal r1, we performed in situ hybridization for Rbpj and Phox2a/Phox2b, two early-differentiating markers for LC neurons, and found that Rbpj-expressing cells were overlapped with those expressing Phox2a and Phox2b (data not shown), suggesting that Rbpj is expressed in the LC primordium.
Next, to inactivate Rbpj in the LC progenitors, we generated RbpjWnt1 conditional knockout (CKO) and RbpjEn1 CKO mice (see Materials and Methods section). Cre expression in both Wnt1-Cre and En1Cre/+ mice occurs as early as E8.5 in the dorsal r1 (Li et al., 2002; Trokovic et al., 2003), thus enabling us to knockout Rbpj in the region that gives rise to LC neurons prior to their generation. To examine whether LC neurons were derived from Wnt1-Cre- or En1-Cre-expressing progenitors, we took advantage of Wnt1-Cre;Rosa26 and En1Cre/+;Rosa26 reporter mice that express β-galactosidase (β-gal) in the Cre expression domain. X-gal staining of whole-mount embryos and coronal sections revealed the presence of β-gal+ cells in the dorsal r1 at E10.5 (Fig. 1D,E, arrowheads) and in the LC at P10 (Fig. 1F,G, boxed areas), respectively. Furthermore, double immunolabeling with TH and β-gal showed that 87.2±6.1% of TH+ neurons in the LC co-expressed β-gal in Wnt1-Cre;Rosa26 reporter mice (Fig. 1H–H′′, J), and 95.8±3.8% in En1Cre/+;Rosa26 reporter mice (Fig. 1I–J). These results show that Wnt1- or En1-driven Cre expression is present in LC neurons, suggesting that these two Cre mouse lines can be used to inactivate Rbpj expression in the LC by crossing with Rbpj floxed mice.
Rbpj inactivation induces overproduction of LC neurons
In situ hybridization showed that the expression of Rbpj was absent in the r1 region of RbpjWnt1 CKO or RbpjEn1 CKO mice compared with wild-type littermates (Fig. 2A–C). To examine the effects of Rbpj deletion on LC development, we performed TH immunohistochemistry and Dbh and Phox2b in situ hybridization, and found a dramatic increase in the number of LC neurons at E17.5 and postnatal day (P) 0 (Fig. 2G–R). The littermate mice with other genotyping (e.g. Wnt1-Cre;Rbpjflox/+ or En1Cre/+;Rbpjflox/+) showed no increase of LC neurons or other detectable defects (data not shown) and they were used as controls in LC neuron counting. Statistical results showed that the numbers of neurons in one LC in RbpjWnt1 CKO mice (1795±88) and RbpjEn1 CKO mice (1950±76) were 2.4 times and 2.6 times higher than that of controls (753±68), respectively (supplementary material Fig. S2). These findings were further confirmed by examining the expression of L-amino acid decarboxylase (L-AADC) (Eaton et al., 1993), monoamine oxidase type A (MaoA) (Arai et al., 1997), calcitonin gene related protein (CGRP) (Holm et al., 2003; Ma et al., 2003), vesicular glutamate transporter 1 (vGluT1) (Barr and Van Bockstaele, 2005), brain-derived neurotrophic factor (BDNF) (Holm et al., 2003) and c-Ret (Holm et al., 2002) at E17.5 or P0 (supplementary material Fig. S2). To determine whether the LC itself was defective in Wnt1-Cre or En1Cre/+ mice, LC neurons of these two Cre mice were examined by TH immunostaining at P10. Our results revealed a normal morphology and neuron number of the LC (Fig. 2D–F), suggesting that the increase of LC neurons in CKO mice was not caused by defective development of LC in Wnt1-Cre or En1Cre/+ mice, which were used to inactivate Rbpj in the dorsal r1.
To determine the developmental stages at which excessive LC neuron induction occurs, we examined the expression of the early differentiation markers Phox2a and Phox2b at E9.5–E10.5, and the late differentiation markers Dbh and TH from E10.5 to E16.5. Compared to wild types, Phox2a in situ hybridization revealed an increase in the number of immature LC neurons in both RbpjWnt1 CKO and RbpjEn1 CKO embryos (Fig. 3A–C; supplementary material Fig. S5A,E). This early stage increase was confirmed by Phox2b in situ hybridization (Fig. 3E–G; supplementary material Fig. S5B,F). By contrast, in Wnt1-Cre mice Phox2a- or Phox2b-positive LC progenitors were comparable to those in wild-type controls (Fig. 3D,H), again showing normal development of LC neurons in this Cre line mouse. At E10.5, E12.5 and E16.5, both TH immunostaining and Dbh in situ hybridization revealed a significant increase in the number of LC neurons in both RbpjWnt1 CKO and RbpjEn1 CKO embryos, in comparison with wild types (Fig. 3I–Z). Thus, Rbpj deletion leads to dramatically enhanced generation of LC neurons at both early and late embryonic stages.
Other r1 derivatives are either reduced or unchanged in Rbpj CKO mice
Besides the LC, the dorsal r1 gives rise to a wide array of brainstem and cerebellar cell types, including principal sensory trigeminal nucleus, parabrachial complex, superior vestibular nucleus, and cerebellar primordium which is comprised of deep cerebellar nuclei, external granular cells and Purkinje cells (Kim et al., 2008; Machold and Fishell, 2005; Wang et al., 2005). Thus, we next asked whether Rbpj inactivation also affected the generation of these neuronal populations. We examined Tlx3+ neurons in the principal trigeminal sensory nucleus, Calbindin+ neurons in the parabrachial complex, and Tbr1+ neurons in the superior vestibular nucleus, but did not find obvious differences between RbpjWnt1 CKO and wild-type mice at P0 (supplementary material Fig. S3A–C,E–G). By contrast, Rbpj inactivation led to an absence of cerebellar foliation (supplementary material Fig. S4A–D). This defect was not due to a loss of laminar structures, as the expression of Pax6 (external granular cell marker) and Calbindin (Purkinje cell marker) were relatively normal (supplementary material Fig. S4A–D), although the cell numbers of these cerebellar neurons and part of Tbr1- or Tbr2-positive deep cerebellar neurons were reduced (supplementary material Fig. S4). Considering that the r1 formation is regulated by the isthmic organizer (Liu and Joyner, 2001a; Nakamura et al., 2005) and mid–hindbrain patterning genes, we examined the expression of Wnt1 and Fgf8, two key genes for the inductive activity of the isthmic organizer (Liu and Joyner, 2001a; Nakamura et al., 2005; Wurst and Bally-Cuif, 2001), and the expression of Gbx2 and Otx2, two key genes for setting up the position of mid–hindbrain boundary (Li et al., 2002). Our whole-mount and section in situ hybridization at E10.5 showed that expression patterns of these four genes were not noticeably affected by the absence of Rbpj (supplementary material Fig. S5I–P), suggesting that Rbpj deletion-induced phenotypes of the r1 are independent of isthmic organizer and mid–hindbrain patterning.
Wnt1- and En1-driven Cre expression is also present in the caudal midbrain and ventral r1 (Danielian et al., 1998; Davidson et al., 1988; Kimmel et al., 2000; Trokovic et al., 2003) which give rise to a subpopulation of dopaminergic and serotonergic neurons, respectively. To determine whether these neurons are altered in Rbpj CKO mice, we examined the populations of ventral midbrain-derived dopaminergic neurons and ventral r1-derived serotonergic neurons by double immunostaining of TH and serotonin (5-HT) on the sagittal sections at P0. Compared with wild types, the dopaminergic neurons in the ventral midbrain and the serotonergic neurons in dorsal raphe nuclei were greatly reduced in Rbpj CKO mice (Fig. 4A,B). However, Rbpj mutant did not obviously affected the serotonergic neurons in the median raphe nucleus, which is situated ventrally to the dorsal raphe nucleus (Fig. 4A,B). The P0 coronal sections immunolabeled with TH and 5-HT, or hybridized by L-AADC RNA probe also revealed the similar results in both RbpjWnt1 CKO and RbpjEn1 CKO mice (Fig. 4C–N). In addition, for the dorsal midbrain, we found that the superior colliculus developed defectively and the inferior colliculus was completely lost in Rbpj CKO mice in comparison to wild types, as determined by Brn3 immunostaining at E17.5 (supplementary material Fig. S3D,H). By contrast, the ventral midbrain-derived Islet1+/Parvalbumin+ oculomotor neurons, Parvalbumin+ trochlear neurons and Brn3a+ neurons in the red nucleus were unchanged in the absent of Rbpj (supplementary material Fig. S3I–P). Taken together, these data show that ablation of Rbpj in the r1 and the caudal midbrain results in a dramatic increase of noradrenergic neurons in the LC, while other neuronal cell types generated in these regions are either decreased or unchanged, suggesting that Rbpj seems to function as a selective inhibitor in the LC development.
LC hyperplasia is due to enhanced and prolonged neurogenesis
To determine whether the increase in the number of LC neurons is caused by enhanced neurogenesis, a prolonged generation period, or a combination of both these events, we performed BrdU labeling at different embryonic stages (from E8.5 to E13.5). The proportion of LC neurons born after a single pulse of BrdU was determined by co-immunostaining for BrdU and TH at E17.5. When BrdU was pulsed at E8.5, BrdU+/TH+ co-labeled cells were hardly detected in either wild-type or RbpjWnt1 CKO embryos (supplementary material Fig. S1A,E,I). After applying BrdU at E9.5, on the other hand, we observed over 2.7 times more co-labeled cells in the LC of RbpjWnt1 CKO embryos relative to wild-type controls (supplementary material Fig. S1B,F,I). The generation of LC neurons decreased abruptly after E9.5 in wild-type embryos, as noted by the significant decreases in BrdU labeling of TH+ LC cells after pulsing at E10.5, at which time point BrdU labeling of TH+ LC precursors was rare. In contrast, LC neurogenesis continued in RbpjWnt1 CKO embryos at E10.5, when BrdU labeling of TH+ neurons was approximately 3 times higher than wild-type controls (supplementary material Fig. S1C,G,I). The generation of LC neurons had mostly ended by E11.5 in wild-type embryos, but a number of BrdU+/TH+ neurons were still observed in RbpjWnt1 CKO embryos (supplementary material Fig. S1D,H,I). From E12.5 on, very few LC progenitors incorporated BrdU in either wild-type or RbpjWnt1 CKO embryos (supplementary material Fig. S1I and data not shown). In addition, TUNEL staining and caspase3 immunohistochemistry showed that no increase or decrease in cell death occurred in the LC of RbpjWnt1 CKO or RbpjEn1 CKO mice relative to their wild-type littermates at all embryonic stages examined (data not shown). Thus, these results indicate that LC hyperplasia is due to both enhanced and prolonged neurogenesis in Rbpj CKO mice.
Rbpj deletion does not affect BMP signaling or Nr2f6 expression
The specification of LC neurons is triggered by BMP signaling, and depends on the activity of a number of factors downstream of BMPs, such as Ascl1 and Nr2f6 (Goridis and Rohrer, 2002; Warnecke et al., 2005). To explore the genetic mechanisms underlying the overproduction of LC neurons in the absence of Rbpj, we first examined the expression of Bmps, BMP target genes Msx1–3 and Nr2f6 in Rbpj CKO embryos. Although Bmp5 is expressed in the r1 roof plate of chick embryos where it is critical for LC development (Vogel-Höpker and Rohrer, 2002), our findings showed that among all Bmps examined (Bmp2, Bmp4–7), only Bmp6 and Bmp7 were expressed in the roof plate of the r1 in E10.5 wild-type embryos (Fig. 5A,B, and data not shown). We did not detect any obvious changes to Bmp6 or Bmp7 expression in the roof plate of RbpjWnt1 or RbpjEn1 CKO embryos relative to wild-type controls, although Bmp7 mRNA was upregulated in the mid–hindbrain boundary region of both CKO lines (Fig. 5E,F and data not shown). Consistent with these results, the expression of Msx1, Msx2 and Msx3 in the r1 roof plate was not obviously affected in E9.0–E9.5 CKO embryos (Fig. 5I–R). Similarly, the expression of Nr2f6 in RbpjWnt1 CKO embryos was indistinguishable from wild-type controls at E9.5 (Fig. 5C,D,G,H). These results suggest that BMP signaling and Nr2f6 may not be involved in the overproduction of LC neurons in Rbpj CKO mice.
Rbpj deletion upregulates Ascl1 expression
Next, we examined the changes in the expression of Ascl1, an essential factor for noradrenergic neurogenesis in mice (Hirsch et al., 1998). At E8.5, Ascl1 transcripts were not detected throughout the wild-type and Rbpj CKO neural tube (Fig. 6A,E). The initiation of Ascl1 expression in wide types was first observed at E9.0 in the dorsal mesencephalon but not in the r1 (Fig. 6B, arrowhead). However, at this stage, Ascl1 expression was already evident in the dorsal r1 of RbpjWnt1 CKO embryos (Fig. 6F, arrowhead). At E9.5, the upregulated Ascl1 expression was more obvious in CKO embryos than wild-type controls (Fig. 6C,G,I,M, arrowheads), and however, at E10.5 this enhanced expression was no more obvious in mutant embryos (Fig. 6D,H,J,N, arrowheads). As development progressed, a slightly higher level of Ascl1 expression was maintained in the dorsal r1 of E11.5–E12.5 RbpjWnt1 CKO embryos than wild types (Fig. 6K,L,O,P, arrowheads). Upregulation of Ascl1 in the developing CNS has been reported in several other Rbpj CKO mice, but it should be noted that it is a transient event around E10.5 followed by a rapid and dramatic decrease of Ascl1 expression at later stages (Imayoshi et al., 2008; Lütolf et al., 2002; Machold et al., 2007). This is believed to be caused by the depletion of progenitor pool in the absence of Rbpj (Louvi and Artavanis-Tsakonas, 2006; Machold et al., 2007; Yoon and Gaiano, 2005). However, our data showed a long-lasting upregulation of Ascl1 in Rbpj-deficient r1, suggesting that Rbpj may regulate Ascl1 expression through a distinct mechanism.
Additionally, proneural bHLH transcription factors Ngn1 and Ngn2, critical for both neuronal differentiation and specification, are also expressed in the dorsal r1 in an overlapping and complementary expression pattern with Ascl1 (Liu et al., 2010). Our whole mount and section in situ hybridization revealed that the expression of Ngn1 and Ngn2 in the dorsal r1 of E10.5 RbpjWnt1 CKO embryos was comparable to wild type (supplementary material Fig. S5C,D,G,H). Taken together, our results suggest that in the absence of Rbpj in the dorsal r1, among the determinants for LC development, Ascl1 is specifically upregulated and this might account for the LC hyperplasia.
Rbpj deletion downregulates Hes1 expression
Rbpj-mediated Notch activity promotes the expression of Hes genes, such as Hes1 and Hes5 (Kageyama and Ohtsuka, 1999; Ohtsuka et al., 1999) that in turn repress the expression of proneural genes like Ascl1 (Bertrand et al., 2002). We thus examined whether Hes1 or Hes5 is altered in Rbpj CKO mice. Hes1 transcripts were expressed in the dorsal r1 of wild-type embryos at E9.5 and E10.5, but were reduced in RbpjWnt1 CKO embryos (Fig. 7A,B,F,G, arrowheads). By contrast, Hes5 expression appeared to be unchanged in the dorsal r1 but reduced in the midbrain (asterisks) of Rbpj CKO embryos compared to wild-type controls at E9.5 and E10.5 (Fig. 7C–E,H–J). Therefore, it is seemingly reasonable to propose that Rbpj deletion-induced downregulation of Hes1 expression may be responsible for upregulation Ascl1 and subsequently for LC hyperplasia. However, a line of previous evidence has shown that the LC forms normally in Hes1 or Hes5 mutant mice (Cau et al., 2000; Hatakeyama et al., 2006; Hirata et al., 2001), suggesting that Hes1 downregulation may not fully account for the LC phenotype in Rbpj CKO mice. Taken together, in general our results are consistent with the role of the canonical Notch pathway that Rbpj inactivation decreases Hes1 expression and subsequently upregulates Ascl1 expression, but Rbpj deletion-elicited lasting expression of Ascl1 in the dorsal r1 (Fig. 6) could not be fully explained by the de-repression effect of Hes1 downregulation, raising the possibility that Rbpj may regulate Ascl1 expression through an uncovered mechanism during LC morphogenesis.
Rbpj binds to the Ascl1 promoter
By analyzing the genomic sequence of mouse (AK143210) and human (AC026108) Ascl1 promoter, an alignment of mouse promoter (Arvidsson et al., 2005) with human counterpart was depicted (Fig. 8A). We found that within the 5′ region of both mouse and human Ascl1 promoters, there exists a low-affinity consensus Rbpj-binding motif (TTCTCATA) (Mumm and Kopan, 2000; Nellesen et al., 1999; Yasuhiko et al., 2006) at nt −630/−623 as well as the Hes consensus binding sites, namely, N-box (CACNAG) at nt −613/−608 and C-site (CACGCG) at nt −243/−238 (Fig. 8A). In order to examine whether this potential Rbpj-binding site within Ascl1 promoter is recognized by Rbpj, we performed an EMSA. Nuclear extracts of HEK293 cells transfected with Rbpj showed an obvious shift using a synthetic oligonucleotide probe containing a consensus Rbpj-binding sequence derived from the Epstein–Barr virus C promoter region (Rbpj) or a putative Rbpj-binding site (PR), whereas no shift was observed using a labeled probe flanking the Rbpj-binding site (FR) or containing mutated Rbpj-binding site (Prm; Fig. 8B). In contrast, nuclear extracts of untransfected HEK293 cells showed no shift using any probes above (data not shown). Moreover, the binding specificity of PR probe to Rbpj–DNA complexes was determined by competition assays employing 100-fold molar excess of unlabeled oligonucleotide probes of PR or FR. It is PR- not FR-unlabeled oligonucleotides that can compete with PR probe for the protein-complex formation (Fig. 8C).
To complement the results of the EMSA, we investigated whether Rbpj binds the site at nt −630/−623 of Ascl1 promoter (Fig. 8A) using a ChIP assay. The DNA fragments immunoprecipitated with anti-Rbpj antibody were PCR-amplified with −630 F/R primers which are located between nt −734 and −504 containing the −630/−623 putative Rbpj-binding site, with −243 F/R primers which amplified nucleotides between −368 and −128 containing the −243/−238 C-site, or with Ex1 F/R primers which are within exon 1 between nt +146 and +400. With the −630 F/R primers, a 231 bp amplicon was detected from the input DNA and from DNA fragments immunoprecipitated by anti-Rbpj antibody. By contrast, no PCR-amplifying products were found when using −243 F/R or Ex1 F/R primers (Fig. 8D). Thus, EMSA and ChIP assays reveal that Rbpj binds to the −630/−623 site of Ascl1 promoter in vitro.
To further confirm the binding of Rbpj to Ascl1 promoter, the dual-luciferase reporter assay was performed. As compared with the control, the expression of Hes1 significantly reduced Ascl1 promoter-directed luciferase activity (Fig. 8E; P<0.01). Because Rbpj combines with various co-factors under physiological condition, transfection of native Rbpj often causes the results hard to analyze (Kato et al., 1997; Kuroda et al., 1999), we thus used the constitutively active form of Rbpj (Rbpj-VP16), which shows strong transactivation activity (Waltzer et al., 1995), in the Ascl1 luciferase assay. It showed that Hes1 expression decreased whereas Rbpj-VP16 as well as the Notch1 intracellular domain NICD enhanced the luciferase activity of Ascl1 significantly (Fig. 8E; P<0.05). Co-expression of Hes1 and Rbpj-VP16 or NICD can rescued Hes1-induced decrease in luciferase activity (Fig. 8E; P<0.05), which, however, was still less than that by Rbpj-VP16 or NICD alone (Fig. 8E; P<0.05). These data imply that Rbpj can promote Ascl1 transactivation, though this promotive effect seems to be less powerful than the repressive effect induced by Hes1. In contrast, these Notch components did not affect Phox2a promoter-driven luciferase activity (Fig. 8F). Taken together, our results demonstrate that Rbpj is able to bind to Ascl1 promoter and induces Ascl1 transcription, but this activation can be counteracted by Hes1. The presence of both activating and repressing motifs in the Ascl1 promoter ensures Notch signaling pathway to finely control the LC morphogenesis via regulating Ascl1 expression.
Rbpj directly promotes Ascl1 expression
To investigate the functional significance of the binding of Rbpj to Ascl1 promoter, we performed the in vitro gene expression assay. Hes1 transfection in HEK293T cells reduced the protein and mRNA expression levels of Ascl1 to about 30–35% of the control transfected with empty vector. However, the expression of Rbpj-VP16, which directly upregulates Hes1 transcription (Li et al., 2012), could not fully mimic the repressive effect of Hes1 on Ascl1 expression as shown by a reduction of Ascl1 expression to 77–85% of the control (Fig. 9A,B), suggesting that Rbpj regulates Ascl1 expression via another way in addition to Hes1 repressive pathway. Furthermore, knockdown of Hes1 expression by siRNA significantly increased Ascl1 expression levels to about 135–150% of the control whereas co-expression of Hes1 siRNA and Rbpj-VP16 further increased Ascl1 expression levels to about 170–200% of the control (Fig. 9A,B), supporting the idea that Rbpj per se is able to upregulate Ascl1 expression directly in the absence of Hes1.
To further confirm this promotive effect of Rbpj on Ascl1 expression in vivo, we performed an in ovo electroporation assay. Various vectors were injected into the aqueduct of HH10–11 chicken embryo (Fig. 9C, asterisk). After electroporation, vectors were delivered into the precursor cells of r1, where the LC is born, and the expression of cPhox2a, cPhox2b and cDBH was examined at HH22–23 (Fig. 9D). Compared with the control, Hes1 misexpression remarkably reduced the number of Ascl1/GFP double-labeled cells and the fluorescence intensity (FI) of Ascl1 immunoreactivity. Rbpj-VP16 misexpression also downregulated the expression of Ascl1, but obviously to a lesser degree relative to that induced by Hes1 misexpression. On the other hand, knocking down Hes1 expression by siRNA upregulated Ascl1 expression. Importantly, under the condition of Hes1 knockdown, Rbpj-VP16 misexpression was capable of enhancing Ascl1 expression, and this enhancement is much higher than that induced by only Hes1 knockdown (Fig. 9E,F). Note that after electroporation of Hes1 or Rbpj plasmids, the cells with strong GFP immunoreactivity did not express Ascl1 and only those with weak or moderate GFP staining were positively stained with Ascl1 antibody (Fig. 9E; arrowheads and arrows in the insets). Taken together, these data suggest that Rbpj regulates Ascl1 expression in the two ways: an indirect repressive way via Hes1 and a direct promotive way by direct binding to Ascl1 promoter.
The majority of noradrenergic neurons in the mammalian brain are located within the LC and involve in a variety of important physiological and behavioral processes. However, to date, the determinants regulating the LC development are not yet fully elucidated. In the present study, we provide the first evidence that Rbpj, a nuclear effector of the Notch signaling pathway, involves in the LC neurogenesis possibly via regulating the transactivation of Ascl1, a key factor for noradrenergic neurogenesis.
Conditional deletion of Rbpj causes LC hyperplasia
As early as E8.5, Rbpj expression has initiated throughout the midbrain and hindbrain of the developing mouse neural tube. In both our CKO lines, the expression of Rbpj was inactivated by Wnt1- and En1-driven Cre expression in the r1 and caudal midbrain around the same stages. In addition to LC neurons, many different neuronal populations are derived from these two structures, such as the superior and inferior colliculi, midbrain dopaminergic neurons, serotonergic neurons in the dorsal raphe nucleus and cerebellum. During early embryogenesis, Notch–Rbpj signaling suppresses neurogenesis, thereby maintaining the neuronal progenitor pools. Ablation of Notch–Rbpj signaling promotes premature neuronal differentiation, leading to the depletion of neuronal progenitor pools, and thereby causing hypoplasia (Louvi and Artavanis-Tsakonas, 2006; Yoon and Gaiano, 2005). Thus, it is not surprising that the inferior colliculus was missing, and some neuronal populations, such as midbrain dopaminergic neurons, serotonergic neurons in the dorsal raphe nucleus, neurons in the superior colliculus and cerebellar neurons were reduced in Rbpj CKO mice (Fig. 4 and supplementary material Fig. S4). However, LC neurons in these mice were exceptionally increased at both early and late embryonic stages, as a consequence of potentiated LC neuron generation that was sustained over a longer time window, with no change in cell death (Figs 2, 3, and supplementary material Figs S1, S2). Thus, the exclusive overproduction of LC neurons in the absence of Rbpj indicates that in addition to canonical Notch–Rbpj signaling pathway, Rbpj seems to regulate LC neurogenesis in another novel way.
LC hyperplasia in Rbpj CKO mice is due to enhanced Ascl1 expression
Previous studies have shown that BMPs are required for LC specification in zebrafish and in chick (Guo et al., 1999; Hynes and Rosenthal, 1999; Vogel-Höpker and Rohrer, 2002). We showed here that Bmp6 and Bmp7 were expressed in the r1 roof plate of mice; however, their expression levels, as well as those of their downstream target genes (i.e. Msx1–3), remained unchanged in this region of Rbpj CKO mice (Fig. 5). Moreover, the expression of Nr2f6, a gene recently identified as being required for LC specification, was also expressed normally in the dorsal r1 when Rbpj was deleted. Thus, it is likely that BMP signaling and Nr2f6 do not account for the overproduction of LC neurons in the absence of Rbpj expression.
Inactivation of canonical Notch signaling leads to the downregulation of Hes gene expression (Machold et al., 2007; Okamura and Saga, 2008), such as Hes1 and Hes5 (Kageyama and Ohtsuka, 1999; Ohtsuka et al., 1999). Hes genes in turn repress the expression of proneural genes like Ascl1, which is required for the development of LC neurons (Hirsch et al., 1998). Consistently, our results demonstrated that Hes1 expression was downregulated (Fig. 7) and Ascl1 expression was greatly upregulated (Fig. 6) in the dorsal r1 of Rbpj CKO mice. Thus, this enhanced Ascl1 expression is likely to result from a de-repression effect of Hes1 expression downregulation induced by Rbpj deletion. However, a group of previous studies have shown that Hes1 knockout mice show no obvious changes in the LC (Hirata et al., 2001), and similar results are also reported in Hes5 mutant mice (Cau et al., 2000). Moreover, Hes1;Hes5 double mutant mice, in which the size, shape, and cytoarchitecture of the CNS are severely disorganized (Hatakeyama et al., 2006; Kim et al., 2008), show the reduced but not increased cell number of the LC. Importantly, our results also revealed a lasting expression pattern of Ascl1 in Rbpj-deficient dorsal r1, and this phenomena has not reported in the developing CNS of other Rbpj CKO mice (Imayoshi et al., 2008; Lütolf et al., 2002; Machold et al., 2007). Taken together, the upregulation of Ascl1 expression could not be simply explained by Rbpj-deletion-induced downregulation of Hes1 expression, and it is likely that Rbpj may regulate Ascl1 expression in the dorsal r1 through an uncovered mechanism.
Rbpj directly binds to the Ascl1 promoter and controls its expression
By analysis of the sequence of Ascl1 gene, we found a potential Rbpj-binding motif in the Ascl1 promoter region where consensus binding sequences for Hes1 (e.g. N-box and C-site) are present. The EMSA and ChIP assays confirmed the binding of Rbpj to this putative DNA site, and dual-luciferase reporter assay revealed that Ascl1 transcriptional activity can be activated by Rbpj and repressed by Hes1 (Fig. 8). In this sense, our previous study also demonstrates that Rbpj and Hes1 can simultaneously regulate Foxp3 transcription in a bi-phasic manner (Ou-Yang et al., 2009). Our in vitro and in vivo gene expression assays provided the functional evidence that Rbpj per se is capable of promoting Ascl1 expression, though it also repressed Ascl1 expression indirectly in the presence of Hes1 (Fig. 9). Additionally, we noted that the promotive effect of Rbpj on Ascl1 expression is more evident in dual-luciferase reporter assays than that of in vitro gene transfection and in ovo electroporation assays. This may be explained by the fact that in the luciferase reporter system, exogenous Rbpj can directly bind to Ascl1 promoter with no or less affecting endogenous Hes1 expression.
Based on our present results and the working flow of canonical Notch signaling pathway, we hypothesize a cooperative model of Rbpj and Hes1 in the neurogenesis of LC neurons. In wild-type mice, once Notch signaling is activated, Rbpj not only induces the transcription of Hes1 but also activates Ascl1 transactivation; Hes1 then binds to Ascl1 promoter and concomitantly represses its transcription. In other words, Rbpj indirectly represses Ascl1 expression via Hes1, while it also directly promotes Ascl1 expression. When Rbpj is deleted, Hes1 expression is greatly downregulated, leading to the upregulation of Ascl1 expression via a de-repressive mechanism. Simultaneously, loss of Rbpj also results in an inability of Rbpj directly induced Ascl1 expression, and this can counteract to some extent Hes1 downregulation-induced Ascl1 upregulation. The net effect of these two events is that Rbpj inactivation causes Ascl1 upregulation and subsequently LC overproduction for a relative long period. Taken together, our findings demonstrate a new role of Rbpj in regulating Ascl1 expression. We propose that the direct promotive effect and indirect repressive effect on Ascl1 expression is present simultaneously during the LC morphogenesis.
Interestingly, Rbpj deletion-induced excessive expression of Ascl1 in the developing CNS only occurs in the early neural tube, which is always followed by a dramatic downregulation of Ascl1 expression as the development progressed, and this has been believed to be caused by the depletion of the progenitor pool (Imayoshi et al., 2008; Louvi and Artavanis-Tsakonas, 2006; Lütolf et al., 2002; Machold et al., 2007; Yoon and Gaiano, 2005). However, this common phenomenon is not the case in our Rbpj CKO mice. In addition, Rbpj and Ascl1 are widely co-expressed in the ventricular zone of the neural tube. However, LC neurons were the only cell type produced in excess in our Rbpj CKO mice. Therefore, it is likely that there exists other cofactor(s) operating with Rbpj specifically in LC primordium to prevent the depletion of progenitor pool and maintain Ascl1 expression. In this sense, recent studies have shown that Rbpj forms a complex with the transcription factor Ptf1a, which is required for specification of GABAergic interneurons in spinal cord (Hori et al., 2008) and of acinar cells during early pancreatic development (Masui et al., 2007). Further studies are needed to identify the cofactor(s) which is specifically expressed in LC progenitors and cooperates with Rbpj in controlling the development of LC neurons.
Materials and Methods
Mouse breeding and genotyping
Wnt1-Cre, En1Cre/+, Rbpjflox/flox and Rosa26 reporter mice were generated and genotyped as previously described (Danielian et al., 1998; Han et al., 2002; Kimmel et al., 2000; Soriano, 1999). To inactivate Rbpj expression in the dorsal r1, we crossed Wnt1-Cre or En1Cre/+ mice with Rbpjflox/flox mice to obtain Wnt1-Cre;Rbpjflox/flox or En1Cre/+;Rbpjflox/flox progeny (referred to as RbpjWnt1 CKO or RbpjEn1 CKO, respectively). The morning of the day on which the vaginal plug appeared was designated as embryonic day (E) 0.5. In each set of experiments, at least three embryos or pups in each experimental group and an equal number of control mice were used. Animal care procedures were reviewed and approved by the Animal Studies Committee at the Tongji University School of Medicine, Shanghai, China.
BrdU labeling, X-gal staining and TUNEL staining
For BrdU birthdating experiments, a single BrdU pulse (60 µg/g of body weight; Sigma, St Louis, MO, USA) was delivered intraperitoneally to timed-pregnant females on post-coitum day 8.5, 9.5, 10.5, 11.5, 12.5 and 13.5, and embryos were collected at E17.5. Sections were processed for TH and BrdU double immunostaining (see below) to localize BrdU-positive cells in the LC. X-gal staining and TUNEL staining of frozen sections were performed as described previously (Shi et al., 2008; Zheng et al., 2009).
In situ hybridization and immunostaining
Whole mount and section in situ hybridization were performed as previously described (Shi et al., 2008; Shi et al., 2010). The following mouse antisense RNA probes were used: c-Ret (Ohsawa et al., 2005), Dbh (Morin et al., 1997), Fgf8 (Wassarman et al., 1997), Gbx2 (Liu and Joyner, 2001b), Hes1 (Lee et al., 2005), Hes5 (Ohtsuka et al., 1999), Ngn1, Ngn2 (Sommer et al., 1996), Otx2 (Martinez-Barbera et al., 2001), Phox2a (Tiveron et al., 1996) and Phox2b (Pattyn et al., 1997). Other probes used in this study, including BDNF (NM_007540; 0.80 kb), Bmp2 (NM_007553; 0.60 kb), Bmp4 (NM_007554; 0.56 kb), Bmp5 (NM_007555; 0.45 kb), Bmp6 (NM_007556; 0.52 kb), Bmp7 (NM_007557; 0.43 kb), chicken Phox2a (cPhox2a, Z49262; 0.34 kb), chicken Phox2b (cPhox2b, XM_001234150; 0.54 kb), chicken DBH (cDBH, XM_415429; 0.73 kb), L-AADC (NM_016672; 1.0 kb), MaoA (NM_173740; 0.43 kb), Ascl1 (NM_008553; 1.0 kb), Msx1 (NM_010835; 0.63 kb), Msx2 (NM_013601; 0.65 kb), Msx3 (NM_010836; 0.60 kb), Nr2f6 (NM_010150; 0.58 kb), Rbpj (NM_009035; 0.90 kb), Wnt1 (NM_021279; 0.84 kb) and vGluT1 (NM_182993; 0.72 kb), were generated by PCR amplifying cDNA templates prepared from E12.5 mouse or HH22 chicken embryos. For immunofluorescent staining, the following primary antibodies were used: rabbit anti-Ascl1 (1∶200; Abcam, Cambridge, UK), rabbit anti-β-gal (1∶1000; Novus Biologicals, Littleton, CO, USA), mouse anti-BrdU (1∶200; Calbiochem, Darmstadt, Germany), goat anti-Brn3 (1∶300; Santa Cruz Biotech, Santa Cruz, CA, USA), goat anti-Brn3a (1∶200; Santa Cruz), rabbit anti-Calbindin (1∶1000; Chemicon, Temecula, CA, USA), rabbit anti-CGRP (1∶1000; Biogenesis, Kingston, NH, USA), rabbit anti-Caspase3 (1∶500; Cell Signaling Technology, Beverly, MA, USA), rabbit anti-GFP (1∶2000; Molecular Probes, Eugene, OR, USA), mouse anti-Islet1 (1∶100; Developmental Studies Hybridoma Bank, Iowa City, IA, USA), rabbit anti-Parvalbumin (1∶1000; Swant, Bellizona, Switzerland), rabbit anti-Pax6 (1∶200; Covance, Princeton, NJ, USA), rabbit anti-serotonin (5-HT) (1∶1000; Sigma), rabbit anti-Tbr1/2 (1∶2500, a gift from R. Hevner, University of Washington, Seattle, WA, USA), mouse anti-TH (1∶5000; Sigma), rabbit anti-TH (1∶1000; Chemicon) and rabbit anti-Tlx3 (1∶2000, a gift from T. Müller and C. Bierchmeier, Max-Delbrück-Center for Molecular Medicine, Berlin, Germany). Species-specific secondary antibodies conjugated to Cy2 or Cy3 (1∶1000; Jackson ImmunoResearch, West Grove, PA, USA) were used to detect primary antibodies. Sections were observed under a Nikon 80i or a Zeiss LSM510 confocal microscope.
Electrophoretic mobility shift assay EMSA and chromatin immunoprecipitation assay
The EMSA was used to study wherever there is a direct binding of Rbpj to Ascl1 promoter, and it was performed as previously described (Song et al., 2003). Nuclear extracts were prepared from HEK293 cells transfected with Rbpj (Santa Cruz). Four biotinylated oligonucleotide probes were used for EMSA: a consensus binding sequence for Rbpj (5′-AAACACGCCGTGGGAAAAAATTTGG-3′) derived from the Epstein–Barr virus C promoter region (Lu and Lux, 1996); the sequence from Ascl1 promoter containing the putative Rbpj-binding motif (PR) (5′-GACTCAAGTTCTCATACAGAGAG-3′); the mutated PR site (PRm) (5′-GACTCAAGTTCgacTACAGAGAG-3′); and the sequence from Ascl1 promoter flanking PR site (FR) (5′-TGGGTGTCCCATTGAAAAGGCGG-3′). The control was HEK293 cells not transfected with Rbpj. To confirm DNA sequence specificity of the protein–DNA complex formation, the competition experiments with 100-fold molar excess of unlabeled oligonucleotides were performed.
For in vivo ChIP assay, E9.0–E9.5 wild-type mouse embryos were rapidly dissected in ice-cold PBS and the r1 with isthmus organizer was isolated. The samples were cross-linked in 1% formaldehyde at room temperature for 30 min. After stopped by adding 150 mM glycine, the tissues were sonicated in SDS lysis buffer. The debris was removed by centrifugation and cleared lysates were used for immunoprecipitation by a ChIP assay kit (Millipore, Billerica, MA, USA). The specific DNA-bound transcription factor complexes were precipitated with an anti-Rbpj antibody (Santa Cruz). After the proteins were removed from DNA by digestion with proteinase K, the purified immunoprecipitated DNA was subjected to PCR with −630 F/R primers (5′-AGCTGAATGGAACAGCAGTGGCAACC-3′ and 5′-CCCTTTCTTCTCTCCGCAGTAACTCC-3′) and −243 F/R primers (5′-ACCCCAAGTCCAGGAGTTATTTGCC-3′ and 5′-CCCACCTCCTCAGCTCCCTCCCTCT-3′), which contain and flank the potential Rbpj-binding site, respectively. The DNA sample was also amplified using Ex1 F/R primers (5′-TGGAGCAAGGGAGAGCGGGCGCAAG-3′ and 5′-TGTCAGGCTGCAGCGGAGGAAGGTG-3′) located within exon 1 of mouse Ascl1 gene for a control.
Plasmid preparation, dual-luciferase reporter assay and in vitro gene expression assay
Expression plasmids encoding Hes1, NICD, and Rbpj-VP16 (a constitutively active form of Rbpj) and an siRNA expression plasmid targeting Hes1 (Hes1 siRNA) were prepared as described previously (Shi et al., 2011). Ascl1- and Phox2a-luciferase reporter plasmids were constructed by subcloning the 5′-upsteam elements of mouse Ascl1 (Arvidsson et al., 2005) or Phox2a (Benjanirut et al., 2006) into the pGL3-enhancer vector (Promega, Madison, WI, USA).
For dual-luciferase reporter assay, HEK293T cells were seeded at 1.0×105 cells per well in 24-well plates in DMEM supplemented with 10% FBS. The following day, cultures were transiently co-transfected with a total 600 ng of DNA containing the pGL3-Ascl1 or pGL3-Phox2a reporter plasmid (400 ng each) and the expression vector encoding Hes1, NICD or Rbpj-VP16 (200 ng each), using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). The empty pGL3 vector was used as the control. Thirty-six hours after transfection, the cells were lysed with 100 µl/well of Passive Lysis Buffer and subjected to a luciferase assay using the Dual-Luciferase System (Promega). In all experiments, 5 ng/well of Renilla luciferase expression vector pRL-SV40 was used as an internal control and luciferase activity was normalized to the internal control activity. For each experimental condition, assays were performed in quadruplicate for each of at least three independent transfections.
For in vitro gene expression assay, HEK293T cells as prepared above were transfected with the expression plasmid (encoding Hes1 or Rbpj-VP16) or Hes1 siRNA plasmid by Lipofectamine 2000 (Invitrogen). For co-transfection, Rbpj-VP16 and Hes1 siRNA plasmids in a 1∶1 ratio were mixed before transfection. The pCAGGS-IRES-EGFP vector was used as the control. Forty-eight hours after transfection, the cells were subjected to western blotting and quantitative real-time reverse-transcription-PCR (RT-PCR) assays as described previously (Zhang et al., 2012). For western blotting assay, rabbit anti-Ascl1 (1∶1000; Abcam) and mouse anti-β-tubulin (CoWin, Tianjing, China) antibodies were used. The results were normalized to β-tubulin expression, which was detected on the same blot, and further normalized to the control. For real-time RT-PCR, the following primer sequences were used: Ascl1, F: 5′-AGATGAGCAAGGTGGAGACGCT-3′, R: 5′-GGAGTAGGACGAGACCGGAGAA-3′; GAPDH, F: 5′-ACCCATCACCATCTTCCAGGAG-3′, R: 5′-GAAGGGGCGGAGATGATGAC-3′. The data were normalized to GAPDH expression and further normalized to the control.
In ovo electroporation
In ovo electroporation was performed as previously described (Shi et al., 2011). In brief, after fertilized chicken eggs were incubated to Hamburger and Hamilton (HH) stage HH10–11, 0.5 µl of expression plasmid (1.0 µg/µl in sterile PBS) or Hes1 siRNA plasmid was injected into the midbrain aqueduct of chicken embryo with glass capillaries (see schematic in Fig. 9). For co-electroporation, Rbpj-VP16 and Hes1 siRNA vectors in a 1∶1 ratio were mixed before the injection. After injection, embryos were pulsed five times (20 V for 50 ms) at 1 s intervals under an Electro Square Porator ECM830 (BTX Harvard Apparatus, Holliston, MA, USA). Electroporated embryos were incubated for another 48 h to stage HH22–23, and then harvested for immunohistochemistry and in situ hybridization.
TH-, BrdU β-gal- and Ascl1-positive cells in the LC were counted in every five section. For each set of comparison at least three experimental mice (RbpjWnt1 CKO or RbpjEn1 CKO) and littermate wild-type controls were included. All data were analyzed using OriginPro7.5 software and presented as means ± s.e.m. Comparisons were made using an unpaired Student’s t-test and statistical significance was set at P<0.05.
We thank Jia-Yin Chen, Li Wang and Xiao Zhang for technical support, A. P. McMahon and A. Joyner for providing Wnt1-Cre and En1Cre/+ mice, R. Hevner for anti-Tbr1/2 antibody, and T. Müller and C. Bierchmeier for anti-Tlx3 antibody.
This work was supported by the National Natural Science Foundation of China [grant numbers 31030034, 31170801, 81101026, 31100788]; and Ministry of Science and Technology of China [grant numbers 2011CB510005 and 2012CB966904].