The Nucleosome Remodeling Factor ISWI Functionally Interacts With an Evolutionarily Conserved Network of Cellular Factors

Abstract

ISWI is an evolutionarily conserved ATP-dependent chromatin remodeling factor playing central roles in DNA replication, RNA transcription, and chromosome organization. The variety of biological functions dependent on ISWI suggests that its activity could be highly regulated. Our group has previously isolated and characterized new cellular activities that positively regulate ISWI in Drosophila melanogaster. To identify factors that antagonize ISWI activity we developed a novel in vivo eye-based assay to screen for genetic suppressors of ISWI. Our screen revealed that ISWI interacts with an evolutionarily conserved network of cellular and nuclear factors that escaped previous genetic and biochemical analyses.

THE eukaryotic cell has evolved regulatory mechanisms to induce structural changes to chromatin in response to environmental and cellular stimuli. Chromatin covalent modifiers catalyze specific post-translational modifications of the histones' amino- and carboxy-terminal tails (Kouzarides 2007), while chromatin remodeling complexes use the energy of ATP hydrolysis to change nucleosome positions or to incorporate histone variants into chromatin (Eberharter and Becker 2004; Saha et al. 2006). These chromatin modifications, occurring without a change in DNA sequence, set different chromatin functional states and constitute the epigenetic marks of our genome (Imhof 2006; Martin and Zhang 2007; Sala and Corona 2009). Although it is expected that a cross talk should exist between ATP-dependent remodelers and covalent modifiers of chromatin, little is known about how these activities are coordinated and integrated with each other in complex chromatin signaling pathways (Strahl and Allis 2000; Jenuwein and Allis 2001).

ISWI is the catalytic subunit of several ATP-dependent chromatin remodeling complexes, highly conserved during evolution and essential for cell viability (Corona and Tamkun 2004). ISWI-containing complexes play central roles in DNA replication, gene expression, and chromosome organization (Dirscherl and Krebs 2004). ISWI uses the energy of ATP hydrolysis to catalyze nucleosome spacing and sliding reactions (Corona and Tamkun 2004). In Drosophila, loss of ISWI function causes global transcription defects and leads to dramatic alterations in higher-order chromatin structure, including the apparent decondensation of both mitotic and interphase chromosomes (Deuring et al. 2000; Corona et al. 2007). Recent findings indicate that ISWI controls chromosome compaction in vivo, in part through its ability to promote the association of the linker histone H1 with chromatin (Corona et al. 2007; Siriaco et al. 2009).

In vitro and in vivo studies carried out in several model organisms have also shown the involvement of ISWI-containing complexes in a variety of nuclear functions including telomere silencing, stem cell self-renewal, neural morphogenesis, and epigenetic reprogramming occurring during nuclear transfer in animal cloning (Dirscherl and Krebs 2004; Xi and Xie 2005; Parrish et al. 2006). Inactivation of ISWI also interferes with the ras pathway (Andersen et al. 2006), and loss of ISWI function seems to be associated with a subset of melanotic tumors and the human multisystemic disease Williams–Beuren syndrome (Mellor 2006).

The variety of functions associated with ISWI is probably connected to the ability of other cellular and nuclear factors to regulate its ATP-dependent chromatin remodeling activity (Corona et al. 2002; Ferreira et al. 2007; Hogan and Varga-Weisz 2007). Due to the broad spectrum of functions played by ISWI in vivo, it is likely that chromatin factors, nuclear enzymatic activities, and a variety of histone modifications may influence its activity in vivo. To identify new regulators of ISWI function, we developed an eye-based assay to identify ISWI genetic interactors in the higher eukaryote Drosophila melanogaster (Corona et al. 2004). Loss of ISWI function, by eye-specific misexpression of the dominant negative allele ISWI^K159R, produces catalytically inactive ISWI that is incorporated into native complexes, giving rise to rough and reduced eye phenotypes in otherwise healthy flies (Deuring et al. 2000; Burgio et al. 2008). In a previous study, we used this in vivo eye assay to conduct an unbiased genetic screen for mutations in genes that dominantly modify phenotypes resulting from the misexpression of ISWI^K159R in the eye (Burgio et al. 2008).

The screen produced the first genetic interaction map for the ATP-dependent chromatin remodeler ISWI in a higher eukaryote (Burgio et al. 2008). The characterization of the network of factors we isolated revealed unanticipated roles for ISWI in the cell as well as novel mechanisms by which its activity could be regulated (Burgio et al. 2008). Interestingly, one class of mutants isolated in the ISWI^K159R screen included chromatin components and nuclear enzymatic activities that could regulate ISWI function by covalently modifying chromatin factors or ISWI itself (Burgio et al. 2008; Sala et al. 2008). The biochemical characterization of the genetic interactions recovered between ISWI^K159R and genes encoding for chromatin covalent modifiers, established that eye-based genetic screens in flies could be a powerful tool for the in vivo dissection of chromatin-remodeling signaling pathways occurring in the nucleus (Corona et al. 2004; Armstrong et al. 2005; Burgio et al. 2008; Sala and Corona 2009). Moreover, the ISWI^K159R screen established that ISWI function could be modulated in vivo by a variety of cellular factors that have escaped previous biochemical analyses (Burgio et al. 2008).

The nonsaturating ISWI^K159R F₁ screen was designed to specifically isolate enhancers of ISWI (Burgio et al. 2008). To identify novel factors working in antagonism with ISWI, we developed a new in vivo assay that allowed us to screen for genetic suppressors of eye phenotypes caused by true loss-of-function ISWI alleles. We took advantage of the Ey-Gal4, UAS-Flip (EGUF) approach to produce flies with eyes composed exclusively of mitotic clones that have lost ISWI function (Stowers and Schwarz 1999). Loss of ISWI in the eye caused reduced rough eyes, eye color variegation, and loss of cell identity. We employed the ISWI-EGUF eye phenotypes to set up a dominant modifier screen to isolate factors antagonizing ISWI activity in vivo. Employing classic gene network bioinformatics analysis, we combined the results of our screen with those obtained in two others screens conducted in Drosophila and in Caenorhabditis elegans, where an ISWI allele and its worm ortholog were isolated (Andersen et al. 2006; Parrish et al. 2006). The combination of genetic and bioinformatics approaches employed resulted in the identification of an evolutionarily conserved network of modifiers of ISWI eye phenotypes, which included several potential antagonists of ISWI function. Our analysis revealed new roles for ISWI in cell cycle progression as well as unanticipated mechanisms by which its activity could be regulated, shedding new light into the evolutionarily conserved physiological function of ISWI family members in cell cycle regulation.

MATERIALS AND METHODS

Drosophila stocks and genetic crosses:

Flies were raised at 25° on K12 medium (Genovese and Corona 2007). Unless otherwise stated, strains were obtained from Bloomington, Szeged, and Drosophila Genetic Resource Center Stock Centers and are described in FlyBase (http://www.flybase.org).

ISWI-EGUF eye characterization:

The EGUF approach (Stowers and Schwarz 1999) was employed to obtain, by mitotic recombination, flies with eyes entirely composed of homozygous wild-type, ISWI², brm², and kis¹ clones (ISWI-EGUF). The control wild-type EGUF adults FRT42D, +/FRT42D, GMR-hid; EGUF/+ were obtained by crossing yw; P{neoFRT}42D males with yw; P{neoFRT}42D,GMR-hid2R,CL2R,y⁺/CyO; EGUF virgins. ISWI-EGUF adults FRT42D, ISWI²/FRT42D, GMR-hid; EGUF/+ were obtained by crossing yw; P{neoFRT}42D,ISWI²,sp/SM5,Cy,sp males with yw; P{neoFRT}42D,GMR-hid2R,CL2R,y⁺/CyO; EGUF virgins. ISWI-EGUF adults bearing a copy of wild-type ISWI (Deuring et al. 2000), FRT42D, ISWI²/FRT42D, GMR-hid; ISWI⁺/EGUF were obtained by crossing yw; P{neoFRT}42D,ISWI²,sp/+;P{w⁺,ISWI⁺}/+ males with yw; P{neoFRT}42D,GMR-hid2R,CL2R,y⁺/CyO;EGUF virgins. ISWI-EGUF adults bearing a copy of the enzymatic inactive ISWI^K159R (Deuring et al. 2000), FRT42D, ISWI²/FRT42D, GMR-hid; ISWI^K159R/EGUF were obtained by crossing yw;P{neoFRT}42D,ISWI²,sp/+;P{w⁺,ISWI^K159R}/+ males with yw; P{neoFRT}42D,GMR-hid2R,CL2R,y⁺/CyO;EGUF virgins. brm-EGUF adults EGUF/+; FRT79B, brm²/FRT79B, GMR-hid were obtained by crossing yw; P{neoFRT}79B,brm²/TM6,Hu,Tb males with yw; EGUF/CyO; P{neoFRT}79B,GMR-hid3R,CL3R,y⁺/TM3,Sb virgins. kis-EGUF adults FRT40A, kis¹/FRT40A, GMR-hid; +/EGUF were obtained by crossing yw; P{neoFRT}40A,kis¹/CyO males with yw; P{neoFRT}40A,GMR-hid2L,CL2L,y⁺/CyO; EGUF virgins. ISWI-EGUF adults bearing a copy of the acf1[1] allele (Fyodorov et al. 2004), FRT42D, ISWI²/FRT42D, GMR-hid; acf1[1]/EGUF were obtained by crossing yw; P{neoFRT}42D,ISWI²,sp/+; acf1[1]/+ males with yw; P{neoFRT}42D,GMR-hid2R,CL2R,y⁺/CyO;EGUF virgins. ISWI-EGUF adults bearing a copy of the E(bx)^Nurf301-2 allele (Badenhorst et al. 2002), FRT42D, ISWI²/FRT42D, GMR-hid; E(bx)^Nurf301–2/EGUF were obtained by crossing yw; P{neoFRT}42D,ISWI²,sp/+; E(bx)^Nurf301-2/+ males with yw; P{neoFRT}42D,GMR-hid2R,CL2R,y⁺/CyO; EGUF virgins.

ISWI-EGUF PiggyBac genetic screen:

New PiggyBac insertions were generatad by crossing small batches of ∼20 pHer{3xP3-ECFP,αtub-piggyBacK10} jumpstarter males with ∼20 yw,pBac{3xP3-EYFP,p-tTA-K10}; P{neoFRT}42D,ISWI²,sp/CyO mutator bearing virgins. In the F₁, males yw,pBac{3xP3-EYFP,p-tTA-K10}; P{neoFRT}42D,ISWI²,sp/pHer{3xP3-ECFP,atub-piggyBacK10} carrying the ISWI² allele and both mutator and jumpstarter elements, recognizable by the double CFP/YFP eye fluorescence and the absence of the Cy marker, were crossed with isogenic w¹¹¹⁸ virgins. In the F₂, single males w¹¹¹⁸; P{neoFRT}42D,ISWI²,sp/pBac{3xP3-EYFP,p-tTA-K10} if the mutator hopped on the second chromosome or w¹¹¹⁸; P{neoFRT}42D,ISWI²,sp/+; pBac{3xP3-EYFP,p-tTA-K10}/+ } if the mutator hopped on the third chromosome, recognizable by the presence of the YFP and absence of the CFP eye fluorescence, were crossed with yw; P{neoFRT}42D,GMR-hid2R,CL2R,y⁺/CyO; EGUF virgins. In the F₃, males and females P{neoFRT}42D,GMR-hid2R,CL2R,y⁺/P{neoFRT}42D,ISWI²,pBac{3xP3-EYFP,p-tTA-K10},sp; EGUF/+ for second chromosome mutator insertions and P{neoFRT}42D,GMR-hid2R,CL2R,y⁺/P{neoFRT}42D,ISWI²,sp; EGUF/pBac{3xP3-EYFP,p-tTA-K10} for third chromosome mutator insertions were scored for their ability to dominantly modify (enhance or suppress) ISWI-EGUF eye phenotypes. Flies carrying only one FRT42D site linked with the GMR-hid transgene could be easily recognized because they cannot undergo eye-specific mitotic recombination, thus generating slightly pigmented adult eyes without ommatidia (Figure 2A) (Stowers and Schwarz 1999). Interacting mutator insertions were retested and balanced on both chromosomes. The PiggyBac mutator line pBac{3xP3-EYFP, p-tTA-K10} (Horn et al. 2003), containing the tetracycline transactivator gene (tTA) instead of the yeast GAL4 transactivator, was employed to avoid interference with the EGUF system.

Figure 2.—

ISWI-EGUF genetic screen strategy. (A) The use of distinguishable fluorescent transformation markers in the adult eyes allows us to follow the mutator (YFP) and the jumpstarter (CFP) elements independently. (B and C) Crossing scheme for the identification of new PiggyBac mutator (B) and third chromosome EP insertions (C) dominantly modifying ISWI-EGUF eye phenotypes. For both the PiggyBac mutator and the EP screen, in the absence of mitotic recombination, the presence of the GMR-hid transgene generates adult eyes without ommatidia but slightly pigmented (Stowers and Schwarz 1999), allowing their unambiguous distinction from the mitotic recombinant experimental class. The presence of w⁺-marked EP insertions allowed the scoring of the mitotic recombinant experimental class in the EP screen. Mut, mutator; Jump, jumpstarter.

ISWI-EGUF EP genetic screen:

The EP collection on the third chromosome (Rorth et al. 1998) was tested by crossing males carrying a balanced or unbalanced EP insertion with yw; P{neoFRT}42D,ISWI²,sp/CyO virgins. In the F₁, yw; P{neoFRT}42D,ISWI²,sp/+;EP/+ males, recognizable by the loss of the Cy and EP balancer markers, were crossed with yw; P{neoFRT}42D,GMR-hid2R,CL2R,y⁺/CyO; EGUF virgins. In the F₂, male and female yw; P{neoFRT}42D,ISWI²,sp/P{neoFRT}42D,GMR-hid2R,CL2R,y⁺; EGUF/EP were scored for their ability to dominantly modify (enhance or suppress) ISWI-EGUF eye phenotypes.

ISWI-EGUF genetic screen scoring system:

The ISWI-EGUF eye phenotypes are highly penetrant (100%; n = 1000) and show a very low expressivity (< ∼15% of progeny show phenotypes less or more severe than the standard ISWI-EGUF eye phenotypes; n = 1000). Nevertheless, both the PiggyBac and the EP ISWI-EGUF screens were conducted in F₃ and F₂, respectively (see Figure 2), to increase the chance to identify true ISWI-EGUF modifiers. A PiggyBac or an EP insertion was considered an Enhancer of ISWI [En(ISWI)] if >50% of the experimental class showed an eye with a more severe phenotype (reduced eye size and increased roughness), when compared to the control ISWI-EGUF eye. A PiggyBac or an EP insertion was considered a Suppressor of ISWI [Su(ISWI)] if >50% of the experimental class showed an eye with a less severe phenotype (increased eye size and reduced roughness), when compared to the control ISWI-EGUF eye. PiggyBac and EP insertions giving in the experimental class both En(ISWI) and Su(ISWI) were termed bimodal, Bi(ISWI).

ISWI-EGUF secondary screens:

To eliminate false ISWI-EGUF interactors, distinct secondary screens were conducted on the basis of the type of insertions (PiggyBac vs. EP transposons), the chromosome arm hit (2L, 2R, 3L, or 3R), and the type of interaction class scored (enhancer vs. suppressor). PiggyBac En(ISWI)'s mapping the 2L, 3L, and 3R chromosomes were tested for their inability to cause eye defects in the absence of an FRT42D, ISWI² chromosome (supporting information, File S1 and Figure S4). PiggyBac En(ISWI) and Su(ISWI) mapping the 2R chromosome, because of the presence of the FRT42D sites, produce eyes with the 2R chromosome in homozygosis. Therefore, false En(ISWI)'s were eliminated because they failed to enhance ISWI-EGUF eyes in the presence of an ectopic copy of the wild-type ISWI⁺ transgene (Figure S4). On the other hand, false Su(ISWI)'s were identified by their inability to suppress ISWI-EGUF eyes in the presence of an extra copy of the GMR-hid transgene not linked to the FRT42D recombination site (Figure S4). Finally, En(ISWI) and Su(ISWI) isolated with the EP screen were eliminated for their failure to enhance or suppress ISWI-EGUF eye phenotypes in the absence of the FRT42D, ISWI² chromosome (Figure S4).

iPCR and candidate allele analysis:

We mapped ISWI-EGUF modifiers to 99 potential protein-coding loci (Table S1), combining iPCR data available in FlyBase (www.flybase.org) for the EP interacting insertions with the iPCR sequencing data we generated for the PiggyBac interactors, using standard protocols (www.fruitfly.org/about/methods/inverse.pcr.html). Some ISWI-EGUF modifiers (i.e., mbf1, ttk, eff, stg, and cpo) were validated by testing other alleles of these genes, available from public stock centers, using the genetic screen scheme used for testing EP insertions. Mutations in genes corresponding to “neuronal morphogenesis,” “multiple cell fate,” and “connecting” nodes were obtained from public stock centers and tested for their ability to interact in the ISWI-EGUF eye assay, using the genetic screen scheme used for testing EP insertions, and in the ISWI^K159R eye assay according to Burgio et al. (2008).

Bioinformatic and cell cycle analyses:

For the BioGrid analysis (Breitkreutz et al. 2008), each En(ISWI), Su(ISWI), and Bi(ISWI) ISWI-EGUF interaction was associated with a single gene on the basis of the insertion DNA sequence data available in FlyBase (www.flybase.org) and the iPCR analysis we conducted on the PiggyBac insertions. The gene ontology data and all the genetic and physical interactions existing between the ISWI-EGUF modifiers were obtained from the BioGrid website (www.thebiogrid.org) and were represented in a graphical format using the Osprey software (Breitkreutz et al. 2003) (http://biodata.mshri.on.ca/osprey/servlet/Index). For Gene Ontology (GO) analysis the FatiGO data mining tool (Al-Shahrour et al. 2007) and the latest gene annotation and Gene Ontology provided by FlyBase were used. The C. elegans genes isolated in the multiple cell fate screening (Andersen et al. 2006) were converted to fly orthologs using the WormBase database (www.wormbase.org).

For the GO analysis the gene annotation and the Gene Ontology provided by FlyBase were used with Ontologizer (www.ohloh.net/p/ontologizer) to determine overrepresented GO terms. The parent–child method of Ontologizer, which takes into account the parent–child relationships of the GO hierarchy, was applied and the P-values were adjusted using Westfall–Young single-step multiple testing correction. A corrected P-value threshold of 0.1 was used as a cutoff for reporting significant matches. To compute statistical significance of the frequencies of GO-component terms hypergeometric distributions were calculated on the basis of the occurrence of the indicated terms and corrected for multiple testing by the Bonferroni correction method.

Cell cycle profiles from imaginal discs and brain cells were obtained according to Collesano and Corona (2007), using a BD FACSCanto flow cytometer (BD Biosciences), with a laser wavelength set at 488 nm (∼6000 events were recorded for each cell cycle profile). Cell cycle profiles were quantitatively analyzed using the ModFit software. Homozygous ISWI², wun^PBacF48 and ISWI², casp^PBacG75α brains and total imaginal discs used for the cell cycle analysis were dissected from third instar larvae, obtained from stocks balanced with the T(2:3),CyO, Tb balancer and selected for the absence of the Tb marker. Imaginal eye discs subjected to cell cycle analysis were obtained by crossing ISWI²; eyGAL4/T(2:3), CyO, Tb virgins with ISWI²; eyGAL4/T(2:3), CyO, Tb or ISWI²; eff^EP3627/T(2:3), CyO, Tb or ISWI²; ttk^EP3314/T(2:3), CyO, Tb or ISWI²; mbf1^EP3684/T(2:3), CyO, Tb males.

RESULTS AND DISCUSSION

An in vivo assay to identify factors antagonizing ISWI activity:

Suppressors of ISWI function are a very important class of genes to identify, as they may give us a complete picture of factors regulating chromatin remodeling reactions in vivo. We previously showed that the misexpression of ISWI^K159R in the eye-antennal discs caused defects that could be enhanced by null alleles of ISWI, indicating that ISWI^K159R eye phenotypes resulted from the specific reduction of ISWI function in the eye. However, due to the conditions used (cross temperature and choice of Gal4 driver), we generated mild eye phenotypes that probably prevented us from identifying suppressors of ISWI^K159R (Burgio et al. 2008). On the other hand, we could not formally exclude that the misexpression of ISWI^K159R might also have had some unspecific dominant effects that compromised eye development, preventing us from recovering suppressors of ISWI^K159R (Burgio et al. 2008).

Because of the limitations derived from the previous eye assay and the potential interfering effect of the dominant negative allele ISWI^K159R, we decided to develop a new in vivo eye assay. To isolate an antagonist of ISWI function we generated flies with eyes composed exclusively of clones that had lost ISWI activity using the EGUF approach (Stowers and Schwarz 1999). The EGUF method uses the eye-specific eyGAL4 driver in combination with the UAS-FLP transgene to express the site-specific recombinase FLP in mitotically active eye precursor cells. When homologous chromosomes containing FRT's recombination sites are present in these cells, FLP-mediated site-specific mitotic recombination occurs. Since one of the two FRT recombination sites is distally linked with the dominant photoreceptor cell lethal transgene GMR-hid, after mitotic recombination all photoreceptor cells bearing one or two copies of GMR-hid will die because of the eye-specific expression of the cell death gene hid during metamorphosis. Thus, this technique allows the generation of adult eyes that are homozygous for a specific mutation in an otherwise heterozygous adult fly (Stowers and Schwarz 1999).

Using the EGUF approach we generated flies with eyes homozygous for the ISWI² null allele (compare Figure 1A with 1B) (Deuring et al. 2000). Loss of ISWI in the eye by the EGUF approach caused reduced rough eyes, eye color variegation, and loss of cell identity (ISWI-EGUF eye phenotype) (compare Figure 1B). The ISWI-EGUF eye phenotypes are specifically caused by the mitotic recombination of the ISWI² allele occurring in the developing eye-antennal discs (Figure S1) and are characterized by defects in the retina structure consisting of the loss of ommatidia boundaries and orientation and a reduced number of photoreceptors (Figure 1B and Figure S2). Indeed, some photoreceptors appear to undergo a process of degeneration probably contributing to the observed ommatidia loss in the ISWI-EGUF adult eye (Figure S2).

Figure 1.—

Loss of ISWI function by the EGUF approach causes specific eye phenotypes. Drosophila adult eyes obtained with the EGUF mitotic clonal approach (Stowers and Schwarz 1999) bear in homozygosis a wild-type 2R chromosome (A), an ISWI² allele (B), an ISWI² allele in the presence of one extra copy of the wild-type ISWI⁺ gene (C), or a copy of ISWI^K159R defective in its ATPase activity (D), a brm² (E) or kis¹ allele (F), an ISWI² allele in the presence of one copy of acf1[1] (G), or E(bx)^Nurf301-2 (H) alleles. The white arrowhead indicates the eye color variegation, while the black arrowhead indicates loss of cell identity defects in which bristles grow in eye territories normally occupied by photoreceptors. The ISWI-EGUF eye phenotype is caused by the progressive depletion, during eye development, of the ISWI mRNA/protein pool present before mitotic recombination in ISWI² heterozygous mother cells. As a consequence, ISWI progressive loss of activity during eye development, occurring after mitotic recombination, could be accelerated by loss of factors positively regulating its activity, like acf1 and E(bx), thus explaining the enhancement of ISWI-EGUF eye defects we observe. On the other hand, the presence of an extra copy of the wild-type ISWI⁺ gene can complement eye defects caused by the recombination-dependent ISWI genetic loss.

The ISWI-EGUF eye phenotypes are very specific for the loss of ISWI activity since they can be fully suppressed by an ectopic copy of wild-type ISWI ⁺ (Deuring et al. 2000), under the control of its natural promoter (Figure 1C), but not by a copy of ISWI^K159R (Deuring et al. 2000) defective in its ATPase activity (Figure 1D). These data strongly suggest that the ISWI-EGUF eye defects are specifically caused by the loss of ISWI enzymatic activity and that ISWI-EGUF eye phenotypes could be in principle suppressed. Interestingly, we observed that loss of the brm chromatin remodeling factor, the Drosophila ortholog of the yeast SWI2/SNF2 protein, and kis, another ATP-dependent chromatin remodeling factor (Becker and Horz 2002), results in eye phenotypes distinct from ISWI-EGUF (compare Figure 1B with 1E and 1F), highlighting the specificity of the ISWI-EGUF phenotype. Remarkably, mutations in acf1 and E(bx), two genes encoding for known physical interactors of ISWI (Badenhorst et al. 2002; Corona and Tamkun 2004; Fyodorov et al. 2004), enhance ISWI-EGUF eye defects (compare Figure 1B with 1G and 1H), further indicating that the ISWI-EGUF phenotypes are specific for loss of ISWI activity and that they could also be used to recover dominant enhancers.

Previous work has shown that individuals that are homozygous for the ISWI² allele survive until late larval development, due to the high maternal contribution of ISWI (Corona et al. 2007; Burgio et al. 2008). Interestingly, loss of ISWI function in very late developing larvae (21 days) also caused global polytene chromosome condensation defects (Figure S3), highly reminiscent of chromosome condensation defects observed when misexpressing ISWI^K159R (Corona et al. 2007; Burgio et al. 2008). Therefore, we reasoned that similar chromatin organization defects in ISWI² null cells in the eye discs could also directly or indirectly contribute to the observed ISWI-EGUF eye defects, thus facilitating the genetic isolation of factors antagonizing ISWI chromatin remodeling activity.

A genetic screen for suppressors of ISWI:

We decided to exploit the ISWI-EGUF eye phenotypes as an in vivo assay to conduct a dominant modifier genetic screen to isolate factors suppressing ISWI activity. The rationale of this screen was that mutations in genes that dominantly modify ISWI-EGUF eye phenotypes are likely to encode for regulators or effectors of ISWI function in vivo. To circumvent the known limitations of P-element-based mutagenesis (i.e., preference to integrate into hotspots) we decided to use a PiggyBac-based insertional mutagenesis approach (Hacker et al. 2003). To identify and stably establish novel insertion lines without the need of balancers, we used two independent and distinguishable fluorescent markers to track a PiggyBac mutator line (marked with YFP) and a jumpstarter (marked with CFP), carrying a source of PiggyBac transposase (Figure 2A) (Hacker et al. 2003). Since each strain can be followed independently by fluorescence of adult fly eyes (Figure 2A), we did not use balancer chromosomes to track new insertions during the screening process. In particular, the dominant YFP fluorescence was used as a visible marker to identify novel insertions in both larval and adult stages, thus facilitating stock keeping. Furthermore, the PiggyBac transposon insertions also marked the mutated gene, thus helping us in mapping and cloning the mutations of interest.

To identify potential loss-of-function interactions suppressing ISWI-EGUF eye phenotypes, we screened ∼2000 newly generated insertions of the transposable mutator element PiggyBac (Figure 2, A and B) (Hacker et al. 2003). After crossing the jumpstarter and the mutator lines, flies carrying both elements as well as the FRT42D, ISWI² chromosome were identified on the basis of the eye-specific CFP and YFP fluorescence and the absence of the CyO balancer (Figure 2, A and B). In the next generation, the CFP-marked jumpstarter was crossed out to allow the stable inheritance of the YFP-marked autosomal insertion (Figure 2, A and B). Finally, males carrying second and third chromosome stable PiggyBac insertions were crossed with the EGUF line to score for ISWI interactions (Figure 2, A and B). In a second complementary approach, we screened potential gain-of-function interactions using the EP collection on the third chromosome that may lead to the GAL4-dependent overexpression of the gene downstream the insertion site (Rorth et al. 1998) (Figure 2C). All PiggyBac and EP interacting lines were retested and rescreened in specific secondary screens to eliminate false positives (Figure S4).

Combining iPCR data for the EP interacting insertions available on FlyBase with iPCR sequencing data we generated for the PiggyBac interactors, we mapped ISWI-EGUF modifiers to 99 potential protein-coding loci (Table S1, A and B). We found 21 PiggyBac (∼1% of total insertions screened) and 78 EP (∼8% of total insertions screened) interactors (Table S1, A and B). Our scoring strategy defined three different classes of modifiers of ISWI-EGUF eye phenotypes: enhancers, En(ISWI), suppressors, Su(ISWI), and few mutations giving a bimodal population comprising both enhancers and suppressors in the same progeny that we called Bi(ISWI) (Figure 3A and Table S1, A and B). One easy way to explain the bimodal type of interaction is that we scored two independent mutations, one enhancing and the other suppressing, coming from the same EP line tested. Indeed, for the bimodal interactions we isolated we confirmed that the presence of two independent mutations (probably arising from a second nonannotated segregating independent insertion or lesion) were likely responsible for the bimodal genetic interaction initially scored (Table S1 C).

Figure 3.—

ISWI genetically interacts with a wide range of cellular components. The 99 potential protein-coding loci corresponding to ISWI-EGUF dominant modifiers are clustered in concentric circles as nodes, colored according to their interaction class (A), current gene ontology (GO) categories, as indicated in the key (B), and their intersection with ISWI interactors isolated in the ISWI^K159R screen (Burgio et al. 2008) (C). The edges represent known physical and genetic interactions identified with the experimental system indicated in the key.

To understand the biological processes regulated by the ISWI-EGUF modifiers we isolated, we conducted a GO analysis employing the latest protein annotations available on FlyBase. Our GO analysis revealed that ISWI interacts with a network of cellular and nuclear factors involved in a variety of biological functions not previously linked with known ISWI activities (Figure 3B).

Furthermore, to gain an integrated view of the interaction network potentially existing between the genes isolated in our screen, we annotated the known genetic and physical interactions existing among the ISWI-EGUF modifiers with the help of the BioGrid database, containing the information of all known physical and genetic interactions reported for a variety of model organisms, including D. melanogaster (http://www.thebiogrid.org). Our analysis revealed that some of the ISWI-EGUF modifiers are known to interact with each other, forming small interaction networks, emphasizing the functional correlation existing between some of the factors isolated in our screen (Figure 3).

Finally we compared the list of ISWI-EGUF modifiers we recovered with the one we obtained with our previous ISWI^K159R eye assay (Burgio et al. 2008). This analysis revealed that the two screens resulted in the isolation of a number of common modifiers (∼25%; Figure 3C, Table S1 B, and Figure S5). However, the ISWI-EGUF screen allowed us also to isolate cellular and nuclear factors that have escaped from our previous ISWI^K159R genetic approach, including several Su(ISWI)'s (Figure 3C and Table S1 B).

ISWI interacts with factors antagonizing its activity in vivo:

One of the main goals of the ISWI-EGUF screen was to isolate factors encoding activities that could antagonize ISWI function in vivo. Our genetic screen resulted in the isolation of several modifiers of ISWI-EGUF eye phenotypes behaving as suppressors (Figure 3A and Table S1). As expected, some ISWI suppressors are associated with enzymatic activities regulating chromatin function like trx, E(Pc), the poly-ADP-ribose polymerase tankyrase, and the class I ubiquitin-conjugating (E2) enzyme effete (eff) (Figures 3A and 4A). Indeed, eff is thought to play essential roles in telomere function (Cenci et al. 2005). Eff biochemically interacts with the zinc-finger transcriptional repressor encoded by the tramtrack gene (ttk) (Badenhorst et al. 1996; Tang et al. 1997), whose mutations also suppress ISWI-EGUF eye phenotypes (Figures 3A and 4A). Another gene that suppresses the ISWI-EGUF eye is multiprotein bridging factor 1 (mbf1) (Figures 3A and 4A). Mbf1 is a highly conserved protein in archaea and eukaryotes (de Koning et al. 2009) that allows cells to maintain adequate activity of the cell proliferation transcription factor AP-1 under oxidative stress (Jindra et al. 2004). Interestingly, Mbf1 also enhances transcription by forming a bridge between distinct regulatory DNA-binding proteins and the TATA-box-binding protein (TBP) (de Koning et al. 2009). Moreover, Mbf1 is predicted to have a methyl-CpG binding domain (inferred from sequence similarity with UniProtKB: Q9Z2E1) that would in principle allow the selective binding of methylated cytosine/guanine DNA, potentially linking ISWI activity with DNA methylation in flies. To validate the specificity of some of the genetic interactions identified, we tested whether other alleles of ttk, eff, and mbf1 also behaved as Su(ISWI)'s. Indeed, some of the alleles tested for ttk, eff, and mbf1 suppress ISWI-EGUF eye phenotypes (Figure 4B).

Figure 4.—

Loss of ISWI function in the eye can be dominantly suppressed by mutations in mbf1, ttk, eff, and wun. (A) Eye phenotypes resulting from an eye homozygous for ISWI² (ISWI-EGUF eye) carrying an EP insertion mapping the mbf1, ttk, and eff genes or a PiggyBac insertion in the wun gene. (B) To validate the genetic interactions we scored, other alleles mapping mbf1, ttk, eff, and wun were tested in the ISWI-EGUF eye assay.

ISWI interacts with an evolutionarily conserved network of cellular factors:

In at least two other screens independently conducted in flies and worms ISWI was picked up as a genetic interactor. An RNA interference screen identified ISWI as one of the factors involved in the proper morphogenesis of Drosophila sensory neuron dendrites (Parrish et al. 2006). In another genetic screen for factors regulating the expression of vulval cell fates in C. elegans, mutations in ISW-1, the worm ortholog of fly ISWI, suppressed the multivulva phenotype caused by the hyperactivation of the Ras pathway (Andersen et al. 2006).

To gain some insights into the evolutionarily conserved network of regulation of ISWI we decided to identify known functional connections existing between genes scored in the four different ISWI-based genetic screens (ISWI^K159R, ISWI-EGUF, the fly neuronal morphogenesis, and the worm multiple cell fate screens). We used the BioGrid to search for known genetic or physical interactions existing between the genes identified in the fly neuronal morphogenesis and the worm multiple cell fate screens and the modifier genes we picked in both the ISWI-EGUF and the ISWI^K159R eye screens (Andersen et al. 2006; Parrish et al. 2006; Burgio et al. 2008). Our analysis identified 1 big (93 nodes) and 12 small networks comprising ISWI genetic interactors isolated in the four different genetic screens analyzed (Figure 5A, Figure S6, and Table S2, A–D). Next, to verify whether the 12 small networks (comprising a total of 33 nodes with a network size ranging from 2 to 5 nodes) were functionally related to each other, we again queried the BioGrid to search for factors bridging any of the genes belonging to the 12 networks. Interestingly, this analysis identified 41 connecting nodes linking the 33 genes included in the 12 small networks, defining a second big network (Figure 5B, Figure S6, and Table S2 E). The bioinformatic analysis allowed us to predict that ISWI genetically and physically interacted with a series of factors, in part shared among the four different genetic screens analyzed (purple, yellow, green, and blue nodes in Figure 5) and in part novel (red nodes in Figure 5), that may have escaped ISWI-based screens.

Figure 5.—

Evolutionarily conserved network of regulation of ISWI. BioGrid analysis is shown of known genetic and physical interactions existing between the genes identified in the fly “neuronal morphogenesis” and the worm “multiple cell fate” screens and the modifier genes we picked in both the ISWI-EGUF and the ISWI^K159R eye screens (Andersen et al. 2006; Parrish et al. 2006; Burgio et al. 2008). Our analysis predicted (A) 1 big and (B) 12 small interaction networks comprising ISWI genetic interactors isolated in the four different genetic screens analyzed, linked by “connecting” nodes corresponding to interactors not isolated in previous ISWI related screens. The big network comprises 93 nodes while the 12 small networks are constituted of 33 original nodes and 41 connecting new nodes. (C) Alleles of genes corresponding to 63% of the neuronal morphogenesis, 90% of the multiple cell fate, and 50% of the connecting nodes genetically interacted with the ISWI-EGUF or the ISWI^K159R eye assays. These interaction frequencies are much greater than the frequencies of ISWI interaction we normally get with eye-based screens (usually in the ∼1–11% range), suggesting that the gene network analysis we conducted increased our ability to predict ISWI interactors. The edges represent known physical and genetic interactions identified with the experimental system indicated in the key. ISWI-EGUF or ISWI^K159R interacting genes are highlighted in boldface type and have a bigger node size.

To verify whether the predicted interacting nodes were true ISWI genetic interactors, we tested multiple alleles for each of the neuronal morphogenesis and the worm multiple cell fate screens and the connecting nodes, for their ability to interact in the ISWI-EGUF or the ISWI^K159R eye assays. Remarkably, alleles of genes corresponding to 63% of the neuronal morphogenesis, 90% of the multiple cell fate, and 50% of the connecting nodes genetically interacted with at least one of the ISWI-EGUF or ISWI^K159R eye assays (Figure 5C and Table S3). Our data strongly indicate that the nucleosome remodeling factor ISWI functionally interacts with an evolutionarily conserved network of cellular factors, predicted by the gene network analysis we conducted.

Loss of ISWI causes cell cycle defects that can be suppressed by Su(ISWI)'s:

A GO analysis on the evolutionarily conserved network of ISWI interactors suggested a significant enrichment of regulators of cell cycle and signal transduction (i.e., Stg, Abl, Rbf1, E2F, and Ras85D) (Figure 5 and Table S2). In particular, an analysis conducted with Ontologizer (to quantify GO terms representation) showed that cell cycle regulation is an overrepresented category within the combined interacting nodes shown in Figure 5 (P-value = 0.03; Table S2). Loss of ISWI activity has been linked to different aspect of cell cycle regulation connected, for example, to the development of melanotic tumors (Mellor 2006), the regulation of the germline stem cell self-renewal, and the Rb pathway (Dirscherl and Krebs 2004; Xi and Xie 2005; Parrish et al. 2006). Moreover, our data show that loss of ISWI in the eye-antennal discs causes an eye phenotype characterized by patches of ommatidia that are missing, disorganized, dedifferentiated, or variegated (Figure 1B and Figure S2; Burgio et al. 2008), suggesting a possible role for ISWI in cell viability and the control of differentiation.

To directly test whether loss of ISWI causes cell cycle defects, we analyzed isolated cell populations from imaginal discs and larval brains, obtained ex vivo from wild-type (w¹¹¹⁸) and ISWI mutant larvae using a new method we recently developed (Collesano and Corona 2007). We decided to analyze imaginal discs because they contain highly proliferating cells and also because we conducted the two ISWI screens in eye-antennal imaginal discs. However, we also extended our analysis to brain tissues to test potential ISWI cell cycle defects also in highly differentiating cells (Collesano and Corona 2007). Isolated cell populations were directly analyzed by flow cytometry and cell cycle profiles were subjected to quantification (Figure S7) (Collesano and Corona 2007).

Our analysis revealed significant differences in cell cycle profiles between wild-type and ISWI mutant larvae. Loss of ISWI in total imaginal or in eye imaginal discs caused a marked decrease of both G₁ and G₂/M peaks as well as a dramatic increase of the pre-G₁ peak (red and purple arrows in Figure 6, B and C; Figure S7, E and G). We reasoned that the increase in the pre-G₁ peak reflects DNA fragmentation due to cell death likely caused by a G₁ or G₂/M block caused by ISWI mutant disc defects (i.e., alterations in chromosome condensation, replication, or gene expression). Interestingly, the pre-G₁ defect is specific for proliferating ISWI mutant discs and it is not detected in ISWI mutant differentiating brain cells that are a mixture of differentiated and cycling cells, with a significant contribution of cells in G₁ (Figure 6A and Figure S7 C). On the other hand, ISWI mutant brain cells show a reproducible approximately twofold increase in the G₂/M cell population and a shorter S phase when compared to wild type (green and blue arrows in Figure 6A and Figure S7 B), probably reflecting a G₂/M block or a faster S phase [as previously shown in SL2 cells silenced for the ISWI regulator Acf1 (Fyodorov et al. 2004)].

Figure 6.—

Cytofluorimetric analysis of ISWI mutant cells. Cell populations derived ex vivo from wild-type and ISWI mutant neuroblasts (A) and total imaginal (B) and eye-antennal imaginal (C) discs were analyzed by flow cytometry. ISWI imaginal disc cells showed a significant decrease of G₁ and G₂/M peaks (red arrows) and an increase in the pre-G₁ peak (purple arrow). On the other hand, when compared to their wild-type counterpart, ISWI mutant neuroblasts showed a small but reproducible increase in the G₂/M peak (green arrow) and a faster S phase (blue arrow). These cell cycle defects can be in part or completely suppressed by some of the Su(ISWI)'s we isolated.

Our cell cycle analysis suggests that highly proliferating cells tend to be more sensitive to loss of ISWI activity while differentiating cells resist ISWI loss by shortening the S phase and accumulating in G₂/M. To check if ISWI cell cycle defects could be suppressed by some of the Su(ISWI)'s we identified in the ISWI-EGUF screen, we conducted the same analysis on double ISWI, Su(ISWI) mutants that survived to third instar larval stage, where brains and imaginal discs could be dissected. For these reasons, we limited our cell cycle analysis to the EP lines: eff^EP3627, ttk^EP3314, mbf1^EP3684, and the newly generated PiggyBac insertion wun^PBacF48, a gene encoding for a lipid phosphate phosphatase guiding germ cell migration (Table S1 A and Figure 4). As an internal positive control to evaluate the level of cell cycle suppression we could get in the double ISWI, Su(ISWI) mutant discs and brains, we also included in our analysis an allele of casp (casp^PBacG75α), a false positive Su(ISWI) we isolated in our ISWI-EGUF screen. The casp gene encodes for a homolog of the mammalian Fas-associating factor 1, an apoptotic signaling factor that acts downstream of the Fas signal transduction pathway. Due to its role in promoting apoptosis, loss of casp is expected to suppress cell death and cell cycle defects linked to apoptotic signaling present in ISWI mutant cells.

Indeed, the false positive Su(ISWI) casp^PBacG75a mutant suppressed the shortening of the S phase observed in ISWI mutant brain cells (Figure 6A and Figure S7 B). Interestingly, wun^PBacF48 Su(ISWI) also suppressed the ISWI S-phase defect (Figure 6A and Figure S7 B); however, both casp^PBacG75a and wun^PBacF48 failed to suppress the ISWI G₂/M increase observed in differentiating brain cells (Figure 6A and Figure S7 B). These results indicate that the specific S-phase shortening and the G₂/M increase observed in brain cells are very likely independent noncoupled ISWI defects.

Interestingly, wun^PBacF48 and casp^PBacG75a can also suppress ISWI pre-G₁ defects observed in imaginal disc cells (Figure S7, D and E) and restore a cell cycle profile similar to wild type (Figure 6B). Remarkably, while the Su(ISWI) EP lines eff^EP3627 and ttk^EP3314 in the presence of the eyGAL4 driver weakly suppress eye disc-specific ISWI pre-G₁ and G₁–G₂/M defects, the mbf1^EP3684 Su(ISWI) strongly suppresses these defects (Figure S7, F and G; Figure 6C).

Concluding remarks:

The ISWI-EGUF screen allowed us to isolate new ISWI genetic interactors that increased our understanding of the complex network of factors regulating ISWI nuclear signaling pathways. The ISWI-EGUF screen revealed that ISWI interacts with an evolutionarily conserved network of cellular and nuclear factors that escaped previous genetic and biochemical analyses, indicating the participation of ISWI in a variety of biological processes not linked to date with known ISWI functions. Moreover, the combination of gene network bioinformatics tools with classic fly genetics approaches has been instrumental to isolate some of the factors that might have been missed in the not saturating ISWI-based screens. The unexpected functional interactions we found between ISWI and factors playing central roles in cell cycle regulation have a high potential to shed light on the mechanistic aspects of cell cycle progression directly or indirectly regulated by ISWI and more generally by chromatin remodelers. Although the molecular basis of the genetic antagonism existing between ISWI and the variety of modifiers we identified will need further characterization and many of the isolated modifiers may act indirectly (i.e., by regulating the perdurance, the expression, or regulators of ISWI), we believe that the network of ISWI enhancers and suppressors we isolated represents an invaluable selected cohort of genes that could be potentially assayed in any biological process where an ISWI-dependent functional assay is available.