The enhancer of trithorax and polycomb gene Caf1/p55 is essential for cell survival and patterning in Drosophila development

The development of complex multicellular organisms requires the precise patterning of a wide range of appendages, organs, tissues and cell types. To a large extent, this task is accomplished by the reiterative use of a limited number of conserved pathways and proteins. However, we are only beginning to understand how pathways and proteins can be used repeatedly and still achieve developmental diversity. Chromatin remodeling is an important mechanism that allows cells to stably yet reversibly lock in repressed or active chromatin states, thereby restricting the fate of the cell during development (Kadonaga, 1998; Schulze and Wallrath, 2007; Vermaak et al., 2003).

Drosophila Chromatin Assembly Factor 1 (Caf1, also known as p55; RbAp48 – FlyBase) is a 55 kDa protein containing seven WD repeats and α-helical regions in the N and C termini, and binds directly to histone H4 (Song et al., 2008; Tyler et al., 1996). The Caf1 gene has homology to two human genes, retinoblastoma associated protein 46 (RbAp46; RBBP7 – Human Gene Nomenclature Database) and retinoblastoma associated protein 48 (RbAp48; RBBP4 – Human Gene Nomenclature Database) (Tyler et al., 1996; Verreault et al., 1996). Caf1 shares 87% and 84% amino acid identity with human RbAp48 and RbAp46, respectively, and is therefore likely to prove an excellent tool for studying the functional and developmental roles of the human proteins (Tyler et al., 1996). Drosophila Caf1, first identified as a component of the Chromatin Assembly Factor 1 complex (CAF-1) which acts in nucleosome assembly following DNA replication, is a component of many chromatin remodeling complexes (Tyler et al., 1996). Caf1 is also found as a component of the Nucleosome Remodeling Factor (NURF) and, like its human homologs RbAp46 and RbAp48, is a component of retinoblastoma (RB)-containing complexes (Korenjak et al., 2004; Martinez-Balbas et al., 1998; Qian and Lee, 1995; Qian et al., 1993; Taylor-Harding et al., 2004).

Caf1 is also a member of the Polycomb group (PcG) complex Polycomb Repressive Complex 2 (PRC2), along with Extra Sex Combs (ESC), Suppressor of Zeste 12 [SU(Z)12] and Enhancer of Zeste [E(Z)] (Muller et al., 2002; Tie et al., 2001). The role of PcG proteins has been best characterized in silencing of Hox genes, although PcG silencing also occurs at many non-Hox loci (Bello et al., 1998; Chen and Rasmuson-Lestander, 2009; Dura and Ingham, 1988; Netter et al., 1998; Pelegri and Lehmann, 1994; Schuettengruber et al., 2007). PRC2 silences genes by effecting histone modifications, particularly trimethylation of histone H3 at lysine 27 (Czermin et al., 2002) (for a review, see Schuettengruber et al., 2007). Beyond its important role in development, PRC2 has also been implicated in cancer pathways. For example, in mammals, aberrant epigenetic modification of PRC2 target genes is associated with colorectal cancer (Widschwendter et al., 2007).

With participation in such diverse complexes, Drosophila Caf1 is likely to have important roles in multiple aspects of development. Until now, mutant or reduced expression phenotypes of Caf1 homologs have been reported in yeast, Arabidopsis and C. elegans, but no mutations in Drosophila Caf1 have been identified and genetically characterized (Bouveret et al., 2006; Guitton and Berger, 2005; Hennig et al., 2003; Jullien et al., 2008; Lu and Horvitz, 1998; Ruggieri et al., 1989). Furthermore, reduced expression of the mammalian Caf1 homologs RbAp46 and RbAp48 is observed in human cancers, and in vitro studies suggest both genes may act as tumor suppressors, but the lack of Caf1 mouse models hinders our understanding of its potential role in human disease (Guan et al., 2001; Guan et al., 1998; Ishimaru et al., 2006; Kong et al., 2007; Li et al., 2003; Pacifico et al., 2007; Thakur et al., 2007).

Chromatin remodeling adds a unique level of transcriptional regulation to the signaling pathways that control development by stably repressing genes in a heritable manner, and may therefore be of particular importance to genes and pathways that are re-used in development. The Drosophila senseless (sens) gene is one example of a gene that is reiteratively employed. Specifically, Sens functions in a number of processes in both the embryo and larval imaginal tissues, including the compound eye. In the third instar eye disc, Sens is necessary for differentiation of the R8 photoreceptor, the founding cell of each ommatidium (Frankfort et al., 2001). Continued expression of Sens is necessary for maintenance of R8 fate during pupal development (Xie et al., 2007). Sens is also necessary for formation of interommatidial bristles (Domingos et al., 2004; Frankfort et al., 2004; Morey et al., 2008).

Recently, we performed a screen in Drosophila for dominant modifiers of a disorganized eye phenotype caused by overexpression of senseless (Pepple et al., 2007). Here, we report the identification of the complementation group S(ls)3, comprising three loss-of-function alleles, as the first mutations in Drosophila Caf1. In the current study, we describe functional and developmental consequences of Caf1 gain- and loss-of-function in Drosophila. Our observations indicate that severe loss of Caf1 expression results in cell death, consistent with key roles for Caf1 in chromatin remodeling complexes necessary for basic cellular functions. However, at intermediate levels of wild-type Caf1 expression, we observe homeotic transformations and eye phenotypes consistent with a major role for Caf1 in Polycomb silencing. We present evidence suggesting that the genetic interaction between Caf1 and sens is mediated by Polycomb silencing. These results suggest that Caf1 is essential not only for basic cellular functions and survival, but also for maintenance of cellular identity and normal body plan.

Caf1 alleles generated in our laboratory were reported previously as S(ls)3 (Pepple et al., 2007). The Caf1 Genomic Rescue construct (Caf1GR) was generated by recombineering a fragment from BACR32A03 into pACMAN (Venken et al., 2006). pACMAN vector was a gift from Hugo Bellen (Jan and Dan Duncan Neurological Research Institute, Houston, TX, USA). The UAS-Caf1 construct was generated by insertion of the Caf1 cDNA (LD33761, Drosophila Genomics Resource Center) into pUAST-attB (Bischof et al., 2007). Transgenic flies were obtained by injection of Caf1GR or UAS-Caf1 into the second chromosome insertion site of the VK1 line (Venken et al., 2006). Other fly stocks used include w¹¹⁸, lz-GAL4, UAS-GFP and UAS-sens/FM7 (ls). eyFLP; FRT82B p{w⁺} cl/TM6B, Dp^49FK-1 c¹/SM5, cn¹, w; Dp^a1bw1 sp¹/CyO, w; E2f2^76Q1 cn¹ bw¹/CyO, E2f⁰⁷¹⁷²/TM3, E(z)⁷³¹, Rpd3⁰⁴⁵⁵⁶ ry⁵⁰⁶/TM3 ry^RK Sb¹ Ser¹, esc¹, UAS-pb and UAS-Antp were obtained from the Bloomington Drosophila Stock Center. Caf1^short, Caf1^med and Caf1^long alleles were recombined onto the FRT82B chromosome. Mosaic heads were generated by crossing mutant alleles on FRT chromosomes with eyFLP; FRT82B p{w⁺} cl/TM6B, obtained from the Bloomington Drosophila Stock Center (Newsome et al., 2000).

Three rounds of P-element mapping were performed to localize S(ls)3 mutations to a 50 kb region as described (Zhai et al., 2003). Sequencing was performed by the MD Anderson DNA Analysis Facility and Macrogen. Sequence analysis was performed using Sequencher software (Genecodes).

Tangential sections of the adult retina were performed as described (Tomlinson and Ready, 1987). Images were acquired with a Zeiss Axioplan 2 microscope, Zeiss Axiocam digital camera and Axiovision software. Adobe Photoshop software was used to resize images and adjust brightness and contrast.

Samples were prepared as described previously (Pepple et al., 2007).

The following antibodies were used: rabbit anti-Caf1 (1:1000, AbCam); guinea-pig anti-Sens (1:5000, a gift from Hugo Bellen); mouse anti-active-Caspase 3 (1:1000, R&D Systems); rat anti-Elav (1:500, Developmental Studies Hybridoma Bank); and rabbit anti-H3K27me3 (1:200, Lake Placid). The following secondary antibodies were all used at 1:500: Alexa-conjugated goat anti-rabbit secondary antibody (Molecular Probes), goat anti-guinea pig Cy3 (Jackson ImmunoResearch), goat anti-rat Cy3 (Jackson ImmunoResearch) and goat anti-rat Cy5 (Jackson ImmunoResearch). For histone methylation staining experiments, discs were dissected and stained as described previously (Fan and Bergmann, 2010). For all other staining experiments, eye-antennal imaginal discs were dissected and stained, and images acquired, as described previously (Pepple et al., 2007). Adobe Photoshop and Gimp software were used to process image brightness, color, contrast and noise, and to merge channels.

Overexpression of UAS-sens under control of lozenge-GAL4 (hereafter abbreviated ls) results in eyes with a disruption of the regular hexagonal array of ommatidia in the adult retina. This disruption is caused, in part, by formation of extra interommatidial bristles (Pepple et al., 2007). Previously, we have reported the isolation of a three-member lethal complementation group, S(ls)3, that was discovered in a screen for mutations that dominantly modify ls (Pepple et al., 2007). Each of the three alleles of S(ls)3 dominantly suppresses the extra bristles and disrupted ommatidial array of ls to a similar degree (Fig. 1B-I), indicating a role in the Sens pathway. We performed P-element recombination mapping (Zhai et al., 2003) to narrow the lethal mutations to a 50 kb region containing 15 genes. Sequencing of all exons from this region revealed mutations in the gene Caf1 (CG4236) in all three alleles (Fig. 1A). Caf1 encodes a 430 amino acid protein with seven WD repeats (Song et al., 2008; Tyler et al., 1996). Caf1^long contains a G-to-A transition that changes a conserved amino acid from glycine to asparagine. The mutation in Caf1^med truncates the protein in its third WD repeat. The Caf1^short mutation, which would truncate the protein in the first WD repeat, occurs in the first exon and probably results in a null allele due to nonsense-mediated decay (Valencia-Sanchez and Maquat, 2004).

To test whether the lethality of the three mutants comprising S(ls)3 is due to mutations in Caf1, we generated flies containing a 12 kb genomic rescue construct, designated Caf1GR, designed to ensure all regulatory regions of Caf1 are included (not shown). This resulted in the inclusion of several other genes: Rpb7, CG31344, Art3, mRPS10 and the 3′ end of CG12241. The Caf1GR construct rescues the lethality of the most severe truncation allele, Caf1^short, in combination with either Caf1^med or Df(3R)ED5664, a deficiency that uncovers Caf1. Rescued flies are viable and appear phenotypically normal, indicating that the lethality of Caf1^short is due to mutations within the 12 kb region.

We attempted to rescue flies homozygous or trans-heterozygous for Caf1 alleles with expression of a UAS-Caf1 construct under the control of Ubiquitin-GAL4 (Ubi-GAL4). In a wild-type background, Ubi-GAL4/UAS-Caf1 flies are viable and phenotypically normal. Ubi-GAL4/UAS-Caf1; Caf1^short/Caf1^long flies are also viable and are phenotypically normal, suggesting that lethality of the mutant alleles is due to mutations in Caf1 (data not shown). Both Ubi-GAL4/UAS-Caf1; Caf1^short/Caf1^med and Ubi-GAL4/UAS-Caf1; Caf1^short/Caf1^short flies raised at 25°C survive to become pharates, though few adults eclose. Pharates dissected live from pupae and the few adults that do eclose appear phenotypically normal (data not shown). Given the differential ability of UAS-Caf1 to rescue different allelic combinations at 25°C, we reasoned that raising Ubi-GAL4/UAS-Caf1; Caf1^short/Caf1^med flies at 18°C would allow us to analyze the effects of reduced levels of Caf1. All flies of this genotype raised at 18°C have disorganized kidney-shaped eyes, as well as homeotic transformations of the aristae to leg-like structures that occasionally terminate in claws. Other defects, including duplications of scutellar bristles and wing defects, are sometimes seen.

We performed a series of temperature shift experiments and determined that the eye and antennal phenotypes are dependent on the temperature during the third larval instar (data not shown). Ubi-GAL4/UAS-Caf1; Caf1^short/Caf1^med flies raised at 25°C during third instar have normal eyes and antennae (Fig. 2A,B,E). By contrast, flies raised at 18°C have small, disorganized eyes (Fig. 2C,D,F). In Fig. 2D, the transformed arista terminates in a claw (arrow). Tangential sections of adult eyes from animals raised at 18°C display abnormalities in the arrangement of rhabdomeres, including loss of small central rhabdomeres (Fig. 2G, arrowheads). These defects are probably due to insufficient levels of Caf1 given the temperature-sensitive nature of the GAL4-UAS system (for a review, see Duffy, 2002) and are reminiscent of phenotypes caused by overexpression of Hox genes (Bello et al., 1998). Hox genes are major targets of the PcG complexes, which are responsible for establishing and maintaining silencing of Hox genes in regions where their expression would interfere with proper anterior-posterior patterning (Schuettengruber et al., 2007). In addition to eye and antennal defects, wing defects are seen in many animals raised at 18°C during larval development; an example of a fly with reduced wings and scutellar bristle duplication is shown in Fig. S1 in the supplementary material.

Larvae homozygous for Caf1^med or Caf1^short appear grossly normal but mature slowly and survive to first or second instar. Many Caf1^long larvae survive to third instar, but die shortly after forming a disorganized pupa. Homozygous Caf1 germline mutant clones are not recovered. To assess the role of Caf1 in Drosophila development and overcome the problem of larval lethality, we attempted to make clones using heat shock-Flippase (hs-FLP). Caf1^long clones survive well in the Drosophila eye (data not shown). However, hs-FLP clones of Caf1^short and Caf1^med do not survive in adults, suggesting cell death or proliferation defects. To recover clones of Caf1^short and Caf1^med, we used the ey-FLP;FRT cell-lethal system, which generates eyes composed mostly of homozygous mutant tissue and has the advantage of marking any remaining heterozygous ommatidia with w⁺ (Newsome et al., 2000). We used two approaches to analyze mutant eyes: light microscopy (LM), which reveals the gross structure of the eye and can differentiate between homozygous and heterozygous tissue but does not reveal the fine details of ommatidial architecture; and scanning electron microscopy (SEM), which shows the detailed structure of the eye but does not allow identification of clones.

Eyes mutant for Caf1^long have large clones of mutant tissue with only a subtle disruption in the ommatidial array, leading to a slightly disorganized appearance (Fig. 3B,F,J). Misplaced bristles are occasionally seen in these eyes and many bristles are shorter than wild type. By contrast, Caf1^short eyes contain very little mutant tissue (white tissue in Fig. 3K). The remaining eye, composed mostly of heterozygous tissue, is small, has irregular ommatidial structure, and is almost completely devoid of bristles (Fig. 3C,G,K). This suggests impaired cell viability or proliferation in Caf1^short homozygous tissue. Consistent with these results, no adult Caf1^short clones are recovered in the thorax using Ubx-FLP; FRT 82B Sb⁶³ P{w⁺y⁺}, whereas large wild-type or Caf1^long clones are recovered in most animals of the correct genotype with this technique (data not shown). Approximately 20% of flies with ey-FLP-induced Caf1^short or Caf1^med clones have extra or missing appendages on the head, including the ocelli, antennae and maxillary palps. In some cases, structures are only partially duplicated, leading to a branched appearance. An example of a fly with a duplicated antenna is shown in Fig. 3D and enlarged in Fig. 3H. Although the original publication of the eyFLP; FRT p{w⁺} cl chromosome described the use of an eye-specific version of ey-GAL4, with our stock, we have observed clones outside of the eye proper, including the peripodial epithelium and the antennal disc (Newsome et al., 2000). Therefore, although adult clones outside the eye field are not marked, these duplicated structures are probably due to Caf1 clones outside the eye field.

To assess the effects of varying levels of Caf1 expression, we generated an ey-GAL4,UAS-Caf1 recombinant chromosome to overexpress Caf1 in the eye antennal disc. These flies are viable when raised at 18°C, 25°C and even 30°C. The eyes of ey-GAL4/UAS-Caf1 flies raised at 25°C are disorganized and slightly smaller than wild type (Fig. 3M), and resemble the Caf1^long phenotype (Fig. 3J). Additionally, the eyes of these flies have reduced numbers of interommatidial bristles (Fig. 3N). When raised at 30°C, the eyes are smaller and more disorganized, probably owing to the increased activity of GAL4 at higher temperatures in Drosophila (Duffy, 2002). Eyes and heads of ey-GAL4, UAS-Caf1 homozygotes raised at 25°C have an even more severe phenotype reminiscent of Caf1 complete loss-of-function: the eyes are extremely small and ommatidial structure is highly disrupted. Occasionally, ectopic structures are seen in these eyes (Fig. 3O,P). Outside the eye, duplications and deletions of other head structures are seen frequently in ey-GAL4, UAS-Caf1 homozygotes, similar to Caf1^short clones (data not shown).

Even when using the ey-FLP; FRT cl technique, Caf1^short clones are small during third instar, and the spacing of developing R8 photoreceptors is disrupted (see Fig. S2 in the supplementary material). Caf1^short clones generated with this technique are rare at 6 hours of pupal development and almost completely absent by 24 hours (data not shown). To determine whether the small size, loss of mutant tissue and disorganization of Caf1^short eyes is due in part to apoptosis, we stained third instar eye discs with an antibody against active Caspase 3 (Fan and Bergmann, 2010). Normally, programmed cell death occurs in the Drosophila eye disc during pupal development to remove excess cells during differentiation of secondary and tertiary pigment cells, and formation of the hexagonal lattice (Frohlich, 2001). At late third instar, control eye discs using the eyFLP; FRT 82B P{w+} cl chromosome show very little Caspase 3 staining (Fig. 4A,C). Consistent with their large adult size, third instar ey-FLP; FRT P{w⁺} cl/FRT Caf1^long eye discs also show very low levels of active Caspase 3 (Fig. 4D,F). Elav staining in these eye discs shows a nearly regular array of developing photoreceptors, similar to that of control discs (compare Fig. 4E with 4B). By contrast, eye discs with homozygous clones for Caf1^short are small and have extensive Caspase 3 staining (Fig. 4G,I). Notably, the majority of Caspase 3 staining in discs harboring Caf1^short clones is observed anterior to Elav staining, implying that most cell death occurs anterior to the furrow in cells that would normally be proliferating. The arrangement of Elav-positive photoreceptor clusters is also irregular (Fig. 4H). Similar to ey-FLP; FRT82B Caf1^short/FRT82B P{w⁺} cl eye discs, significant apoptosis marked by Caspase 3 staining occurs in the anterior region of the eye disc in ey-GAL4, UAS-Caf1 third instar eye discs (Fig. 4J-L). Increased apoptosis is also observed in the Ubx-FLP; FRT82B Caf1^short/FRT82B ubi-GFP wing disc (see Fig. S3 in the supplementary material).

The homeotic transformations observed in Caf1 mutant flies partially rescued by Caf1 cDNA expression (Fig. 2) suggest that PRC2 function may be impaired when levels of Caf1 are low, resulting in ectopic expression of one or more Hox genes. To test whether overexpression of Hox genes could account for dominant suppression of the ls phenotype by mutations in Caf1, we crossed ls flies to UAS lines for several Hox genes, including proboscipedia (pb), abdominal-A (abd-A), Deformed (Dfd), Ultrabithorax (Ubx), Sex combs reduced (Scr), Antennapedia (Antp) and labial (lab). Overexpression of Hox genes alone in the eye also leads to a disorganized eye phenotype; each of the above UAS lines, when crossed to lz-GAL4 alone, also generates disorganized eyes (data not shown). Therefore, suppression of ls can only occur if levels of Hox and Sens activity functionally cancel each other out. Consistent with our hypothesis, ectopic expression from two lines, UAS-Antp and UAS-pb, results in strong suppression of the ls disorganized eye phenotype (Fig. 5B,C,F,G) compared with ls crossed to UAS-GFP (Fig. 5A,E).

As Caf1 is a component of PRC2, we hypothesize that suppression of ls by Caf1 mutations is due to loss of PRC2 function and predict that mutations in other PcG genes should also dominantly suppress the ls phenotype. To test this prediction, we crossed ls flies to mutations in PRC1 and PRC2 genes. The PRC2 complex mutations Su(z)12², Su(z)12³, Su(z)12⁴, Rpd3³⁰³, Rpd3⁰⁴⁵⁵⁶, E(z)⁷³¹, esc¹ and esc² each suppressed the ls phenotype, as did PRC1 mutations Polycomb³ (Pc³) and Pc¹⁵ (Fig. 5D,H; data not shown). We also tested mutations in genes encoding factors from other Caf1-containing complexes. Mutations in three members of the dREAM complex were suppressors of ls (see Fig. S4 in the supplementary material), consistent with previous observations of cooperation between PcG and dREAM complex members (Dahiya et al., 2001; Kotake et al., 2007; Tonini et al., 2004).

We tested for a genetic interaction with the PRC1 gene Pc (Fig. 5I). Males heterozygous for the Pc¹⁵ allele display ectopic sex combs on the second and sometimes third legs, consistent with homeotic transformation of second or third legs to first leg. Crossed to wild type, we observe that an average of 82% of male Pc¹⁵/+ flies have an ectopic sex comb on at least one posterior leg. By contrast, only 12.3% of Pc¹⁵/Caf1^short and 30% of Pc¹⁵/Caf1^long double heterozygote male flies have one or more ectopic sex combs. Thus, Caf1 is a strong dominant suppressor of the Pc¹⁵ phenotype. Although this result seems consistent with a Trithorax Group-like activity for Caf1, mutations in other bona fide PcG members sometimes interact genetically in a similar manner; these data suggest that like these PcG members, Caf1 can be placed in the Enhancer of Trithorax and Polycomb (ETP) group (for a review, see Fedorova et al., 2009). Furthermore, ETP-like activity of Caf1 is not surprising, given its aforementioned participation in multiple chromatin remodeling complexes. To confirm that the interaction is specific to Pc, we repeated this analysis with the Pc³ allele, which also results in ectopic sex combs on second and sometimes third legs. Eighty-five percent of Pc³/+ males exhibit ectopic sex combs, compared with 64% of Pc³/Caf1^long flies (P<0.06; 170 Pc³/+ and 88 Pc³/Caf1^P9 flies counted over six replicates). There was no significant difference between Pc³/+ and Pc³/Caf1^long. Thus, although mutations in Caf1 suppress Pc³ less strongly than Pc¹⁵, the genetic interaction is consistent between the two alleles.

The PRC2 complex is associated with trimethylation of histone 3 at lysine 27 (H3K27me3) (Czermin et al., 2002) due to the methyltransferase activity of the E(z) protein. The H3K27me3 mark is associated with repression of gene transcription. Our previous results suggest that loss of Caf1 results in a reduction in PRC2 activity, and therefore predicts a reduction in global levels of H3K27me3. To test this, we generated Caf1^short and Caf1^long clones in the eye disc using ey-FLP and stained third instar discs with an antibody against H3K27me3 (Fig. 6). The H3K27me3 mark is reduced in clones of both Caf1 alleles compared with surrounding heterozygous and wild-type tissue. A similar reduction of H3K27me3 in Caf1 clones is seen in the wing disc (see Fig. S5 in the supplementary material). Taken together, these results suggest that at least some of the phenotypes and genetic interactions we observe with partial or complete loss of Caf1 are due to loss of PRC2 activity and H3K27 trimethylation.

Several lines of evidence suggest that the participation of Caf1 in PcG complexes may account for many of the phenotypes we observe in flies with altered expression of Caf1. First, Caf1 loss-of-function clones in the eye have phenotypes ranging from slight disorganization and bristle defects (Fig. 3B,F,J) to almost complete loss of homozygous tissue in adults (Fig. 3C,G,K), and incomplete rescue of Caf1 results in adult eyes that are small and disorganized (Fig. 2C,F). Clones of many PcG genes have similar phenotypes in the eye. Loss of E(z) or Pc causes mild defects in differentiation in the third instar disc, but clones fail to survive in adults (Brook et al., 1996; Janody et al., 2004). An analogous situation occurs in Caf1^short clones, where expression of Elav, which marks differentiating neurons, is present in Caf1 clones at third instar but Caf1^short tissue is largely missing in the adult. Derepression of Hox genes could account for these phenotypes, as ectopic expression of many Hox genes in the eye field causes small disorganized eyes in adults (Plaza et al., 2001).

Second, flies with incomplete rescue of Caf1 display a range of homeotic phenotypes, notably transformation of arista to leg (Fig. 2C,D). Similar homeotic transformations, including antenna-to-leg transformations, are a hallmark of mutations in PcG genes (Denell, 1973; Lewis, 1978; Wang et al., 2006). We also observe a genetic interaction between Caf1 and the PRC1 gene Pc, as mutations in Caf1 are able to dominantly suppress the homeotic transformation of second or third leg to first leg in Pc¹⁵/+ males (Fig. 5).

Third, the disrupted patterning of Caf1^short mutant heads may also result from PcG dysfunction (Fig. 3). It is possible that these patterning defects are non-cell autonomous and may be an indirect result of widespread apoptosis in the eye disc (Fig. 4). Normally, when an imaginal disc is injured, remaining cells proliferate and assume correct identities, leading to a perfectly patterned adult structure (McClure and Schubiger, 2007) (for a review, see Bergmann and Steller, 2010). This type of regeneration requires that some determined cells must change their fates and involves substantial chromatin remodeling. Under specific circumstances, the disc can regenerate with incorrect patterning, leading to duplication, deletion or transformation of structures, a phenomenon referred to as transdetermination (McClure and Schubiger, 2007). Levels of many PcG transcripts are increased in transdetermining imaginal discs, and heterozygous mutations in PcG genes can enhance transdetermination in regenerating imaginal discs (Klebes et al., 2005). Therefore, one interpretation of the patterning defects in Caf1^short mutant discs is that under the stress of widespread apoptosis, the remaining heterozygous tissue is haploinsufficient for the chromatin remodeling activity required to properly regenerate and pattern the injured disc. Consistent with this interpretation, no extra or missing appendages are observed in flies with Caf1^long clones, which show less active Caspase 3 staining at third instar.

Finally, in Caf1 mutant tissue, we observe a reduction in levels of the H3K27me3 mark, which is associated with inactive chromatin and PRC2 activity (Fig. 6). These data are consistent with a disruption of PRC2 function as a result of loss of Caf1 and represent the first in vivo evidence that Caf1 is an essential member of this chromatin remodeling complex in an animal model.

One obvious question arises from the current study: why were multiple Caf1 alleles identified in a screen for modifiers of the sens overexpression phenotype? Moreover, it is surprising that mutations in Caf1 were not identified in previous Drosophila modifier screens involving PcG or Rb pathway members (Ambrus et al., 2009; Janody et al., 2004; Steele et al., 2009). We propose that the link between sens and Caf1 is due to the role of Caf1 in PcG-mediated silencing.

Recent evidence suggests that Sens and Hox proteins can compete for binding at overlapping sites at an enhancer of the rhomboid (rho) locus (Li-Kroeger et al., 2008). When the Hox protein Abdominal-A (Abd-A) binds, transcription of rho is activated, whereas binding by Sens leads to repression of rho. In the embryo, this mechanism acts as a molecular switch to allow differentiation of either chordotonal organs (under control of Sens) or hepatocyte-like cells called oenocytes (by the action of Abd-A). We propose that a similar mechanism underlies the suppression of the Sens overexpression phenotype (Fig. 7). We hypothesize that one or more targets of Sens in the eye contain similar overlapping sites that can be bound by either Sens or a Hox protein. During normal development, these loci are bound by neither Sens nor Hox in undifferentiated cells posterior to the furrow, as no Hox genes are known to be widely expressed in the eye field (Hueber and Lohmann, 2008). In the absence of both types of factors, these loci are transcriptionally active, and are necessary to ultimately attain the proper fates of the cells in which they are expressed. When Sens is overexpressed, as in ls, Sens binds to its recognition site in the downstream loci, repressing transcription. Repression of these genes initiates a cascade leading to a change in cell fate; for example, some of the cells that would normally become secondary or tertiary pigment cells now become bristle precursors, giving rise to the extra bristles of ls. However, when one copy of Caf1 is lost, a slight derepression of the Hox genes occurs due to loss of PcG activity. Hox proteins are now able to compete with Sens for the overlapping binding sites, tipping the balance towards activation of downstream genes and attainment of normal cell fate – effectively suppressing ls. The ability of ectopic expression of pb and Antp in the eye to suppress ls is consistent with this hypothesis. Suppression of ls by Hox proteins is particularly significant given that ectopic expression of Antp alone in the eye field leads to a small and disorganized eye (Bello et al., 1998; Plaza et al., 2008; Plaza et al., 2001). As Sens activity is exquisitely sensitive to Hox proteins, especially in the eye, our screen for modifiers of a sens overexpression phenotype was therefore ideal for identifying mutations in Caf1.

Previous studies have explored pro-apoptotic roles of Hox proteins and anti-apoptotic roles of Sens. It is therefore possible that one effect of Hox gene derepression in ls eyes suppressed by Caf1 may be restoration of an apoptotic fate in cells that would otherwise form bristle precursors due to ectopic Sens. Abd-A expression in the abdomen during normal third instar larval development leads to apoptosis of proliferating neuroblasts of the central nervous system, and ectopic expression of other Hox genes can also cause neuroblast apoptosis (Bello et al., 2003). Accordingly, survival of neuroblasts is dependent on PcG activity to repress Hox gene expression (Bello et al., 2003). Furthermore, expression of Sens is necessary in the Drosophila embryonic salivary gland to prevent apoptosis (Chandrasekaran and Beckendorf, 2003). Thus, one possible mechanism for suppression of ls by Caf1 mutations is that in the ls eye, Sens may promote the ectopic bristle fate partly by repressing apoptotic genes in cells normally fated to die, whereas in the ls eye suppressed by mutations in Caf1, ectopic Hox proteins may promote apoptosis and prevent bristle formation. We were unable to detect increased Ubx expression by antibody staining; however, a very small increase in one or more Hox proteins may be all that is necessary to change the transcriptional state of downstream loci and prevent the ectopic bristles and other defects in the highly sensitized ls eye – especially in the eye field, where no Hox genes are known to be highly expressed. Furthermore, the fact that multiple Hox proteins can recognize the same DNA binding site offers the possibility that the competitive effect of each Hox protein type on genes with overlapping Sens/Hox binding sites would be additive (Hueber and Lohmann, 2008). Therefore, although loss of one copy of Caf1 may only cause a small derepression of any one Hox gene, mild derepression of many Hox genes collectively can lead to strong repression of the ls phenotype.

Biochemical evidence suggests that Caf1 is a member of multiple complexes that effect gene regulation through chromatin remodeling, suggesting that it is a vital component of the cell's arsenal of chromatin modifying factors (Henikoff, 2003). Although our results suggest that disruption of PRC2 function may be the most important consequence of Caf1 gain- or loss-of-function, many phenotypes we observed in Caf1 mutant tissue are also reminiscent of mutations in members of other complexes previously shown to contain Caf1. It is not surprising that all three alleles of Caf1 in the current study are homozygous lethal, and that Caf1^short cells have poor viability, considering that Caf1 has been found in the NURF and CAF-1 complexes, which have fundamental roles in nucleosome assembly and spacing (Bulger et al., 1995; Kamakaka et al., 1996; Martinez-Balbas et al., 1998; Tyler et al., 1996; Verreault et al., 1996). The apoptosis we observe in eyes with Caf1^short clones is also consistent with a role for Caf1 in the dREAM (Drosophila Rbf, E2F2, and Myb-interacting proteins) complex. Members of the E2f family of transcription factors can complex with Dp proteins and bind short recognition sites to activate transcription (Brehm and Kouzarides, 1999; Classon and Harlow, 2002; Korenjak and Brehm, 2006; Macaluso et al., 2006). When Rb binds the E2f-Dp complex, transcription is repressed. Like Caf1 homozygotes, homozygous rbf1 null flies die in early larval development (Du and Dyson, 1999). Fully rbf1-deficient embryos display increased apoptosis, a phenotype reminiscent of the increased active Caspase-3 staining seen anterior to the morphogenetic furrow in eye discs with Caf1^short clones (Fig. 4).

The mammalian homologs of sens, Growth Factor Independence 1 (Gfi1) and Gfi1b are essential to the development of multiple cell types and have been implicated as oncogenes (Duan et al., 2005; Gilks et al., 1993; Gilks et al., 1995; Grimes et al., 1996; Hochberg et al., 2008; Karsunky et al., 2002; Kazanjian et al., 2004; Liao et al., 1995; Person et al., 2003; Schmidt et al., 1998; Wallis et al., 2003; Yucel et al., 2003) (for reviews, see Duan and Horwitz, 2003; Hock and Orkin, 2006). Therefore, the possibility that Caf1 links Sens with the activity of PcG complexes through parallel, competing pathways has implications for both Drosophila development and the activity of Gfi1 family members in human development and disease, and warrants additional study beyond the scope of the present work. Our results underscore the importance of Caf1 to diverse processes, including cell survival and tissue identity, and highlight the participation of Caf1 in multiple chromatin remodeling complexes. Further studies are needed to fully assess the importance of Caf1 in Drosophila development, as well as its developmental role in other chromatin remodeling complexes.

We thank Mardelle Atkins for helpful discussions and reading of the manuscript, Georg Halder for expertise with sequencing, Richard Atkinson and Jodie Polan for advice on confocal microscopy, and Kenneth Dunner (Cancer Center Core Grant CA16672) for assistance with scanning electron microscopy. This work was supported by the Retina Research Foundation(G.M.), and National Institutes of Health grants R01 EY011232 (G.M.), R01 GM068016 (A.B.), R01 GM074977 (A.B.), R01 GM081543 (A.B.), T32 CA009299 (A.E.A.) and T32 EY007102 (K.L.P.). Deposited in PMC for release after 12 months.