The Drosophila HMG-domain proteins SoxNeuro and Dichaete direct trichome formation via the activation of shavenbaby and the restriction of Wingless pathway activity

The Drosophila ventral epidermis is the classic model system in which to study the mechanisms that regulate the specification of two distinct cell types: cells that secrete cuticle with trichomes (denticles) and cells that secrete naked cuticle. Many studies have focused on the patterning mechanisms that govern the establishment of alternating fields of denticle-producing versus smooth cells(Fig. 1, cartoon) (for reviews,see Hatini and DiNardo, 2001; Sanson, 2001). These studies have identified the transcription factor shavenbaby (svb) as the most downstream target of the signaling cascades that pattern the ventral epidermis (Payre et al.,1999). Recent studies have shown that svb is necessary and sufficient to direct denticle formation in a cell-autonomous manner by regulating the expression of genes whose products are involved in epidermal cell-shape remodeling (Chanut-Delalande et al., 2006). In each segment of the ventral epidermis, svbis expressed in a discreet domain of six rows of cells. The spatial limits of this domain reflect the inputs of Der- and Wg-pathway activities on the transcriptional regulation of svb expression: Der pathway activity stimulates and Wg pathway activity represses svb expression(Szuts et al., 1997).

The expression of svb in a discreet and invariable domain poses a question that appears in many developmental contexts: how do individual cells integrate opposing extrinsic information such that the response of the cell is non-stochastic and invariable? Circumstantial evidence suggests that svb is not a direct target gene of the Wg signaling cascade. Our study shows that the high mobility group (HMG)-domain protein SoxNeuro (SoxN)has a dual role in the establishment and maintenance of a discreet domain of svb expression. First, SoxN expression is activated by Der- and repressed by Wg-pathway activities. SoxN is necessary for the expression of svb and is sufficient to cell-autonomously activate svbexpression even in the presence of high levels of Wg signaling. Hence, SoxN represents a molecular link between these signaling cascades and the expression of svb. Moreover, we show that svb is required for the maintenance of but not for the establishment of the late epidermal SoxN expression. This indicates a reciprocal regulatory relationship between svb and SoxN. Second, many vertebrate Sox proteins have been shown to antagonize Wg pathway activity. We present evidence that this function is conserved in Drosophila SoxN. Hence, the spatial limits of both SoxN expression and Wg pathway activity are determined by a negative-feedback loop. These results suggest that the expression of svb in a discreet domain is resolved at the level of SoxN. Furthermore, we show that a closely related HMG-domain protein, Dichaete, is co-regulated with SoxN and has a redundant,albeit weaker, function in the activation of svb expression and in the restriction of Wg pathway activity.

Oregon R flies were used as the wild-type stock. The EMS-induced alleles SoxN^U6-35 and SoxN^GA1192 have been described previously (Buescher et al.,2002; Overton et al.,2002). SoxN^U6-35 contains a C to T transition that results in a nonsense codon at amino acid position 133; SoxN^GA1192 contains a C to T transition that results in nonsense codon at amino acid position 172. The putative translation products of both mutant alleles do not contain the HMG domain. The following Gal4 driver lines and UAS-constructs were used: scaGal4 (which drives expression in the ventral ectoderm from early stage 9 onward) and wgGal4 were obtained from the Bloomington Stock Center. enGal4 (Brand and Perrimon, 1993), armVP16Gal4(Sanson et al., 1996) was kindly provided by J. P. Vincent (MRC National Institute for Medical Research,London, UK); UAS-SoxN(Overton et al., 2002). pangolin² (synonymous with dTCF²)(van de Wetering et al., 1997)and wg^cx4 (Baker,1988) were obtained from the Bloomington Stock Center. D^r72 was kindly provided by S. Russell(Soriano and Russell, 1998). UAS-Egfr^Act, UAS-der^DN and UAS-wgHA were kindly provided by J. Bateman and J. Ng (both from King's College London, London, UK), respectively. The amorphic allele svb¹ was kindly provided by F. Payre(Payre et al., 1999). UAS-SoxN-YFP was generated by Gateway cloning of the complete SoxN coding sequence into the pTVW vector.

Cuticles of first-instar larvae were prepared as described(Buescher et al., 2004).

Embryos were collected, fixed and immunostained as previously described(Yang et al., 1997). Primary antibodies were polyclonal mouse anti-SoxN (1:1000)(Buescher et al., 2002);polyclonal rabbit anti-Dichaete (1:1000)(Nambu and Nambu, 1996);monoclonal mouse anti-Engrailed/Invected (also known as Engrailed, En) (4D9;developed by C. Goodman). The Engrailed/Invected hybridomas were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the National Institute of Child Health and Human Development (NICHD) and maintained by the University of Iowa, Iowa City, IA. Histochemical detection was performed with Jackson ImmunoResearch HRP-conjugated secondary antibodies and visualized by the glucose-oxidase-DAB-nickel method, as described(Shu et al., 1988). Fluorescent secondary antibodies were obtained from Jackson ImmunoResearch(Cy3- and FITC-conjugated anti-mouse and anti-rabbit antibodies).

RNA in situ hybridization was carried out as described previously(Tautz and Pfeifle, 1989). For the generation of the svb RNA probe, a svb-specific DNA fragment of approximately 1.5 kb length was amplified from genomic DNA using PCR and subsequently used as a template for in vitro transcription with T7 RNA polymerase.

The wild-type ventral larval cuticle exhibits a segmentally re-iterated pattern of six rows of denticles separated by stretches of naked cuticle(Fig. 1A,A′). We have previously reported that SoxN mutant animals show an aberrant cuticle phenotype (Buescher et al.,2002). The present study was performed using two independent,putatively amorphic alleles of SoxN, which have essentially identical mutant phenotypes: SoxN^GA1192(Buescher et al., 2002) (also see Materials and methods) and SoxN^U6-35(Overton et al., 2002). SoxN homozygous mutant embryos die late in embryogenesis, after the secretion of cuticle. The SoxN mutant phenotype is characterized by a moderate loss of denticles and a corresponding increase of naked cuticle(Fig. 1B,B′). To understand the role of SoxN in the formation of denticles, we first examined the expression pattern of SoxN. During embryogenesis, SoxN protein expression undergoes dynamic changes (Buescher et al.,2002; Cremazy et al.,2000). To determine the precise boundaries of SoxN expression within segments, we double-stained wild-type embryos with anti-SoxN and anti-En, the latter providing a stable landmark throughout embryogenesis. From stage 5 to late stage 8, SoxN is expressed throughout the ventral ectoderm(Fig. 1C)(Buescher et al., 2002; Cremazy et al., 2000; Overton et al., 2002). From stage 9 onwards, SoxN protein decays; this is first observed in a narrow stripe just anterior to the En domain, so that, at late stage 9, SoxN protein remains present in the entire epithelium with the exception of the Wg expression domain (Fig. 1D). After stage 9, SoxN protein expression further decays throughout the segment. From early stage 12 onwards, SoxN expression is re-initiated in segmental stripes of 6-cell width at the lateral periphery of the ventral epidermis. Subsequently, these stripes expand along the dorsoventral (D/V) axis and, from stage 13 onwards, span the ventral epidermis. SoxN remains expressed in 6-cell-wide stripes until the end of embryogenesis(Fig. 1E). However, we noticed that, after stage 13, the level of SoxN expression is further upregulated in the two posterior-most rows of cells (Fig. 1F, arrows). En and SoxN stripes partially overlap so that the anterior-most row of SoxN expression co-localizes with the posterior row of En expression (Fig. 1F). These results indicate that, during the embryonic stages in which epidermal cell fate is specified, SoxN expression is restricted to those cells that will differentiate to produce denticles. Most notably, the SoxN and svb expression domains completely coincide(Fig. 1, cartoon).

The expression domain of svb is established by the Der- and Wg-pathway activities (Szuts et al.,1997). This prompted us to examine whether the late expression of SoxN is regulated by the same signaling cascades. In wg null mutant embryos (wg^cx4) SoxN expression did not decay after stage 8 and remained ubiquitous until the end of embryogenesis(Fig. 1G). In fact, the aberrant expression patterns of SoxN and svb in wgmutant embryos were very similar (compare Fig. 1G with 1H). However,derepression of SoxN in a wg mutant background did not result in uniform levels of SoxN expression throughout the ventral epidermis. Instead,the level of ectopic SoxN expression remained lower than that of SoxN within its endogenous domain, suggesting that the establishment of wild-type levels of SoxN expression might require additional stimulating input(s). Onset and location of SoxN expression in epidermal stripes suggests that the activation of the Der pathway via its ligand Spi might stimulate SoxN expression. Activation of the Der pathway throughout the ventral epidermis with the armadillo (arm)-VP16 driver(armVP16Gal4-UAS-Egfr^Act) indeed resulted in a posterior expansion of SoxN expression by two to three rows of cells (Fig. 1I). Conversely,reduction of Der pathway activity throughout the ventral epidermis(armVP16Gal4-UAS-Egfr^DN) resulted in a strong loss of SoxN expression (Fig. 1J). Taken together, our data indicate that the late expression of SoxN in epidermal stripes is positively regulated by Der- and negatively regulated by Wg-pathway activities. Hence, the co-expression of SoxNand svb reflects their regulation by the same signaling cascades.

Our observations raise the issue as to whether the expression of SoxN and svb are regulated in parallel by the Der- and Wg-pathway activities, or whether Spi and Wg might regulate the expression of SoxN, which could in turn activate the expression of svb. To address this, we first examined whether loss of SoxN results in a loss of svb expression. In stage-14 SoxN mutant embryos, we found that svb expression was strongly reduced compared with wild-type, indicating that SoxN is necessary for the expression of svb (Fig. 2A,B). To determine whether SoxN is sufficient to cause ectopic svbexpression/denticle formation, we used a number of Gal4 lines that drive expression in different parts of the ventral epidermis, and observed that misexpression of SoxN resulted in ectopic denticle formation. For example, misexpression of SoxN in the anterior row of the En stripe(enGal4-UAS-SoxN) resulted in ectopic denticle formation within this row of cells (Fig. 2C). Misexpression of SoxN with the armVP16Gal4 driver(armVP16Gal4-UAS-SoxN-YFP) resulted in ectopic denticle formation in most of the ventral epidermis except for in narrow stripes anterior to the row 1 denticles. In these regions, ectopic denticle formation remained sparse. The ectopic denticle formation reflected the ectopic expression of svb in broadened segmental stripes, which were separated by narrow gaps with little or no svb expression(Fig. 2E). Although the armVP16Gal4 driver has been described as directing the expression of transgenes in a ubiquitous manner, staining of armVP16Gal4-UAS-SoxN-YFP embryos with anti-GFP indicated that this was not the case. Instead, we observed YFP-immunoreactivity in broad stripes, which were separated by narrow stripes with little or no YFP (Fig. 2F). This suggests that the ectopic expression of svb/ectopic denticle formation in armVP16Gal4-UAS-SoxN-YFP embryos reflects the distribution of SoxN-YFP.

To examine whether the SoxN-mediated formation of ectopic denticles is strictly cell-autonomous, we misexpressed SoxN-YFP with the wgGal4 driver. We observed two to three rows of ectopic denticles in wgGal4-UAS-SoxN-YFP larvae(Fig. 2G). Ectopic denticle formation was found be restricted to those cells that showed YFP-immunoreactivity (Fig. 2H),indicating that SoxN directs the formation of denticles in a cell-autonomous manner. Vertebrate Sox proteins have been shown to antagonize Wnt pathway activity by sequestering nuclear β-catenin (see below)(Akiyama et al., 2004; Haremaki et al., 2003; Mansukhani et al., 2005; Zorn et al., 1999). To determine whether the SoxN-mediated formation of denticles within the wg expression domain might be due to a reduction of Wg pathway activity, we co-expressed wg together with SoxN(wgGal4-UAS-SoxN-YFP-UAS-wgHA). We observed ectopic denticle formation in cells that expressed high levels of wg, as shown by anti-hemagglutinin (anti-HA) staining(Fig. 2I). These results indicate that SoxN activates the expression of svb downstream of Wg pathway activity.

In wg null mutant embryos, svb is expressed throughout the ventral epidermis and the resulting cuticle is covered by a lawn of denticles (Fig. 3A)(Payre et al., 1999). Our results suggest that ectopic denticle formation in wg mutants might result from the derepression of SoxN, which, in turn, activates the expression of svb. Removal of SoxN function in a wg mutant background resulted in cuticles with few or no ectopic denticles, indicating that ectopic denticle formation in wg mutants is, at least in part,dependent on SoxN (Fig. 3B). We confirmed this result by analyzing the function of SoxN in the absence of pangolin (pan, dTCF), the nuclear effector of Wg signaling(van de Wetering et al.,1997). pan² mutant larvae were covered by a lawn of denticles (Fig. 3C). Reduction of SoxN function (pan²;SoxN/+) resulted in fewer ectopic denticles, whereas removal of all SoxN function(pan²;SoxN) resulted in a SoxN mutant cuticle phenotype (Fig. 3D,E). Hence,SoxN acts genetically downstream of wg to activate the expression of svb.

As shown above, SoxN expression is activated by Der pathway activity. This suggests that SoxN might act downstream of Spi to activate svb. To confirm this hypothesis we analyzed the epistatic relationships of the Der pathway, SoxN and svb. Previous studies had shown that activation of the Der pathway throughout the ventral epidermis results in the formation of ectopic denticles (Payre et al.,1999; Szuts et al.,1997). We confirmed this result using the scabrous(sca)-Gal4 driver to misexpress an activated form of the epidermal growth factor receptor (Egfr)(scaGal4-UAS-Egfr^Act). scaGal4drives the expression of transgenes throughout the ventral epidermis from stage 9 onwards. Activation of the Der pathway throughout the ventral epidermis resulted in the formation of three to four rows of ectopic denticles posterior to the wild-type row-6 denticles(Fig. 3F). Removal of SoxN function in embryos misexpressing Egfr^Act(SoxN;scaGal4-UAS-Egfr^Act) strongly reduced ectopic denticle formation, indicating that the ability of Der pathway activity to direct denticle formation is at least in part dependent on SoxN function (Fig. 3G).

Taken together, our results show that SoxN directs the formation of denticles downstream of Der- and Wg-pathway activities and upstream of svb. Hence, SoxN links the Der- and Wg-signaling cascades with svb, expression of which itself is necessary and sufficient to direct denticle formation.

The expression of SoxN in epidermal stripes of 6-cell width is established by stage 13 and is maintained until the end of embryogenesis. As shown above,the establishment of the SoxN expression domain results from the opposing activities of the Der and Wg pathways. However, the maintenance of SoxN expression until the end of embryogenesis might depend on additional factors. From late stage 13 onwards, SoxN and svb are co-expressed. Taking into account that Svb is a transcription factor, it is conceivable that Svb acts to support the late expression of SoxN. To analyze whether Svb plays a role in the maintenance of SoxN expression, we stained svb¹ mutant embryos with anti-SoxN antibody. At embryonic stage 13, SoxN expression in wild-type and svb¹ mutant embryos showed no appreciable differences, indicating that svbfunction is not required for the establishment of the SoxN expression domain(Fig. 4A,B). This result is consistent with our observation that the epidermal expression of SoxN precedes and is necessary for the expression of svb. However, in stage-16 svb¹ mutant embryos, the expression of SoxN was strongly reduced as compared with wild-type embryos, indicating that svbfunction is necessary for the maintenance of SoxN expression(Fig. 4C,D). Taken together,our results demonstrate a reciprocal regulatory relationship between SoxN and svb: SoxN is required for the establishment of the expression of svb; in turn, svb is necessary to maintain the expression of SoxN.

As described above, loss of SoxN does not cause a complete loss of svb expression and denticle formation. Likewise, in SoxN;wgand pan;SoxN double mutants, some denticles are still formed. This suggests that SoxN cannot be the only factor that activates svbexpression/denticle formation. Dichaete, an HMG-domain protein closely related to SoxN, has been shown previously to be expressed in the ventral epidermis and to have a partially redundant function in the establishment of neural fate(Buescher et al., 2002; Overton et al., 2002). We wondered whether Dichaete might also have a function in the activation of svb expression/denticle formation. Previous studies have shown that,during the embryonic stages 5 to 9, Dichaete expression partially overlaps that of SoxN (Overton et al.,2002). We sought to determine whether there is any overlap between SoxN and Dichaete expression during later stages of embryonic development. Double-staining of wild-type embryos with anti-SoxN and anti-Dichaete antibodies revealed that SoxN and Dichaete co-localized in the ventral epidermis from stage 12 until the end of embryogenesis(Fig. 5A). Furthermore, like SoxN, Dichaete expression was negatively regulated by Wg pathway activity: in wg null mutant embryos (wg^cx4), Dichaete was derepressed and found throughout the ventral epidermis, although its expression levels were not uniform (Fig. 5B). In addition, we found that Dichaete expression is positively regulated by Der pathway activity: misexpression of Egfr^Act(armVP16Gal4-UAS-Egfr^Act) resulted in an expansion of Dichaete expression; whereas a reduction of Der pathway activity(armVP16Gal4-UAS-der^DN) caused a partial loss of Dichaete expression (Fig. 5C and data not shown). As shown above, during the late stages of embryogenesis, the maintenance of SoxN expression depends on svb. To determine whether the late expression of Dichaete requires svb function, we stained svb¹ mutant embryos with the anti-Dichaete antibody. At embryonic stage 13, no significant difference was observed between wild-type and svb¹ mutant embryos(Fig. 4E,F). However, at stage 16, Dichaete expression was found to be reduced in svb¹mutant embryos as compared with wild type(Fig. 4G,H). This indicates that svb is required for the maintenance but not for the establishment of the Dichaete expression domain. Taken together, these results show that, from mid-embryogenesis onwards, SoxN and Dichaete are co-regulated and, as a result, co-expressed.

Previous studies have shown that loss of Dichaete results in an aberrant cuticle pattern (Nambu and Nambu,1996). However, these defects mainly reflect the early function of Dichaete in segmentation rather than a specific function in the regulation of svb expression. To determine whether Dichaete has a more specific function in regulating the expression of svb/denticle formation, we lowered the level of Dichaete function in a SoxN mutant background(SoxN;Dichaete/+). We observed that this change strongly enhanced the SoxN mutant phenotype, because denticle formation was nearly completely abolished (Fig. 5D-F). This indicates that Dichaete has a function in the regulation of svb expression that is revealed in the absence of SoxN. This conclusion is supported by the observation that misexpression of Dichaete in the ventral epidermis results in ectopic denticle formation (P.M.O. and S. Russell, unpublished results).

The results described above do not provide much insight into the issue of how opposing extrinsic information is integrated such that a sharp posterior border of svb expression is achieved. Instead, our results shift the problem from the regulation of svb expression to the regulation of its activators - SoxN and Dichaete. However, additional regulatory functions of SoxN and Dichaete might provide a solution to this problem. Experimental evidence in vertebrate systems has shown that members of the Sox protein family negatively regulate the activity of the Wg pathway in different developmental contexts (Akiyama et al.,2004; Haremaki et al.,2003; Mansukhani et al.,2005; Zorn et al.,1999). Sox proteins were shown to reduce Wg signaling by sequestering nuclear β-catenin and preventing its binding to TCF factors. Nuclear β-catenin acts as a transcriptional co-activator at Wg target genes by associating with dTCF (Cadigan and Nusse, 1997). A possible role for Drosophila Sox proteins in antagonizing Wg signaling has not been explored so far.

This prompted us to examine whether SoxN and Dichaete might have a function in restricting Wg pathway activity. Due to the function of SoxN and Dichaete in the regulation of svb expression, analysis of the cuticle phenotype, the classic read-out for Wg signaling, does not provide a suitable model. Instead, we analyzed the expression of the en gene, which is positively regulated by Wg signaling. During the embryonic stages 9 and 10,the maintenance of En expression in epithelial stripes requires Wg signaling and the width of the En stripe is a read-out of the strength of Wg signaling. Staining of late-stage-10 SoxN mutant embryos revealed an expansion of the En stripe from the wild-type 2-cell width to a 3- to 4-cell width at the ventral midline (compare Fig. 6A with 6C). This observation demonstrates that, in SoxN mutant embryos, Wg pathway activity is increased. Conversely, misexpression of SoxN (armVP16Gal4-UAS-SoxN) resulted in a partial decay of en expression from stage 10 onwards, indicating a reduction in Wg pathway activity (Fig. 6B). These results show that SoxN negatively regulates Wg pathway activity in the ventral epidermis. To determine whether Dichaete can antagonize Wg pathway activity, we reduced Dichaete function in a SoxN mutant background (SoxN;D/+) and examined late-stage-10 embryos with the anti-En antibody. We observed an expansion of the En stripe from a 3- to 4-cell width in SoxN single mutant embryos to a 5- to 6-cell width in SoxN;D/+ embryos at the ventral midline(Fig. 6D). This result indicates that Dichaete and SoxN have redundant functions in restricting Wg pathway activity.

In the embryonic ventral epidermis of Drosophila, two alternative cell fates are specified: smooth cells and trichome-producing cells. These binary cell fates are distinguished by the expression of svb, the most-downstream effector of epidermal morphogenesis.svb is necessary and sufficient to cell-autonomously direct trichome formation(Payre et al., 1999). The expression of svb is regulated by the opposing gradients of two signaling molecules: Spi, which activates, and Wg, which represses, svb expression. svb is expressed in segmentally reiterated,epidermal stripes, which invariantly encompass six rows of cells. This raises the question of how is opposing extrinsic information integrated to establish a distinct domain of svb expression with a sharp posterior border?

In this study, we demonstrate that the HMG-domain proteins SoxN and Dichaete represent a molecular link between the expression of svb and the upstream Der- and Wg-signaling cascades. We show that SoxN and Dichaete are expressed in the ventral epidermis at the time when epidermal cell fates are specified. The late phase of SoxN and Dichaete expression is stimulated by Der- and repressed by Wg-pathway activity. These regulatory mechanisms result in the expression of SoxN and Dichaete in those six rows of cells within each abdominal segment that differentiate to produce trichomes. SoxN and, to a lesser extent, Dichaete, are necessary and sufficient to activate the expression of svb. Furthermore, our results show that the well-described repression of svb by Wg is due to the repression of SoxN, which, in turn, results in the loss of svb activation. Likewise, the Spi-mediated activation of svb expression relies on the activation of SoxN, which, in turn, activates svb. This indicates that the competition of Der- and Wg-pathway activities for the specification of trichome-producing versus smooth cell fates is resolved at the level of SoxN and Dichaete.

These results do not provide much insight into the issue of how opposing extrinsic information is integrated such that a sharp posterior border of svb expression is achieved. Instead, they raise the question of how is a sharp posterior border of SoxN and Dichaete expression established/maintained? Our findings suggest that this is achieved by a combination of negative- and positive-feedback loops(Fig. 6, cartoon). First, we provide evidence that SoxN and Dichaete negatively regulate Wg pathway activity. This negative-feedback loop provides a likely mechanism for the establishment and maintenance of a sharp posterior border of SoxN and Dichaete expression. The issue arises of how robust this system might be in the face of fluctuating levels of Wg pathway activity. The efficiency with which SoxN and Dichaete restrict Wg pathway activity will crucially rely on the levels of SoxN and Dichaete protein. In this context, it is noteworthy that the levels of SoxN protein, but not Dichaete, are several-fold higher in the two posterior-most rows of the SoxN stripe compared with the anterior four rows(Fig. 1F). The regulatory mechanisms that underlie the different levels of SoxN expression are currently unclear. Second, we provide evidence that the maintenance of SoxN and Dichaete expression is supported by a positive-feedback loop: svb, the expression of which is activated by SoxN and Dichaete, is itself required for the maintenance of SoxN and Dichaete expression(Fig. 6, cartoon). Together,these mechanisms contribute to an invariant read-out of cell identity from opposing Der- and Wg-pathway activities.

In Drosophila, SoxN and Dichaete are necessary and sufficient to activate the expression of svb, which in turn directly regulates the expression of genes involved in trichome morphogenesis(Chanut-Delalande et al. 2006). Is a function in hair formation of the Sox proteins conserved in other species, including vertebrates? A previous study has shown that the mouse Sox9 protein is required for the differentiation of hair-producing epidermal cells and acts genetically downstream of sonic hedgehog pathway activity(Vidal et al., 2005). This study did not address whether Sox9 regulates the expression of movo1(Ovol1), the mouse ortholog of svb(Dai et al., 1998). Nevertheless, the demonstrated roles of SoxN, Dichaete and Sox9 raise the exciting question of do Sox proteins have an essential function in the activation of an epidermal differentiation program that is conserved across species as distantly related as mice and flies?

A study by Chao et al. confirms our observation that SoxN controls Wg signaling (Chao et al., 2007). This study suggests that SoxN does not antagonize Wg signaling by sequestering armadillo.

We thank Joseph Bateman, John Nambu, Julian Ng, Francois Payre, Steven Russell, Jean-Paul Vincent, the Bloomington Stock Center and the Developmental Hybridoma Bank for providing materials. We also thank the anonymous reviewers whose comments have made an invaluable contribution to this manuscript. W.C. is supported by the Wellcome Trust.