Drosophila p27Dacapo expression during embryogenesis is controlled by a complex regulatory region independent of cell cycle progression

Unrestrained cell proliferation results in an exponential increase of cell numbers. An accurate regulation of cell proliferation is therefore required in particular in multicellular organisms where tissue-specific limits for cell numbers must be obeyed in the presence of homeostatic mechanisms that keep nutrient levels constant within the body. Cell proliferation in higher organisms, therefore, is kept in check by dedicated regulatory processes rather than by nutrient limitations as in unicellular organisms. Cyclin-dependent kinases (Cdks) are intracellular targets of these regulatory mechanisms that limit cell proliferation (Edgar and Lehner, 1996). In particular the D- and E-type cyclin-Cdk complexes that regulate the activity of the retinoblastoma tumor suppressor protein and thereby E2F transcription factors are thought to have a central role (Harbour and Dean, 2000).

How are D- and E-type cyclin-cdk complexes regulated during development? Drosophila provides an attractive system for analyses in vivo. Surprisingly, D-type cyclin-Cdk complexes do not have an absolutely essential role. Flies that lack Cdk4, the single known kinase partner of Drosophila Cyclin D, can develop to the adult stage (Meyer et al., 2000). By contrast, Cyclin E and its kinase partner Cdk2 are essential (Knoblich et al., 1994; Lane et al., 2000), and the control of Cyclin E/Cdk2 activity is crucial for a timely arrest of cell proliferation in the dorsolateral embryonic epidermis. In this tissue, the majority of cells exit from the cell cycle after the sixteenth round of mitosis. During the final division cycle, transcription of Cyclin E is downregulated (Knoblich et al., 1994). In parallel, expression of dacapo (dap), which encodes a Drosophila CIP/KIP-type Cdk inhibitor specific for Cyclin E/Cdk2 complexes, is activated. p27DAP accumulation is required to inhibit Cyclin E/Cdk2 after the final division because some residual Cyclin E protein remains after the last mitosis. In dap mutants, therefore, epidermal cells fail to arrest after mitosis 16. Instead of stopping in G1, they progress through an additional division cycle (de Nooij et al., 1996; Lane et al., 1996). This same phenotype is also observed after overexpression of Cyclin E (Knoblich et al., 1994).

Even though inhibition of Cyclin E/Cdk2 is clearly essential for a timely arrest of the epidermal cell proliferation, this complex does not function as a master regulator of cell proliferation. After persistent Cyclin E overexpression in the epidermis, only a single extra division cycle is observed. The downregulation of Cyclin A and Cyclin B transcripts, and presumably that of other cell cycle genes as well, which is observed when epidermal cells exit from the cell cycle during wild-type embryogenesis, is not prevented by Cyclin E overexpression (Knoblich et al., 1994).

The upregulation of dap expression which starts during wild-type development in G2 before the final mitosis is the first known event of the regulatory processes that terminate cell proliferation in the embryonic epidermis. An analysis of the mechanism that dictates this precise onset of dap expression therefore should provide further insights into how cell proliferation is terminated on time. We have addressed whether the onset of dap transcription is dependent on completion of the preceding cell proliferation program. We find that dap upregulation is not advanced after a premature arrest of cell proliferation. In addition, we find that regulators like Cyclin E, Cyclin D/Cdk4 and E2F, which are known to stimulate cell proliferation (Neufeld et al., 1998; Datar et al., 2000), do not interfere with the upregulation of dap transcription when overexpressed in the epidermis. We have also addressed whether dap expression is an up- or downstream event in the cellular program that arrests proliferation. In the latter case, dap transcription might be expected to be regulated by a single enhancer that is used in every tissue whenever cell proliferation is arrested. However, the dap regulatory region is composed of a complex array of enhancers that integrate tissue- and stage-specific signals. The dap regulatory region therefore is involved in the integration of the information whereby the decision to terminate cell proliferation is reached.

Genomic clones of Drosophila melanogaster dap were isolated from a lambda DASH library using a dap cDNA probe (Lane et al., 1996). Six clones were plaque purified. A 10 kb XbaI-NotI fragment containing 7 kb upstream region and the transcribed region up to the NotI site located immediately before the stop codon was subcloned and sequenced. Except for minor differences, which presumably reflect polymorphisms, our sequence agrees with the subsequently published genome sequence (Adams et al., 2000). The start of our cDNA (Lane et al., 1996) is at position 125303 of the genome sequence (Accession Number, AE003832.2) and will be designated as 1 in the following description. Additional cDNAs characterized as expressed sequence tags by the Drosophila genome project start 14 bp or further downstream of this position.

The dap-g and dap-gm constructs were made using the vector CaSpeR 4. The 3′ region in the dap-g series was derived from the dap cDNA starting at the NotI site immediately before the stop codon at position 2742 and extending beyond the poly A tail to an EcoRI site in the linker added during cDNA library construction. In the dap-gm series, a fragment encoding six copies of the Myc epitope was used to replace the region from this NotI site to the XhoI site at position 2954 within the 3′ region. The dap-gm series therefore coded for a full-length p27DAP protein C-terminally fused to Myc epitopes. Genomic fragments that start at –6377 (dap-1g and gm), –4472 (dap-2g), –3939 (dap-3gm), –2460 (dap-4gm), –2023 (dap-5gm) and –1611 (dap-6g and gm), and extend to the NotI site at position 2742 were added to the 3′ regions.

lacZ reporter constructs (dap-l series) we constructed with pC4PLZ (Wharton and Crews 1993), which resulted in expression of β-galactosidase with a nuclear localization signal. A polymerase chain reaction (PCR) fragment containing the region from –2460 to –1250 was inserted into the BamH1 site of pC4PLZ to yield dap-1l. Additional fragments (–3970 to –1928 in dap-2l, –2603 to –1928 in dap-3l, –3304 to –2553 in dap-4l, –3970 to –3240 in dap-5l, –3830 to –3240 in dap-6l, –3970 to –3455 in dap-7l, –3970 to –3544 in dap-8l, –3970 to –3651 in dap-9l, 3894 to –3586 in dap-10l, –3894 to –3494 in dap-11l, –3771 to –3494 in dap-12l, –3670 to –3354 in dap-13l, –3670 to –3240 in dap-14l and –3558 to –3240 in dap-15l) were inserted into the EcoRI and SpeI sites. dap-15l-A8 is identical to dap-15l except that the region from –3455 to –3394 was replaced with a SpeI site by an inverse PCR strategy.

Genomic clones of Drosophila virilis dap were isolated by screening a lambda EMBL3 library with a Drosophila melanogaster dap cDNA probe at low stringency. Five independent clones were isolated. A 9.0 kb fragment extending 5 kb from the putative transcriptional start site into the 5′ region and 1 kb from the putative stop codon into the 3′ region was sequenced (GenBank accession number AY061931).

Several independent lines of each transgene were analyzed. The expression pattern described here were detected with at least three independent insertions. The analyzed mutations were stg^7B, vvl^GA3, trh¹⁰⁵¹², pros¹⁷, CycA^neo114 and Df(2R)599-5 which deletes Cyclin B in combination with blue balancer chromosomes. prd-GAL4 was used to express UAS-Abd-B (Castelli-Gair et al., 1994), UAS-CycE (gIII.1), UAS-CycD (III.1), UAS-Cdk4 (III.2), UAS-E2F (5A) and UAS-DP (8C). UAS-CycD and UAS-Cdk4 were present on a recombinant chromosome and expressed simultaneously, as well as UAS-E2F and UAS-DP.

In situ hybridization and immunolabeling was performed as described previously (Lane et al., 1996). Antibodies against β-galactosidase (Cappel), Myc (mouse monoclonal 9E10), Even-skipped (kindly provided by M. Frasch), Eagle (Dittrich et al., 1997), Prospero (mouse monoclonal MR1A) (Spana and Doe 1995), Futsch (mouse monoclonal 22C10) and p27DAP (Lane et al., 1996) were used. Double labeling of mRNAs by in situ hybridization and β-galactosidase by immunofluorescence was achieved using tyramide signal enhancement (NEN) as described (Knirr et al., 1999). A Zeiss Axiophot equipped with a cooled CCD camera (Photometrics) or a Sony video camera and a Leica TCS SP was used for microscopic analyses.

EMSA was performed essentially as described (Shen et al., 1997). ABD-B was generated in a reticulocyte lysate (TNT, Promega) using an Abd-B-r cDNA fragment (Kuziora and McGinnis, 1988) in a pCITE-2a(+) vector (Novagen). Reticulocyte lysate (5 μl) was used in each assay. Mock incubations were performed with the same amount of reticulocyte lysate programmed with the empty pCITE-2a(+) vector. Complementary oligonuclotides were annealed after end-labeling of one of the two strands and used for binding reactions. Oligonucleotides generating the A8 region (Fig. 4) from wild-type Drosophila virilis and melanogaster dap with the sequence 5′-GTGTTTTTTATGCCGTTTCTCACCAAAAATCAGAGTCATAAATGCATT-3′ and 5′-CGCTGCTTTATGGCGTTTCTCACCCGATGAAGTCGAGTCATAAATGCGTT-3′, respectively, were used. In addition, we used oligonucleotides with the Drosophila melanogaster A8 region in which the two core motifs of ABD-B protein binding sites (indicated in bold) were altered from TTAT to GGCG.

In the embryonic epidermis, dap transcripts start to accumulate during G2 before the final mitosis 16. Within the epidermis, the pattern of dap transcript accumulation anticipates the pattern of mitosis 16. It is first observed in the region of the tracheal pits and in the prospective posterior spiracle region (Fig. 1A), then in the dorsal epidermis and finally also in the ventral epidermis. To determine whether dap expression is dependent on progression through previous divisions, we analyzed string (stg) mutant embryos. In these embryos, cell proliferation is prematurely arrested in G2 before mitosis 14 (Edgar and O’Farrell, 1989). In situ hybridization indicated that the accumulation of dap transcripts was not delayed in stg embryos, although signal intensities were not as strong as in stg⁺ sibling embryos (data not shown) (de Nooij et al., 2000). The lower signal intensities presumably reflect the fact that stg embryos display only a quarter of the normal cell density at the time when dap transcription starts. To confirm that dap expression is not inhibited when epidermal cells are arrested prematurely, we analyzed Cyclin A Cyclin B double mutant embryos. In these embryos, cell proliferation is prematurely arrested in G2 before mitosis 15 (Knoblich and Lehner, 1993). Nevertheless, accumulation of dap transcripts occurred normally in these embryos (Fig. 1, compare A with C).

In contrast to mutations in stg and Cyclin A and B, which result in a premature cell cycle arrest, overexpression of Cyclin E triggers an additional division cycle, as also observed in dap mutants (Knoblich et al., 1994; de Nooij et al., 1996; Lane et al., 1996). To address whether Cyclin E overexpression inhibits dap transcription, we analyzed embryos carrying prd-GAL4 and UAS-Cyclin E. In these embryos, Cyclin E is overexpressed in alternating segments of the epidermis. However, accumulation of dap transcripts started normally throughout the entire epidermis (Fig. 1E). We conclude therefore that the extra division cycle which occurs in the UAS-Cyclin E-expressing segments (Knoblich et al., 1994) does not result from inhibition of dap expression, and we assume that p27DAP protein levels are simply insufficient to bind and inhibit all of the Cyclin E/Cdk2 complexes present in the overexpressing regions.

Although overexpression of Cyclin E had no effect on the onset of dap expression, it resulted in a prolonged presence of dap transcripts. We still observed signals in UAS-Cyclin E expressing regions at a stage when the non-expressing regions were no longer labeled with the dap probe (Fig. 1F). However, this difference between expressing and non-expressing regions emerged well after the onset of the extra division cycle triggered by UAS-Cyclin E expression.

Using prd-GAL4 and appropriate UAS transgenes, we analyzed whether dap expression is inhibited by other regulators known to stimulate cell proliferation in imaginal discs. However, overexpression of E2F1/DP and Cyclin D/Cdk4 had no effect on the onset of dap transcription (data not shown).

As dap expression appeared to occur independent of cell cycle progression, we studied the role of cell fate determinants. We focused on the prominent early dap expression observed in the forming tracheal pits before mitosis 16. The transcription factors encoded by ventral veins lacking (vvl) and trachealess (trh) are expressed within the prospective tracheal pit regions and are known to co-operate for the specification of tracheal cell fate (Boube et al., 2000). The characteristic early dap expression in tracheal pits was not detected in vvl embryos (Fig. 2C) and it was severely decreased in trh embryos (Fig. 2E).

The pattern of stg expression anticipates and determines the embryonic cell division pattern (Edgar and O’Farrell, 1989; Edgar and O’Farrell, 1990). stg expression in the embryonic epidermis before mitosis 16, therefore, is highly similar to the pattern of dap expression which also precedes this terminal mitosis 16 (compare Fig. 2A with 2B). To analyze whether stg and dap expression before mitosis 16 are mechanistically coupled, we also analyzed the distribution of stg transcripts in vvl and trh embryos. Interestingly, while the characteristic early expression of stg was not observed in trh embryos (Fig. 2F), it was normal in vvl mutants (Fig. 2D). As expected, progression through mitosis 16 also occurred in the normal pattern in these vvl mutants. The absence of the characteristic early dap expression in tracheal pits of vvl embryos, therefore, is not preceded by a change in the proliferation program. These findings indicate that the control of stg and dap expression is mechanistically distinct. Moreover, they suggest that regulators of developmental fates control dap expression directly and not indirectly by controlling the cell proliferation program via other cell cycle regulators.

In the epidermis, cell cycle exit is preceded by dap expression. Is cell cycle exit in all tissues preceded by dap expression? Does the same rule apply during nervous system development? Development of the central nervous system starts with delamination of single neuroblasts from the neuroectoderm. These neuroblasts divide asymmetrically. One daughter cell continues with asymmetric neuroblast divisions. The other, the ganglion mother cell (GMC), divides just once to produce two post-mitotic neurons. As in the epidermis, where dap expression starts before the terminal division, dap might be induced in GMCs to prevent further proliferation after its division. To evaluate this idea, we double-labeled embryos with antibodies against p27DAP and Prospero, a pan-neural transcription factor. Prospero is localized to the cell membrane of neuroblasts and segregated asymmetrically into GMCs during neuroblast divisions (Hirata et al., 1995; Knoblich et al., 1995; Spana and Doe, 1995). In GMCs, Prospero translocates to the nucleus. Because of this asymmetric segregation, Prospero is more than a useful marker for GMCs. It is also an attractive candidate transcription factor that might activate dap expression in all GMCs. We double-labeled embryos during the stages where the first GMC divisions are initiated. During these stages, we did not detect p27DAP in Prospero-positive GMCs (Fig. 3A,B). This result indicates that cell cycle exit in the early CNS neurons is not accompanied by p27DAP accumulation before the final division, as in the epidermis.

However, we readily observed p27DAP expression in the Prospero-positive MP2 neuroblast (Fig. 3A,B). MP2 is an exceptional neuroblast that accumulates Prospero in the nucleus. Moreover, MP2 divides just once to produce two postmitotic neurons (Doe et al., 1988; Spana and Doe, 1995). The final division of this unusual neuroblast therefore is preceded by dap expression, as in the epidermis. However, anti-p27DAP labeling of prospero mutant embryos, indicated that dap expression in MP2 is not the result of Prospero translocation into the nucleus. (Fig. 3D).

At later stages, the pattern of anti-p27DAP labeling observed in the CNS was complex and highly dynamic. Anti-p27DAP signals of non-uniform intensity were detected in both Prospero-positive as well as Prospero-negative cells (Fig. 3C).

dap expression in epidermis and CNS appears to be controlled by developmental regulators of cell fate rather than by coupling with cell cycle progression. To confirm this suggestion, we undertook a molecular dissection of the dap regulatory region. We first constructed a series of dap transgenes (dap-g series), which included 5′ regulatory sequences ranging from 2 to 6 kb in size (Fig. 4) and tested whether they were able to complement the recessive semi-lethality of the null allele dap⁴ (Lane et al., 1996). The largest transgene, dap-1g, and many of the shorter transgenes, fully complemented dap⁴, and the rescued flies were found to be fertile.

As we do not know whether all aspects of the dap expression program are required for full viability, it remained a possibility that these rescuing transgenes did not include the complete regulatory region. Transgene expression was analyzed by in situ hybridization and immunofluorescence in the background of dap⁴, which carries a substantial intragenic deletion (Lane et al., 1996). In addition, we also constructed transgenes that expressed a dap fusion with six copies of the Myc epitope tag at the C terminus (dap-gm series). As observed with the non-tagged transgenes, the largest transgene, dap-1gm, and many of the shorter transgenes, fully complemented dap⁴. Expression of the dap-gm transgenes could be analyzed by immunolabeling with anti-Myc and compared with the endogenous dap expression by double labeling with anti-p27DAP.

Our analyses indicated that dap-1g and dap-1gm were expressed during embryogenesis in the same pattern as endogenous dap, with one exception. dap-1g and dap-1gm are not expressed efficiently in the prospective amnioserosa. Some regulatory sequences required for this aspect of expression are presumably either downstream or more than 6 kb upstream or of the dap transcription unit.

We were most interested in identifying the regulatory region which directs expression in the epidermis where the dap phenotype has been characterized most extensively. dap-1g, dap1-gm, dap-2g and dap-3gm were expressed in the epidermis at the appropriate developmental stage (Fig. 5C, and data not shown). By contrast, shorter transgenes (dap-4gm, dap-5gm, dap-6g and dap6-gm) did not direct the strong upregulation during the final division cycle of the epidermal cells (Fig. 5D, and data not shown). However, all the transgenes still retained at least some aspects of the expression pattern in the peripheral and central nervous system (Fig. 5E,F and data not shown). These results indicated that expression in different embryonic tissues is directed by distinct regulatory regions (Fig. 4).

Analysis of the dap⁹ allele further confirmed this notion. dap⁹ was isolated after mobilization of the P element insertion present in dap¹. dap⁹ has little effect on viability. However, dap⁹ homozygotes have phenotypic abnormalities (bristle duplications) very similar to the few homozygous escapers obtained with dap-null alleles. Molecular analysis indicated that dap⁹ lacked the sequences from –1246 to +10 (Fig. 4). Even though the region surrounding the putative transcription start site is deleted in dap⁹, it is not a null allele. Expression in the embryonic epidermis and in some cells of the peripheral nervous system before germband retraction was found to be normal in dap⁹ homozygotes (data not shown). However, after germband retraction, expression in the nervous system was severely affected (Fig. 5H), confirming a proximal location of important regulatory sequences required for aspects of CNS expression (Fig. 4).

To define the sequences that direct the expression in the embryonic epidermis in more detail, we constructed a series of lacZ reporter constructs (dap-l series). A summary of the results obtained with these transgenes is described in Fig. 4, and expression patterns are illustrated in Fig. 6. The 1.2 kb genomic fragment present in dap-1l was found to drive expression in tracheal pits, CNS and PNS (data not shown). The adjacent 2 kb genomic fragment present in dap-2l resulted in expression in the epidermis (Fig. 6A), the gut (Fig. 6B) and in some cells of the peripheral nervous system (Fig. 6C). Further subdivision of the dap-2l fragment revealed multiple independent enhancers. Expression in the peripheral nervous system was obtained with a subregion dap-3l (Fig. 6D,F). A subregion dap-4l directed expression mainly in the interstitial cell progenitors (icp) of the gut (Fig. 6H) and subregion dap-5l in epidermis and gut (Fig. 6J,K). A further analysis of the 700 bp dap-5l subregion again resolved independent regulatory elements (Fig. 4 and Fig. 6M-O) emphasizing the modularity of the dap regulatory region.

To identify conserved sequence elements in the dap regulatory region, we isolated a dap homolog from Drosophila virilis. The gene product shares 65% amino acid sequence identity with p27DAP from Drosophila melanogaster. The results of a nucleotide sequence comparison is illustrated in Fig. 4. This comparison revealed several blocks of high sequence similarity not only within the coding region but also within the regulatory region. The most distal of these blocks (A8) is located within the region containing regulatory elements controlling expression in the epidermis. By comparing the expression of transgenes with and without this block (dap-15l and dap-15l-A8) we examined its significance (Fig. 4 and Fig. 6N,O). The results indicate that the A8 block is required for the early high level expression in a region of the epidermis within the abdominal segment A8. Wild-type dap expression occurs early and at high levels within this prospective posterior spiracle region. Interestingly, the A8 block contains binding sites for the homeodomain transcription factor encoded by Abdominal-B (Abd-B) (Fig. 4; Fig. 7) which is expressed within the posterior abdominal region and required for the specification of posterior spiracles (Celniker et al., 1989; Hu and Castelli-Gair, 1999). prd-GAL4 driven expression of UAS-Abd-B induced expression of the endogenous dap gene (Fig. 6P), dap-15l (Fig. 6Q) but not of dap-15l-A8 (Fig. 6R).

These results indicate that ABD-B is involved in the control of dap expression within the abdominal segment A8. Moreover, they indicate that dap expression is regulated by multiple control elements even within a tissue like the embryonic epidermis. While the regulatory sequences present in the dap-12l construct drive relatively uniform expression throughout the epidermis, the conserved block A8 adds an earlier onset within abdominal segment A8. Another element in dap-1l adds the early onset within the region of the tracheal pits. In summary, our analyses demonstrates that the dap regulatory region is composed of a complex array of stage and tissue-specific enhancers.

The precise onset of dap transcription during the final division cycle is the first event known to be required for the timely cell proliferation arrest in the embryonic epidermis (de Nooij et al., 1996; Lane et al., 1996). We have addressed the regulation of this precise onset of dap transcription and we have determined whether the mechanism used to terminate cell proliferation in different tissues is always identical.

Proximal parts of the dap regulatory region control expression in the peripheral and central nervous system. Even though we have not dissected this regulation in detail, it appears that the expression in the nervous system is controlled by several independent cis-regulatory elements. The 1.3 kb deletion present in dap⁹ eliminates many but not all aspects of dap expression in the nervous system. Similarly, the 0.7 kb fragment in the dap-3l reporter construct is sufficient to direct expression in some but not all of the peripheral nervous system lineages.

In addition, our limited analyses of dap expression during CNS development in wild-type embryos demonstrates that cell cycle exit is not always preceded by dap expression. When the first GMCs are generated during embryonic CNS development and progress through their terminal division cycle, we cannot detect p27DAP in these cells, while expression in the unusual MP2 neuroblast is readily observed. Interestingly, we find that this dap expression in the MP2 neuroblast occurs also in prospero mutant embryos. All our findings therefore argue against the suggestion favored by previous studies, which argues that the timely arrest of cell proliferation in GMC progeny might depend on the induction of dap expression by the transcription factor Prospero. Prospero remains cytoplasmic in neuroblasts and is asymmetrically segregated during neuroblast divisions into GMCs where it accumulates in cell nuclei (Hirata et al., 1995; Knoblich et al., 1995; Spana and Doe 1995). Prospero acts as an inhibitor of cell proliferation in the late embryonic CNS (Li and Vaessin, 2000). Therefore, the idea that nuclear Prospero might trigger dap expression in GMCs to bring about the G1 arrest observed in the two neurons generated by GMC division appeared very attractive. Moreover, in principle this hypothesis is also suggested by the correlation that MP2, an exceptional neuroblast that behaves like a GMC in that it divides just once to produce two postmitotic neurons, accumulates Prospero in the nucleus (Hirata et al., 1995; Knoblich et al., 1995; Spana and Doe 1995) and expresses dap (this work).

As our findings argue against a mechanism that operates generally in all GMCs to prevent further cell cycle progression after the terminal division by Prospero-mediated induction of dap expression, we might have to consider more complex mechanisms that might even vary in different neuroblast lineages. The regulation of dap expression in the nervous system certainly entails such complexity. We emphasize, however, that our work does not exclude a dap-independent, general cell cycle arrest mechanism that operates in all GMCs and is perhaps even triggered by nuclear Prospero.

At least in some CNS lineages, dap expression is required to limit cell proliferation. dap mutants have excess Even-skipped and Eagle-positive neurons in the CNS at stage 14 (data not shown). The proliferation-limiting function of dap therefore is not restricted to the embryonic epidermis, where it has been most convincingly demonstrated. However, apart from the absence of DAP in early GMCs of the CNS, there are additional observations indicating that G1 arrest is not always accompanied by induction of dap expression. dap expression is detectable neither in the zone of non-proliferating cells formed in third instar wing imaginal discs (M. E. Lane and C. F. L., unpublished) nor in the region just anterior to the morphogenetic furrow in eye imaginal discs (de Nooij et al., 1996; Lane et al., 1996) where cells become arrested in G1.

Our findings concerning dap expression in the embryonic epidermis further support and extend the notion that dap expression is controlled by a complex array of cis-acting elements. The initial epidermal expression in regions where the tracheal pits invaginate requires different regulatory sequences than the subsequent expression in the remainder of the epidermis. Similarly, the early expression in the prospective posterior spiracle region in the abdominal segment A8 requires a specific region which includes binding sites for the HOX-protein ABD-B, which acts at the top of a regulatory cascade directing posterior spiracle development (Hu and Castelli-Gair, 1999). The ABD-B-binding sites are required for the early expression of dap reporter constructs in the prospective posterior spiracle region and they are conserved in the dap homolog of Drosophila virilis. Transcription factors that specify developmental fates during Drosophila embryogenesis are therefore likely to regulate dap expression directly in the same way as ABD-B.

The idea that developmental regulators govern dap expression is consistent with our finding that the onset of dap transcription is not dependent on completion of the embryonic cell proliferation program. The accumulation of dap transcripts occurs at the correct stage also in the epidermis of stg and Cyclin A Cyclin B double mutants (de Nooij et al., 2000) (this work) where cells arrest prematurely. Moreover, the onset of dap transcription in the epidermis is not affected by overexpression of positive regulators of cell proliferation like Cyclin E, Cyclin D/Cdk4 and E2F1/DP (Knoblich et al., 1994; Neufeld et al., 1998; Datar et al., 2000; Meyer et al., 2000).

Our results therefore demonstrate that the activation of dap expression in the embryonic epidermis is not coupled to cell cycle progression. It is also not a downstream event of a cellular program, which is realized invariably whenever cell proliferation is arrested in various tissues and developmental stages. Instead we show that the control of dap expression involves a complex regulatory region composed of many cis-acting elements which direct distinct aspects of the intricate expression pattern. Specific combinations of transcription factors which specify various developmental fates appear to act directly at the dap regulatory region.

During development of multicellular organisms, cell proliferation evidently has to be coordinated with other processes (pattern formation, morphogenesis and growth). In principle, coordination could be achieved by developmental control of a single essential cell cycle regulator, while all others might simply be governed by feedback coupling to cell cycle progression. Analyses in Drosophila embryos have clearly demonstrated that multiple cell cycle regulators are controlled by developmental signals. Apart from our work on dap, previous studies have revealed very similar findings in the case of string (Edgar et al., 1994; Lehman et al., 1999) and Cyclin E (Jones et al., 2000). The embryonic expression of both genes is controlled largely independent of cell cycle progression by many independent enhancers within extensive cis-regulatory regions.

How general is the massive parallel input of developmental regulators on many elements of cell cycle control? This developmental control of various cell cycle regulators is presumably particularly widespread during the postblastoderm cell division cycles of Drosophila embryogenesis because these cycles represent the final phase of cell proliferation, which is used for fine-tuning of cell numbers according to tissue-specific needs. The bulk of cells are generated earlier during the syncytial cycles. During the earlier syncytial cycles, nuclear losses are regulated by compensatory divisions and the termination of the extremely rapid syncytial cycles occurs when a particular ratio of nuclei to cytoplasm is reached (Edgar et al., 1986; Yasuda et al., 1991). By contrast, during the later postblastoderm cell division cycles, cell losses are no longer regulated by compensatory divisions although apoptosis can remove excess cells (Busturia and Lawrence, 1994; Namba et al., 1997; Li et al., 1999). Like embryogenesis, imaginal development starts also with an exponential and regulative phase of cell proliferation, which is followed by highly patterned cell divisions during the final stages allowing for tissue-specific fine-tuning in combination with apoptosis (Edgar and Lehner, 1996). The regulation of the initial proliferative imaginal cell cycles might differ substantially from the final embryonic postblastoderm cycles for additional reasons. Maternal provisions largely uncouple embryonic cell cycles from growth. By contrast, the initial imaginal cycles are presumably governed by growth to a critical cell size which triggers a sequence of coupled cell cycle transitions with minimal direct influence of developmental fate regulators.

Experimental support for this suggestion has been obtained in the case of string. A relatively small promoter proximal fragment of the complex string regulatory region is sufficient to drive expression during imaginal cell proliferation (Lehman et al., 1999). Therefore, during the initial phase of imaginal cell proliferation, the control of string expression appears to be comparatively simple and coupled to cell cycle progression. Similarly, a restricted fragment from the dap regulatory region has been shown to drive Cyclin E-stimulated expression during the imaginal cell proliferation (de Nooij et al., 2000). The expression of dap during the initial imaginal proliferation, therefore, might be coupled to periodic Cyclin-E/Cdk2 activity.

Although the modes of cell cycle regulation during the final embryonic postblastoderm and the initial imaginal cell cycles are likely to be distinct, we expect these differences to be gradual, because the intricate developmental inputs on many cell cycle regulators that govern the late embryonic cycles have presumably evolved gradually as well, starting from a primordial growth-coupled control mode. At least some cell cycle regulators in different imaginal discs therefore might also be controlled directly by specific fate regulators and not exclusively by growth-coupled cell cycle progression. Conversely, it is not excluded that some coupling of dap expression with cell cycle progression is also operating in certain embryonic tissues. The often surprising variability in the details of cell cycle control described in different mammalian cell lines might reflect similar tissue- and stage-specific differences. We expect that the processes that terminate cell proliferation in different tissues of multicellular organisms will display a maximal regulatory diversity. The control of dap expression in Drosophila embryos is clearly highly stage and tissue specific.

We thank R. Blackman for the Drosophila virilis library, B. Edgar and J. Casanova for fly stocks, and F. Karch for the Abd-B cDNA. We acknowledge the contributions by C. Horn, H. Jacobs, B. Muhs and A. Wolf during the early stages of the work. We are very grateful to E. Wimmer for comments on the manuscript. This research was supported by the Deutsche Forschungsgemeinschaft (DFG LE 987/1-2, 1-3 and 2-1).