Multiple modes of engrailed regulation in the progression towards cell fate determination

Abstract

The engrailed gene product of Drosophila specifies the fate of a subset of cells in each segment. Our studies of engrailed regulation suggest that fate determination is an elaborate, multistep process. At the time in embryogenesis when the engrailed-dependent cell fate is probably determined, four modes of control act in an overlapping progression to govern engrailed expression. After activation by pair-rule genes, both an extracellular signal, wingless, and autoregulation are required for engrailed expression. Autoregulation graduates to wingless independence, but is transient, and is superseded by an engrailed-independent mode of maintenance.

Two levels of commitment have been defined for the choice of cell fate¹. Early in development, fate can be instructed by environmental cues. This fate specification is flexible and allows fate to change in response to extracellular signals. But cells can undergo a transition from this type of specification to one which is determined and therefore irreversible. Because a determined cell maintains a particular fate independent of its environment, the transition to a determined state requires that development comes under cell-intrinsic, rather than cell-extrinsic control. The identification of selector genes in Drosophila provides an opportunity to study the basis of both types of cell-fate control. Selector genes encode regulators whose function defines the fate of a cell, and whose continued expression stably maintains cell fate^2,3. Where the choice of cell fate is initially dependent on the environment, selector gene expression must be sensitive to extracellular signals. By contrast, once cell fate is determined, expression of these genes must be governed exclusively by cell-intrinsic factors.

The engrailed (en) gene is a selector gene that distinguishes two populations of cells in each developing embryonic segment, the progenitors of the anterior and posterior compartments^3,4. This distinction can be seen as a lineage restriction when individual embryonic cells are marked by mitotic recombination^5,6. These give rise to clonal patches in the adult that never span the boundary between the anterior and posterior compartments, demonstrating that, from the time of marking, the progeny of a given cell only contribute to one compartment. This restriction requires expression of en by posterior cells^3,4,7. But it seems that the onset of localized en expression in early embryos is not sufficient to irreversibly determine posterior compartment identity because en requires at least one cell-extrinsic input at this time.

This input is provided by the wingless (wg) gene, expressed in segmentally repeated stripes of cells that lie next to en-expressing cells^8,9. The wg protein, which is homologous to the product of the proto-oncogene Wnt-1 (ref. 10), is secreted and appears to be taken up by en-expressing cells⁹. In wg mutant embryos, en expression is initiated in a normal striped pattern but soon stops, indicating that continued en expression in wild-type embryos requires wg in neighbouring cells^11,12. Here we show that en expression becomes independent of this extracellular influence and comes to rely on positive autoregulation. But autoregulation is transient and so does not supply a mechanism for determination of the en-cell fate. We show that autoregulation is followed by a still-later-acting mode of en control. These transitions may be the progression from a stage in which cell fate is influenced by interactions with neighbours to a later stage in which cells are stably determined.

Transition to wg independence

If anterior and posterior compartments become determined, then the selector gene en must become independent of the extracellular wg influence. Previous experiments with a temperature-sensitive(ts) wg allele demonstrated that, during embryogenesis, there are changes in the requirement for wg in body patterning¹³. To test whether wg becomes dispensible for en expression, wg^ts embryos were shifted from permissive to restrictive temperature at half-hour intervals beginning at cellular blastoderm (2.5 h). After further development at restrictive temperature, the embryos were fixed and stained with anti-en antibodies (Fig. 1). When wg^ts embryos are shifted to nonpermissive temperature before 4 h of development, en expression decays prematurely in the thoracic and abdominal ectoderm (compare mutant expression in Fig. 1a with wild type in Fig. 2b). Loss of en parallels that seen in wg null mutants^11,12. (We do not consider the regions of the embryo where en expression does not involve wg control: the head, the first three en stripes and the central nervous system (CNS) in each segment.) Between 4 and 5 h, inactivating wg has progressively less effect on en expression. For example, when wg is inactivated at about 4.5 h, a few cells in each segment persist in expressing en (Fig. 1b). After 5 h, removing wg function has no apparent effect on en expression (Fig. 1c). The striped pattern is indistinguishable from that of wg^ts embryos raised at permissive temperature. We conclude that en expression becomes independent of the wg signal at about 5 h of development. Thus, different regulators maintain en past this time.

FIG. 1.

Expression of en becomes wg-independent after 5 h. Embryos are stained with an anti-en antibody¹⁴. Magnification 140×, anterior is left and ventral is down in all figures except when noted. Embryos are shifted from 18 °C to 29 °C (at different times, but age of embryos is corrected for effect of temperature and is given in hours after egg laying (AEL) as if embryos were raised at 25 °C. a,wg^IL114ts/wg^CX4 embryo grown at nonpermissive temperature (29 °C) from ∼3.5 to 7.5 h AEL. Expression of en is indistinguishable from that seen in a wg null embryo at this stage¹¹,¹². Only expression that is not under wg control persists: in the head, the first four ectodermal stripes (1–4), and a segmentally repeated subset of CNS cells. Expression has disappeared from the lateral ectoderm posterior to the fourth en stripe (for example, the normal position of stripe 7 is indicated by an arrow). In wild-type embryos at this stage, en striped expression is easily detectable (see Fig. 2b). b,wg^IL114ts/wg^CX4 embryo grown at nonpermissive temperature from ∼4.5 to 8 h AEL. Some ectodermal cells continue to express en, creating a partial stripe in each segment (for example, stripe 7 is indicated by an arrow). c,wg^IL114ts/wg^CX4 embryo grown at nonpermissive temperature from ∼5 to 8 h AEL. The en stripes are almost wild-type (for example, stripe 7 is indicated by an arrow). Some stripes are incomplete dorsally. This is the same phenotype seen when these embryos are grown exclusively at permissive temperature (18 °C). Immunocytochemistry was done as in refs 14, ¹⁸. The wg^IL114ts is described in ref.13 and was obtained from the Tübingen stock collection. The wg^CX4 is described in ref. 8.

FIG. 2.

Expression of en stops prematurely in en mutants. Embryos (120×) are stained with an anti-en antibody. a, Wild-type embryo, ∼4.5 h AEL. b, Wild-type embryo, ∼6.5 h AEL. c, An en^CX1/en^CX1 embryo ∼4.5 h AEL. The mutant cytoplasmic protein is expressed in a wild-type pattern. d, An en^CX1/en^CX1 embryo, ∼6.5 h AEL. The mutant protein has disappeared from the ectoderm at this stage. For example, the normal position of stripe 7 is indicated by an arrow (compare with c). Some cells of the CNS in each segment continue to express the en^CX1 allele. The diffuse signal in other areas of the embryo is yolk autofluorescence. Loss of en^CXI is not the result of defects in the cis-acting control region because this allele is expressed normally when heterozygous with wild-type en (data not shown). The en^CX1 stock was obtained from R. Holmgren.

Requirement for en

Studies of expression of mutant en alleles suggest that en undergoes autoregulation. The en^CXI allele is a rearrangement (R. Holmgren, personal communication) that truncates the en transcription unit (data not shown) and produces a cytoplasmic protein. For this reason, and because en^CXI shows a severe en phenotype (S.D., unpublished observation), we consider this a nonfunctional protein. We can monitor expression of this allele using an anti-en antibody and distinguish it from the nuclear wild-type en product because en^CXI is cytoplasmic. The initial activation of en^CXI expression occurs normally (compare Fig 2a with c). But in embryos homozygous for en^CXI, striped expression in the ectoderm stops during full germ-band extension (Fig. 2d), a time when en is easily detectable in wild-type embryos^7,14 (Fig. 2b). Thus, the normal persistence of en expression past this stage is directly or indirectly dependent on the action of en itself.

A second mutant en allele, en^IM, has a stop codon in the homeodomain that truncates the protein upstream of the DNA binding motif (Gln at position 497 mutated to the stop codon TAG; J. Little and P.O'F., unpublished data). RNA in situ hybridization shows that en^IM mutant embryos fail to maintain en transcripts (data not shown). The behaviour of these mutant alleles demonstrates that maintaining en expression requires en function.

Autoregulation of en

To test whether en activates its own expression we constructed transgenic flies that carry the en gene driven by a heat-shock promoter (Fig. 3a). Embryos that carry this construct (hs–en) produce en transiently in all cells after a brief heat shock (Fig. 3b). We examined the effect of this protein in cells that do not normally express en. Inducing global en expression by heat shock had a different effect early, during the wg-dependent period for en, than it did later, at wg-independent stages.

FIG. 3.

Globally induced en activates the endogenous en gene in novel cells. Embryos (115×) are stained with an anti-en antibody. b–e, Embryos between 6 and 7 h AEL. a, Schematic diagram of hs–en. The en sequences consist of a 2-kb EcoRI–SnaBI fragment of the en cDNA clone c-2.4 (ref. 32). Transcription of en sequences (stippled box) is driven by the hsp70 heat-shock promoter (hsp), contained in a 1.3-kb Sphl/Pstl fragment from pUCHSP 70 (provided by E. Gavis). The polyadenylation sequences (pA+) are from SV40. For transformation, the above sequences are inserted into the pCaSpeR vector³³ containing P element ends (ovals) and the white gene (w+) as a selectable marker. b, An hs–en embryo, heat shocked at ∼6 h, then immediately fixed and stained. The en expression induced by heat shock is detectable in all cells. c, An hs–en embryo, heat shocked at ∼3 h AEL, aged 3.5 h, then fixed and stained. The en stripes are abnormally broad in the ventral and lateral regions compared with wild type (compare distance between arrowheads in c and d). Inset, 326× magnification of part of one stripe showing that the en antigen is confined to the nucleus. d, Wild-type embryo, showing the normal pattern of en expression. e, An en^CX1/+;hs–en embryo, heat shocked at ∼3 h AEL, aged 3.5 h, then fixed and stained. The pattern of cytoplasmic en^CXI antigen is similar to the pattern of en in c. Inset is a 326× magnification of part of one stripe showing that the cytoplasmic antigen, characteristic of en^CX1, is produced in novel cells after heat shock. Both the wild-type nuclear and mutant cytoplasmic en antigens are expressed in these cells, although the nuclear signal is weak relative to the cytoplasmic signal. f, An hs–en embryo, heat shocked at ∼5 h then fixed after 3 h recovery at 25 °C. Striped en staining is nearly wild type. There are isolated cells in the anterior compartment that express en (small arrowheads), as well as staining in the amnioserosa (large arrowhead), neither of which is seen in wild type.

METHODS. Transgenic lines were established²⁹ using the p-wings clipped helper and Df(1)w^67c2,y as host. For heat-shock experiments, embryos were collected and aged at 25 °C on grape agar plates. At the appropriate stage, plates were put in moist chambers at 37 °C for 35 min. They were then returned to 25 °C for recovery. The novel expression pattern was detectable as soon as the en expressed from the heat-shock promoter decayed to background levels, ∼45 min after heat shock. Similar results were seen with three different hs–en lines, one on chromosome III and two on chromosome II.

Our standard early heat-shock protocol consists of heat-shocking embryos at 3–3.5 h of development. On recovery, the globally induced en expression resolves into stripes that are abnormally broad as a consequence of new en expression in the ventral and lateral regions (compare Fig. 3c and d). This new expression pattern is stable and persists to at least 12 h of development when cells begin to differentiate (data not shown).

To verify that the novel en pattern results from activation of the endogenous en gene, we used en^CXI as a marker for activity of the en promoter. Embryos with one copy of en^CXI and one copy of wild-type en are phenotypically wild type except for the cytoplasmic accumulation of truncated en^CXI protein (compare insets in Fig. 3c and e). When heat-shocked, en^CXI heterozygotes that carry a hs–en element express the cytoplasmic en protein ectopically, resulting in broadened stripes (Fig. 3e). Therefore, the pulse of en produced from the heat-shock promoter results in transactivation of the en promoter.

If stable en expression after heat shock results from activation of an autoregulatory circuit, then maintaining ectopic en expression should require a functional endogenous en gene. Indeed, in en^CXI homozygous mutant embryos the early heat-shock protocol does not give stable ectopic expression. The en^CXI mutants with or without the hs–en insert are nearly indistinguishable, as both lose ectodermal en expression (not shown). Thus, maintenance of expression from the endogenous en promoter, once activated in new cells by heat-shock-produced en, requires the activity of the endogenous en gene. We conclude that en is subject to positive autoregulation. Because en is a transcription factor¹⁵, one simple model for autoregulation is that en directly acts at its own promoter. But autoregulation might be indirect and could even involve cell-extrinsic signalling.

Another consequence of heat shock is repression of wg. Although the wg RNA pattern is almost normal immediately after heat shock (data not shown), on 1 h of recovery from heat shock, wg expression is drastically reduced (compare Fig. 4a and b). At this time, en is expressed in the broadened stripes shown in Fig. 3c. The stripes of wg RNA have completely disappeared by 2 h after heat shock (Fig. 4c). The wg decay is still consistent with its requirement for en expression; wg RNA is detectable until the end of the wg-dependent period as defined with wg^ts.

FIG. 4.

Decay of wg RNA after heat shock of hs–en embryos. Embryos (125×) are hybridized with a digoxygenin-labelled³⁰ wg probe. a, Wild-type embryo, ∼5 h AEL, showing the normal pattern of wg RNA expression³¹. b, An hs–en embryo, heat shocked at ∼3 h, then allowed to recover at 25 °C for 1 h. Although the patches of wg expression in the head are of normal intensity (arrowhead, compare with a), the striped expression in the ectoderm is dramatically reduced (for comparison, arrow points to the stripe that corresponds to the arrow in a). Signal persists mostly in the ventral ectoderm. c, An hs–en embryo, heat shocked at ∼3 h and allowed to recover for 2 h at 25 °C. There is normal wg staining in the head (arrowhead), but all striped ectodermal RNA has decayed (arrow shows the normal position of the stripe shown by arrows in a and b).

METHODS. Embryos were fixed and probed for wg RNA as in ref.30. The wg probe was made from a 1.3-kb genomic EcoRI–HindIII fragment containing the fourth and fifth exons (provided by N. Baker). In a collection of embryos from parents that were heterozygous for the hs–en insert, only 3/4 showed premature loss of wg RNA after heat shock, consistent with the loss resulting from hs–en, due to either heat-shock-produced en or stable ectopic en.

The loss of wg after heat shock requires the presence of hs–en, thus it appears that wg is repressed by either the pulse of en from the heat-shock promoter or ectopic expression of the endogenous en gene. The repression of wg by en when the two genes are expressed in the same cell may have a function in normal development. This mechanism could assure that the expression of these two genes is established in immediately adjacent but nonoverlapping stripes.

Activation of autoregulation by wg

Ectopic en expression can only be induced during the wg-dependent window (Fig. 3f, described below). We therefore considered the possibility that the initiation of autoregulation requires wg input by examining the effect of hs–en in wg-mutant embryos (hs–en; wg null). After recovery from standard early heat shocks, en expression decays in hs–en; wg null embryos just as it does in wg mutant embryos without hs–en (data not shown). Therefore, wg function is required for autoregulation at this stage.

As en expression normally requires wg input transiently in development, this experiment does not exclude the possibility that the wg signal is only required before autoregulation is initiated. Therefore, we tested whether the wg requirement is concurrent with autoregulation using wg^ts embryos carrying hs–en (hs–en; wg^ts). Heat shock has two effects on hs–en; wg^ts embryos raised at permissive temperature: it provides a pulse of en to all cells and simultaneously removes wg function. We carried out standard early heat shocks in the wg-dependent window, then let the embryos develop further at a temperature that is restrictive for the wg^ts allele. As with heat shock in a wg null background, hs–en; wg^ts embryos failed to give stable ectodermal en expression (data not shown). Thus, in early embryogenesis, the cell-extrinsic wg signal is required during or after the initiation of autoregulation.

Autoregulation independent of wg

Although en autoregulation requires wg early, normal en expression continues after the end of the wg-dependent period. To discover whether there is a later stage of autoregulation that is wg-independent, we heat-shocked wg mutant embryos aged beyond the wg-dependent period. Recall that, after about 5 h, wg mutant embryos show no en expression in the ectoderm at any position posterior to the fourth en stripe^11,12 (Fig. 5a). Surprisingly, in hs–en; wg null embryos, a late heat-shock protocol induced stable, albeit variable, en expression in the ectoderm (Fig. 5b). Thus, wg function is required for autoregulation only early in development; there is a later stage of autoregulation that acts independently of wg (Fig. 8a). Perhaps a different positive regulator substitutes for wg later, or alternatively, wg may counteract an early repressor of en that is not present late (Fig. 8b).

FIG. 5.

Late autoregulation is wg independent. Embryos (140×) are stained with an anti-en antibody. a, A wg^CX4/wg^CX4 embryo⁸, ∼8 h AEL. There is no striped en expression in the ectoderm posterior to the third en stripe. For example, the arrow points to the normal position of en stripe 7. The repeated staining in the ventral region is in the CNS¹¹. b, A wg^CX4/wg^CX4;hs–en embryo, ∼8 h AEL. Embryo was heat shocked at ∼5 h then fixed after 3 h recovery at 25 °C. There are partial en stripes in positions that have no signal in a wg mutant without heat shock. For example, the arrow points to the partially rescued stripe 7. The staining phenotype varied. The embryo shown is representative of about half of the wg^CX4/wg^CX4;hs–en embryos heat shocked at this stage. The others had relatively fewer ectodermal cells staining. As with wild-type hs–en, there is staining in cells of the amnioserosa (arrowhead).

METHODS. Embryos were derived from parents that were heterozygous for wg^CX4 on chromosome II and hs–en on chromosome III. Thus, 3 out of 16 of the embryos are homozygous for the wg mutation and carry at least one copy of hs–en. Consistent with this, 3 out of 16 of the embryos show the phenotype we interpret as wg^CX4/wg^CX4;hs–en, that is dramatically fewer than the wild-type number of en cells staining, due to lack of wg, and en staining in the amnioserosa, due to activity of the hs–en insert.

FIG. 8.

Integrating four modes of en regulation during embryogenesis. a, Approximate periods of action of the four modes of en regulation and their temporal overlap. The time line indicates hours AEL. 1, The start of the pair-rule period is defined as the time of the earliest appearance of en expression⁷,¹⁴. The end is defined as the time when en expression stops in wg mutants¹¹, presumably a condition in which only pair-rule function activates en. 2, The start of wg influence is defined as the earliest time that heat shock stably induces wg-dependent ectopic en expression. The end of the wg period is defined as the latest time at which raising the temperature of wg^ts embryos has an effect on en expression. 3, The start of autoregulatory influence is taken as the earliest time that heat shock stably induces ectopic en. The end of the period during which en autoregulation operates is inferred from two observations: first, that en-responsive sequences cannot drive lac-Z expression past this time, and second, that heat shock does not activate expression in en mutants past this time. Stippling indicates the time during which autoregulation requires wg input. 4, The fourth mode of regulation (late control) begins when heat shock first results in stable expression in en mutants. It is not known whether this regulatory phase is responsible for maintaining en indefinitely or whether still later transitions occur. b, Two models to account for the onset of wg-independent en autoregulation. Cells spanning two segment primordia are schematized as eight open circles. Labelled cells represent those that normally express en or wg. Coactivation model: cells directly exposed to the wg coactivation signal are competent to autoregulate during the wg-dependent period. Autoregulation is allowed in whichever of these cells en is activated, either by pair-rule gene activity¹⁸,¹⁹ or by heat shock. Expression of en becomes wg-independent when another coactivator (A) takes over. This new coactivation is confined to posterior compartment cells, and en expression is thereby restricted to the posterior compartment after late heat shock. Antirepression model: wg prevents a general repressor of en (R) from acting on wg cells and their neighbours during the wg-dependent period. Autoregulation is allowed in whichever of these cells en is activated, either by pair-rule gene activity¹⁸,¹⁹ or by heat shock. Later, the wg requirement is relieved because late repression (R′) is confined to the anterior compartment. Autoregulation of en can now act in the normal en-expressing cells in the absence of the wg signal. Late heat shock cannot overcome the late repression in the anterior compartment and so does not induce ectopic autoregulation. It should be noted that, although we suggest that ectopically induced en expression occurs in the wg-expressing cells, we have not yet experimentally identified which cells induce en after heat shock. Additionally, wg is still expressed at late stages, but has not been indicated in the schematic because it is dispensible for en expression.

The second stage of en autoregulation is restricted to fewer cells. The pattern generated in hs–en; wg null embryos after late heat shock is different from that seen with earlier heat shocks in wild-type hs–en embryos. Although late heat shock induces en in a variable number of cells, the cells are distributed in a narrow stripe. These stripes line up with underlying en-expressing cells of the CNS, strongly suggesting that these are posterior compartment cells. It appears that, once en expression becomes wg-independent, only the cells of the posterior compartment are capable of en autoregulation (Fig. 8b). We will return to this point below.

Restricting autoregulation

Autoregulation of en does not operate in all cells, as globally induced en resolves into patterned expression after heat shock (compare Fig. 3b and c). Two possibilities, which are not mutually exclusive, might explain this restriction. First, necessary coactivators may be localized. Second, there may exist repressors that block autoregulation in some cells.

We have identified at least one repressor that restricts autoregulation. In embryos mutant for the segment polarity gene naked (nkd), en is expressed in a broadened stripe pattern¹² (Fig. 6a, compare with Fig. 3d (wt)). Thus, nkd directly or indirectly represses en in certain cells of the embryo. We tested whether nkd activity restricts en autoregulation using nkd embryos carrying hs–en (hs–en; nkd). After the standard early heat shock, these embryos stably express en in almost all ectodermal cells, except the head and telson (Fig. 6b). This global expression involves more cells than the sum of broadened en expression in nkd (Fig. 6a) and broadened en in hs–en (Fig. 3c), at least in the lateral and dorsal regions. Therefore, nkd function restricts early en autoregulation. But autoregulation is restricted by different factors later in development. Although early heat shocks of hs–en; nkd embryos produce global en expression, later in development nkd mutants are indistinguishable with and without heat shock (data not shown).

FIG. 6.

Autoregulation is repressed by nkd. Embryos (150×) are stained with an anti-en antibody. a, A nkd^7E/nkd^7E embryo, ∼5 h AEL, showing abnormally broad en stripes compared with wild type¹². The broadening is due to abnormal en expression in cells posterior to the normal en stripe in nkd mutants (S.D., unpublished observation). b, A nkd^7E/nkd^7E;hs–en embryo heat shocked at ∼3 h and fixed after recovering for 2 h at 25 °C. Nearly all ectodermal cells stain except areas of the head and telson. In addition, some isolated areas in the ventral–lateral ectoderm do not express en. These cells vary somewhat in number and position from embryo to embryo, and thus do not appear to be cells of one specific type.

METHODS. The nkd^7E/nkd^7E;hs–en embryos came from parents that were heterozygous for nkd^7E on chromosome III and homozygous for hs–en on chromosome II. Thus, 1/4 of the embryos were homozygous for nkd and carried hs–en, and 1/4 of the embryos showed global staining in the ectoderm after recovery from heat shock. The nkd^7E stock was obtained from the Tübingen stock collection.

The pattern to which en autoregulation is restricted after heat shock also changes during development. As shown above, late autoregulation, detectable as reinduction of en in wg mutants (Fig. 5b), is not activated in cells outside the normal en domain. This is also evident from a comparison of early and late heat shocks in wild-type embryos. Although early heat shocks in wild type induce stable en expression in cells outside the normal en stripe, heat shock after 4 h does not generate sable ectopic en expression, demonstrating a block to autoregulation in cells outside the posterior compartment (Fig. 3f). Using a similar construct, Poole and Kornberg reported the effect of heat-shock-produced en on the larval body pattern¹⁶. Only early heat shocks disrupted the pattern, consistent with our observation that the en expression pattern is undisturbed by later heat shocks.

Autoregulation becomes dispensible

In addition to the two stages of autoregulation, we have identified one further mode of en regulation. Between 1 and 2 h after the transition to wg independence, en expression also becomes independent of en function. The evidence for later, en-independent regulation comes from late heat shocks of en mutants carrying hs–en. The early heat-shock protocol does not induce stable expression of the endogenous en locus in en^CXI homozygous embryos (see above), presumably because the mutant en protein cannot autoregulate. But later heat shocks (between about 4 and 7 h) result in reactivation of the en^CXI locus (Fig. 7a). The mutant en protein is expressed in a pattern of partial stripes in cells at the position expected for en expression. The position of the original en cells can be identified by a stable lac-Z protein expressed under the control of en regulatory sequences¹⁷. Though the initial expression of the endogenous en^CXI gene decays (Fig. 2d), the more stable β-galactosidase persists to the time when heat shock has reinduced en^CXI expression. The newly induced en^CXI expression is in a subset of the cells marked with lac-Z (data not shown).

FIG. 7.

A fourth mode of en control follows autoregulation. a, An hs–en;en^CXI/en^CXI embryo (110×), heat shocked at ∼5 h, aged 3 h, then fixed and stained with an anti-en antibody. The cytoplasmic en^CXI protein is expressed in partial stripes in each segment. en^CXI mutants normally show no expression by this stage. The number of cells that are rescued by heat shock varies, depending in part on the age at heat shock. This embryo is representative for a heat shock done at 5 h. Most rescued cells lie in the lateral and dorsal ectoderm and are a subset of the cells that initially express en^CXI in early embryos (see text). b, An en–lac-E;hs–en embryo (75×, ventral view, anterior left) heat shocked at ∼3 h, aged 2 h, then fixed and hybridized with a digoxygenin-labelled³⁰ lac-Z probe. Stripes of lac-Z RNA are abnormally broad compared with non-heat shocked control embryos (compare distance between arrowheads flanking the stripes in b and c). c, An en–lac-E embryo (75×, ventral view, anterior left), fixed at ∼5 h and hybridized with a digoxygenin-labelled³⁰ lac-Z probe. The en–lac-E contains 2.4 kb of en promoter/upstream sequences and the en first intron driving lac-Z expression¹⁷. Stripes of lac-Z RNA correspond to the position of en expression, although the level of lac-Z expression from this construct varies among the stripes.

METHODS. a, Embryos were derived from parents that were heterozygous for en^CXI on chromosome II and hs–en on chromosome III. Thus, 3/16 of the embryos are homozygous for the en mutation and carry at least one copy of hs–en. Consistently, 3/16 of the embryos show the phenotype we interpret as en^CXI/en^CXI;hs–en, that is cytoplasmic stain in ectodermal stripes with fewer than the normal number of cells staining in each stripe. b, c, Embryos were fixed and probed for lac-Z RNA as in ref. 30. The en–lac-E stock was obtained from J. Kassis.

The late en^CXI expression initially requires en function as it is reactivated by the pulse of en produced by heat shock. As reactivation is in the appropriate cells, it seems that the factors that restrict autoregulation to these cells act independently of en function. Heat-shock-produced en quickly decays, but en^CXI expression is maintained throughout embryogenesis in the absence of en function. This identifies a late-acting, en-independent maintenance programme for en expression.

Autoregulation not only becomes dispensible for en maintenance, but seems to stop functioning near the time that these new regulators come into play. As described above, heat shock of en mutant embryos between 4 and 7 h can induce expression that lasts well beyond 7 h. But heat shocks done still later, after 7 h, do not initiate new en expression. The inability of the hs–en protein to activate the endogenous en^CXI allele shows that en autoregulation is not sufficient to activate en after 7 h.

A second indication that en autoregulation stops near 7 h comes from looking at the period of expression of a construct in which en regulatory sequences drive lac-Z expression in the en pattern¹⁷ (Fig. 7c). The construct appears to include an en autoregulatory element. When wild-type embryos carrying both the en–lac fusion gene and hs–en are given an early heat shock, expression of lac-Z is induced in novel cells in parallel with induction of the endogenous en gene (Fig. 7b). Hence, this construct responds to en autoregulation. If en autoregulation operates continuously, expression of this en-responsive construct should always parallel expression of en. But in wild-type embryos, lac-Z expression stops at about 7 h¹⁷ even though en expression persists. Taken together, the data indicate that en autoregulation does not function beyond 7 h and that en expression is maintained past this time by new regulators.

Progression to determination

Regulation of en expression is unexpectedly complex during embryogenesis. Over a 4 h period, there are four distinguishable modes of en activation (Fig. 8a). These regulatory programmes act in an overlapping progression. The pair-rule regulatory programme initiates striped en expression at cellular blastoderm^18,19, and we infer that it remains active until about 4.5 h, as en expression is maintained until this time in wg mutants^11,12. Even before 4.5 h, wg provides an activating signal from the neighbouring cell. This is inferred from the earliest time that wg-dependent ectopic en induction can be observed, at about 3 h. In contrast to the redundancy between pair-rule and wg control, en autoregulation is dependent on the wg signal during their period of overlap.

The initial dependence of autoregulation on the extracellular wg signal indicates that the activation of en expression does not immediately constitute a determined state. Instead, the period of wg dependence seems to represent a phase in development during which the choice of cell fate is responsive to neighbouring cells. Such a stage might provide a proofreading step to correct mistakes in the position of en activation. This would not be possible if en autoregulation were sufficient for cell-intrinsic stabilization of expression as soon as en is activated.

Temperature shift experiments suggest that the end of wg involvement in en expression is about 5 h. Recently, others have arrived at a similar conclusion (A. Besjovec and A. Martinez Arias, manuscript submitted). Despite the initial dependence on the wg signal, the en autoregulatory loop continues to function beyond the wg-dependent period, as demonstrated by the induction of en expression in wg mutants after late heat shocks. There are two likely explanations for the transition to wg independence (Fig. 8b). First, if wg acts positively, wg-independent en autoregulation could arise because another positively acting factor takes over at 5 h. Alternatively, wg might be required to counteract an early acting repressor of en autoregulation. If the repressor normally decays at 5 h, the requirement for wg would be lifted. Whether wg acts positively or counteracts negative regulators of en autoregulation, the factors that influence en autoregulation do change near the time that en becomes wg-independent. This is indicated by the observation that late autoregulation is active in only a subset of the cells that are capable of early autoregulation.

Evidence for positive autoregulation has been obtained for several genes involved in the determination of cell fate, such as the Drosophila sex-determination gene, Sex-lethal²⁰, the homeotic gene, Deformed²¹, and a vertebrate myoblast determination gene, myoD²². Here, we have shown that the selector gene en, responsible for the determination of posterior compartments^3,4, undergoes positive autoregulation. Surprisingly, en autoregulation does not persist throughout the period that en is expressed. We define the end of autoregulation as the time after which heat shock cannot induce the endogenous allele in en mutants, and the time that an en-responsive reporter gene loses expression in a wild-type background. As autoregulation is transient, it does not supply a mechanism for stable determination of the en cell fate.

We have detected a yet-later-acting tier of en maintenance that is independent of autoregulation (Fig. 8a). After the wg-dependent period, the transient production of functional en through heat shock promotes the stable expression of the en^CXI mutant protein in homozygous embryos. Thus, after an autoregulatory phase, en is controlled by new regulators that maintain expression independent of en function. This program may provide maintenance of expression and not activation as en^CXI expression is not reactivated at late stages without heat shock. Although it has not yet been tested directly, it may be this level of en regulation that provides the stable, cell intrinsic control that characterizes a determined cell. Candidates for later acting en regulators, both positive and negative, have been identified among the trithorax and Polycomb gene classes. Mutations in these genes have been found to affect maintenance of expression patterns of homeotic selector genes^23–25, and members of the Polycomb class have been shown to affect en expression as well (D. Moazed and P.O'F., unpublished data)^26–28

The classically defined stages of reversible and irreversible cell fate commitment lead us to expect changes in the mode of en regulation during development. But the elaborate progression that we have uncovered is difficult to rationalize in all its detail. Although the transition to wg independence suggests a move away from cell-extrinsic control of identity, the subsequent transition, from autoregulatory activity to a still later stage of control, has no ready explanation. Identifying the later-acting regulators and the time at which the en cell fate becomes irreversibly determined may help explain these transitions.