|Home | About | Journals | Submit | Contact Us | Français|
The Drosophila GAGA factor (GAF) has an extraordinarily diverse set of functions that include the activation and silencing of gene expression, nucleosome organization and remodeling, higher order chromosome architecture and mitosis. One hypothesis that could account for these diverse activities is that GAF is able to interact with partners that have specific and dedicated functions. To test this possibility we used affinity purification coupled with high throughput mass spectrometry to identify GAF associated partners. Consistent with this hypothesis the GAF interacting network includes a large collection of factors and complexes that have been implicated in many different aspects of gene activity, chromosome structure and function. Moreover, we show that GAF interactions with a small subset of partners is direct; however for many others the interactions could be indirect, and depend upon intermediates that serve to diversify the functional capabilities of the GAF protein.
The Drosophila GAGA factor (GAF) is an unusually versatile DNA binding protein that functions in remarkably diverse range of regulatory contexts. GAF was first identified as a transcriptional activator in in vitro transcription experiments with the Ultrabithorax (Ubx) and engrailed (en) genes. It bound to GAGAG motifs in the promoter region and stimulated transcription [1–3]. Consistent with a function in transcriptional activation, mutations in the gene encoding GAF, Trithorax-like (Trl), were shown to dominantly enhance the haploinsufficiency of the Ubx gene . Moreover, the Trl mutations also dominantly enhanced position effect variegation (PEV) . While these findings suggested that GAF functions as a conventional transcriptional activator, in vitro chromatin assembly experiments pointed to a rather different and unexpected role. When GAF was included in chromatin assembly assays using a plasmid containing the hsp70 gene as the DNA template, it was found to mediate the formation of a nucleosome free region spanning the GAF binding motifs in the hsp70 promoter . The GAF factor helped recruit chromatin remodeling complexes to the template, and then functioned to exclude nucleosomes from the exposed promoter sequence . Amongst the remodeling complexes that are thought to function together with GAF are PBAP, NURF and FACT [7–10]. A role in the formation/maintenance of nucleosome free regions of chromatin in vitro is recapitulated in vivo in transgene experiments with the hsp26 and hsp70 genes [11,12]. In addition to ensuring that promoter sequences are accessible, GAF is thought to play a more direct role in transcription by regulating promoter pausing [13–15]. These are not, however, the only known biological activities of the GAF protein. It has also been implicated in Polycomb group (PcG) dependent silencing [16–18], chromosome condensation and segregation during mitosis  and boundary activity . Consistent with these multiple functions, GAF binding sequences are found in promoters, enhancers, Polycomb response elements (PREs) and boundary elements, while chromatin immunoprecipitation experiments localize GAF protein to these elements in vivo [21–26].
It is not yet understood how GAF carries out this diverse array of functions. The GAF protein itself has a relatively simple structure. It has an N-terminal BTB/POZ domain, a central C2H2-type zinc finger and several alternative glutamine rich (Q) C-terminal domains. The single zinc finger domain is responsible for DNA binding to the GAGAG pentanucleotide . As there is little apparent flexibility in the DNA recognition properties of GAF, a plausible idea is that its different activities depend upon the ability of the GAF protein to interact either directly or indirectly with multiple partners. There is already evidence supporting this possibility. The GAF BTB/POZ domain has been shown to mediate protein-protein interactions and it participates in the formation of homo-oligomers and hetero-oligomers with other BTB/POZ proteins [28–30]. These BTB/POZ proteins include Tramtrack (Ttk) [28,29,31]; Mod(mdg4) [29,32]; Pipsqueak (Psq) [29,33] and Batman (Lolal) [29,34,35]. The GAF BTB/POZ domain has also been shown to contribute to interactions with non-BTB/POZ proteins, for example SAP18, a component of the Sin3-HDAC corepressor complex . Finally, the alternative C-terminal domains could expand the range of possible GAF partners.
Despite the identification of a number of GAF partners, the scope of the GAF interacting protein network is unknown and its relationship to the diverse nuclear functions of the GAF protein remains poorly understood. In the studies reported here we have used a combination of immunoprecipitation and mass spectrometry to identify the proteins associated with GAF in nuclear extracts. Our data support the idea that the functional versatility of the GAF protein arises, at least in part, from its ability to associate with multiprotein complexes that have dedicated, but distinct functions in transcriptional regulation and chromosomal architecture. These complexes include the SWI/SNF and ISWI subfamilies of ATP-dependent chromatin remodelers, Polycomb (PcG) and Trithorax (TrxG) Group proteins, condensin, cohesion, and boundary/insulator-associated factors. Finally, several GAF network proteins were confirmed to interact directly with the GAF protein.
Animal handling for the antibody production was carried out strictly according to the procedures outlined in the NIH (USA) Guide for the Care and Use of Laboratory Animals. The protocols used were approved by the Committee on Bioethics of the Institute of Gene Biology, Russian Academy of Sciences. All procedures were performed under the supervision of a licensed veterinarian, under conditions that minimize pain and distress.
Rabbits were purchased from a licensed specialized nursery, Manihino. Soviet chinchilla rabbits used in the study are not endangered or protected. Only healthy rabbits, certified by a licensed veterinarian were used. The rabbits were individually housed in standard size, stainless steel rabbit cages and provided ad libitum access to alfalfa hay, commercial rabbit food pellets, and water. The appetite and behavior of each rabbit was monitored daily by a licensed veterinarian. Body weight and temperature of each rabbit were evaluated prior to and daily following immunization. No animals became ill or died at any time prior to the experimental endpoint. At the end of the study period all rabbits were euthanized by intravenous injection of barbiturate anesthetics.
Antibodies against GAF (1–519 aa of isoform PC) , Iswi (1–96 aa of isoform PA), Bap55 (1–162 aa of isoform PA), Adf1 (full length of isoform PC), Su(z)12 (448–798 aa of isoform PA), E(z) (8–184 aa of isoform PA), eIF3-S8 (399–553 aa of isoform PA) were raised in rabbits.
Antigens for antibody production were expressed as a 6 × His-tagged fusion proteins in Escherichia coli, affinity purified on Ni Sepharose 6 Fast Flow (GE Healthcare) according to the manufacturer’s protocol and injected into rabbits following the standard immunization procedure. Antibodies were affinity-purified from serum on the same antigene as was used for immunization and tested by Western blotting and immunoprecipitation (IP) to confirm their specificity.
Other antibodies (used for the co-IP and EMSA experiments) were generously donated by Anton Golovnin (Mod(mdg4) common, Mod(mdg4)-PT (67.2) and Pzg/Z4), Elissa Lei (Mod(mdg4)-PT (67.2)), Oksana Maksimenko (E(y)2), Rakesh Mishra (Batman), Carl Wu (GAF), and David Gilmour (GAF).
GAGA factor-associated proteins were isolated from the nuclear extracts of 0–12 hour Drosophila wild-type Oregon embryos (Bloomington Stock Center #5) prepared as described in [38,39] by immunoaffinity purification. The GAF antibody was coupled to the protein A Sepharose beads (Sigma) using DMP (Sigma) according to a published protocol  and loaded onto a column. Protein A sepharose with no antibody served as the negative control. The column was equilibrated with HEMG buffer (25 mM Hepes-KOH pH 7.6, 12.5 mM MgCl2, 0.1 mM EDTA, 10% glycerol, 1 mM DTT, and complete protease inhibitor mixture Roche) containing 150 mM NaCl (HEMG-150). Nuclear extract in the amount of 10 mg (by protein) was loaded on to the column containing the GAF antibody-protein A sepharose beads and incubated for two hrs. The column was washed extensively with HEMG-500 plus 0.1% Nonidet P-40 and then eluted with 0.1 M glycine at pH 2.5. Eluted proteins were analyzed mass spectrometry.
For analytical IP and co-IP the same procedure was followed. For a typical analytical experiment 100 μg of the nuclear extract was incubated with agarose A (mock control) or anti-GAF antibodies with or without DNaseI (0,5 U, Thermo Scientific). The co-IP’d proteins were detected by Western blotting.
In brief, proteins were reduced with 5 mM Tris(2-carboxyethyl)phosphine hydrochloride (Sigma-Aldrich, St. Louis, MO, product C4706) and alkylated with 2-chloroacetamide. Proteins were digested for 18 hr at 37°C in 2 M urea 100 mM Tris pH 8.5, 1 mM CaCl2 with 2 ug trypsin (Promega, Madison, WI, product V5111). Analysis was performed using an Agilent 1200 quaternary pump and a Thermo LTQ Orbitrap Velos using an in-house built electrospray stage . Protein and peptide identification and protein quantitation were done with Integrated Proteomics Pipeline—IP2 (Integrated Proteomics Applications, Inc., San Diego, CA. http://www.integratedproteomics.com/). Tandem mass spectra were extracted from raw files using RawExtract 1.9.9  and were searched against UniProt Drosophila melanogaster database with reversed sequences using ProLuCID [43,44]. The search space included all fully-tryptic and half-tryptic peptide candidates for the tryptic digest with static modification of 57.02146 on cysteine. Peptide candidates were filtered using DTASelect .
Nuclear extracts from 6–18 hour Drosophila wild-type Oregon embryos, utilized for EMSAs were prepared using methods adopted from previously published procedures . In the EMSAs, the binding reactions were performed in a 20 μL volume consisting of 25 mM Tris-Cl (pH 7.4), 100 mM KCl, 1 mM EDTA, 0.1 mM dithiothreitol, 0.1 mM PMSF, 0.03 mg/ml bovine serum albumin, 10% glycerol, 0.25 mg/ml poly(dA-dT)/poly(dA-dT) and 20 μg of protein derived from nuclear extract. The reaction mixtures containing the 32P labeled GAGA4 probe (see  for description of probe) were incubated for 30 minutes at room temperature with or without 20 μg of nuclear extracts (and plus antibodies as indicated) and loaded onto a pre-cleared 4% acrylamide:bis-acrylamide gels in 0.5 x TBE–2.5% glycerol gel. Binding reactions were electrophoresed at 180V for 3–4 hr with a 0.5 x TBE-2.5% glycerol running buffer at 4°C, dried and imaged using a Typhoon 9410 scanner and Image Gauge software and/or X-ray film.
Two-hybrid assays were carried out using yeast strain pJ69-4A, plasmids, and protocols obtained from Clontech. For growth assays, plasmids were transformed into pJ69-4A by the lithium acetate method as described by the manufacturer and were plated on nonselective media lacking tryptophan and leucine. After 3 days of growth at 30°C, plates were replicated on selective media: (1) lacking tryptophan, leucine, and histidine in the presence of 1mM or 5 mM 3-aminotriazole (3-AT); (2) lacking tryptophan, leucine, histidine and adenine. Each assay was repeated at least twice and growth was compared after 2, 4 and 7 days. Based on the extent of growth the interactions were scored as strong (+++), intermediate (++) or weak (+).
The full-length GAF isoform PC 1-519aa or GAF fragments 1–131 aa (BTB domain), 1–316 aa, 1–389 aa were cloned into pGBT9 vector (Clontech) to make fusions with the GAL4 DNA-binding domain. Other cDNA were cloned into pGAD24 vector to make fusion with the GAL4 activating domain. The BTB/POZ domain containing fragments were used for: Lola (CG12052) PA 1–120 aa; CG41099 PB 33–164 aa; Mri (CG1216) PA 86–210 aa; Bab2 (CG9102) PA 194–313 aa; CG8924 PB 1–146 aa. The full-length cDNAs were used in case of CP190 (CG6384 PA) 1–1096 aa; Batman (Lolal, CG5738 PA) 1-127aa; Sry-delta (CG17958 PA) 1–433 aa; L(3)neo38 (CG6930 PA) 1–380 aa; Pzg/Z4 (CG7752 PA) 1–996 aa; MEP-1 (CG1244 PA) 1–1152 aa; Su(Hw) (CG8573 PA) 1–941 aa; CG2199 PB 1–733 aa; Pita (CG3941 PA) 1–681 aa. The cDNA fragments were used for Row (CG8092 PA) 10–1281 aa; CG (CG8367) PF 31–467 aa.
We used immunoaffinity purification (IP) to identify proteins in nuclear extracts that are associated with GAF. Nuclear extracts were prepared from 0–12 hr wild type embryos (Figure A in S1 File) and incubated with a GAF polyclonal antibody that had been coupled to Protein A Sepharose beads. The polyclonal antibody was raised against the full-length 519 amino acid GAF isoform. After extensive washing the bound protein was eluted from the Protein A Sepharose beads with 0.1 M glycine at pH 2.5  (IP, Figures B and C in S1 File) and analyzed by mass spectrometry [41,42]. These experiments were done in triplicate. Numerous proteins were detected in the samples isolated from the GAF antibody affinity beads that were absent in the sample prepared from the control beads (S2 File). Altogether proteins encoded by 2421 genes were present in at least one of GAF IP sample; however, slightly less than half (1202) were found in all 3 replicates (S3 File). Of those that were detected in all three experiments, only 903 had p-values <0.05 (S3 File). GO analysis of this latter group indicated that most of the proteins had biological functions likely to be relevant to GAF including chromosome organization, transcription, DNA replication, and mRNA processing and export (Figure D in S1 File). Also most (but not all) of the GAF associated proteins identified in previous studies were found in the immunoaffinity purified samples (S4 File). The different protein complexes and proteins that were found associated with GAF in nuclear extracts are discussed further below.
In agreement with previous studies [7,10], proteins derived from a collection of ATP-dependent chromatin remodeling complexes are found in the GAF IP samples. However, unlike these previous studies, we were able to detect most if not all of the known components of these complexes (Table 1, and see Table A in S5 File for detailed data). Moreover, with only a few exceptions, the protein components of these complexes had p-values <0.05. For the two SWI/SNF, PBAP and BAP , subfamily members we detected all of the (7) proteins that are common to both complexes. The co-purification of GAF with Bap55 was confirmed by co-IP using Bap55 specific antibodies (Figure E in S1 File). We also detected the three proteins, Polybromo, Bap170 and SAYP, that are specific to the PBAP complex, as well as the single BAP specific subunit, Osa [47,48]. The fact that all of the components of PBAP and BAP are found GAF associated would provide a compelling argument that GAF interacts with intact PBAP/BAP complexes, rather than with some type of sub-complex variants.
The same is true for the ISWI subfamily complexes NURF, ACF and ToRC, and the CHD subfamily complex dNURD. There are four known NURF complex proteins, Iswi, E(bx)/NURF301, NURF-38 and Caf1-55/p55 [9,50], plus an associated protein Pzg/Z4 . All five of these proteins are detected in the GAF immunoaffinity sample (Table 1, and Table A in S5 File). Similarly, known subunits of the ACF complex , Iswi and Acf1, and ToRC complex , Iswi and CtBP and Tou, are present. The CHD family remodeler, dNURD, has eight subunits, Mi-2, MEP-1, MTA1-like, Simj/p66, HDAC1, Caf1-55, MBD-like, CDK2AP1  and Ttk as associated protein [54,55]. All of these proteins are immunoaffinity purified from nuclear extracts with the GAF antibody. The association of GAF with Iswi (NURF, ACF) and Pzg/Z4 was additionally confirmed by co-IP (Figure E in S1 File).
While these findings provide strong evidence that GAF associates with multiple remodeling complexes, these interactions need not be direct and could be mediated by accessorial proteins. For example, like GAF, Ttk has an N-terminal BTB/POZ domain and a C-terminal zinc finger DNA binding domain. The Ttk BTB/POZ is known to interact directly with the dNURD subunits, MEP-1 and Mi-2 [54,55]. It also interacts directly with the GAF BTB/POZ domain [28,29,31]. The same is true for GAF interactions with the NURF complex. Previous studies have shown that the zinc finger protein, Pzg/Z4 interacts not only with dNURF , but also with GAF . Moreover, as described further below, this GAF:Pzg/Z4 interaction is direct. Thus, GAF association with both the dNURD and NURF remodeling complexes could be mediated in vivo by intermediary zinc finger DNA binding proteins, rather than by GAF itself.
In addition to its role in the formation and maintenance of nucleosome free regions over promoters, GAF protein has been implicated at other steps in the transcription process, including recruitment of transcription factors and Pol II to promoters, promoter pausing and mRNA biogenesis. In S6 File we have listed the proteins that are subunits of factors that function in transcription and/or mRNA biogenesis. While subunits from RNA Pol II, Mediator, TFIID and NELF can be detected in the GAF immunoaffinity sample, the number of peptides and the spectral counts for the proteins in these complexes are typically rather low and not reproduced well in biological replicates. Moreover, unlike the chromatin remodeling factors, proteins known to be components of these complexes (e.g, Rbp4 and Rbp12 for RNA Pol II) aren’t detected. Given that these factors frequently co-localize with GAF at promoters or just downstream, it seems likely that GAF association either reflects their co-localization on the DNA or is not especially robust.
The TREX-THO complex functions in transcription elongation, mRNA biogenesis and mRNA export [56–59]. Supporting the findings reported in a previous study  we detected all of the subunits of TREX–THO complex (Tho2 and Hpr1) in all GAF immunoaffinity samples (S6 File). GAF was also shown to associate with the FACT complex proteins Ssrp and Dre4 [7,8]. FACT acts subsequent to transcription initiation to release RNA polymerase II from a nucleosome-induced block . Only one of the two FACT complex proteins, Ssrp was detected in the GAF immunoaffinity samples, and both the peptide and spectral counts were quite low (S6 File). Thus, at least under our conditions, FACT does not seem to be stably associated with GAF.
Polycomb-dependent silencing in Drosophila is typically mediated by special cis-acting elements called Polycomb Response Elements (PRE). These elements recruit the PRC1, PRC2 and PhoRC complexes, which then function to silence genes in the neighborhood [61–65]. Like promoters and other regulatory elements, PREs are nucleosome free and have recognition sites for DNA binding proteins such as Pho, Cg, Zeste, Psq, Adf1, Grh, Dsp1 and Spps, which have been implicated in PcG silencing [63,66,67]. Like these DNA binding proteins, GAF also binds to PRE sequences in vitro and in vivo and has been shown to be important for their PcG dependent silencing activity [16–18,68]. Several additional non-DNA binding PcG proteins that are required for homeotic gene silencing but have not been assigned to PcG complexes have also been described, including Sxc/Ogt , Dom , and Batman (Lolal) .
Of these factors, we find that the Adf1 and Zeste DNA-binding proteins are present in all IP samples with p-values of < 0.01 reflecting a stable association with GAF. The presence of Adf1 was confirmed by Western blotting (Figure E in S1 File). With respect to the three major Polycomb group complexes, all of the PRC2 subunits are present in all three GAF immunoaffinity purifications and have p-values of <0.05 (Table 2, and Table B in S5 File). To further verify GAF- PRC2 association, we probed for the E(z) and Su(z)12 subunits of the PRC2 complex in GAF IPs using Western blotting (Figure E in S1 File).
Also present in the GAF immunoaffinity sample are components of the PhoRC and PRC1 complexes. However, the peptide and spectral counts for PhoRC and PRC1 proteins are quite low and most of the PRC1 components are not present in all three immunoaffinity purifications. Additionally, Pc isn’t even detected (Table 2, and Table B in S5 File). These findings would be consistent with the idea that GAF—PRC1/PhoRC association is not especially stable and is likely to be indirect either via other proteins (for example, Batman or by interactions with accessory factors such as the putative acetyltransferase l(1)G0020 or the helicase p90. It is also worth noting that in the case of PRC2, which seems to be GAF associated, the peptide counts are less than that observed for, for example, components of the PBAP/BAP complexes (Table 1, and Table A in S5 File).
We also detected several known TrxG factors in the GAF immunoaffinity purified sample. Besides members of BAP/PBAP complexes (Brm, Mor, Snr1, Osa, SAYP) [48,73–75] listed in Table 1, these include: Fs(1)h [76,77], Ash2 , Vtd, Kis , Lid , Bre1  (Table 2).
GAF-binding sites were shown to be required for the insulator activity of chromatin domain boundary elements [80,81] and the GAF protein localizes to many known or putative insulators in vivo . One of the boundary elements that requires GAF for its insulator activity is the Fab-7 boundary from the Drosophila Bithorax complex [81,82]. There are six GAGA motifs in the major Fab-7 nucleosome free region (nuclease hypersensitive site HS1). Four of these are located in a 293 bp long fragment called dHS1 that has developmentally restricted insulator activity in transgene assays, functioning (primarily) from mid-embryogenesis through to the adult stage. Recent studies have shown that a large ~700 kD complex named LBC binds to probes containing three of the four GAGA motifs in dHS1 . In each case, the minimal recognition sequence for the LBC is more than 65 bp. Supershift experiments with antibodies directed against known insulator proteins together with gel filtration experiments indicate that the LBC complex contains GAF, Mod(mdg4) and E(y)2. Thus unlike the other complexes that interact with GAF (e.g., PABP/BAP or NURF remodeling complexes) either directly or indirectly, GAF is an integral component of the LBC. As would be expected if the LBC were pulled down by the GAF antibody, both Mod(mdg4) and E(y)2 are present in the affinity purified sample (Table 3, and Table C in S5 File).
As noted in the introduction, previous studies have shown that the N-terminal BTB/POZ domains of the GAF and Mod(mdg4) mediates both self-interactions and interactions between these two proteins. The GAF BTB/POZ domain tends to form homodimers, while the Mod(mdg4) BTB/POZ domain forms octamers. Interestingly, there are 31 predicted isoforms of Mod(mdg4). All share the N-terminal BTB/POZ protein interaction domain, but have unique C-terminal FLYWCH zinc finger DNA binding domains. Since GAF-Mod(mdg4) interactions are mediated by the BTB/POZ domains, the LBC could contain a complex mixture of Mod(mdg4) isoforms. Consistent with this possibility, we detect a total of 14 Mod(mdg4) isoforms in the GAF IP samples (S7 File). As we don’t know the relative abundance of the different Mod(mdg4) isoforms in 0–12 hr embryonic nuclear extracts, we can’t exclude the possibility that there are yet other GAF-associated Mod(mdg4) isoforms in embryos. Also while we cannot be certain that all of the isoforms detected are derived from LBC complexes, it seems likely that this is the case as they all have the Mod(mdg4) BTB/POZ domain that is responsible for interactions with GAF.
One of the Mod(mdg4) isoforms in the GAF IP is PT [(Mod(mdg4)67.2 or 2.2] which has been implicated in su(Hw) dependent insulator function [84,85]. We first confirmed that this specific isoform is present in GAF IP by co-IP (Figure E in S1 File). At the next step, to confirm that Mod(mdg4)PT is present in the LBC we did EMSA experiments with a probe, GAGA4, that spans one of the dHS1 GAGA motifs (#4). As shown in Fig 1, the GAGA4 probe generates a characteristic slowly migrating shift when incubated with embryonic nuclear extracts. The yield of this shift is slightly reduced when control rabbit serum is included in the reaction mix, while it is not greatly affected by control rat serum. As previously described, LBC supershifts are observed when antibodies against GAF and E(y)2 are included in the reaction mix. In contrast, antibodies against another BTB/POZ domain protein, Batman, which physically interacts with GAF in vitro and in vivo ([29,34,35], see below) don’t generate supershifts. This finding would argue that the LBC probably doesn’t contain Batman and that there must be some other, as yet unidentified, GAF:Batman complex. Importantly, antibodies specific for the PT isoform of Mod(mdg4) generate a supershift, indicating that this isoform is present in LBC complexes.
In addition to the LBC, GAF appears to be associated with other factors known to be involved in boundary activity. These include Pzg/Z4, CP190, and Su(Hw) and the boundary associated factors Dref and Chro (Table 3, and Table C in S5 File). Moreover, key subunits of the chromosome architectural complexes, cohesin and condensin are also present in the GAF immunoaffinity purified sample (Table 3, and Table C in S5 File).
Though many proteins are found associated with GAF after immunoaffinity purification, only a subset are expected to interact directly with GAF. Instead, many will be present in the purified sample because they linked to GAF through common interactions with intermediary proteins. To begin classifying the proteins in the GAF immunoaffinity sample as either direct or indirect interactors, we took advantage of the yeast two-hybrid assay.
Previous studies have shown that the BTB/POZ domain of GAF interacts with several other BTB/POZ domain proteins. Thus we expected to find GAF associated BTB/POZ domain proteins. Altogether 11 different BTB/POZ proteins were present in all 3 GAF IP samples (S8 File). These includes four previously identified direct partners of GAF: Ttk [28,29,31], Batman [29,34,35], Psq/BTB-V [29,33,49], and Mod(mdg4) [29,32,83]. The BTB-domains of these four proteins have been shown to interact with the BTB-domain of GAF [28,29,31–33]. Due a high degree of sequence similarity these and two other proteins found in all GAF immunoaffinity samples, Lola/BTB-IV and Bab2/BTB-II, have been classified as Ttk Group BTB/POZ domain proteins [29,86]. In addition, there are two other proteins, CG8924 and Rib, in all GAF immunoaffinity samples that have BTB/POZ domains which display a high degree of sequence similarity to members of the Ttk group. Finally, there are three GAF associated proteins, CP190, CG41099, Mri, that have a more distantly related BTB/POZ domain (S9 File).
For the yeast two-hybrid experiments, we fused the GAL4 DNA binding domain to full length GAF 519 aa isoform or to an N-terminal GAF fragment 1–131 aa that contains the BTB/POZ domain. To test for interactions, we then fused the BTB/POZ domains of the GAF associated BTB/POZ proteins to the GAL4 activation domain. Since direct interactions have already been demonstrated for four of these BTB/POZ domain proteins (see above and Fig 2), we selected only one, Batman, as a positive control. The full-length CP190 protein that previously failed interact with GAF  was used as a negative control.
As was found previously for a subset of the Ttk-group BTB/POZ members , all of the new proteins in this group interact directly with GAF through BTB/POZ domains (Fig 2). Most of these proteins have several different isoforms, all of which share their BTB/POZ domain. This means that the complexes containing GAF and these Trk-group proteins band could contain different isoform combinations. The other BTB/POZ protein besides Mod(mdg4) that has an extraordinarily large number of isoforms is Lola, which has around 20 distinct isoforms. Like Mod(mdg4) (see above) most of these isoforms have a unique C-terminal DNA binding domain. Altogether we were able to identify four different Lola isoforms in GAF immunoaffinity purified samples (S10 File).
While all members of the Ttk-group can interact directly with GAF, this is not true for the two other BTB/POZ domain proteins found associated with GAF. As indicated in Fig 2, we were unable to detect direct interactions between the BTB/POZ domains of CG41099, Mri proteins and GAF using the yeast two hybrid assay.
Another class of proteins that could potentially interact directly with GAF is C2H2-type Zinc finger domain proteins. Like GAF, these proteins are also expected to bind directly to DNA. Altogether a total of 26 GAF associated proteins that have C2H2-type Zinc fingers were present in all three of the GAF immunoaffinity samples (S8 File). As discussed above, a few of these resemble GAF in that they also have a BTB/POZ domain (Lola/BTB-IV, Ttk, CP190). To extend our analysis, we selected 9 of non-BTB/POZ domain proteins to test for direct interactions with GAF (see Fig 3). Included in this list are the boundary proteins, Su(Hw), Pita [87,88], and two proteins that have been shown to associate with GAF in extracts, but not tested for direct interaction, Pzg/Z4 and CG2199 .
For these experiments we fused cDNAs encoding these proteins in frame to the GAL4 activation domain. Of the 9 proteins that we tested, only three (Pzg/Z4, CG2199 and MEP-1) were able to interact directly with full-length GAF in our two-hybrid assay (Fig 3). Since previous studies have shown that the BTB/POZ domain of GAF mediates its interactions with other BTB/POZ proteins, we wondered whether this domain is also responsible for interactions with proteins that lack a BTB/POZ domain. To address this question we generated a series of GAF truncations and then tested whether they retained the ability to interact with these three C2H2-type Zinc finger domain proteins. We found that two of the proteins, Pzg/Z4 and CG2199, appear to interact with GAF via its BTB/POZ domain (Fig 3). However, the MEP-1 requires a small region of the GAF protein that spans the single C2H2-type Zinc finger domain (Fig 3). This finding raises the possibility that the GAF Zinc finger domain not only has DNA binding activity but can also participate in protein-protein interactions.
The GAF protein has extraordinarily diverse array of functions in Drosophila. Amongst the many functions that have so far been identified are generating nucleosome free regions of chromatin, transcriptional activation and elongation, PcG and heterochromatic silencing, chromatin domain boundaries and mitosis. The studies reported here provide a plausible explanation for how GAF can have so many different, and in some instances seemingly contradictory functions. Using immunoaffinity purification combined with mass spectrometry we have isolated proteins that are associated with GAF in nuclear extracts prepared from 0–12 hr embryos. For many of known GAF functions we have identified GAF-associated proteins and/or multi-subunit complexes that have the appropriate activities. For example, we detect both of SWI/SNF chromatin remodeling complexes, the CHD family complex, and the ISWI family complexes NURF, ACF and ToRC. Moreover, in contrast to previous studies all of the known subunits for these complexes are present in the GAF immunoaffinity sample, providing compelling evidence that GAF interacts with fully functional remodeling complexes. The same seems to be true for TREX-THO, cohesion and condensin I complexes. For some of the GAF associated proteins we have shown that interactions with GAF are direct. Finally, it is likely that the GAF associated proteins we have identified is still an incomplete list. There are a number of reasons for this. One is that not all complexes will remain intact during the affinity purification. Another limitation is that of antibody accessibility. Even though we used polyclonal antibodies, there may be complexes in which key GAF epitopes are not readily accessible, thus substantially reducing the yield of these protein/complexes in the immuonaffinity purified sample.
Supplementary figures A-E.
GAF IP Tables A-C.
We thank Elissa Lei, Anton Golovnin, Oksana Maksimenko, Rakesh Mishra, Carl Wu, and David Gilmour for the gift of antibodies. This work was carried out using infrastructure of Center for collective use of the Institute of gene Biology, Russian academy of science: "Biology of the living cell and drug biomedical nanotransporters".
This work was supported by the Russian Science Foundation (14-24-00166) to PG; Russian Federation Ministry of Education and Science grant to the Gene Biology Institute (14.B25.31.0022); NIH (GM043432) to PS; Russian Foundation for Basic Research (16-34-60214) to YS; National Institute of General Medical Sciences (8 P41GM103533) to JJM and JRY.
All relevant data are within the paper and its Supporting Information files.