The F-box protein Cdc4/Fbxw7 is a novel regulator of neural crest development in Xenopus laevis

Background The neural crest is a unique population of cells that arise in the vertebrate ectoderm at the neural plate border after which they migrate extensively throughout the embryo, giving rise to a wide range of derivatives. A number of proteins involved in neural crest development have dynamic expression patterns, and it is becoming clear that ubiquitin-mediated protein degradation is partly responsible for this. Results Here we demonstrate a novel role for the F-box protein Cdc4/Fbxw7 in neural crest development. Two isoforms of Xenopus laevis Cdc4 were identified, and designated xCdc4α and xCdc4β. These are highly conserved with vertebrate Cdc4 orthologs, and the Xenopus proteins are functionally equivalent in terms of their ability to degrade Cyclin E, an established vertebrate Cdc4 target. Blocking xCdc4 function specifically inhibited neural crest development at an early stage, prior to expression of c-Myc, Snail2 and Snail. Conclusions We demonstrate that Cdc4, an ubiquitin E3 ligase subunit previously identified as targeting primarily cell cycle regulators for proteolysis, has additional roles in control of formation of the neural crest. Hence, we identify Cdc4 as a protein with separable but complementary functions in control of cell proliferation and differentiation.


Background
During the development of multi-cellular organisms, cells receive signals and must elicit the appropriate response. This involves changes in the level and activity of proteins, and targeted proteolysis represents a rapid and irreversible mechanism to block protein function. During regulated proteolysis, proteins are targeted for degradation by covalent attachment of the 76 amino acid protein ubiquitin, and the polyubiquitin chains assembled on the target protein serve as signals for degradation by the 26S proteasome. Transfer of ubiquitin onto target proteins is catalyzed by a hierarchical multi-enzyme cascade. An E1 (ubiquitin activating) enzyme forms a thioester linkage with the carboxyl terminus of ubiquitin, in an ATP-dependent process. Ubiquitin is then transferred to an E2 (ubiquitin conjugating) enzyme. E3 (ubiquitin ligase) enzymes recruit distinct substrates, allowing ubiquitin transfer, and confer specificity on the ubiquitin proteasome system. RING (Really Interesting New Gene) E3s are the largest class of E3 ligases, and the human genome encodes approximately 400 proteins with a RING domain [1]. Conserved cysteines and histidines coordinate two zinc ions in the RING domain, which is important for the recruitment and activation of E2 enzymes. Skp1-Cul-lin1-F-box (SCF) E3 ligases are a large class of modular RING E3 ligases that have the RING component Roc1 (also known as Rbx1 and Hrt1). Cullin1 forms a scaffold to recruit the E2 (via Roc1) and the F-box protein (via binding of the F-box to Skp1) [2,3]. The F-box component of these E3 ligases is variable, and different F-box proteins recruit different substrates via carboxy-terminal domains, allowing SCF ligases to target a huge number of substrates [4].
In mammals, there are three isoforms of Cdc4: alpha (α), beta (β) and gamma (γ). These are produced by alternative splicing of three unique 5' exons to ten common 3' exons, such that the resulting proteins differ only at their amino termini [9,10]. In mammals, known Cdc4 substrates include c-Myc, c-Jun, Cyclin E, Notch intracellular domain, c-Myb, sterol regulatory element binding proteins (SREBPs) and steroid receptor coactivator-3 (SRC3) [9,[11][12][13][14][15]. Given these substrates, it is perhaps unsurprising that Cdc4 has been shown to be a haplo-insufficient tumor suppressor gene [16]. This list of substrates also suggests that Cdc4 could regulate developmental events, and attempts to generate knockout mice led to an embryonic lethal phenotype [17]. We became interested in a role for Cdc4 during neural crest development in particular because several of its substrates have been implicated in the development of this tissue, for example, c-Myc and Notch intracellular domain [18,19].
The neural crest is a unique population of cells, arising at the neural plate border in response to bone morphogenetic protein, Wnt and fibroblast growth factor signaling (for reviews, see [20,21]). Neural crest cells are initially multipotent, but subsequently undergo an epithelial to mesenchymal transition and migrate throughout the embryo, where they give rise to a wide range of derivatives (for reviews, see [22,23]). These include the neurons and glia of the peripheral nervous system, the autonomic nervous system, cartilage, bone, connective tissue, cardiac cells and melanocytes. The induction of the neural crest is often defined according to the expression of neural crest specifier genes, including the transcriptional repressors Snail2 and Snail [24,25]. A number of proteins involved in neural crest development display dynamic expression patterns, and it is becoming apparent that several are targets of the ubiquitin proteasome system. For example, Snail2 is degraded by the F-box protein Partner of paired (Ppa) [26].
Here we describe identification of the Xenopus laevis homologues of Cdc4, which are highly conserved at the sequence level, and are also functionally equivalent in terms of their ability to degrade Cyclin E. Two isoforms of Xenopus Cdc4 (xCdc4α and xCdc4β) are found to be dynamically expressed throughout early Xenopus development, with particular enrichment in neural crest and neural crest-derived tissues. Inhibition of xCdc4 activity, using dominant negative F-box mutants, blocks neural crest development, without affecting cell division or cell survival, nor affecting development of the other tissues in which they are expressed. Thus, Cdc4 directly and specifically regulates neural crest formation, independent of a previously described ability to regulate the cell cycle.

Results
X. laevis encodes two isoforms of Cdc4: xCdc4a and xCdc4b The pseudotetraploid genome of X. laevis presents unique challenges to identifying genes reported in other model systems. In contrast, Xenopus tropicalis is possessed of a diploid genome, making it well suited for genetic manipulation and bioinformatic analysis. The existence of X. tropicalis and X. laevis in the same genus -therefore sharing a high level of evolutionary conservation between their respective genes -suggested to us a method of harnessing the sequenced genome of X. tropicalis to identify potential orthologs of human Cdc4 (hCdc4) present in X. laevis. BLAST of hCdc4α (GenBank accession number AY049984) against the X. tropicalis genome revealed a sequence on scaffold 60:1694203-1694241 with strong nucleotide complementarity to the first exon of hCdc4α. Further downstream (scaffold 60:1696322-1,729,489) were exons 2 to 11, whose sequence corresponded to the conserved regions found within hCdc4.
PCR primers targeting the identified xCdc4 sequence were used on oligo d(T) primed mRNA derived from stage 20 X. laevis embryos. We performed 5' and 3' rapid amplification of cDNA ends (RACE) to isolate a full length product. Although slightly truncated, the protein was most similar to the β isoform of hCdc4 and has been designated xCdc4β (Additional file 1). Analysis of the aligned regions demonstrates that xCdc4β was 98% identical at the amino acid level and 83% identical at the nucleotide level to hCdc4β. In comparison to other X. laevis F-box proteins, xCdc4β was 27% identical and 43% similar to β -TRCP (β -Transducin repeat containing protein; GenBank accession number M98268) [27], and 17% identical and 20% similar to Skp2 (GenBank accession number DQ228920) [28].
A second isoform of xCdc4 was similarly cloned from stage 7 X. laevis embryos (GenBank accession number DQ666345). Exon 1 was found on scaffold 60:1646293-1646793. Analysis demonstrated nearly perfect conservation at the amino acid level to the common set of exons, 2 to 11, that occur in hCdc4α and hCdc4β (data not shown). Additionally, the second identified xCdc4 contained an amino terminal exon most similar to the one present in human Cdc4α and is referred to hereafter as xCdc4α. Furthermore, exon 1 of xCdc4β is spliced to the same downstream exons, 2 to 11, used by xCdc4α. Alignment of their amino acid sequences showed strong conservation; differences in amino acid identity being conserved by use of residues with similar characteristics. Furthermore, both the isoform specific nuclear localization signal (amino acids 11 to 14) and common nuclear localization signal (amino acids 169 to 172) found in hCdc4α are present in xCdc4α, which is suggestive of a similar pattern of localization and regulation [29].
From the region of the gene corresponding to the second exon, xCdc4α and xCdc4β are 99.8% identical (1,623 out of 1,626 nucleotides identical). A single amino acid substitution was made (changing Gly649 to Asp) to the xCdc4α coding sequence used here, as this was conserved in all sequence orthologs. A X. laevis ortholog of hCdc4γ was not detected in either stage 7 or stage 20 embryos.
xCdc4a and xCdc4b proteins are expressed throughout development, including prominent expression in the early neural crest Next, the temporal expression of xCdc4α and xCdc4β was examined during X. laevis development. Staged embryo lysates were separated by SDS PAGE and western blotted using antibodies to detect the two different isoforms, xCdc4α and xCdc4β. Anti-Cdc4 3B7 (referred to as anti-Cdc4α) detected both endogenous and in vitro translated xCdc4α ( Figure 1A). Although unable to detect endogenous levels of xCdc4β, this antibody weakly detected overexpressed xCdc4β (data not shown). Anti-Cdc4 3A9 (referred to as anti-Cdc4β) detected endogenous and in vitro translated xCdc4β ( Figure 1B) Anti-Cdc4β also detected xCdc4α, although less effectively than anti-Cdc4α (data not shown). Thus, western blotting demonstrates that both xCdc4 isoforms were expressed at all stages tested, from fertilized egg until stage 28.
The spatial distribution of xCdc4 transcripts during X. laevis embryogenesis was also examined by in situ hybridization (ISH) using a probe that would recognize both isoforms of xCdc4. Prior to stage 15, xCdc4 transcripts were expressed in a diffuse pattern in the ectoderm (data not shown). At stage 15, while more diffuse background staining remained, transcripts accumulated at the highest level in the neural crest, forming at the preplacodal ectoderm ( Figure 1C, ppe) at the border of the epidermis and neural plate and in particular in the presumptive optic placode ( Figure 1C, pop). At late neurula stages, xCdc4 transcripts are detected in the trunk ( Figure 1D, E, TNc) and branchial neural crest ( Figure 1D, b) as well as in the optic placode ( Figure 1E, Opt P), and this staining pattern persists until stage 20 ( Figure 1D, E). While continuing to be expressed in the neural crest derivatives (cranial neural crest (Cc) and branchial arches (Ba) at later stages ( Figure 1F-G)) xCdc4 was additionally expressed in the myotome (Figure 1G, H, My). From stage 26, staining of the posterior somites (s) and the neural crest-derived branchial arches (Ba) persists, while staining in the fin mesenchyme (Fin), another neural crest-derived tissue, becomes prominent ( Figure 1I-K). Expression in anterior placodes also becomes pronounced at stage 32 ( Figure 1K, arrowheads). Staining was not observed with the sense control probe ( Figure 1L-N). In summary, xCdc4 transcripts were detected in the neural crest and placodes, neural crest-derived tissues, and additionally in the brain and somites. We compared the expression of xCdc4 with other markers of neural crest and placodes: Snail2 (also known as Slug), Snail, Sox10, Six1, Opl and Pax3 at stages 15/16, 18/19, 21/22, 26/27 and 32/33 [25,30-34] (Figures 2 and 3). At neurula and early tailbud stages, although xCdc4 shows a distinct expression pattern from each of these markers, its expression overlapped both in placodal regions and at the border of the epidermis and neural plate, where early neural crest arises ( Figure 2). Similarly, at later stages xCdc4 expression in the branchial arches (Ba), pharyngeal pouches (Pp), cranial ganglia (Cg) and olfactory placode (Olf P) overlapped most significantly with the neural crest marker Snail and placodal marker Six1 (Figure 3).
Surface expression of xCdc4 was clear by whole mount ISH and was most prominent in neural crest and placodal regions. However, we saw less distinct staining more widely across the embryo and in other tissues, such as the myotome and potentially the neural tube (for example, Figure 1G), especially when the staining period was lengthened. Whole mount ISH is most suited to detect expression in tissues near to the embryo surface. To look more closely at xCdc4 expression in deep tissues, we performed ISH on sections. When allowing the ISHs to develop for 2 days, xCdc4 expression was seen throughout dorsal ectodermal and mesodermal tissue of the embryo, including in the neural plate/tube, epidermis, myotome and the notocord, with somewhat weaker staining in the epidermis (Figure 4A, B, G-I), while no staining was seen in the sense control ( Figure  4E, F, J-L), demonstrating specificity. In agreement with our whole mount staining ( Figure 1D), as the neural tube was closing, expression of xCdc4 appears strongest in areas lateral to the neural plate ( Figure 4B) that also express Snail2 ( Figure 4D), and so are developing neural crest. While other tissues also expressed xCdc4, staining of the neural crest was clear in all embryo sections examined ( Figure 4A, B, G-I, arrows).

xCdc4 degrades Cyclin E and xCdc4ΔFbox mutants act as dominant negatives
Mammalian Cdc4, and in particular Cdc4α, is known to target Cyclin E for ubiquitin-mediated proteolysis [35][36][37][38]. We wanted to determine whether xCdc4 could target Cyclin E for degradation in X. laevis embryos, both by overexpressing the xCdc4 protein and by knocking out its function. To block xCdc4 function, we initially tried to prevent translation of xCdc4 mRNAs by microinjection of antisense morpholino oligonucleotides. Although xCdc4-directed morpholinos specifically inhibited translation of synthetic mRNA encoding xCdc4α or xCdc4β, demonstrating their functionality, microinjection into embryos had only a small effect on xCdc4α protein levels and no detectable effect on xCdc4β levels, as detected by western blot (data not shown). Asterisk indicates premigratory cranial crest; b, branchial crest; Cc, cranial neural crest; di, diencephalons; dlp, dorsolateral placodal area; Hg, hatching gland; h, hyoid crest; m, mandibular crest; me, mesencephalon; Nc, neural crest; Olf P, olfactory placode; Opt P, optic placode; plp, presumptive lens placode; pOlf P, presumptive olfactory placode; ppe, preplacodal ectoderm; pop, presumptive optic placode; s, somites; tel, telencephalon; TNc, trunk neural crest; yellow arrowhead, neural plate border region of the prospective rhombencephalon. Anterior view, dorsal up, stages as indicated. Morpholinos are only effective when protein levels depend on new translation of mRNAs. Our developmental western blot ( Figure 1A, B) indicated that both xCdc4α and xCdc4β are supplied in the egg as maternal stockpiles, and these would not be affected by this antisense strategy.
xCdc4 is a member of the F-box E3 ligase family, and as such, requires an intact F-box to interact with the rest of the SCF complex to target proteins for proteolysis. Deletion of the F-box allows substrate binding but prevents recruitment to the ubiquitination machinery, and this strategy has been successfully used many times to create a dominant negative construct (for example, [9,39]), whose overexpression results in specific substrate hyper-stability. Therefore, we used an F-box deleted dominant negative form of xCdc4 to block the function of endogenous xCdc4 protein.
Cyclin E overexpression in X. laevis embryos results in a loss of DNA at early cleavage stages. This leads to a reduction in the rate of cell cleavage, followed by apoptosis of the affected cells at mid-gastrulation [40]. To determine whether xCdc4α and xCdc4β could both act as E3 ligases for Cyclin E in vivo, and to demonstrate that xCdc4ΔFbox has lost this activity, we compared their ability to inhibit Cyclin E-mediated apoptosis after micro-injection into X. laevis embryos, as compared with the anti-apoptotic protein BclXL. As expected, injection of Cyclin E either alone or with green In addition to the expression in both superficial and deep layers of the ectoderm, there is also expression of xCdc4 in the notochord (n) and in somitogenic mesoderm (sm). fluorescent protein (GFP) mRNA resulted in a slowing of blastomere cleavage detectable at stage 9, followed by apoptosis at mid-gastrulation stage 11 ( Figure 5B, H). Co-injection of Cyclin E and BclXL resulted in the appearance of larger blastomeres, but these failed to undergo apoptosis later, as expected ( Figure 5E, K). Injection of both xCdc4α and xCdc4β along with Cyclin E suppressed the number of embryos with enlarged cells and largely prevented apoptosis ( Figure 5A, C, D, I, J), while xCdc4αΔFbox and xCdc4βΔFbox proteins were ineffective at suppressing these phenotypes ( Figure 5A, F, G, L, M). Thus, xCdc4α and xCdc4β can act as E3 ligases in vivo when overexpressed and require the Fbox for this function.
Cdc4 is thought to target only a phosphoform of Cyclin E for destruction [35][36][37][38]. To look directly at the effect of xCdc4 on degradation of Cyclin E protein by western blot, 1 ng of RNA encoding amino-terminal FLAG Cyclin E was injected into fertilized eggs along with mRNAs encoding xCdc4 (using the α isoform) or GFP, as an injection control ( Figure 6), and embryos were allowed to develop to stage 10.5. Embryo lysates were prepared and immunoblotting was performed to detect the FLAG epitope on Cyclin E.
Cyclin E migrated as multiple bands in GFP-injected embryos ( Figure 6A, lane 2), and these are likely to be phospho-forms of the protein [41]. When co-injected with Cyclin E mRNA, xCdc4 expression resulted in a reduction in the abundance of predominantly the slower-migrating phospho-forms of Cyclin E protein ( Figure 6A, lane 3). Densitometry analysis, normalizing to actin expression in the same samples, confirmed that phosphorylated forms of Cyclin E are degraded by xCdc4. This is in agreement with previous results, which demonstrated a requirement for Cyclin E phosphorylation to direct Cdc4-mediated degradation [9,38,42]. These results verify that xCdc4 is functionally orthologous to mammalian Cdc4 in its ability to degrade phospho-forms of Cyclin E.
To confirm that the F-box deletion mutant of xCdc4 (xCdc4ΔFbox) possessed dominant negative activity, its ability to inhibit Cyclin E degradation mediated by wildtype xCdc4 was assessed. Cyclin E mRNA was coinjected with xCdc4 along with one of the following: xCdc4ΔFbox, xSkp2ΔFbox, an F-box mutant of the related SCF E3 ligase Skp2, or GFP ( Figure 6A, lanes 2 to 5). As expected, overexpression of either xSkp2ΔFbox or GFP had no effect on phospho-Cyclin E degradation by xCdc4. In contrast, degradation of Cyclin E by xCdc4 was inhibited by co-injection of xCdc4ΔFbox (lane 5), confirming that this mutant acts as a dominant negative form of xCdc4 in the embryos. When phospho-Cyclin E levels were normalized to the levels in embryos injected with Cyclin E and GFP, xCdc4 and GFP co-injection led to a 60% reduction in Cyclin E levels ( Figure 6B). When xCdc4 was co-injected with xSkp2ΔFbox, a similar reduction in Cyclin E levels was observed (70%). However, coexpression of xCdc4ΔFbox with xCdc4 partially rescued the degradation of Cyclin E. On average, there was a 30% reduction in Cyclin E levels compared to Cyclin E-and GFP-injected embryos. In addition, we also found that xCdc4 targeted endogenous Cyclin E for degradation (data not shown), again predominantly reducing the slower-migrating phospho-form of the protein.
At these early developmental stages, Cyclin E is expressed throughout the ectoderm [43], so may act as a target for xCdc4. Since Cdc4 is known to regulate the stability of other proteins that regulate cell proliferation (for a review, see [10]), we investigated whether xCdc4 modulated cell cycle progression in the early embryo. As xCdc4α and xCdc4β share the same substrate binding regions, xCdc4βΔFbox (hereafter known as xCdc4ΔFbox) would be expected to inhibit the activity of both xCdc4 isoforms, and was used throughout in this study. Indeed, xCdc4αΔFbox gave similar results to xCdc4βΔFbox (data not shown).

xCdc4ΔFbox does not affect cell cycle in the embryonic ectoderm
To investigate the effect of xCdc4 activity on cell proliferation, mRNA encoding xCdc4β or xCdc4ΔFbox was injected unilaterally into two-cell-stage embryos with GFP mRNA injected as a control. Whole mount staining using an antibody against phosphorylated histone H3 (phH3) was used to examine the number of mitotic cells [44]. No change in the number of mitotic cells in the neural folds was observed following overexpression of xCdc4ΔFbox or xCdc4 (Figure 7). To quantify this, the number of mitotic cells in the neural folds was counted on the injected and uninjected sides, and the percentage difference was calculated for each embryo. The average percentage change for xCdc4ΔFbox (-2.8 ± 3.9%) and xCdc4 (-9 ± 1.2%) overexpression was not significantly different from that found after GFP overexpression (-8.3 ± 1.5%, n = 84 to 91). Therefore, overexpression of xCdc4ΔFbox did not perturb the cell cycle. Indeed, we saw no difference in the size of the blastomeres on the injected versus the uninjected side of the embryo (data not shown and Figure 7), confirming this conclusion.  (Figures 1 and 2). In order to examine the effect of inhibition of xCdc4 on neural crest development, xCdc4ΔFbox was overexpressed in developing embryos and compared to xCdc4 and GFP injections as controls. After injection of mRNAs into one cell of two-cell embryos, and subsequent development to neural plate stage 15, whole mount ISH for the neural crest markers Snail2, Snail and c-Myc were performed. Overexpression of xCdc4ΔFbox inhibited formation of the neural crest as measured by decreased Snail2 and Snail expression (Figure 8A, D), while wild-type xCdc4 and GFP had very little effect ( Figure 8B, C, E, F). In addition, c-Myc, a very early marker of neural crest [28], was also reduced by xCdc4ΔFbox ( Figure 8G), but not by xCdc4 or GFP ( Figure 8H, I).
To quantify the reduction in expression of neural crest markers brought about by overexpression of xCdc4ΔFbox, the area of staining of Snail2, Snail and c-Myc on the injected side of each embryo was measured and expressed as a ratio of the area on the uninjected side. A ratio less than 1 indicated a reduction in staining on the injected side compared to the uninjected side. The mean ratio was then calculated for at least two independent experiments, and compared with the ratio for GFP-injected embryos (which had minimal effect).
Overexpression of xCdc4ΔFbox blocked neural crest development, as determined by expression of Snail2, Snail and c-Myc, resulting in an average 57% reduction in Snail2 staining on the injected side of the embryo, compared with embryos injected with GFP alone. Overexpression of xCdc4 had essentially no effect on Snail2 staining (n = 58 to 122). Similarly, overexpression of xCdc4ΔFbox reduced Snail and c-Myc staining on the injected side of the embryo by 80% and 67% respectively, when compared to embryos injected with GFP alone (n = 30 to 52 for both).
To confirm that xCdc4ΔFbox was acting as a dominant negative, we determined whether co-injection with wild-type xCdc4 could rescue the reduction in neural crest. In this experiment, xCdc4ΔFbox significantly reduced Snail2 and Snail expression in 34% and 67% of embryos, respectively, compared to only 12% and 6%, respectively, when the wild-type xCdc4 was co-injected with xCdc4ΔFbox (n = 18 to 21; Figure 9). This rescue demonstrates that xCdc4ΔFbox can act as a dominant negative in the presence of wild-type xCdc4. Indeed, as endogenous xCdc4 is likely to be expressed at a much lower concentration than that expressed from microinjected mRNA, xCdc4ΔFbox is likely to be much more effective at inhibiting the endogenous xCdc4 protein.
As our sections revealed rather broad expression of xCdc4 in ectodermal and mesodermal derivatives, we investigated whether xCdc4ΔFbox affected patterning and/or specification of other tissues. We saw that anterior-posterior patterning and mesoderm development of the embryos were unaffected by overexpression of either wild-type xCdc4 or xCdc4ΔFbox, as determined by Otx2, Engrailed2, Krox20, MyoD, Heavy chain myosin (HCM) and epidermal keratin expression (Additional files 2 and 3). Moreover, Sox2, a marker of neural plate neuronal precursors, and Neural beta-tubulin (NbT), a marker of neuronal differentiation, were also unaffected ( Figure 8S; Additional file 3). This suggested that the effect of xCdc4ΔFbox overexpression on neural crest development was not dependent on either early patterning events or secondary effects resulting from induction of mesoderm, and was indeed specific to the neural crest. This effect is also specific to xCdc4ΔFbox; the Skp2ΔFbox mutant had no effect on Snail2 expression (Additional file 4).
We investigated whether xCdc4 activity was required for expression of other regulators of neural crest formation lying upstream of Snail2 and Snail (Figure 8). At stage 15, Pax3 (53%) and Msx1 (100%) showed significant reduction in the presence of xCdc4ΔFbox but not wild-type xCdc4, indicating that xCdc4 plays a broad role in controlling expression of regulators of neural crest. Interestingly, xCdc4ΔFbox inhibited neural crest expression of Opl in some embryos (62%), but did not have a significant effect on its placodal expression (Figure 8M). This is supported by the failure of xCdc4ΔFbox to affect the expression of the placodal marker Six1 (Additional file 5). Thus, xCdc4 does not play an essential role in specification of all the tissues in which it is expressed, but does have an essential role in formation of the neural crest.
xCdc4 is expressed broadly and diffusely before stage 15, and while expression is seen in other tissues, transcripts concentrate in the placodes and neural crest derivatives as the neural tube closes (Figures 1, 2, and  4). The first signs of neural crest specification occur earlier, so we investigated whether xCdc4 was required for establishment or maintenance of neural crest identity. To this end, we investigated the effect of xCdc4ΔFbox expression on earlier marker expression in the prospective neural crest ( Figure 10). After overexpression of xCdc4ΔFbox, we saw a reduction in Snail2 (44%; n = 32) and Snail (58%; n = 26) expression by stage 13/14 that was not seen with the wild-type protein (both 100% no significant reduction; n = 25 and n = 35, respectively), indicating that xCdc4 has a role early in neural crest specification, and is not solely required for its maintenance.
To determine the effect of inhibiting xCdc4 on later neural crest derivatives, we allowed embryos injected in both dorsal cells of four-cell-stage embryos to develop until stage 43, and then investigated the differentiation of melanocytes and formation of cartilage, both neural crest-derived tissues. We saw that embryos injected with GFP had a normal number and distribution of melanocytes. In contrast, after injection of xCdc4ΔFbox, embryos showed significant reduction and an abnormal distribution of melanocytes ( Figure 11; n = 31 to 57, P << 0.0001)) compared with the GFP-injected controls. A much smaller decrease in the number of melanocytes was also observed following injection with xCdc4 when compared with the GFP control (P < 0.01). As xCdc4ΔFbox-injected embryos looked abnormal, with a significant reduction of head morphology, we examined the effect of xCdc4ΔFbox on cartilage formation by staining the embryos with alcian blue. We found that most xCdc4ΔFbox-injected embryos showed compacted head cartilage with weaker cartilage staining compared to xCdc4-or GFP-injected embryos (data not shown).
We also noted that embryos that were expressing xCdc4ΔFbox show other developmental abnormalities, including oedema and a high frequency of reduced eyes ( Figure 11A). These phenotypes, which we have not been characterized further, may be secondary to loss of neural crest or may result from a later requirement of  xCdc4 activity in other tissues where it is expressed, for example, the eye field ( Figure 1I). As injected mRNA is not thought to persist much beyond tailbud stages, these defects most likely result from a requirement for xCdc4 at these earlier stages.
Since xCdc4ΔFbox did not result in an increase in cycling cells that could disrupt neural crest differentiation, we investigated the possibility that xCdc4ΔFbox results in apoptosis of cells destined to become neural crest. TUNEL (terminal deoxynucleotidyltransferasemediated dUTP-biotin nick end labeling) staining was used to detect apoptotic cells following mRNA injection. No difference in TUNEL staining was observed following overexpression of xCdc4ΔFbox ( Figure 12A-D). Approximately 70% of embryos had TUNEL positive cells (n = 100 to 110), but in no case were differences in TUNEL staining observed, either in the neural folds or in the embryo as a whole.
We were surprised at the low frequency of TUNELpositive cells detected in this assay, although embryos that had been wounded were used as a positive control and were highly TUNEL-positive (data not shown). Therefore, an alternative assay for apoptosis was also used. We took advantage of the fact that hBclXL can block apoptosis in X. laevis [19,40]. We checked whether Snail2 expression in xCdc4ΔFbox injected embryos could be rescued by co-expression of hBclXL. xCdc4ΔFbox, xCdc4 or GFP (1 ng) were injected unilaterally into two-cell-stage embryos. In addition, 1 ng of hBclXL or GFP were co-injected as appropriate. Whole mount ISH for Snail2 was performed on stage 18 embryos, and Snail2 staining in xCdc4ΔFbox/hBclXL- (D)The average number of melanocytes for each condition was calculated and the mean ± standard error of the mean from two independent experiments are shown (n = 31 to 57). There was a significant decrease in the number of melanocytes following injection with xCdc4ΔFbox when compared to GFP or xCdc4 (***P << 0.0001), and a smaller decrease following injection with xCdc4 when compared to injection with GFP (*P < 0.01). injected embryos was compared to xCdc4ΔFbox/GFPinjected embryos (n = 57 to 82; Figure 12E-J). In these experiments, xCdc4ΔFbox/GFP-injected embryos displayed an average 32% reduction in Snail2 staining, compared to GFP-injected embryos (mean ± SEM ratios were 0.66 ± 0.05 and 0.97 ± 0.02, respectively). However, no rescue of Snail2 staining was seen when hBclXL was co-injected with xCdc4ΔFbox, with an average 41% reduction in Snail2 staining on the injected side compared to embryos injected with GFP/hBclXL. Thus, loss of neural crest after inhibition of xCdc4 activity is not due to tissue loss by apoptosis.
In summary, we have identified two homologues of Cdc4 in X. laevis that show dynamic expression in the early embryo, in particular in the developing neural crest and its derivatives. While unexpectedly having no effect on cell cycling, blocking xCdc4 function by overexpression of a dominant negative form of the protein resulted in inhibition of specification and formation of the neural crest. This had long-term consequences for the formation of neural crest derivatives such as melanocytes and cartilage. Thus, regulated and specific proteolysis by xCdc4 plays an essential function in early development distinct from its well-established role in regulating cell division.

Discussion
This work has identified a novel role for the F-box protein xCdc4 in neural crest development in X. laevis. Cdc4 orthologs in vertebrates have attracted considerable interest due to the plethora of substrates they degrade, and the fact that Cdc4 is a haplo-insufficient tumor suppressor protein [10,16]. However, the embryonic lethality of Cdc4 knock-out mice precluded a detailed analysis of the role of this protein in development [17].
X. laevis Cdc4 is likely to be orthologous to vertebrate Cdc4 for several reasons. Firstly, xCdc4 is highly conserved compared to human Cdc4, both in terms of sequence and apparent gene organization. Secondly, xCdc4α and xCdc4β are capable of degrading Cyclin E, a known substrate of mammalian Cdc4. This validates the use of X. laevis to study the developmental function of Cdc4. In contrast, recovered Cdc4 -/mouse embryos show profound vascular defects [17], precluding study in other developmental processes. Developing amphibian embryos do not require a functional vasculature until later embryonic stages, making this an excellent system for studying other functions of Cdc4.
In this work, we report the isolation of two xCdc4 isoforms, xCdc4α and xCdc4β. The genes encoding them are almost identical from the region corresponding to the second exon (1,622 out of 1,625 nucleotides identical). This strongly suggested that the structure of the Xenopus gene was conserved compared to mammals. In humans and mice, the Cdc4 locus encodes three isoforms of Cdc4; designated α, β and γ. These are produced by alternative splicing of three unique 5' exons to ten common 3' exons, yielding proteins that differ only at their amino termini. Although our data suggest that this gene structure is preserved, no xCdc4γ isoform was detected during this work. In addition, in silico analysis, using the Ensembl genome browser, failed to detect sequences corresponding to the γ specific exon.
To check if xCdc4 was functionally orthologous to vertebrate Cdc4, the ability of xCdc4 to degrade Cyclin E was examined. xCdc4 overexpression reduced the abundance of Cyclin E, compared to embryos coinjected with GFP, preferentially reducing the abundance of the slower migrating hyper-phosphorylated form of Cyclin E ( Figure 6A). This is in agreement with previous findings that Cdc4 only interacts with phosphorylated Cyclin E [35][36][37][38]. xCdc4 also degrades Cyclin E in vivo, while xCdc4ΔFbox does not ( Figure 6).
The spatio-temporal expression of xCdc4 was examined by immunoblotting for xCdc4 isoforms, and by whole mount ISH (Figure 1). Using antibodies that detected xCdc4, both isoforms were expressed at all stages tested. xCdc4α migrated at an apparent molecular weight of 100 kDa, despite having a predicted molecular weight of 79 kDa ( Figure 1A). This is in keeping with a previous report that showed that human Cdc4α migrated with an apparent molecular weight of 110 kDa, although its predicted molecular weight was 80 kDa [9]. In contrast, xCdc4β consistently migrated slightly faster than its predicted molecular weight of 62 kDa ( Figure  1B).
Although expressed across the embryo, the neural crest and placodes were the first sites where xCdc4 transcripts accumulated most prominently (Figures 1, 2 and  4). Subsequently, transcripts were detected in the branchial arches and fin mesenchyme, which are both neural crest derived tissues, as well as in the somites and brain, and co-localized in many areas with other known markers of neural crest and placodes (Figures 1 and 2). In order to examine the function of xCdc4 during development, gain and loss of function experiments were performed. Translation-blocking morpholinos did not inhibit expression of endogenous xCdc4α and xCdc4β (data not shown). One possible explanation for this was the maternal expression of the proteins. Therefore, an F-box deletion mutant was used as an alternative way to inhibit the function of xCdc4. xCdc4ΔFbox acted as a dominant negative inhibitor of xCdc4's ability to degrade Cyclin E ( Figure 6) because it retains its ability to bind to substrate but cannot recruit the substrate to the rest of the SCF complex. The specificity of Fbox proteins resides in their substrate binding site, and F-box deletion mutants, which bind specifically to their substrate but not to the Skp1 component of the SCF E3 ligase complex, have been widely employed as reagents to block degradation of specific F-box targets. Confirming this, xSkp2ΔFbox did not inhibit xCdc4-mediated degradation of the protein. As xCdc4 is expressed strongly in the neural crest, we investigated the effect of overexpression of xCdc4ΔFbox on development of this tissue.
Inhibition of xCdc4, using xCdc4ΔFbox, inhibited neural crest development, as determined by Snail2 and Snail ISHs (Figures 8, 9 and 10). This inhibition of neural crest development occurred at an early stage, upstream of c-Myc, reducing expression of a number of proteins acting as regulators of Snail2, Snail and c-Myc ( Figure 8). This indicates that xCdc4 is required for establishment of neural crest identity, rather than simply maintenance of that identity. Although expressed more broadly in ectodermal and mesodermal derivatives, we did not detect a requirement for xCdc4 function in specification or differentiation of other tissues, nor in patterning of the neural tube (Additional files 2, 3 and 5), strongly indicating a tissue-autonomous role in the neural crest.
The most well characterized role of Cdc4 is in regulating the cell cycle, and we investigated whether xCdc4 regulates cell cycling in the early X. laevis embryo. xCdc4ΔFbox overexpression did not perturb the cell cycle as determined by phH3 staining in these early embryos (Figure 7). The results from phH3 staining are consistent with previous observations that cell proliferation has a minor role in neural crest development; neural and neural crest induction has been reported to proceed normally when cell division was blocked from mid-gastrula stages onwards [19,45].
xCdc4ΔFbox does not inhibit neural crest development through activation of an apoptotic program. The evidence for this is threefold: xCdc4ΔFbox overexpression did not lead to a loss of β-gal staining expressed from co-injected mRNA; there was no change in TUNEL staining ( Figure 12A-D); and finally, co-injection of hBclXL did not affect the ability of xCdc4ΔFbox to block neural crest development ( Figure 12E-J). X. laevis embryos can undergo apoptotic cell death from stage 10.5 onwards [46][47][48]. TUNEL analysis during embryonic development showed that only 50 to 60% of embryos had TUNEL-positive cells in the ectoderm prior to stage 18 [49]. After this stage a higher percentage of embryos were TUNEL-positive, but individual embryos displayed less TUNEL staining. In terms of neural crest development, Sox10 and Id3 depletion have been reported to increase TUNEL staining in the neural crest, but this was coupled to a reduction in cell proliferation [50,51]. However, it has been reported that inhibition of the cell cycle reduces clearance of apoptotic cells, meaning that cell cycle inhibition in these embryos may have led to increased TUNEL staining [52].
Previous studies have reported increased levels of apoptosis occurring in neural folds rather than other areas of the embryonic ectoderm [47,49,53], and a number of genes involved in neural crest development possess anti-apoptotic activity (for example, [53,54]). It is possible that apoptosis regulates neural crest development in a stage-specific manner. For example, it may be important in defining the neural crest boundary, rather than for neural crest induction. However, as xCdc4 acts at an early stage of neural crest development, regulation of apoptosis may not be involved.
The identification of Cdc4 as a regulator of neural crest development adds to existing evidence that the ubiquitin proteasome system plays a role in the development of this tissue. Overexpression of dominant negative Cullin1, which is predicted to inhibit all SCF E3 ligases, expanded the neural crest in X. laevis. This was primarily due to stabilization of the Wnt pathway component β-catenin, although other substrates are likely to have been involved [55]. Recently, the F-box protein Ppa was shown to regulate neural crest development by degradation of Snail2 in X. laevis. Overexpression of Ppa inhibited neural crest development [26]. The related protein Snail is degraded by the F-box protein β-TRCP in tissue culture cells [56]. Thus, a number of proteins involved in neural crest development display dynamic expression patterns, and regulation by ubiquitinmediated proteolysis is emerging as a method to achieve this through activity of different E3 ligases that target distinct substrates.
What substrates is Cdc4 targeting to influence neural crest formation? c-Myc regulates neural crest development [19,57] and can be degraded by Cdc4 [11,29], but we saw that xCdc4ΔFbox decreased c-Myc mRNA (Figure 8), indicating that xCdc4 also acts upstream of c-Myc expression. Indeed, xCdc4 acts upstream of early regulators of neural crest, such as Pax3 and Msx1 (Figure 8). We hypothesized that xCdc4 was degrading a negative regulator of neural crest development. c-Jun is another known target of Cdc4 [12] and is expressed at the right time and place to be regulating neural crest formation, although its exact role is poorly understood [58,59]. We saw that overexpression of c-Jun did indeed inhibit Snail2 expression (Additional file 6). However, when we tested whether xCdc4 could target c-Jun protein for degradation in embryos by co-injection of mRNAs encoding these proteins, and performing immunoblotting against HA-tagged c-Jun, we saw no effect of xCdc4 on c-Jun levels (data not shown). From these results, it is not clear whether c-Jun is a target for Cdc4 in this developmental context, and in any case, an essential role for c-Jun in regulating neural crest formation in X. laevis has not been clearly demonstrated.
Identification of E3 ligase targets is challenging, and co-factors required for protein degradation may be active only in certain contexts. For example, it has been noted that Cdc4-mediated degradation of c-Jun in tissue culture cells required co-expression of Glycogen synthase kinase 3β (GSK3β) [60]. It will be important to now identify the in vivo targets of Cdc4 that regulate formation of the neural crest if we are to have a fuller understanding of the role of selective protein degradation in the development of this tissue.

Conclusions
Here we identify xCdc4 as a novel regulator of neural crest development in X. laevis, acting early in neural crest formation, potentially by regulation of c-Jun. These results demonstrate that Cdc4's role as a tumor suppressor protein may extend beyond its ability to regulate the cell cycle to an ability to directly regulate tissue differentiation.

Plasmids and constructs
Expression of constructs was verified by immunoblotting embryo lysates after microinjection of capped mRNA.
X. laevis Cdc4a (xCdc4a) was similarly cloned using cDNA derived from stage 7 embryos. Primers used to obtain the full length coding sequence were: forward, 5'-GCTGGCTTTTGGAAATGAATCAGG-3'; reverse, 5'-CTTCACTTCATGTCCACATCAAAGTCC-3'. xCdc4a was cloned into pGEM-T Easy (Promega, Madison, WI, USA) downstream of the SP6 promoter. The GenBank accession number is DQ666345. A single point mutation was introduced into this sequence (nucleotide G1946A in the coding sequence, resulting in G649D in the protein), to introduce an Asp residue that is conserved amongst vertebrates, to correct what was likely a cloning error; xCdc4α (G) was much less potent at inhibiting formation of neural crest compared to xCdc4α (D) (data not shown). XCdc4α (D) also degraded Cyclin E more efficiently than xCdc4α (G) (data not shown). Site directed mutagenesis was performed using a QuikChange® Multi Site Directed Mutagenesis kit (Stratagene, La Jolla, California) according to the manufacturer's instructions. xCdc4aΔFbox was produced by an identical method to that described for xCdc4bΔFbox. X. laevis Skp2ΔFbox has been described elsewhere [39].
X. laevis c-Jun in pCS2+ (GenBank accession number AJ243954) was a kind gift of Professor Walter Knochel (Institute of Biochemistry, University of Ulm, Germany).

In situ hybridization
Whole mount ISH was performed using a BioLane™ HTI in situ robot (Holle and Huttner (Tubingen, Germany).
The washes and composition of solutions were as described in [62], with some modifications in the protocol. The RNase step was omitted, and embryos were blocked with 2% Blocking Reagent (Roche) and 20% heat inactivated lamb serum in maleic acid buffer. Incubation with 1:5,000 anti-digoxigenin was performed in the same solution. The color reaction was terminated using PBS washes and embryos were re-fixed in MEMFA. Embryos were bleached as described in [62]. xCdc4 in pBSK+ was linearized with PstI and transcribed T3 for antisense, and linearized with NotI and transcribed T7 for sense. The following probes were used: c-Myc, MsxI [53], Opl/Zic1 [32], Pax3 [30], Snail, Six1 [34], Sox2, Sox10 [31] and Snail2 [25].
To quantify the area of Snail2 expression, Openlab™ software (Improvision/Perkin Elmer, Waltham, MA, USA) was used to select the area of Snail2 staining on the injected (I) and uninjected (U) side, using all injected embryos unless they were damaged. For each experiment the average ratio of I/U was determined for each injection to determine the change in Snail2 staining on the injected side, a ratio of <1 indicating Snail2 reduction. For at least two experiments, a mean ratio ± standard error of the mean (SEM) for each injection was calculated by taking the mean of the average ratios, and a Student's t-test performed. The ratios were compared to GFP-injected control embryos.
ISH on sectioned embryos was performed as described in [63]. Stage 15 to 22 embryos were fixed in MEMFA, embedded in a paraffin and beeswax solution and sectioned using a Leica microtome. The ISH was performed on 14-μm sections using xCdc4 and Snail2 probes. The slides were mounted using Aquamount (BDH-Merck, VWR International, West Chester, PA, USA).

Analysis of melanocyte distribution
Embryos were injected with 0.5 ng of xCdc4ΔFbox, xCdc4 or control GFP mRNA into each of the two animal dorsal cells of four-cell-stage embryos. At stage 43, the embryos were fixed in MEMFA and dehydrated overnight in ethanol. For the examination of melanocytes, individual embryos were photographed, the anterior region selected and the number of melanocytes counted using Image J software. For each condition, the number of melanocytes was averaged, the mean ± SEM was calculated and a Student's t-test performed.

Western blotting
Embryos were lysed in 100 mM NaCl, 5 mM EDTA, 0.1% Triton X-100 and 50 mM β-glycerophosphate. Cleared supernatants were mixed with an equal volume of 2× SDS gel loading buffer (100 mM Tris pH 6.8, 4% SDS, 20% glycerol and 0.2% bromophenol blue), and dithiothreitol was added to a final concentration of 100 mM. One embryo equivalent per lane was loaded. In order to quantify protein levels, blots were imaged by using infrared fluorescence of appropriately tagged secondary antibodies and quantified by using a LiCOR Biosciences (Lincoln, Nebraska, USA) Odyssey scanner and software.
Additional file 1: Alignment of X. laevis Cdc4 isoforms to hCdc4a and hCdc4b. ClustalW alignment of human (HU) and X. laevis (Xl) Cdc4α and Cdc4β. Xl Cdc4α shows 88% identity to HU Cdc4α. Similarly, Xl Cdc4β shows 86% identity to HU Cdc4β. Black boxes denote nuclear localization signals for Xl and HU Cdc4α; red box indicates F-Box motif; green boxes outline WD40 repeat region. Domain assignments were made using the Pfam program at the Sanger Centre [64]. The putative dimerization domain of Cdc4 (based on [646566] is shown as a black dashed line. Grey regions indicate identity while blue signifies similar amino acids. A single point mutation was introduced into xCdc4α (nucleotide G1946A in the coding sequence, resulting in G649D in the protein) to introduce an Asp residue (denoted by red text at position 649) that is conserved amongst vertebrates, to correct what was most likely a cloning error. Click here for file [ http://www.biomedcentral.com/content/supplementary/1749-8104-5-1-S1.JPEG ] Additional file 2: xCdc4ΔFbox does not affect anterior-posterior axis patterning. We injected 1 ng of xCdc4ΔFbox (A,B,E,F,I,J) or xCdc4 mRNA (C,G,K) into one cell of two-cell-stage embryos; 1 ng of GFP mRNA (D,H, L) was injected as a control and β-gal mRNA was co-injected as a lineage tracer. Stage 16 to 18 embryos were stained for the forebrain/ anterior midbrain marker Otx2 (A-D), the midbrain/hindbrain junction marker En2 (E-H), or the hindbrain marker Krox20 (I-L). Representative embryos from three independent pooled experiments are shown (n = 38 to 79; anterior view, dorsal up, injected side right). Numbers represent the percentages of embryos displaying each phenotype and white arrows highlight differences in expression of the markers on the injected side. Click here for file [ http://www.biomedcentral.com/content/supplementary/1749-8104-5-1-S2.JPEG ] Additional file 3: xCdc4ΔFbox does not affect development of the myotome, the epidermis or primary neurons. xCdc4ΔFbox, xCdc4 or control GFP mRNA (1 ng) was injected into one cell of two-cell-stage embryos. β-gal was co-injected as a lineage tracer. ISH was performed on stage 13 to 15 embryos for MyoD (A-C) and epidermal keratin (EK) (G-I), on stage 15 embryos for neural b-tubulin (NbT) (J-L) and at stage 18 to 20 for heavy chain myosin (HCM) (D-F). Representative embryos for the indicated injection are shown (dorsal view, anterior up, injected side on the right). Numbers are the percentage of embryos with normal phenotypes (n = 59 to 109). Click here for file [ http://www.biomedcentral.com/content/supplementary/1749-8104-5-1-S3.JPEG ] Additional file 4: The F-box protein xSkp2 does not affect neural crest development. xSkp2 or xSkp2ΔFbox mRNA (2 ng) was injected into one cell of two-cell-stage embryos. GFP mRNA (2 ng) was injected as a control. β-gal mRNA was co-injected as a lineage tracer. ISH for Snail2 was performed on stage 18 embryos. As an additional control, ISH was performed in parallel for NbT on stage 15 embryos. Additional file 6: c-Jun is a negative regulator of neural crest development. c-Jun mRNA (0.5 ng or 1 ng) was injected into one cell of two-cell-stage embryos. GFP mRNA was injected as a control, and β-gal mRNA was injected as a lineage tracer (light blue unilateral staining). Whole mount ISH was performed for Snail2 (A-C) or c-Myc (D-F) expression. The area of Snail2 staining on the injected side was expressed as a ratio of the area on the uninjected side. The mean ± SEM ratio (n = 37 to 87) is shown for each injection. Representative embryos are shown (anterior view, dorsal up, injected side right). The average percentage reduction in Snail2 staining on the injected side, compared to embryos injected with GFP, is shown for Snail2 ISHs.