Hedgehog and Wingless stabilize but do not induce cell fate during Drosophila dorsal embryonic epidermal patterning

Organizing centers, first discovered by Hans Spemann and Hilde Mangold(Spemann and Mangold, 1924),consist of groups of cells that coordinate patterning by inducing the fate of their neighbors. According to the `organizer model', the inductive signal carries information that dictates both the position and the nature of the cell. Studies in various systems have led to the identification of a number of signaling pathways, mediated by secreted ligands that bind to specific receptors and appear to play instructive roles in determining cellular identity and patterning, in conjunction with cellular history that provides competence to interpret signals in specific ways. Some ligands are thought to execute these functions in a concentration-dependent fashion, determining different cell identities at different levels of ligand concentration. A key feature of this model is that signals induce cell identities. Throughout this discussion, by `induction' we mean the eliciting of a cell state different from that in which a cell existed prior to reception of the signal.

Many secreted signals and the components of their signaling pathways have been identified in genetic screens conducted in the Drosophila embryo(Nusslein-Volhard and Wieschaus,1980; Perrimon et al.,1989; Perrimon et al.,1996). This amenable genetic system provides an effective paradigm to investigate patterning. In particular, various types of denticles and hairs on the larval cuticle provide markers to assess patterning. Distinct models have emerged to explain ventral and dorsal epidermal patterning. Wg and Hh are thought to control the patterning of the ventral epidermis by acting on secondary secreted signals, such as the EGF ligand Spitz, that constitute a system of induction by relay (Alexandre et al., 1999). In the dorsal epidermis, Hh has been proposed to induce directly and pattern several cell types posterior to the Hh-expressing cells, one of them being distant from the source of secretion(Heemskerk and DiNardo, 1994). Because the data were interpreted to indicate that Hh induces two different cell types at different threshold concentrations, Hh was described as a morphogen. From this first proposal of Hh acting as a morphogen, the Drosophila embryo has provided an elegant system with which to investigate how inductive mechanisms control patterning.

Positioning of the ligand secreting cells is the primary determinant of cell identity in these models. A single row of cells per segment expresses Wg(Baker, 1987), followed posteriorly by two rows that express Engrailed (En)(Fjose et al., 1985; Kornberg et al., 1985) and secrete Hh (Tabata and Kornberg,1994; Taylor et al.,1993). The Wg and En cells form what has been referred to as the parasegmental organizer, by analogy to the Spemann organizer and because they form a boundary that does not coincide with the segmental boundary(Martinez-Arias and Lawrence,1985). The purpose of the segmentation cascade is to specify the location of the parasegmental organizer that in turn was proposed to induce cell fates that give rise to terminally differentiated cells.

The segmentation cascade involves the sequential action of several gene families that define transient domains along the anteroposterior axis(Nusslein-Volhard and Wieschaus,1980). The pair rule genes control the positioning of the Wg- and Hh-secreting cells: sloppy paired (slp) activates wg transcription (Cadigan et al.,1994) and even skipped and fushi tarazu activate the transcription of en (DiNardo and O'Farell, 1987; Lawrence et al., 1987; Macdonald et al., 1986). Positive feedback between the two signal secreting cells ensures their mutual stabilization at the correct location(van den Heuvel et al., 1993). The position of the parasegmental boundary is therefore defined by the pair rule genes, and maintained by Wg and Hh. According to the current model, the spatial information carried by the pair rule genes is superseded by that provided by En, Wg and Hh, which control intrasegmental patterning. En, Wg and Hh are therefore called segment polarity genes. In this model, signaling by Wg and Hh simultaneously provides spatial information and cell fate specification.

To develop a more precise understanding of this system, we analyzed the development of the groove cells. They differentiate immediately posterior to the En- and Hh-expressing cells, and their differentiation is Hh dependent(Larsen et al., 2003). We identified Odd skipped (Odd) as a marker of groove cell identity, which allowed us to monitor temporally the influence of Wg and Hh on the development of the groove cells and their neighbors. We report data most consistent with the idea that the groove identity is specified several hours before grooves differentiate. Wg patterns the En stripe of cells by maintaining En,antagonizing a transition towards a more posterior identity. The range of Wg signaling determines the width of the En-expressing domain, and therefore the position of the segmental boundary. The cells beyond the reach of Wg, and therefore beyond the boundary, adopt the groove identity. In a distinct subsequent step, Hh, which is secreted from the En cells, is required to refine the groove cell pattern by blocking another transition towards a more posterior identity. We propose that, in this tissue, contrary to the cell fate induction model, signaling acts by blocking transitions of cell identity rather than by instructing cell identity. Cell diversity is generated by a progression of cell identity from one to the next, towards a default fate. The patterning signals act by maintaining spatially selected cells in specific intermediate identities until differentiation occurs.

We used standard techniques of fluorescence detection and used Leica microscopes SP2 and SP5. 3D image computations were performed with Leica Application Suite Advanced Fluorescence software. Cell tracking was carried out by painting cells with Adobe Photoshop. The images were exported as a movie with iPhoto.

Ore^R was taken as wild type. opa⁸,run³, odd⁵, odd⁰¹⁸⁶³, hh^AC,ptc^IN, nkd² and UAS-yan^act are from the Bloomington Drosophila Stock Center. Trh-lacZ(1eve1) was obtained from M. Krasnow, Cadherin-GFP from H. Oda, En-Gal4 and UAS-hh from the Perrimon laboratory.

Anti-Odd was a generous gift from Jim Skeath, anti-Ptc from Pascal Therond and anti-Slp from John Reinitz. Anti-aPKC was from Santa Cruz Biochemical(sc-216) and mouse anti-β-galactosidase was from Promega. Monoclonal antibodies against En, Ena and Dlg, developed by C. Goodman, and Wg, Cadherin,Crb and Yan developed, respectively, by S. Cohen, T. Uemura, E. Knust and G. Rubin were obtained from the Developmental Studies Hybridoma Bank.

To analyze the determination of cell identity in the embryonic trunk, we examined the development of a highly specialized cell type, displaying specific morphological features and traceable molecular markers. During their morphogenesis, the cells located at the bottom of the segmental grooves(Fig. 1A-C) adopt a characteristic rectangular shape and their cross-sectional area at the level of the adherens junctions decreases to become about 1.5 times smaller than their neighbors (Fig. 1E,E″). Several markers allow their unambiguous identification. The apical proteins Crumbs (Crb)(Tepass et al., 1990)(Fig. 1C,D) and atypical Protein Kinase C (aPKC) (Wodarz et al.,2000) (Fig. 1E‴) are enriched at the apical circumference of the groove cells at embryonic stage 12, when the epithelium starts folding (see Fig. S1 in the supplementary material). When the groove is fully formed, at stage 13,the actin-binding molecule Enabled (Ena) is enriched at the junctions separating groove cells, whereas it is detected only at the vertices in non-groove cells (Grevengoed et al.,2001) (Fig. 1E′). In the fully deepened grooves at stage 14, the dorsal epidermis contains a single row of about 19 groove cells that terminates with a groove slit, a several cell wide cluster of 8-10 cells in the ventral part of the lateral epidermis. No groove cells are present in the ventral epidermis(Fig. 1D)(Martinez-Arias, 1993). Here,we focus on the patterning of the single dorsal row of groove cells.

An additional marker, Odd(Nusslein-Volhard and Wieschaus,1980), is specifically expressed in all groove cells(Fig. 1B-D,F). Odd is a zinc-finger transcription factor (Coulter et al., 1990) known for its pair rule function(Coulter and Wieschaus, 1988),but has also been reported to have gap and segment polarity properties(Saulier-Le Drean et al.,1998). Odd is expressed in seven stripes at blastoderm stage, and in 14 stripes when the germ band elongates(Coulter et al., 1990). Significantly, we can follow the Odd protein expression pattern in fixed embryos from early stages (Fig. 2A) until groove morphogenesis. To determine whether Odd expression correlates with groove cell identity, we analyzed groove formation in pair rule mutants, in which anteroposterior patterning defects result in overall spatial disorganization of cell types. In both odd paired(Fig. 2B,B′,B‴)and runt (Fig. 2C,C′) mutant embryos, we find that Odd-expressing cells also express Ena (Fig. 2B″) and Crb (Fig. 2C′) in the characteristic groove cell patterns, and form an indentation in the epidermis (Fig. 2B′), although a full groove is not always present. In odd hypomorphic mutant embryos, the remaining Odd protein is strictly correlated with remaining groove cells(Fig. 2D,D′). Similarly,in other conditions that modify the number or position of Odd cells, there is a strict one-to-one correspondence between Odd expression and groove (see Fig. 4C″,F″). Notably, in odd hypomorphic mutants, segments lacking Odd expression do not make grooves. We conclude that Odd expression faithfully marks groove cell identity after stage 12. As the expression of Odd precedes groove morphogenesis by several hours, we wondered whether the 14 Odd stripes present at germ band extension represent an earlier specification of the groove cell identity.

We wished to determine whether the groove cells derive from these early Odd populations. At full germband extension, Odd is detected in two groups of cells, both posterior to the En stripe. One is in the dorsal domain, and the other in the ventrolateral domain, with the tracheal placode in between(Fig. 3A). No other cell types are present between the En, Odd and tracheal cells. As the tracheal placodes invaginate, significant morphogenetic rearrangement of the epidermis must occur. Analysis of fixed embryos suggests that the dorsal Odd cells may give rise to the dorsal groove cells and the ventral ones to the segmental slit. Using Cadherin-GFP (Oda and Tsukita,1999) to track cells by live imaging, we found that the epidermal surface lost as the tracheal placode invaginates is replaced by convergence of the cells dorsal and ventral to the placode(Fig. 3B,C; see Movie 1 in the supplementary material). Although we do not have a live marker for Odd expression, two cell types could be identified by their morphological features: the tracheal cells that invaginate at stage 10 and the groove cells that adopt their rectangular shapes and align at the bottom of the grooves. Tracing backwards, the progenitors of both the tracheal cells and the groove cells could be identified, revealing that the single row of dorsal groove cells originates exclusively from cells dorsal to the placode, in the region where Odd is expressed, but not from cells anterior to the placode, where En is present. We conclude that the groove cells originate from the Odd cells that were previously dorsal to the tracheal placode. These Odd-expressing cells are likely to be pulled by the invaginating tracheal cells to occupy the space left by their invagination. Strikingly, the En cells respect the segmental boundary and are not pulled posteriorly by the invaginating trachea,indicating that the segmental boundary is effective before segmental groove morphogenesis and before the border between the En-and Odd-expressing cells straightens.

This result contradicts a previous report that the groove cells originate from the posterior-most En cells of each segment(Larsen et al., 2003). In the previous study, progenitors of En cells were marked using En-Gal4 to drive expression of cytoplasmic horseradish peroxidase (HRP), and HRP staining was observed adjacent to groove cell nuclei, suggesting that the groove cells may have derived from cells that had recently expressed En. We repeated this experiment, using En-Gal4 to drive UAS-APC2::GFP to mark the cytoskeleton of En cells. We found that at stage 14, many En cells extend processes labeled with APC2::GFP that can reasonably be mistaken to be within the Odd cells (see Fig. S2 in the supplementary material). Second, a minority of cells posterior to the En stripe express GFP (see Fig. S3 in the supplementary material), but that expression is not restricted to the groove cells (see Fig. S4 in the supplementary material), and time lapse imaging shows that it does not decrease with time, implying a non-physiological expression due to the Gal4 system rather than residual expression of GFP from a cell that has lost En expression. We conclude that the groove cells originate from the Odd-expressing cells and not the En cells.

Interestingly, a cell division occurs during the morphogenetic movement accompanying tracheal invagination, and we observed that some sisters of Odd-expressing groove cells did not become groove cells(Fig. 3B,C; see Movie 1 in the supplementary material), indicating that the groove cells are selected from among a larger Odd-expressing precursor population. Examination of Odd staining in fixed embryos confirms that the number of Odd-expressing cells decreases: two rows of Odd cells are present during germ band retraction, and only one remains at stage 14. By following lacZ expression in Odd-lacZ embryos, we verified that only the cells that end up adjacent to the En stripe retain Odd expression, while their posterior neighbors lose Odd expression during stage 13, keeping only the perduringβ-galactosidase (Fig. 3F,F′,F″). Pair rule mutants provide additional evidence that maintenance of Odd expression depends on the En cells. In these mutants, the territories of En and Odd are no longer in regular stripes(Fig. 3D,D′,E); yet, as in wild type, Odd is expressed in regions adjacent to En cells, still in broad territories several cells wide at stage 12, but soon afterwards, exclusively in single cell wide stripes that are immediately adjacent to the En cells.

We conclude that Odd marks the progenitors of the groove cells before reorganization of the epidermis but that the Odd pattern is refined,maintaining expression only adjacent to the En cells, narrowing to a single row of differentiated groove cells. These data indicate that Odd is not a marker of cell competence, as such a marker would remain expressed in cells just posterior to the single row of groove cells at stage 14. Instead, Odd marks a population of cells that, if maintained in their present state, are destined to become groove cells. However, as we will argue below, this cell identity is unstable and requires maintenance. In the absence of a maintenance signal, which occurs in a single cell wide stripe, we suggest that cells progress towards a different identity.

The presence of Odd in the ancestors of the groove cells suggests that the groove identity may be specified long before the differentiation of the groove. We therefore examined the earlier specification of Odd-expressing cells and its effect on groove differentiation. Regulation of the seven stripes of Odd expression before germ band elongation has been studied, but nothing is known about the mechanisms controlling the 14 stripes present during germ band elongation. Two Odd homologues, Bowl and Drumstick, are also expressed in 14 stripes located posterior to the En stripes, and have been shown to be repressed by Wg signaling(Hatini et al., 2005). We therefore asked whether Wg controls the formation of the 14 Odd stripes. We find that Odd expression expands at the expense of En expression when Wg function is impaired, indicating that Wg inhibits Odd expression as the 14 stripes of Odd are established (Fig. 4A,B). At the end of germ band elongation, the 14 Odd stripes appear broader in wg^IL114 mutants, and, as expected, only a few En cells remain. naked mutant embryos, in which Wg signaling is derepressed (Zeng et al.,2000), display the opposite phenotype(Fig. 4D,E): the En stripes are wider, and only a few Odd cells are present. Thus, Wg signaling controls the allocation of cells into the En and Odd identities. Wg therefore patterns the position of the segmental boundary that lies between these two cell types, and the boundary is respected during trachea invagination, as Odd cells are displaced along the boundary, but cells on either side do not pass from one side to the other (Fig. 3B,C).

The regulation of Odd by Wg influences groove development. The later expression of Odd in wg mutants is highly dynamic. As En expression disappears, the broad domains of Odd, in 14 stripes at stage 9, give way to a pattern in which all the dorsal and ventrolateral cells of the trunk express Odd, forming two longitudinal stripes spanning the length of the trunk at stage 11 (Fig. 4B). However, by stage 12, Odd has disappeared, except in close proximity to the few small groups of En cells that remain. These exceptional small En/Odd territories express groove markers in the Odd cells and form localized indentations centered on the Odd cells (Fig. 4C-C″). nkd mutants also display a few remaining Odd cells, and these too differentiate as groove cells. In these mutants, no grooves are present at the border of the En cells, except when they abut Odd cells (Fig. 4F-F″),supporting our hypothesis that both early specification of Odd cells, as well as their subsequent maintenance, is relevant to groove development. Wg therefore patterns the segment boundary by maintaining the En identity among cells within the reach of the Wg signal, while cells beyond the reach of Wg adopt the groove identity and become part of the next segment.

We next tested whether Wg controls only the choice between maintaining En and adopting the groove identity or whether it can also repress groove differentiation at later stages as proposed previously(Larsen et al., 2003; Piepenburg et al., 2000). To do this, we analyzed Pnr-Gal4; UAS Wg embryos, in which Wg is ectopically expressed in the dorsal epidermis. Pnr-Gal4 drives expression beginning at stage 10-11 (after tracheal invagination and before groove morphogenesis), until late embryonic stages. Ectopic Wg is therefore delivered after the En/Odd transition has occurred and includes the full period of groove morphogenesis. The ectopic expression is similar in strength to the endogenous Wg expression in the ventral epidermis(Fig. 4I). These embryos show no sign of Odd weakening, and the grooves are not affected(Fig. 4I′,I″). Therefore, Wg does not repress Odd expression or groove identity at later stages, indicating that Wg blocks the specification of the Odd-expressing groove precursor cells during a specific time window. As we demonstrate below,this mechanism is distinct from the mechanism that later refines the population of Odd-expressing groove precursors.

Despite its broad expression in wg mutants at stage 10, Odd is maintained after stage 12 strictly next to the residual En cells, consistent with the idea that refinement of Odd expression to a single cell stripe depends on the En cells. As Hh is secreted by En cells(Tabata and Kornberg, 1994; Taylor et al., 1993), and is required for groove formation (Larsen et al., 2003), we investigated whether it controls this refinement of Odd expression. In hh null mutants, Odd expression is normal or slightly wider than wild type until stage 10(Fig. 5A,A′), and the dorsal and ventral Odd populations are present on either side of the tracheal placode. At stage 11, Odd expression decreases compared with controls, and disappears by stage 12 (Fig. 5B,B′). When the germ band retracts, no Odd is detected in the epidermis and no grooves form. The effects of too much Hh activity were first assayed in patched (ptc) mutants; Ptc is the Hh receptor and functions as an antagonist of signaling(Chen and Struhl, 1996; Ingham et al., 1991). In ptc mutants, in which Hh signaling is activated in all non-En cells,Odd is expressed at stage 12 in stripes broader than wild-type stripes(Fig. 5C,C′). The refinement step does not occur, and from stage 13 through stage 15, broad stripes of Odd persist. A similar result was obtained when Hh was overexpressed in the En domain (Fig. 5D,D′). Thus, while Hh absence causes an excessive reduction of the Odd-expressing population, Hh excess prevents the refinement of the Odd population, instead resulting in the maintenance of what appears to be the full population of Odd cells previously established when Wg signaling patterns the width of the En stripes.

We next asked whether the additional Odd-expressing cells, present when Hh signaling is inappropriately maintained, differentiate into groove cells. In ptc mutants, broad Odd domains are maintained, in which all cells express Crb, aPKC and Ena in the characteristic groove manner and display smaller diameters (Fig. 5G,G′). All the groove cells of a given segment form a single wide groove (Fig. 5G″). Edges of the grooves form regardless of the presence or absence of ectopic En cells that are characteristic of Ptc mutants. Similar wide grooves form when Hh is overexpressed, a situation in which ectopic En cells are never present, ruling out a requirement for direct contact with En cells for groove formation.

The above results imply that, in wild type, Hh may control the refinement of Odd cells to a single stripe by signaling over a range that becomes progressively limited to a single cell by stage 14. Interestingly, cells that maintain Odd expression also express high levels of the Hh target gene and receptor Ptc (Forbes et al.,1993), indicating that they are responding to Hh signal(Fig. 5E-E″), while Ptc expression decreases to its basal level in the cells that lose Odd (Fig. 5F-F‴). The synchronous reduction of Ptc and Odd expression, which correspond,respectively, to the pathway activation and the biological output, shows that this system adapts very quickly to the withdrawal of Hh.

Our data show that Hh directs the refinement of a population of groove cells to a single row of cells. If Hh signaling is disrupted, or if it is expanded, either too few or too many cells maintain the groove identity,resulting respectively in the subsequent absence or the widening of the grooves. This mechanism is independent of, and occurs after the specification of the groove identity controlled by early Wg.

Groove cells are not induced by Hh in non-Odd expressing territories, such as the dorsal and ventral Wg-expressing cells or the lateral cells on the anterior side of the En stripes. In the ventral domain, Hh does not induce groove cells anterior or posterior to the En cells. To rule out that Hh could initiate groove identity if non-Odd cells were present immediately posterior to En cells in the lateral domain, we genetically displaced cells. Activated Yan (Rebay and Rubin, 1995)overexpression compromises the integrity of the En stripe, allowing some cells to cross the segmental boundary and intercalate between the converging Odd populations during tracheal invagination(Fig. 6A,A′). These cells lose En expression, become receptive to Hh and express Ptc at levels similar to their dorsal and ventral neighbors (Fig. 6B-B″). Still, they do not express Odd or adopt a groove identity (Fig. 6C-C″) but instead interrupt the groove. As a control, ectopic expression of activated Yan in Odd cells before or after groove differentiation does not interfere with groove formation (data not shown). As the En cells are the nearest lineage to the Odd cells, this is the most stringent possible test of the ability of Hh to induce groove in non-Odd cells. This strongly suggests that Hh signaling cannot induce groove identity in non-Odd cells.

We have investigated the specification and patterning of the segmental grooves that develop immediately posterior to the Hh-secreting En cells in the lateral and dorsal epidermis. Our findings are most consistent with a model in which Wg and Hh function sequentially, each acting to select from a larger population of cells a group that will maintain an identity, whereas those that do not receive signal progress towards a different identity. This contrasts with the previous model, in which these signals function to induce cell identities. These signals therefore appear to control binary responses, and the responses entail maintenance of pre-established identities rather than induction of new cell identities.

The identification of Odd as a marker of groove cells allowed us to trace the groove cell lineage. We observe that in maintaining the Hh/En cells, Wg signaling also inhibits the development of the Odd-expressing groove cell precursors. Wg therefore not only stabilizes Hh/En cells, but determines the position of the border between En- and Odd-expressing cells that becomes the segmental boundary. This boundary is functional even before becoming straight,forming a barrier to prevent En- and Odd-expressing cells from mixing during tracheal invagination. Because the En cells are specified before Wg action, Wg maintains but does not induce their identity(DiNardo et al., 1988; Martinez-Arias et al., 1988). Our demonstration that Wg patterns the En cells, although it does not specify them, is consistent with the notion that specification and patterning can be regulated by independent mechanisms(Martinez-Arias and Hayward,2006).

How might Wg signal the choice to maintain En identity rather than switch to the Odd-expressing identity? A previous characterization of the En enhancer(Florence et al., 1997) may help understand the molecular mechanism governing the En/Odd identity switch. The En enhancer contains Ftz Ftz-F1-binding sites sufficient to establish reporter expression (see Fig. S5 in the supplementary material). A Ftz Ftz-F1 reporter is expressed in a wide domain and is not refined to a two cell wide stripe as is wild-type En, showing that its expression is maintained independently of Wg. A specific region of the En enhancer is needed to refine expression, and was hypothesized to contain Odd-binding sites, as Odd limits En expression (Coulter and Wieschaus,1988). Wg signaling acts through the TCF transcription factor, and TCF-binding sites are located between the Ftz Ftz-F1-binding sites and the repressing region. The TCF sites do not modify expression of the Ftz Ftz-F1 reporter, but do limit refinement to the normal width stripe when the proposed Odd-binding sequences are present. In the wild-type enhancer, we propose that,through the TCF-binding sites, Wg signaling maintains expression by preventing Odd-mediated repression. Consistent with this interpretation, ectopic Wg expression after the switch from En to Odd expression but before grooves differentiate shows that Wg cannot repress either Odd expression or groove morphogenesis (Fig. 4I′,I″). This might therefore provide a molecular model for the ability of Wg to block a switch in cell identity.

Following En patterning, the remaining Odd cells posterior to the En domain are again partitioned into two domains: the anterior domain retaining Odd and the posterior domain losing Odd expression. Our data show that Hh, which is produced by the En cells, maintains Odd expression within its reach, which in wild type is progressively narrowed to a single row adjacent to the En domain. The Odd cells differentiate into grooves, whereas the cells that lose Odd produce a characteristic cuticular hair at later stages. The close temporal relationship between reduction of Ptc expression and loss of Odd expression suggests a relatively close coupling between Hh signal and Odd expression.

Because Odd expression is associated with the groove cell identity, the maintenance of Odd expression in some of the Odd-expressing cells suggests that Hh maintains the groove identity in cells within reach of the signal. However, because we do not know how to maintain the Odd-associated identity in the absence of Hh signaling, we cannot discern whether Hh is also required to provide a signal in addition to that needed to maintain the groove identity. Whether or not such a hypothetical additional signal might be required, the direct correspondence between Odd maintenance and groove differentiation demonstrates that this hypothetical signal would not provide any information beyond that associated with the binary decision to maintain the Odd/groove cell identity, and would therefore not be instructive in the usual sense of specifying between alternative developmental pathways.

The cell types posterior to the En domain have previously been studied based on the type of cuticle they secrete. The groove cells secrete naked cuticle, whereas their posterior neighbors secrete thick pigmented hairs. Using these markers of cell identity, it was proposed that Hh acts as a morphogen, inducing at high concentration the fate of naked cuticle-producing cells, and at lower concentration the fate of thick pigmented hair-secreting cells (Heemskerk and DiNardo,1994). By monitoring Ptc and Odd expression in the wild-type embryo, we observe elevated levels in the Odd (groove) domain, and basal levels as soon as Hh signaling is withdrawn and the groove identity is abandoned in favor of the thick pigmented hair identity. Our data are therefore more consistent with the interpretation that Hh exceeds the threshold for signaling in the naked cuticle (Odd) domain, but drops below threshold in the thick pigmented hair-secreting domain. Heemskerk and DiNardo also observed an expansion of the thick pigmented hair domain and a decrease in the posterior fine hair domain with elevated Hh levels(Heemskerk and DiNardo, 1994). Because in wild type we observe equal, basal, Ptc expression in this region,we propose that this boundary is not normally patterned by Hh. We hypothesize that with elevated Hh, decreased cell death in the thick pigmented hair domain and competition with the posterior fine hair domain produce this altered pattern. A single Hh threshold would be incompatible with the morphogen model.

Our data are consistent with the idea that cell identities may be transient, and reveal the existence of a pathway by which cell identity progresses. Although it has not been invoked for the embryonic epidermis, the notion of a default pathway of cell identity is not new. For example, in Drosophila, neuroblasts have been proposed to follow a default neural identity pathway unless the glial cell deficient/missing gene diverts them towards the glial identity (Hosoya et al., 1995; Jones et al.,1995; Vincent et al.,1996). In the absence of Wg signaling, dorsal epidermal cells switch from En to groove identity and, subsequently, if Hh is lost, from groove to a third, thick hair identity. It is, as yet, unknown whether these identity switches depend on exogenous signals, or whether they are cell autonomously controlled. Furthermore, how their timing is regulated remains unknown. Wg acts during a specific time window, implying that the patterning signal must persist as long as the transcription factors responsible for the switch are active. The patterning signal has no effect on the pace of the progression. Patterning of the dorsal epidermis is controlled by the persistence and reach of the signal within the time window, when the switch is functional, and not by the concentration of the signal(Fig. 7).

Our observations suggest that cell identity and differentiation are independently controlled. Groove differentiation begins before refinement of the Odd domain is complete, as the two remaining rows of Odd cells begin to express some groove differentiation markers (see Fig. S1 in the supplementary material), showing that differentiation depends on the identity of the cell at that time. As the Odd domain is further narrowed to a single cell stripe,differentiation markers are lost along with Odd expression from the row further from the Hh/En cells. Because grooves develop specifically from cells that express the Odd marker well before groove morphogenesis, and because signaling to maintain Odd precedes morphogenesis, we conclude that morphogenesis requires cell prior competence or specification, and does not result simply from the direct influence of signaling on cellular morphogenesis.

How general is this inhibition of cell identity progression by patterning signals? This mechanism may be relevant to the development of stem cells, the environment of which provides stabilizing signals that maintain stemness. For example, vertebrate homologues of Wg (Wnt), have been proposed to maintain stemness in hematopoietic stem cells (HSCs)(Reya et al., 2003; Willert et al., 2003). It is likely that Wnt signaling blocks the transition from HSC to the multipotent progenitor identity. Importantly, progenitors may need to escape the Wnt signal in order to proceed through hematopoiesis, as transient Wnt stimulation enhances hematopoietic reconstitution, but constitutive signaling induced by transgenic pathway activation prevents the appearance of progenitors, and hematopoiesis fails (Kirstetter et al.,2006; Scheller et al.,2006). It is unclear why similar experiments using retroviral signal activation produced different results(Reya et al., 2003).

Later stages of hematopoiesis appear to be similarly regulated by Wnt signaling. In vivo constitutive activation of the Wnt pathway blocks development of multiple hematopoietic lineages without affecting already differentiated cells (Kirstetter et al.,2006; Scheller et al.,2006). Transitions between blood cell identities may therefore constitute steps controlled by Wg signaling, analogous to the switch we describe between En and Odd identities. Additional work will be needed to test whether hematopoietic stem cell development is controlled by signal-dependent antagonism of switches governing a progression of cell identity changes.

We thank J. Skeath, P. Therond, J. Reinitz and the DSHB for antibodies, and the Bloomington stock center for flies. We thank T. Clandinin for use of his SP2 microscope, and A. Gallet, P. Therond, C. Alexandre, J. P. Vincent, M. Simon, D. Ma and members of the Axelrod laboratory for critical comments and discussions. This work was supported by NIH R01GM59823 (to J.D.A.). S.V. was supported by fellowships from EMBO and the Human Frontier Science Program. N.P. is an investigator of the Howard Hughes Medical Institute.