Suppression of the rbf null mutants by a de2f1 allele that lacks transactivation domain

Transcription

1 Development 127, (2000) Printed in Great Britain The Company of Biologists Limited 2000 DEV Suppression of the rbf null mutants by a de2f1 allele that lacks transactivation domain Wei Du Ben May Institute for Cancer Research and Center for Molecular Oncology, The University of Chicago, 924 E. 57th Street, Chicago, IL, 60637, USA wdu@ben-may.bsd.uchicago.edu Accepted 3 November; published on WWW 20 December 1999 SUMMARY In mammals, a large number of proteins including E2F transcription factors have been shown to interact with the tumor suppressor gene product prb, but it is not clear to what extend the function of prb is mediated by E2F. In addition, E2F was shown to mediate both transcription activation and repression; it remains to be tested which function of E2F is critical for normal development. Drosophila homologs of the RB and E2F family of proteins RBF and de2f1 have been identified. The genetic interactions between rbf and de2f1 were analyzed during Drosophila development, and the results presented here showed that RBF is required at multiple stages of development. Unexpectedly, rbf null mutants can develop until late pupae stage when the activity of de2f1 is reduced, and can develop into viable adults with normal adult appendages in the presence of a de2f1 mutation that retains the DNA binding domain but lacks the transactivation domain. These results indicate that most, if not all, of the function of RBF during development is mediated through E2F. In turn, the genetic interactions shown here also suggest that de2f1 functions primarily as a transcription activator rather than a co-repressor of RBF during Drosophila development. Analysis of the expression of an E2F target gene PCNA in eye discs showed that the expression of PCNA is activated by de2f1 in the second mitotic wave and repressed in the morphogenetic furrow and posterior to the second mitotic wave by RBF. Interestingly, reducing the level of RBF restored the normal pattern of cell proliferation in de2f1 mutant eye discs but not the expression of E2F target genes, suggesting that the coordinated transcription of E2F target genes does not significantly affect the pattern of cell proliferation. Key words: RBF, E2F, Cell cycle, Genetic interaction, Retinoblastoma gene, E2F transcription factors, Drosophila melanogaster INTRODUCTION The Retinoblastoma gene (RB) is a prototype tumor suppressor gene. Individuals with a mutation in RB are predisposed to the development of retinoblastoma and osteosarcoma, with tumors exhibiting loss of the wild-type copy of RB (Fearon, 1997). In mammals, RB is a member of the RB family with two other related proteins p107 and p130. The RB family of proteins are important regulators of cell proliferation and cell differentiation (Weinberg, 1995; Bartek et al., 1997), and function by regulating the activities of their targets through direct protein-protein interactions. A large number of proteins have been identified that can associate with prb, p107 and p130, including the E2F family of transcription factors (Dyson, 1998; Nevins, 1998). The E2F transcription factor was originally identified as the DNA binding activity that regulates the expression of adenovirus E2 promoter, and is composed of one E2F subunit and one DP subunit. In mammals, six E2F subunits and three DP subunits have been identified to date (Helin et al., 1992; Kaelin et al., 1992; Girling et al., 1993; Lees et al., 1993; Beijersbergen et al., 1994; Ginsberg et al., 1994; Ormondroyd et al., 1995; Sardet et al., 1995; Wu et al., 1995; Trimarchi et al., 1998). These E2F and DP proteins can dimerize with each other to bind to E2F sites. E2F sites have been identified in the promoters of two major groups of genes: one group encoding essential enzymes for DNA replication such as PCNA and DNA pol α, and the other group encoding regulators of cell cycle progression such as cyclin E (Lavia and Jansen-Durr, 1999). E2F binding sites are implicated in the cell cycle regulation of both groups of genes. It is possible that the RB family of proteins regulate cell cycle through E2F transcription factors. In addition to E2F, there are a large number of other proteins also shown to interact with and which may be potential targets of prb (Mulligan and Jacks, 1998). A number of these proteins such as MyoD and C/EBP are transcription factors critical for the terminal differentiation of certain tissues. It was shown that the interactions between prb and these proteins were critical for these proteins to be able to activate their target genes (Gu et al., 1993; Chen et al., 1996). At present, it is not clear to what extent the functions of the RB family of proteins are to regulate the function of E2F transcription factors and to what extent they are to regulate the functions of these other proteins. It has been suggested that E2F may mediate the cell

2 368 W. Du proliferation function of prb, while the other proteins may mediate the function of prb in terminal differentiation. E2F transcription factors can function both as transcription activators and as co-repressors of the RB family of proteins. In the transcription activator model, E2F transcription factors are required for the expression of E2F target genes, and the activation of these target genes mediates the biological effect of E2F transcription factors. This model is supported by a number of observations. First, E2F transcription factors have a strong transcription activation domain, capable of activating the transcription of reporter constructs with E2F binding sites (Helin et al., 1992; Kaelin et al., 1992; Shan et al., 1992). Second, the ability of E2Fs to activate transcription strongly correlates with their ability to drive cell cycle progression (Johnson et al., 1993; Shan and Lee, 1994; Qin et al., 1995). Third, analysis of Drosophila embryos with mutations in the E2F homolog, de2f1, showed that de2f1 is required for the coordinate transcription of E2F target genes that encode replication functions, such as dmrnr2, PCNA (Duronio et al., 1995). In the active repression model, the binding of free E2F is not required for expression of some critical E2F target genes, instead, the biological effect of E2F is mediated primarily through the recruitment of prb to these E2F sites to repress the transcription of these genes. This model was first suggested by studies that showed prb can actively repress transcription when targeted to specific promoters (Weintraub et al., 1992; Sellers et al., 1995; Weintraub et al., 1995). In addition, this model is supported by the observations that mutations of E2F binding sites in B-myb, cyclin E and cdc2 promoters causes derepression during G 0 /G 1 (Dalton, 1992; Lam and Watson, 1993; Geng et al., 1996). These two models are not mutually exclusive, however, as it is possible that E2F transcription factors function both to activate the expression of their target genes during S phase and to repress the transcription of their target genes during G 0 /G 1 phase. Loss of E2F transcription factors may cause both a lack of transcription activation during S phase and a lack of repression during G 0 /G 1 phase. The question that remains to be addressed is whether the biological functions of E2F transcription factors are mediated mainly by transcription activation or by active repression during normal development. It is also possible that a specific E2F may preferentially function to activate transcription or mediate active repression. In this regard, it is worth pointing out that the G 0 /G 1 E2F complexes contain mostly E2F-4 and E2F-5, while the E2F complexes containing E2F-1, E2F-2, and E2F- 3 peaks around G 1 /S transition. One way to begin to address these questions is to carry out genetic analysis of the functional relationship between the RB and E2F family of proteins. The analysis of RB/E2F-1 double knock-out mice provided some insight about the contribution of E2F-1 to the phenotypes associated with RB mutation during mouse development (Tsai et al., 1998). RB / mice die in mid-gestation with defects in erythropoiesis, cell cycle control and apoptosis (Clarke et al., 1992; Jacks et al., 1992; Lee et al., 1992). These phenotypes are suppressed in some tissues by removing E2F-1. This indicates that E2F-1 is an important target of prb in these tissues (Tsai et al., 1998). However, the interpretation of these experiments are complicated due to the presence of a large family of E2F and RB proteins and the likelihood of significant functional overlap. For example it is not clear whether the defect observed in E17.0 Rb/E2F-1 double mutant embryos are mediated by the remaining E2Fs or by other targets of prb. Previous studies of mice lacking RB, p107, p130, or different combinations of these mutations indicate that RB, p107 and p130 have overlapping functions during development and in the adult (Clarke et al., 1992; Jacks et al., 1992; Lee et al., 1992; Cobrinik et al., 1996; Lee et al., 1996). In addition, mice lacking E2F-1 develop and reproduce normally, perhaps due to the presence of overlapping functions provided by other E2F family members (Field et al., 1996; Yamasaki et al., 1996). The function of RB/E2F appears well conserved between Drosophila and mammals. Importantly, the number of the RB, E2F and DP are significantly reduced in Drosophila. One RB family protein (RBF), one ddp, and two de2fs have been identified in Drosophila (Dynlacht et al., 1994; Ohtani and Nevins, 1994; Du et al., 1996a; Sawado et al., 1998). As in the mammalian system, two groups of genes have been shown to be regulated by E2F in Drosophila, one encodes replication functions such as PCNA and dmrnr2, and the other encodes regulatory functions of the cell cycle such as cyclin E (Duronio et al., 1998). In embryos with mutations in de2f1 or ddp, the expression of these E2F target genes have not been detected (Duronio et al., 1995; Royzman et al., 1997). In contrast, overexpression of de2f1 induced ectopic expression of dmrnr2 and PCNA (Duronio and O Farrell, 1995; Duronio et al., 1996). In addition, overexpression of de2f1 in developing eye discs also drives G 1 arrested cells into S phase and induces increased apoptosis (Asano et al., 1996; Du et al., 1996b). These effects of overexpression of de2f1 can be suppressed by coexpression of RBF (Du et al., 1996a). Analysis of Drosophila embryos lacking both maternal and zygotic RBF showed that RBF is required for the first G 1 arrest (G117) during embryonic development. Loss of RBF leads to ectopic S phase entry and increased apoptosis (Du and Dyson, 1999). Interestingly, the function of the two Drosophila E2F transcription factors does not appear to overlap significantly. Cotransfection of either de2f1 or de2f2 with a PCNA gene promoter construct into tissue culture cells showed that de2f2 caused repression while de2f1 caused activation (Sawado et al., 1998), suggesting that de2f1 and de2f2 have distinct functions. Thus the RB/E2F pathway is much simpler in Drosophila, without the significant functional overlaps observed in mammals. In addition, a number of mutant alleles of de2f1, ddp and rbf have been generated, which can be used to characterize the functional interrelationship between these genes. To begin to address the fundamental questions about the function of the RB and E2F family of proteins in Drosophila development, we have begun a genetic analysis of the functional relationship between RBF and de2f1. The results presented here show that RBF is required at multiple stages of development, and a critical function of RBF at these different stages is to inhibit de2f1. Furthermore, reducing the level of RBF can partially suppress the phenotypes of de2f1 mutants. These results are consistent with the model that de2f1 functions primarily in transcription activation instead of active repression. In addition, analysis of the rbf and de2f1 null eye discs provides significant insights into E2F activity and cell proliferation during development.

3 Function of RBF and de2f1 during development 369 MATERIALS AND METHODS Fly stocks rbf mutants rbf 14, rbf 11, and the viable weak allele P[w+]wd 120a (renamed rbf 120a in this paper) were described by Du and Dyson (1999). de2f mutants were described previously (Duronio et al., 1995; Seum et al., 1996; Royzman et al., 1999). Heat shock rescue experiments 0-24 hour embryos were collected, the average age of the embryos (12 hours) were used to indicate the time heat shocks were given. Heat shocks were carried out once per day at 37 C for an hour, starting at day 1. Adult flies were counted at day 15. Between 100 to 300 flies were scored for each heat shock time point. rbf 14 /FM7 virgin females were mated with males carrying P[w+,CaSpeRhsRBF-2] on the second chromosome. rbf mutants rescued were identified as males that were not y or B and with a twisted abdomen. The twisted abdomen phenotype is associated with rbf 11 and rbf 14 alleles, but does not appear to be the result of mutations of rbf, nor does it affect viability. This phenotype can be used as a marker to identify rbf 11 and rbf 14 adult males. Similar rescue of rbf mutants was observed when a different heat shock RBF line P[w+,CaSpeRhsRBF-7] was used. P[w+,CaSpeRhsRBF-7] is on the first chromosome. Rescue experiments were carried out either by crossing rbf 14,P[w+,CaSpeRhsRBF-7]/FM7 virgin females with FM7 males, or by crossing rbf 14,P[w+,CaSpeRhsRBF-7]/DP(1:Y) males with rbf 11 /FM7 females. In either case, the respective rbf mutants (rbf 14,P[w+,CaSpeRhsRBF-7]/Y or rbf 14,P[w+,CaSpeRhsRBF-7]/ rbf 11 ) are rescued. In addition these rescued adults have the same adult phenotypes as when RBF was only expressed at the early larval stages. Genetic crosses rbf 14,P[w+,CaSpeRhsRBF-7]/FM7;de2f1 164 virgin females were crossed with rbf 120a /Y; de2f1 91 /TM6B males to rescue the lethality of rbf mutant by lowering the activity of de2f1. This cross was carried out at 25 C with no heat shock provided, a total of 529 adult flies were scored. To test the effect of lowering the activity of de2f1 on rbf null phenotypes, rbf 14 /FM6;de2f1 164 virgin females were crossed with rbf 120a /Y; de2f1 91 /TM6B males. rbf null larvae were identified by larval marker not Tb, not yellow male larvae or by adult structures when they can be seen in the pupae cases. To test the suppression of RBF 14 null mutants by de2f i2 /de2f rm729, rbf 14 /FM6;de2f1 rm729 /TM6B virgin females were crossed with de2f i2 /TM6B males. A total of 694 flies were scored, and 29 rbf 14 /Y;de2f1 rm729 /de2f i2 were observed. This is 50% of the expected amount of rbf 14 /Y;de2f1 rm729 /de2f i2 flies if it is fully viable. Western blots did not detect any RBF protein in these rescued adults. To test the effect of lowering the activity of RBF on the phenotypes of de2f1 mutants, rbf 14 /FM6;de2f1 rm729 /TM6B virgin females were crossed with rbf 120a /Y; de2f1 91 /TM6B males. The non Tb late third instar larvae identified were mostly female, these larvae were presumed to have the genotypes of rbf 14 /rbf 120a ;de2f1 rm729 /de2f1 91, because in control crosses, +/rbf 120a ;de2f1 rm729 /de2f1 91, and de2f1 rm729 /de2f1 91 late third instar larvae were never observed. Scanning electron microscopy Flies were dehydrated through 25%, 50%, 75% and 100% ethanol. The ethanol was exchanged with a mix of 100% ethanol: Hexamethyldisilazane (HMDS) at 1:1 ratio for minutes, followed by at least two changes of 100% HMDS. Samples were transferred to a shallow dish and allowed to dry. They were then mounted onto metal stubs and coated for viewing. BrdU staining, AO staining, in situ hybridization of eye discs BrdU staining Appropriate staged larvae with appropriate markers were picked and dissected in M3 medium, and were placed in 1 mg/ml BrdU for 1 hour at room temperature for incorporation of BrdU. Following incorporation, the brains/discs were washed once in M3 medium and twice in PBS. These brains/discs were then fixed for 30 minutes in 4% formaldehyde/pem (100 mm Pipes disodium salt, 2 mm EGTA, 1 mm MgSO 4), followed by washing three times in 0.3% Triton X- 100/PBS (PBST) for 20 minutes each. BrdU was detected using a mouse anti-brdu antibody (Becton Dickinson, 1:100). Secondary antibodies were from Jackson ImmunoResearch Lab. Acridine Orange staining Acridine Orange (AO) staining was carried out by dissecting the eye discs in 5 µg/ml acridine orange solution. The samples were then rinsed in PBS and mounted on slides for viewing. In situ hybridization of eye discs Eye discs were dissected in PBS and fixed in 4% formaldehyde/pbs. Fixed eye discs were washed 3 times with PBST, dehydrated by adding 0.5 ml 300 mm ammonium acetate, and 0.5 ml ethanol, followed by an ethanol wash. The discs were treated with equal volume of xylene/ethanol for 10 minutes and then washed three times in ethanol, once with methanol, and once with equal volumes of methanol/4% formaldehyde. The eye discs were fixed again in 4% formaldehyde for 20 minutes, followed by 5 washes with PBST. These discs were then pre-hybridized and hybridized at 60 C. RESULTS Requirements of RBF during development Antibody staining or western blot experiments showed that RBF protein was expressed at all the developmental stages (data not shown). To test if RBF is required at multiple stages of development, an rbf null mutant (rbf 14 ) generated using P element local hopping followed by imprecise excision (Du and Dyson, 1999) was analyzed. Egg counts showed that the majority of the rbf null mutant embryos hatched into larvae, and died at early larvae stage (data not shown). To rescue the lethality of rbf 14 mutants, P element transgenic flies expressing RBF cdna under the control of a heat-shock promoter were established. Three individual lines were tested for the ability to rescue the lethality of rbf 14 mutants by providing heat shock once a day throughout development; all three lines were able to rescue the rbf null mutants (Du and Dyson, 1999). To determine when RBF is required during development, we tested whether expressing RBF at specific periods during Drosophila development was sufficient to rescue the lethality of rbf 14 mutants. As shown in Fig. 1B, no viable rbf 14 adult flies were recovered without heat shock to induce the expression of RBF. In contrast, providing one or two pulses of heat shocks in the first 2 days after egg laying (AEL) was sufficient to rescue a small percentage of rbf 14 mutant flies (Fig. 1). Complete rescue of the RBF lethality was achieved by expression of RBF during the first 3 days AEL (Fig. 1). These results indicate that RBF has an essential function during early larval development. Interestingly, the adult flies rescued by expressing RBF during first 3 days AEL showed a number of alterations in adult structures such as the abnormal macrochaetae on the notum and rough eyes (Fig. 2A,B,E,F). Significantly, all these adult phenotypes were rescued when RBF was provided throughout fly development (Fig. 2C,G). Thus both the early larval lethality and the bristle and eye phenotypes observed in the rescued adults are all due to lack

4 370 W. Du days of fly development A B viable rbf mutants/expected rbf % time when heat shock was provided number of heat shocks Fig. 1. Heat shock rescue of rbf mutants. (A) Diagram showing when heat shocks were given during development. The numbers above the line indicate the days after egg laying (AEL). Arrows below the line indicate the developmental time at which heat shocks were provided as described in Materials and Methods. (B) The percentage of rbf mutant flies rescued by different numbers of heat shock. Viable rbf mutant adults rescued as a percentage of the expected number of rbf mutants. The average of two sets of results is shown. The first heat shocks were given at 12 hours AEL as shown in A. of RBF function, indicating that RBF is required at multiple stages during Drosophila development. The requirements of RBF during development can be attributed to its activity to inhibit de2f1 To test whether these observed phenotypes of rbf mutants are due to the excess amount of de2f1 activity, the effect of lowering the activity of de2f1 on the phenotypes of rbf mutants was examined. A de2f1 null allele (de2f1 91 ) and a viable weak allele (de2f1 164 ) were used to reduce the activity of de2f1 as described in the Materials and Methods (Duronio et al., 1995; Seum et al., 1996). Transheterozygotes of de2f1 91 /de2f1 164 developed normally, with no obvious developmental defects. As shown in Fig. 3, while rbf null mutants (rbf 14 ) die at very early larval stage, rbf 14 3rd instar larvae can be readily identified (y + male larvae) when the activity of de2f1 is reduced using transheterozygotes of de2f1 164 /de2f1 91 (Fig. 3B). The identity of these larvae as rbf null mutants was also confirmed by the lack of staining of the internal larval structures (such as salivary gland and gut) by an antibody against RBF (data not shown). The development of the rbf null larvae is delayed by about 2 days compared to the wild-type larvae. Most of the rbf 14 /Y;de2f1 164 /de2f1 91 mutants can develop further into pupae (Fig. 3D). The rbf 14 /Y;de2f1 164 /de2f1 91 larvae and pupae are a little bit smaller than the wild-type ones (see Fig. 3A,B and 3C,D). Some rbf 14 /Y;de2f1 164 /de2f1 91 mutants can even develop into very late stage pupae with adult structures (adult eyes, legs, wings, cuticles and bristles) but cannot completely get out of the pupal case (Fig. 3E and data not shown). No viable adults of rbf 14 /Y;de2f1 164 /de2f1 91 were recovered. The ability to rescue the development of rbf null mutants all the way to pharate adults simply by reducing the level of de2f1 indicates that excess de2f1 is the cause for the early larval lethality of rbf null mutants. In addition, extensive differentiation can occur in the complete absence of RBF as long as the level of de2f1 is reduced. To further investigate if de2f1 is also responsible for the adult eye and bristle phenotypes observed in the rescued rbf 14 adults (Fig. 2B,F), the effect of lowering the activity of de2f1 was analyzed in the rbf mutant background with a heat-shock RBF transgene. The presence of a heat-shock RBF transgene may provide some basal level of RBF expression. As described above, no viable rbf 14,P[w+,hsRBF-7]/Y adult flies (with wildtype level of de2f1 activity) were recovered in the absence of heat shock (Fig. 1B and data not shown). Interestingly, lowering the level of de2f1 was sufficient to rescue the lethality of the rbf 14,P[w+,hsRBF-7]/Y mutants (see Materials and Methods for the cross). About 76% of the expected number of rbf 14,P[w+,hsRBF-7]/Y;de2f1 164 /de2f1 91 were recovered as viable adults. Thus lowering the activity of de2f1 is sufficient to rescue the lethality of rbf 14,P[w+,hsRBF-7]/Y with no heat shock induction. Furthermore, reducing the de2f1 activity also suppressed the bristle and the eye phenotypes observed in the rbf mutant flies rescued by expressing RBF only in the first couple days AEL (compare Fig. 2B,D,F,H). These results indicate that excess de2f1 is also the critical mediator for the rough eye and missing macrochaetae phenotypes observed in rbf mutants. The failure of rbf 14 /Y;de2f1 164 /de2f1 91 to develop into viable adults suggests that RBF may play some other essential roles in addition to inhibiting de2f1 (such as de2f2 or other targets of RBF), or that the de2f1 alleles used may still have too much de2f1 activity. To determine if the expression of E2F target genes is greatly deregulated in the rbf 14 /Y;de2f1 164 /de2f1 91 background, in situ hybridization was carried out using PCNA as a probe. As discussed previously, the expression of E2F target genes such as PCNA, dmrnr2, and DNA pol α are coordinated regulated, (Duronio and O Farrell, 1994; Duronio et al., 1995; Royzman et al., 1997; Du and Dyson, 1999). In third instar larval eye discs, an indentation in the eye disc marks the morphogenetic furrow where cell differentiation initiates. Anterior to the morphogenetic furrow, cells are asynchronously dividing (also referred to as first mitotic wave). In the morphogenetic furrow (and just anterior to the morphogenetic furrow), 8-9 rows of cells are arrested in G 1 phase of the cell cycle. Immediately posterior to the morphogenetic furrow, a subset of cells start neuronal differentiation and the remaining cells enter a synchronous round of S phase (also referred to as second mitotic wave). Posterior to the second mitotic wave, all the cells in the eye disc arrest in G 1 (Wolff and Ready, 1993). As shown in Fig. 3F-3H and 3J-3L, PCNA is highly expressed immediately posterior to the morphogenetic furrow in a region that overlaps with the second mitotic wave in the wild-type eye discs. In addition expression of PCNA is also detected in the anterior part of the eye disc (first mitotic wave). No PCNA expression is detected in (and just anterior to) the morphogenetic furrow or posterior to the second mitotic wave (Fig. 3F,J). The PCNA expression pattern is not significantly altered in de2f1 164 /de2f1 91 eye discs, although the expression in the second mitotic wave is somewhat reduced (Fig. 3G,K).

5 Function of RBF and de2f1 during development 371 Strikingly, in rbf 14 /Y;de2f1 164 /de2f1 91 eye discs, the pattern of PCNA expression is highly abnormal. High PCNA expression is detected in the morphogenetic furrow and anterior to the morphogenetic furrow (Fig. 3H,L). In addition, PCNA expression is observed everywhere posterior to the morphogenetic furrow, with the expression gradually decreasing from the morphogenetic furrow to the posterior end. Notably, PCNA expression can be detected at the very posterior part of the eye disc in rbf 14 /Y;de2f1 164 /de2f1 91 (Fig. 3L). Although it is not demonstrated that the observed deregulated PCNA expression is due to the unregulated de2f1 activity, this observation is at least consistent with the possibility that the de2f1 alleles used have too much activity. Additional de2f1 alleles will be needed to test if rbf null adults can be rescued simply by reducing the activity of de2f1. A new de2f1 allele (de2f1 i2 ) was reported recently (Royzman et al., 1999), de2f1 i2 is a stop codon mutation at amino acid 527, and is expected to produce a truncated protein that can not activate transcription or bind to RBF (Ohtani and Nevins, 1994; Du et al., 1996a). Interestingly, de2f1 i2 is viable but female sterile. To test if eliminating the transcription activation domain of de2f1 is sufficient to suppress the lethality of rbf null mutant, flies with the genotype rbf 14 /Y;de2f1 i2 /de2f1 rm729 were generated and identified as described in the Materials and Methods. Interestingly, about 50% of the expected number of rbf 14 /Y;de2f1 i2 /de2f1 rm729 mutants can survive until the adult stage, and the development of these flies is not significantly delayed compared to the other flies in the same cross. In addition, the surviving rbf 14 /Y;de2f1 i2 /de2f1 rm729 adults showed normal adult eyes and macrochaetae on the notum (Fig. 4A-C, compare with Fig. 2B,F), as well as normal wings and legs (Fig. 4D-I). In summary, the results presented here indicate that RBF is required at multiple stages of Drosophila development, and that all these observed requirements for RBF during development can be attributed to its interaction with E2F. Since de2f1 i2 lacks the transcription activation domain, and de2f1 rm729 behaves as a null allele (Duronio et al., 1995; Brook et al., 1996; Royzman et al., 1997), the eye disc from rbf 14 /Y;de2f1 i2 /de2f1 rm729 provide a suitable model system to investigate whether the expression E2F target genes such as PCNA is deregulated in the absence of RBF and transcription activation function of de2f1. As shown in Fig. 4J,L, no expression of PCNA was detected posterior to the second mitotic wave or in the morphogenetic furrow in wild type eye discs. In contrast, high expression of PCNA was detected immediately posterior to the morphogenetic furrow and anterior to the furrow (in regions overlapping the first and second mitotic wave). Interestingly, PCNA expression was detected in the morphogenetic furrow as well as posterior to the second mitotic wave (Fig. 4K, M) in rbf 14 /Y;de2f1 i2 /de2f1 rm729 eye discs and the level of PCNA expression was relatively constant throughout the posterior or anterior part of the eye disc. These observations are consistent with the idea that in the wild-type eye disc, PCNA expression in the morphogenetic furrow and posterior to the second mitotic wave is actively repressed by RBF. In the absence of transcription activation by de2f1 and repression by RBF, basal level of PCNA expression was detected in these regions. In contrast, high level of PCNA expression was observed immediately posterior to the furrow in the wild-type eye discs but not in the eye discs from rbf 14 /Y;de2f1 i2 /de2f1 rm729, suggesting that de2f1 is required for the high level PCNA expression in this region. In conclusion, these observations support the notion that there are at least three levels of PCNA expression: an activated level when there is active de2f1, a repressed level when there is RBF, and a basal level when there is neither active de2f1 nor RBF. Reducing the level of RBF suppresses the phenotypes of de2f1 mutant The results presented above demonstrate that de2f1 is a critical downstream target of RBF, mediating multiple phenotypes of rbf mutants. However, these results do not address the relative importance of the function of de2f1 in activating transcription versus its function in recruiting RBF to Fig. 2. RBF is required at multiple stages of development, and reducing the level of de2f1 suppressed both the early larval lethality as well as the late differentiation phenotypes of rbf mutants. (A-D) Scanning electron micrographs of adult notum at 90 ; (E-H), Scanning electron micrographs of adult eyes at 190. The genotypes: (A,E) Wild-type, (B,F) rbf 14 /Y;P[w+, hsp70- RBF-2]/+ with two heat shocks during early larval development, (C,G) rbf 14 /Y;P[w+, hsp70- RBF-2]/+ with nine heat shocks, (D,H) rbf 14,P[w+, hsp70-rbf-7]/y;de2f1 164 /de2f1 91 with no heat shock. White arrowheads in B indicate missing or small bristles instead of normal macrochaetae. The fly eye in F has disorganized ommatidia shape, arrangement and a few multiple bristles but these phenotypes were not observed in E,G and H.

6 372 W. Du repress transcription. As depicted in Fig. 5A, there are two generally accepted models for the function of E2F transcription factors. In model I, the biological function of E2F is mediated by its ability to activate transcription. prb functions as a repressor of E2F to repress the E2F-dependent transcription activation. A predication from this model is that E2F and RB function antagonistically. In model II, the biological effect of E2F is mediated by its ability to recruit RB to repress transcription of E2F target genes, thus E2F functions as a corepressor of RB in this model. In Drosophila, E2F transcription factors consist of both de2f1 and de2f2. The two models discussed above give completely different predications regarding the genetic interaction between RBF and de2f1. Considering that in de2f1 mutants, de2f2 can exist in either the free de2f2/ddp from or in a complex with RBF, reducing the level of RBF will decrease the amount of de2f2/ddp/rbf complex and increase the amount of free de2f2/ddp, which will lead to an increase in the expression of E2F target genes either by increased transcription activation or by decreased active repression. If the developmental phenotype of de2f1 mutants is due to lack of transcription activation (model I), then the increase in the expression of the E2F target genes by reducing the level of RBF should suppress the de2f1 mutant phenotypes. On the contrary, if the phenotype of de2f1 is mainly due to lack of active repression (model II), reducing the level of RBF will result in even less repression of these E2F target genes, and thus an enhancement of the de2f1 mutant phenotypes. The phenotypes of de2f1 mutants were first analyzed. Two strong alleles of de2f1 were used: de2f1 91, which has a stop codon mutation at amino acid 31 and de2f1 rm729, which contain a P element insertion 48 nucleotides upstream of the initiator methionine. Both mutants are defective in the expression of de2f target genes such as dmrnr2 and PCNA, and are presumed to be null alleles (Duronio et al., 1995; Royzman et al., 1997). Similar to the previously reported phenotypes of transheterozygotes between de2f1 91 /de2f1 rm7172, transheterozygotes between de2f1 91 /de2f1 rm729 can develop into late second instar or early third instar larvae (Fig. 5C) at a very slow pace (around 13 days AEL) (Royzman et al., 1997). However, no de2f1 91 /de2f1 rm729 pupae can be detected, so this difference may be due to de2f1 rm7172 still having some residual de2f1 activity. To test whether the most critical function of de2f1 during development is to activate transcription or to repress transcription by targeting RBF to specific promoter region, the effect of reducing the level of RBF on the phenotypes of de2f1 91 /de2f1 rm729 was the analyzed. As shown in Fig. 5D, reducing the level of RBF using transheterozygotes of rbf 120a /rbf 14 can suppress the larval lethality of the de2f1 91 /de2f1 rm729 mutants. The rbf 120a /rbf 14 ;de2f1 91 /de2f1 rm729 mutants can develop into large late third instar larvae (Fig. 5D). In addition, significant numbers of rbf 120a /rbf 14 ;de2f1 91 /de2f1 rm729 larvae can pupate (Fig. 5E,F). Some rbf 120a /rbf 14 ;de2f1 91 /de2f1 rm729 can develop into pharate adults with adult eyes, legs and wing structures (Fig. 5G and data not shown). However, the development of the rbf 120a /rbf 14 ;de2f1 91 /de2f1 rm729 mutant larvae is still significantly retarded (around 9-10 days AEL) compared to wild-type larvae (around 5-6 days AEL), with the earliest pupae observed around day 11 AEL. The observed suppression of the de2f1 mutant phenotype by reduction of the level of RBF is consistent with the predication that de2f1 functions principally to activate transcription (model I), but is inconsistent with the prediction that de2f1 functions principally to mediate active repression (model II in Fig. 5A). From these data and the previous published results, the simplest explanation is that the most critical function missing Fig. 3. Suppression of the early larvae lethality of rbf null mutants by reducing the activity of de2f1, and the expression of E2F target gene in the absence of RBF or de2f1. (A-E) images of wild-type or the rescued rbf null larvae or pupae as indicated: (A) A wild-type third instar larva. (B) A rbf 14 /Y;de2f1 164 /de2f1 91 larva. Note that rbf null larvae are a little thinner than the wild-type larvae. (C) A wild-type pupa. (D) A rbf 14 /Y;de2f1 164 /de2f1 91 pupa. The genotype of such pupae can be identified by the pupae and adult structures in the pupal cases (white eyes, male sexcombs, and not Tubby pupae). These rbf 14 /Y;de2f1 164 /de2f1 91 pupae are easily found in the cross. (E) A rbf 14 /Y;de2f1 164 /de2f1 91 pharate adult partially eclosed. The legs of this fly were still moving at the time of photograph. (F-L) The expression of an E2F target gene PCNA was detected by in situ hybridization. The genotypes are: Wild-type (F, J); de2f1 164 /de2f1 91 (G, K); rbf 14 /Y;de2f1 164 /de2f1 91 (H, L); rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 (I). Arrows in J, K and L indicate the morphogenetic furrow.

7 Function of RBF and de2f1 during development 373 Fig. 4. Suppression of the phenotypes of rbf null mutants by de2f1 i2, and the expression of PCNA in the absence of repression by RBF and transcription activation by de2f1 in eye discs. (A,B) Scanning EM image of adult eyes from rbf 14 /Y;de2f1 i2 /de2f1 rm729 flies at 190 (A) and 1000 (B). (C) Scanning EM image of an adult notum from rbf 14 /Y;de2f1 i2 /de2f1 rm729 at 80. (D,H) A wild-type adult wing at low and higher magnification, respectively. (E,I) a rbf 14 /Y;de2f1 i2 /de2f1 rm729 adult wing at low and higher magnification. (F) Wildtype and (G) rbf 14 /Y;de2f1 i2 /de2f1 rm729 adult legs. (J-M) expression of PCNA in eye disc detected by in situ hybridization, arrowheads indicate the morphogenetic furrow. (J and L) Wild-type eye disc at low and high magnifications. (K and M) rbf 14 /Y;de2f1 i2 /de2f1 rm729 eye discs at low and high magnifications, respectively. in de2f1 mutants is the ability of de2f1 to activate transcription, not its ability to recruit RBF to repress transcription. Cell proliferation and cell death in rbf null eye imaginal discs The ability to rescue and identify rbf null mutants at third instar larval stage allowed an analysis of the role of RBF in determining the pattern of cell proliferation in the eye imaginal discs. In wild-type eye discs, cell proliferation and differentiation are coordinated (Wolff and Ready, 1993). Cells are asynchronously dividing anterior to the morphogenetic furrow, and become G 1 arrested in and just anterior to the morphogenetic furrow (about 8-9 rows of cells). Immediately posterior to the morphogenetic furrow, a subset of cells start neuronal differentiation and the remaining cells enter a synchronous round of S phase (the second mitotic wave). Posterior to the second mitotic wave, all the cells in the eye disc arrest in G 1 (Wolff and Ready, 1993). As shown in Fig. 6B,E, BrdU incorporation of the eye discs from de2f1 164 /de2f1 91 is the same as the BrdU incorporation pattern from the wild-type eye disc. In particular, cells in and just anterior to the morphogenetic furrow are arrested in G 1 and the width of the G 1 arrested region (the spacing between the second mitotic wave and the first mitotic wave) are similar (Fig. 6D,E). Interestingly, in eye discs from rbf 14 /Y;de2f1 164 /de2f1 91 mutants, both the first and second mitotic wave are still detected, and cells in the morphogenetic furrow are also arrested in G 1 (Fig. 6C,F). However, there are two notable differences in the pattern of BrdU incorporation in the eye discs from rbf 14 /Y;de2f1 164 /de2f1 91 larvae. First, the G 1 arrested region in the morphogenetic furrow is much narrower, cells just anterior to the morphogenetic furrow (indicated by an open arrowhead in Fig. 6F) are not arrested in G 1 ; instead they become part of the first mitotic wave. This observation indicates that RBF plays an important role in the cell cycle synchronization at G 1 just anterior to the morphogenetic furrow. In contrast, RBF is not required for the G 1 arrest within the morphogenetic furrow. Second, some cells posterior to the second mitotic wave still incorporate BrdU (indicated by a filled arrowhead in Fig. 6F). This is consistent with the previous observations that overexpression of de2f1 in the posterior part of the eye discs induces ectopic S phases (Asano et al., 1996; Du et al., 1996b). These results suggest that RBF is critical for the maintenance of the G 1 arrest in the posterior part of the eye disc. The eye discs from rbf 14 /Y;de2f1 164 /de2f1 91 were a little smaller in spite of the observed increase in the number of cells undergoing DNA replication. Cell death in rbf 14 /Y;de2f1 164 /de2f1 91 eye discs were analyzed using Acridine Orange (AO) staining, a dye that stains apoptotic cells. As shown in Fig. 6G-I, there is very little cell death in eye discs from wild-type or de2f1 164 /de2f1 91 larvae. Interestingly, strong AO staining was observed in cells just anterior to the morphogenetic furrow in rbf 14 /Y;de2f1 164 /de2f1 91 mutant discs (grey arrowhead in Fig. 6I), this is the same region that cells from these discs failed to be arrested in G 1 (compare Fig. 6F with 6D,E). It is possible that this increased cell death eliminates those cells that are not arrested in G 1 just anterior to the morphogenetic furrow. The elimination of these cells may play an important role in cell cycle synchronization in the morphogenetic furrow where differentiation initiates. In contrast, no significant increase in the amount of apoptosis was observed in the posterior end of the eye discs (Fig. 6I).

8 374 W. Du Fig. 5. Reducing the level of RBF suppressed larval lethality of de2f1 mutants, indicating de2f1 functions mainly to activate transcription during development. (A) Two models for the function of E2F transcription factors. In model I, E2F functions as a transcription activator, the biological function of E2F is mediated by its ability to activate transcription. prb functions as a repressor of E2F, the binding of prb to E2F represses the E2F-dependent transcription activation. Thus E2F and prb functions antagonistically in this model. In model II, the binding of E2F together with prb represses the expression of some E2F genes. If this repression mediates most of the biological function of E2F, then E2F functions mainly as a co-repressor of prb in this model. These two different models gave distinct predications for the genetic interactions between E2F and RB (see text for discussions). (B-G) Suppression of de2f1 mutant development by reducing the level of RBF. (B,E) Wild-type larva and pupa. (C) de2f1 rm729 /de2f1 91 larva at 13 days AEL, no late third instar larvae or pupae were observed for this genotype. (D) rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 larva at 10 days AEL. (F) rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 pupa, (G) a rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 pharate adult dissected out from pupal case. The rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 pharate adult has all the adult structures developed inside the pupal case. Taken together, these observations suggest that the loss of RBF and the consequent deregulated E2F activity in and just anterior to the morphogenetic furrow may block normal cell cycle synchronization in cells just anterior to the morphogenetic furrow and induce extensive apoptosis there. The observation that cells within the morphogenetic furrow can still be G 1 arrested in the absence of RBF indicates that there is a RB-independent mechanism for the G 1 arrest in the morphogenetic furrow. The slight deregulation of E2F activity in the posterior part of eye discs (Fig. 3L) correlated with a few ectopic BrdU-stained cells and no significant increased cell death in the posterior part of eye disc from rbf 14 /Y;de2f1 164 /de2f1 91. Reducing the level of RBF restores the normal pattern of DNA replication but not the expression of E2F target genes in de2f1 mutant As shown previously, reducing the level of RBF suppressed the defects in larval development of de2f1 rm729 /de2f1 91 mutants. To further test if reducing the level of RBF also restores the normal pattern of DNA replication, the pattern of BrdU incorporation was examined. As shown in Fig. 7A, there is a specific pattern of cell proliferation in the developing brain of the wild-type third instar larvae (around day 5-6 AEL). There are clusters of cells (corresponding to the optic lobe region of the brain) that are strongly labeled by BrdU. In addition there are discrete clusters of cells in the thoracic region of the brain that are also labeled, these cells are also undergoing DNA replication. As described earlier, de2f1 rm729 /de2f1 91 mutant larvae develop much slower, and can never reach the wild-type third instar larvae size. Interestingly, cells in the thoracic region of the brain of day 13 de2f1 rm729 /de2f1 91 mutant larvae are labeled by BrdU. BrdU incorporation in the optic lobe region of the brain is not detected (Fig. 7B). This suggests that de2f1 activity is required for the proliferation of these cells. Notably, the normal pattern of DNA replication in the optic lobe region is restored in brains of rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 larvae (Fig. 7C). Thus the suppression of the lethality of de2f1 mutants by lowering the activity of RBF is accompanied by a restoration of normal pattern of DNA replication in the developing brain. In addition to restoring the normal pattern of DNA replication in the developing brain, reducing the level of RBF also largely restores the normal pattern of DNA replication in the eye imaginal discs. In wild-type eye discs, BrdU staining was detected in the first and second mitotic wave, but not in the morphogenetic furrow or posterior to the second mitotic wave (Fig. 7D). The eye discs from de2f1 rm729 /de2f1 91 larvae are very small and only have a few cells labeled with BrdU with no specific pattern (data not shown). In rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 eye discs, the overall pattern of DNA replication is normal: BrdU incorporation was observed in the first and second mitotic wave, cells in the furrow and posterior to the second mitotic wave do not incorporate BrdU (Fig. 7D,E). Compared to the wild-type eye discs, eye discs from rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 are smaller, although a larger fraction of cells appear to be labeled with BrdU in the anterior part of the eye disc (see below). The coordinated transcription of the replication function of such genes as PCNA, and dmrnr2 are regulated by de2f1 and are normally expressed as cells enter S phase, thus the expression of E2F target genes normally mirrors the pattern of DNA replication. To test if the pattern of S phases in rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 eye discs corresponds to the expression of E2F target genes, the expression of PCNA was detected by in situ hybridization. Interestingly, only background levels of PCNA expression were detected in rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 eye discs (Fig. 3I). High expression of PCNA was not observed in the area of the first or the second mitotic wave. Thus, the suppression of the de2f1

9 Function of RBF and de2f1 during development 375 Fig. 6. Cell proliferation and cell death in the complete absence of RBF. (A-F) BrdU incorporation in eye discs from third larval instar. (G-I) Acridine Orange staining of eye discs. Magnifications: 25 (A- C), 126 (D-F), and 32 (G-I). Genotypes are: wild-type (A,D,G), de2f1 164 /de2f1 91 (B,E, H), rbf 14 /Y;de2f1 164 /de2f1 91 (C,F, I). Black arrowhead in F shows a few cells posterior to the second mitotic wave incorporating BrdU in eye disc from an rbf 14 /Y;de2f1 164 /de2f1 91 larva. In contrast, no cells were incorporating BrdU posterior to the second mitotic wave in eye discs from the wild-type (D) or de2f1 164 /de2f1 91 (E) larvae. The out-offocus BrdU-positive cells in the furrow or in the posterior part of the eye disc in D, E and F were from the cells in the peripodial membrane. Arrow in F indicates the morphogenetic furrow, which is aligned with the furrow in D and E. Note that while the G 1 arrest region in the morphogenetic furrow of eye discs from wild-type (A,D) or de2f1 164 /de2f1 91 (B,E) larvae are similar in width, the G 1 arrest region is much narrower in eye discs from rbf 14 /Y;de2f1 164 /de2f1 91 (C,F). White arrowhead in F indicates cells just anterior to the morphogenetic furrow are strongly incorporating BrdU. Grey arrowhead in I indicates cells just anterior to the morphogenetic furrow in eye discs from rbf 14 /Y;de2f1 164 /de2f1 91 stained with AO. These cells are in a similar location relative to the morphogenetic furrow as those cells indicated by the white arrowhead in F. mutant phenotype by reducing the level of RBF restores the normal pattern of DNA replication but not the coordinated transcription of the E2F target genes. This background level of expression of E2F target genes, which may cause an extended S phase, due to insufficient amount of DNA replication machinery, may explain the observation that there are a larger proportion of cells incorporating BrdU in the rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 eye discs (Fig. 7D,E). In wild-type eye imaginal discs, cell proliferation and differentiation are coordinated (Wolff and Ready, 1993). During the third larval instar, the morphogenetic furrow initiates at the posterior end of the eye disc and moves toward anterior. Cells anterior to the morphogenetic furrow are dividing asynchronously, which provides a reservoir of cells for differentiation. In the morphogenetic furrow, cells are arrested in G 1 and began to differentiate. Neuronal differentiation initiates just posterior to the morphogenetic furrow. The differentiating photoreceptor cells can be stained by an antibody against Elav. As shown in Fig. 7F, at the end of the third larval instar of wild-type animals, there are about 25 rows of the photoreceptor cells that are stained by the anti- Elav antibody posterior to the morphogenetic furrow, while there are still many undifferentiated cells anterior to the furrow. As described earlier, the development of rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 larvae is slower, so that they reach late third instar around day 10 AEL. Interestingly, in late third instar rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 eye discs, there are also about 25 rows of photoreceptor cells posterior to the furrow that stained with the Elav antibody (Fig. 7G), indicating that development and neuronal differentiation appear to be coordinated normally. However, there are very few cells left anterior to the morphogenetic furrow in rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 eye discs (compared the areas anterior to the furrow in Fig. 7F,G). No significant cell death was detected in rbf 120a /rbf 14 ;de2f1 rm729 /de2f1 91 eye discs (data not shown). Thus the significant decrease in the amount of cells anterior to the morphogenetic furrow is not due to an increase in cell death, but may result from the cell proliferation defect due to insufficient DNA replication machinery in the absence of de2f1. DISCUSSION In vitro binding and overexpression studies have identified a large number of potential targets of prb, including E2F transcription factors, C/EBP, myod, ATF-2, UBF, to list a few. It was not clear to what extend the function of prb is mediated through E2F. In addition, studies of the function of E2F transcription factors showed that E2F could function both as a transcription activator and as a co-repressor of prb. These studies raise the question of whether the main function of E2F is to activate transcription or to repress transcription during development. In Drosophila, there is one RB family homolog and two de2f homologs identified. The identification of several de2f1 and rbf mutant alleles allowed an analysis of the functional relationship between RBF and de2f1 to address some of these questions. The results presented in this paper show that most if not all of the requirements of RBF during development can be attributed to its interaction with E2F, and that de2f1 function primarily to activate transcription during development. Given that de2f1 is just one of many potential targets of RBF, the dramatic suppression of the rbf mutant phenotypes by de2f1 mutants is very unexpected. First, lowering the activity of de2f1 can suppress the early larval lethality of the rbf mutants as well as the developmental phenotypes observed in the adult eyes and bristles. Second, an allele of de2f1 with an intact DNA binding domain but with no transactivation domain or RBF binding domain can suppress the lethality of rbf null mutants, allowing the rbf 14 /Y;de2f1 i2 /de2f1 rm729 flies to develop into viable adults. Furthermore, these suppressed rbf null adults showed normal adult structures. Thus the uninhibited de2f1 in rbf mutants mediates both the lethality

10 376 W. Du as well as the observed eye and bristle phenotypes in adults. These observations provide strong evidence that E2F mediates most, if not all, of the phenotypes of rbf during development. Interestingly, lowering the activity of RBF can also partially suppress the de2f1 null phenotypes. There are at least two possible explanations for this observation. One possibility is that RBF has a function downstream of de2f1; the other possibility is that RBF can affect the de2f1 mutant phenotypes through a parallel pathway (for example RBF may be able to regulate the expression of E2F target genes through another target such as de2f2). The second explanation seems to be more likely for the following reasons. First, in the suppressed de2f1 null mutants, the expression of de2f target genes is not restored, and the larvae growth is still greatly retarded, suggesting that reducing the level of RBF bypassed rather than restored the function lost by de2f1 mutation. Second, the rbf null mutant phenotypes is fully suppressed by a de2f1 mutant that lacks transcription activation and RBF binding domain, demonstrating that RBF functions upstream of de2f1. In summary, these and other published results do not point to a role of RB downstream of E2F. In contrast, loss of RBF indeed caused deregulation of the expression of PCNA even in the absence of transcription activation by de2f1 (Fig. 4J-M), supporting the notion that RBF can regulate the expression of PCNA by targets other than de2f1. E2F transcription factors can function both to activate transcription and to repress transcription by recruiting RB family proteins to specific promoters. Although analysis of de2f1 mutants showed that de2f1 is required for the coordinated expression of replication functions such as PCNA and dmrnr2, it is not clear whether the lack of transcription of these set of genes is the cause of the larval lethality (Royzman et al., 1997). It is formally possible that the lethality of de2f1 mutant is caused by the failure to repress certain critical E2F target genes. As discussed earlier, depending on the function of de2f1 as a transcription activator or a co-repressor of RBF, completely different predications are expected regarding the genetic interaction between RBF and de2f1. The observation that lowering the level of RBF can suppress the larval lethality of de2f1 mutants and allow the de2f1 mutants to develop into pharate adults, suggests that during Drosophila development, the function of de2f1 is mainly to activate transcription and not to recruit RBF to repress transcription. In addition to de2f1, there is another E2F, de2f2, identified in Drosophila: (Sawado et al., 1998). Interestingly, de2f2 can bind to E2F binding sites, but the function of de2f2 appears to be distinct from that of de2f1. Cotransfection of de2f2 repressed the expression from the PCNA gene promoter while cotransfection of de2f1 activated the expression (Sawado et al., 1998). Similar findings were also observed in transfection experiments in which de2f1 strongly activates transcription, while de2f2 fails to activate a reporter with E2F binding sites (W. Du and N. Dyson, unpublished results). These results suggest that de2f2 may function mainly to repress transcription while de2f1 functions mainly to activate transcription. Taken together, these data suggest a model for the function of RBF, de2f1 and de2f2 (Fig. 8). In this model, de2f1 functions mainly to activate transcription of the E2F target genes. RBF negatively regulates the activity of de2f1 to inhibit the expression of E2F target genes. In addition, RBF can also repress the expression of E2F targets genes through other targets of RBF such as de2f2. Thus the expression of E2F target genes will have three different states: activated, when there are free de2f1/ddp to activate transcription; repressed, when there are de2f2/ddp/rbf (and possibly de2f1/ddp/rbf) to repress transcription; and basal, when there is neither activation or repression. At present, it is not clear whether de2f1 also has a function to repress transcription during development, nor is it is clear about the function of free de2f2/ddp. This model explains a number of the previous observations. First, embryos with mutations in de2f1 showed defects in the expression of S phase genes, even though these mutants have Fig. 7. Cell proliferation and differentiation in the absence of de2f1. (A) BrdU staining of a brain from wild-type third instar larva. Black arrows in A and C indicate cells in the optic lobe strongly incorporating BrdU. (B) BrdU staining of a brain from a de2f1 rm729 /de2f1 91 larva at 13 days AEL. Black arrow indicates cells in the optic lobe were not incorporating BrdU. In contrast, white arrows indicate cells in the thoracic region of the brain undergoing DNA replication. (C) BrdU staining of a brain from a rbf 120a /rbf 14 ; de2f1 rm729 /de2f1 91 late third instar larvae at 10 days AEL. (D,E), BrdU incorporation visualizes DNA replication in eye discs from wild-type (D) and from rbf 120a /rbf 14 ; de2f1 rm729 /de2f1 91 (E) flies. (F) Anti-Elav staining of a wild-type eye disc. (G) Anti-Elav staining of eye discs from rbf 120a /rbf 14 ; de2f1 rm729 /de2f1 91. Arrows in F and G indicate the morphogenetic furrow. Note that while there are about 25 rows of developing photoreceptor cells in both the wild-type and the rbf 120a /rbf 14 ; de2f1 rm729 /de2f1 91 eye disc, there are much fewer cells anterior to the morphogenetic furrow in the rbf 120a /rbf 14 ; de2f1 rm729 /de2f1 91 eye disc.