Barrier-to-Autointegration Factor Proteome Reveals Chromatin-Regulatory Partners

Nuclear lamin filaments and associated proteins form a nucleoskeletal (“lamina”) network required for transcription, replication, chromatin organization and epigenetic regulation in metazoans. Lamina defects cause human disease (“laminopathies”) and are linked to aging. Barrier-to-autointegration factor (BAF) is a mobile and essential component of the nuclear lamina that binds directly to histones, lamins and LEM-domain proteins, including the inner nuclear membrane protein emerin, and has roles in chromatin structure, mitosis and gene regulation. To understand BAF's mechanisms of action, BAF associated proteins were affinity-purified from HeLa cell nuclear lysates using BAF-conjugated beads, and identified by tandem mass spectrometry or independently identified and quantified using the iTRAQ method. We recovered A- and B-type lamins and core histones, all known to bind BAF directly, plus four human transcription factors (Requiem, NonO, p15, LEDGF), disease-linked proteins (e.g., Huntingtin, Treacle) and several proteins and enzymes that regulate chromatin. Association with endogenous BAF was independently validated by co-immunoprecipitation from HeLa cells for seven candidates including Requiem, poly(ADP-ribose) polymerase 1 (PARP1), retinoblastoma binding protein 4 (RBBP4), damage-specific DNA binding protein 1 (DDB1) and DDB2. Interestingly, endogenous BAF and emerin each associated with DDB2 and CUL4A in a UV- and time-dependent manner, suggesting BAF and emerin have dynamic roles in genome integrity and might help couple DNA damage responses to the nuclear lamina network. We conclude this proteome is a rich source of candidate partners for BAF and potentially also A- and B-type lamins, which may reveal how chromatin regulation and genome integrity are linked to nuclear structure.


Introduction
The nuclear envelope and lamin filament networks tether chromatin and influence chromatin organization and gene expression at several levels [1,2], through mechanisms that remain obscure. A-and B-type lamins form separate networks of nuclear intermediate filaments, which are anchored to inner nuclear membrane (INM) proteins but also ramify throughout the nuclear interior [3,4]. Lamins interact with chromatin and support most nuclear activities including replication, transcription and DNA damage repair [5][6][7][8]. Mutations in A-type lamins or other lamina components cause more than 15 human diseases (''laminopathies'') including Emery-Dreifuss muscular dystrophy and accelerated aging syndromes (i.e., atypical Werner, Hutchinson-Gilford Progeria Syndrome) [9,10]. Cells expressing mutant lamins have compromised nuclear integrity. For example, in Hutchinson-Gilford progeria cells the mutated lamin A precursor (''progerin'') disrupts the levels or distribution of wildtype lamins, LEM-domain proteins, histone-modifying enzymes and histone posttranslational modifications, and also responses to DNA damage [5,7,11]. Another protein with essential roles in nuclear integrity is BAF, a conserved direct partner for lamins [12,13] and the LEM-domain family of nuclear proteins. All characterized members of this family, including LAP2b, Emerin, MAN1 and Lem2/NET25 at the INM, and LAP2a in the nuclear interior, bind BAF via their LEM-domain, and also bind directly to lamins (reviewed by [14]). Studies in Caenorhabditis elegans and mammalian cells showed that BAF is required for the integrity of the nuclear lamina network, and in combination with LEM-domain proteins and lamins has interdependent structural roles during mitosis and nuclear assembly [13,[15][16][17][18].
We hypothesized that BAF influences these various activities by associating with novel partners. To discover such partners, we used affinity purification and mass spectrometry methods to identify proteins from HeLa cell nuclear extracts that specifically associated with recombinant purified human BAF dimers. This strategy yielded several expected partners (lamins, core histones and LEM-domain protein LAP2), plus many novel candidate interactors including four human transcription factors and several proteins that modify or repair chromatin. Association with endogenous BAF was confirmed in vivo for seven of 11 candidates tested. We further report that BAF and its nuclear membrane partner emerin dynamically interact with DNA damage response proteins of the CUL4-DDB-ROC1 complex, suggesting novel roles for BAF and emerin in genome integrity.

Results
To identify proteins that potentially associate with BAF we affinity-purified proteins from HeLa nuclear extracts using recombinant purified His-tagged human BAF (H 6 BAF) dimers bound to Ni ++ -agarose beads, with agarose beads alone as controls (see Methods). Bound proteins were eluted sequentially with buffer containing 0.3 M NaCl to remove proteins of moderate affinity, then with 1 M NaCl to elute higher-affinity partners ( Figure 1A). An aliquot (10%) of each eluate was resolved by SDS-PAGE and silver-stained for qualitative analysis. Reproducible patterns of BAF-associated bands were seen ( Figure 1B; n = 3). The remaining eluates (90%) from a single experiment were digested with trypsin in solution and all detectable constituent proteins were identified by liquid chromatography-coupled tandem mass spectrometry (LC-MS/MS). Data were analyzed using an in-house Mascot server and Scaffold software, resulting in over 70 significant, unique, and non-redundant protein ''hits'' (Table S1). Proteins identified by Scaffold with at least 89% confidence in the 0.3 M or 1 M NaCl eluates from BAF-beads, and not detected in bead-only controls, are listed in Table 1. Table 2 lists selected proteins that were detected with at least 89% confidence in a BAF-bead eluate(s) and also in a bead-only control(s). A few proteins appear in both tables based on peptides that fit each criterion. The full (100 megabyte) searchable Scaffold dataset is available upon request. We also scaled up sample preparation, and used isobaric tagging for relative and absolute quantitation (iTRAQ) reagents to independently identify and quantify proteins that were specifically enriched on BAF-beads, relative to bead-only controls. For iTRAQ analysis, the four samples were each labeled with a different iTRAQ reagent [35]. This analysis yielded many proteins with significantly enriched binding to BAF-beads, relative to beadonly controls, expressed as enrichment ratios (Table 3).

Independent validation of selected candidates
To evaluate the physiological relevance of this proteome, we chose 11 candidates to test independently for potential association with BAF in vivo. Antibodies for many candidates were unavailable or unsuitable for immunoprecipitation, so we used an epitope-tag strategy. Each candidate was transiently expressed as a FLAG-or GFP-tagged protein in HeLa cells, then immunoprecipitated from whole cell lysates using antibodies against the epitope tag, resolved by SDS-PAGE and western blotted using antibodies specific for either the tag, endogenous BAF or other endogenous proteins (Figures 2, 3, 4).

BAF associates with Requiem in vivo
Requiem, with a Kruppel-type zinc finger and four atypical zincfingers, is a pro-apoptotic transcription factor in myeloid cells homologous to the mouse Ubi-d4 protein [47,48]. FLAG-Requiem reproducibly (n.4) and specifically co-immunoprecipitated both endogenous BAF and endogenous emerin from cells ( Figure 2A, lane 2, aBAF and aEm). These results independently validated Requiem as BAF-associated in vivo, and further suggested potential ternary interactions with emerin. Whether BAF or emerin (or both) bind Requiem directly was not tested.
BAF associates with PARP1 in HeLa and HEK293 cells PARP1, represented by at least 31 distinct peptides, was the most significant protein identified by LC-MS/MS (Tables 1-2). PARP1 ADP-ribosylates many substrates including histones, regulates chromatin structure and has well-characterized roles in DNA damage responses [49]. We therefore independently tested a potential association between FLAG-PARP1 and endogenous BAF in HeLa cells. Controls showed little or no background pelleting of endogenous BAF with FLAG vector alone ( Figure 2B, lane 4, aBAF), and positive co-precipitation with FLAG-H3.1 ( Figure 2B, lane 6, aBAF), as expected. Endogenous BAF coprecipitated reproducibly and at high levels with FLAG-PARP1 under both low stringency (150 mM NaCl, 0.15% TX-100; data not shown), and high-stringency conditions (300 mM NaCl, 0.3% TX-100; Figure 2B, lane 2, aBAF; n.4). A reciprocal experiment using a stable, tetracycline-inducible FLAG-BAF overexpressing cell line (HEK293, human embryonic kidney; [50]) gave similar results: endogenous PARP1 co-immunoprecipitated with FLAG-BAF (data not shown). Further Western blotting of the FLAG-PARP1 immunoprecipitates confirmed the presence of lamins A and C ( Figure 2B, lane 2, aLmnA/C; n.4) under high stringency conditions, as expected [51,52]; and also revealed low levels of the nuclear membrane protein emerin ( Figure 2B, lane 2, aEm; n.4), with higher signals under low stringency conditions (not shown).  These results suggest BAF associates with PARP1 in two independent cell lines: HeLa and HEK293.
Selective association of BAF with RBBP4 in HeLa cells RBBP4 (also known as Chromatin Assembly Factor-1 [CAF-1] p48 subunit or RbAp48) and RBBP7 (RbAp46) are homologous histone chaperones encoded by different genes. Both were specifically identified by LC-MS/MS (Table 1) and at least one was identified by iTRAQ (Table 3). RBBP4 and RBBP7 are WDrepeat proteins that contribute, either together or separately, to several chromatin-modifying complexes including the NuRD complex [40,41] and the CAF-1 complex, required for chromatin assembly after DNA replication and repair [53,54]. We tested for potential interaction between endogenous BAF and either GFP-RBBP4 or GFP-RBBP7 in HeLa cells under normal conditions, or 1 hr after UV-irradiation, since RBBP4 and the CAF-1 complex mediate histone deposition during UV-damage repair. Controls confirmed little association of BAF with GFP alone ( Figure 3A, lanes 1 and 2, aBAF) and positive association with GFP-H1.2 ( Figure 3A, lanes 7 and 8, aBAF), as expected [28]. Endogenous BAF co-precipitated robustly with GFP-RBBP4 in both undamaged and UV-treated cells ( Figure 3A, lanes 3 and 4, respectively, aBAF; n = 3). By contrast, near background levels of BAF co-precipitated with GFP-RBBP7 ( Figure 3A, lanes 5 and 6, aBAF; n = 3). Endogenous emerin co-precipitated very weakly with both GFP-RBBP4 and GFP-RBBP7 ( Figure 3A, lanes 3-6, aEm; n = 3). Control Western blots of whole cell lysates verified that UV-irradiation did not significantly alter the levels of endogenous BAF or emerin proteins, relative to loading control c-tubulin ( Figure 3B, inputs; n = 3). BAF interaction with RBBP4 was further confirmed in reciprocal studies of stable HEK293 cells with tetracycline-inducible expression of FLAG-BAF (as above): FLAG-BAF co-immunoprecipitated endogenous RBBP4 (data not shown). Thus, despite 89.4% amino acid identity between RBBP4 and RBBP7 ( Figure 3C) [55], BAF associated robustly and selectively with RBBP4 in both untreated and UVtreated cells.

BAF and emerin interactions with DDB2 and CUL4 in UV-irradiated HeLa cells
Three proteomic candidates (DDB1, DDB2, XPC; Table 1) are involved in DNA damage repair. DDB1 and DDB2 are key components of the UV-DNA-damage-binding protein complex (UV-DDB) [56], which recognizes DNA-distorting lesions and mediates nucleotide-excision repair, in part by recruiting XPC-RAD23B heterodimers [57]. In addition, DDB1 and DDB2 also associate with Cullin 4 (CUL4)-containing E3 ubiquitin ligase complexes. One such complex, CUL4-DDB-ROC1, is recruited rapidly (5-10 min) by DDB1-DDB2 to sites of UV-induced DNA damage, where it ubiquitylates H3 and H4 causing nucleosome eviction and exposure of damaged DNA [58]. The CUL4-DDB-ROC1 complex then promotes repair by recruiting repair proteins XPC and its partner RAD23B to damage sites [59]. We found that endogenous DDB1 co-immunoprecipitated with FLAG-BAF in tetracycline-induced HEK293:BAF cells (data not shown), independently validating its association with BAF in vivo. Given the above-described roles for DDB1-DDB2 and XPC-RAD23B, and recognizing that CUL4 was not identified in our (unirradiated) HeLa cell proteome, we hypothesized that BAF might be involved in cellular responses to UV-damage. To test this idea we expressed either FLAG-CUL4A, FLAG-DDB2, FLAG-H3.1 (positive control) or FLAG vector alone (negative control) transiently (,24 hrs) in HeLa cells. Cells were then either treated with UV light (20 J/m 2 ) or left untreated as controls, and allowed to recover for the indicated time. Whole cell lysates were then prepared and either Western blotted directly ( Figure 4A, inputs) or after immunoprecipitation with FLAG-conjugated beads ( Figure 4B, pellets). To assess whether cells were effectively damaged by UV-treatment, whole cell lysates were Western blotted with antibodies against endogenous PARP1 to assay production of its characteristic apoptosis-induced cleavage product [60]. PARP1 cleavage was weak or undetectable in untreated cells ( Figure 4A, ''-''; arrowhead), and detectable but faint 60 min after UV-treatment ( Figure 4A, lanes 2, 4, 9, 12, 14, 19; arrowhead; n = 3). PARP1 cleavage was obvious 120 min after UVtreatment ( Figure 4A, lanes 10 and 20; arrowhead), suggesting cells had entered apoptosis. FLAG-pulldowns from untreated cells or from each timepoint after UV-irradiation, were probed with antibodies specific for either FLAG, endogenous DDB1, endogenous emerin or endogenous BAF ( Figure 4B). Endogenous DDB1 coprecipitated with FLAG-DDB2 at all timepoints ( Figure 4B, lanes 5-10, aDDB1; n = 3) as expected, since these proteins form heterodimers [61]. Validating our proteomic results, endogenous BAF co-precipitated with FLAG-DDB2 ( Figure 4B, lanes 5-10,  aBAF; short and long exposures both shown; n = 3) above the FLAG vector control background ( Figure 4B, lanes 1-2, aBAF; n = 3), in both UV-treated and untreated cells. Endogenous emerin also coprecipitated efficiently with FLAG-DDB2 ( Figure 4B, lanes 5-10, aEm; n = 3) in both UV-damaged and undamaged cells. In addition, emerin also co-precipitated with FLAG-CUL4A in undamaged cells, and this association transiently decreased, then increased after UVtreatment ( Figure 4B, lanes 15-20, aEm; n = 3). A similar trend was seen for the endogenous BAF association with FLAG-CUL4A ( Figure 4B, lanes 15-20, aBAF; n = 3). Despite experiment-toexperiment variation in the exact timing of highest association between BAF or emerin and FLAG-CUL4A, endogenous BAF and emerin consistently ''co-peaked'' in their interaction with FLAG-CUL4A after UV-treatment. These results validated CUL4A, DDB2 and DDB1 as partners for BAF in vivo, and suggested their association with BAF is differentially regulated in response to UVdamage. Furthermore, we discovered robust and potentially UVregulated associations of the nuclear membrane protein emerin with both DDB2 and CUL4A. These results suggest novel roles for both BAF and emerin in genome integrity.

Discussion
Many pathways including transcription, replication, DNA repair, signaling and chromosome tethering require an intact nuclear lamina network [2,[5][6][7][8]. The mechanisms of this dependence are major questions in biology with relevance to aging and a spectrum of human diseases. BAF is an essential component of the nuclear lamina network, interacting with both A-and B-type lamins as well as nuclear membrane proteins and chromatin [13]. Our use of two complementary proteomic methods (LC-MS/MS and iTRAQ), and comparisons to negative controls, yielded over 70 potential interactors. Many candidates are likely to bind lamins directly, rather than BAF, and comprise a ''lamin proteome'' within the BAF proteome. BAF has nanomolar to low micromolar affinity for key partners (histones, LEM-domain proteins, lamins) [12,28,62] and can stabilize ''ternary'' interactions between lamin A and emerin [63]. BAF-mediated stabilization of ternary complexes might account for the quantitative enrichments of lamins and histones (and potentially other candidates, e.g., SET/I2PP2A, Huntingtin) detected by iTRAQ analysis. The iTRAQ quantification was important because many nuclear proteins including LAP2, lamins and histones have been reported to bind beads nonspecifically [64], and were also detected by LC-MS/MS in our control eluates ( Table 2). Confidence that this proteome includes novel interactors for BAF was increased by independent validation: seven of 11 tested FLAG-or GFP-tagged candidates, including single-peptide ''hits'' Requiem (reported here) and G9a (Montes de Oca, Andreassen & Wilson, in preparation), co-precipitated endogenous BAF from HeLa cells. An important caveat is that these verifications were based on transient overexpression of epitope-tagged candidates, which might cause artifacts. It will be important in the future to verify the interaction between endogenous BAF and each endogenous partner.

Four human transcription factors that potentially bind BAF
BAF can regulate tissue-specific gene expression in C. elegans by binding and occupying specific promoters [26]. In mammalian retinal cells, BAF is known to bind directly to Crx and related homeodomain transcription factors and represses Crx-dependent gene activation [32]. Whether BAF binds other transcription factors, or regulates specific genes in other human tissues, is largely unexplored. Our proteome identified four transcription factors (p15/SUB1/PC4, NonO, Requiem and LEDGF) with which BAF might associate in human cells. p15/SUB1/PC4 (Table 1; hereafter ''PC4'') is an abundant nuclear protein that binds p53 and also transcriptionally co-activates p53 function [73]. The p53 protein, in turn, activates PC4 transcription as part of a feedback loop [74]. Interestingly, PC4 also binds histones, condenses chromatin [75] and helps repair damaged DNA [76,77]. NonO (aka p54 nrb ) ( Table 3), binds and helps SWI/SNF complexes activate transcription, and might be involved in splicing of the telomerase reverse transcriptase (TERT) gene [78]. Requiem (single-peptide ''hit''; Table 1), validated in this study as BAF-associated in vivo, is a relatively uncharacterized transcription factor that regulates apoptosis in the myeloid lineage [47,48]. LEDGF/p75 (single-peptide ''hit''; Table 1) is a transcription factor that can bind human immunodeficiency virus type 1 (HIV-1) integrase directly, and contributes significantly to the efficiency and location of HIV-1 integration sites in human chromosomes [79]. BAF binds the HIV-1 matrix and Gag proteins directly, and is a component of HIV-1 virions [80]. BAF and emerin were both reported to contribute significantly to the efficiency of HIV-1 integration into primary human macrophages [33]. Two subsequent studies of emerin-null cells failed to support this role for emerin [34,81], and one study reported that siRNA-mediated downregulation of BAF had only modest effects on HIV-1 infectivity in HeLa-P4 cells [34]. Further studies of macrophages and other primary cells are needed to test the proposed contribution of BAF to HIV-1 infection. Our results, which suggest a potential association between BAF and LEDGF, present a new way to study this question in HIV-1-infected primary cells.

BAF association with PARP1
BAF associated robustly in vivo with our best proteomic ''hit'', PARP1. PARP1 transfers ADP-ribose units to itself and other proteins in an NAD + -dependent manner. PARP1 can promote higher-order compaction of nucleosome arrays and repress transcription [37,82], but can also activate transcription and has major roles in DNA repair [83]. Our BAF proteome also included RecQ protein-like (Table 1), a DNA helicase involved in DNA repair that can associate with PARP1 [84]. PARP1 was shown previously to interact with A-type lamins [51,52]; our findings suggest these ternary complexes also include BAF, and a subset might also associate with emerin at the INM. How and when BAF associates with PARP1, and whether the BAF-PARP1 interaction includes RecQ, or requires lamins, are important new questions.

Implications of BAF association with RBBP4
Our independent validation suggests BAF associates preferentially with RBBP4, not RBBP7, under similar immunoprecipitation conditions. RBBP4 and RBBP7 are 89.4% identical, and the majority of the amino acid changes are conservative substitutions ( Figure 3C, grey boxes). Thus, BAF selectivity for RBBP4 over RBBP7 might be due to recognition of residues unique to RBBP4, or to an intermediary protein that discriminates RBBP4 from RBBP7. Understanding the basis for this preference might shed light on how DNA replication and DNA repair depend on the nuclear lamina network [7,85,86]. RBBP4 is a chaperone common to H3.1, H3.3 and CenH3 deposition complexes, which assemble H3 into nucleosomes both during and independently of DNA replication [87,88]. BAF can bind H3.1 directly [28], and has intriguing behavior in primary cells: BAF localizes to both the nucleus and cytoplasm of interphase cells except during S-phase, when it becomes exclusively nuclear [31]. S-phase is prolonged four hours in HeLa cells downregulated for BAF, suggesting BAF somehow facilitates S-phase progression [31], through unknown mechanisms. RBBP4, as a component of the CAF-1 complex, associates indirectly with PCNA (a lamin A-binding protein [86]) to promote chromatin assembly on newly-replicated or newlyrepaired DNA [89]. We propose BAF might dynamically stabilize chromatin complexes that include RBBP4 and PCNA by helping tether these complexes to lamins, analogous to the proposed BAFmediated stabilization of emerin binding to lamin A [63]. Our proteome included another CAF-1 subunit, p60 (Table 1); whether BAF associates with the CAF-1 complex itself remains to be tested. We also recovered both SPT16/p140 (Table 2), the large subunit of the Facilitates Chromatin Transcription (FACT) chaperone complex, and its binding partner SSRP1/FACT p80 (Table 2) which are involved in DNA replication, transcription and repair [90]. RBBP4 (validated) and the FACT complex (not yet independently tested) each suggest mechanisms by which BAF might link chromatin assembly, replication, repair and other activities to the nuclear lamina network.
Proposed novel roles for BAF and emerin in DNA damage response Human cells that transiently express certain laminopathycausing lamin A mutations, show either reduced formation of cH2AX foci after DNA damage [7] or higher proportions of cH2AX-positive cells in the absence of damage [91]. This suggests that DNA damage responses require an intact lamina network. Furthermore, nuclear alpha-II-spectrin, which binds and recruits specific repair proteins to sites of DNA damage [92], also coprecipitates with lamin A and emerin [93] and co-purifies with emerin and lamin A in multi-protein complexes isolated from HeLa cells [94]. Our new results independently implicate both BAF and emerin in cellular responses to DNA damage. In undamaged HeLa cells, both proteins associated with CUL4A and DDB2; and FLAG-BAF also associated with endogenous DDB1 in HEK293:BAF cells (unpublished observations), implying some kind of ongoing molecular surveillance or regulation that involves the nuclear envelope (emerin) and BAF. Our results suggest these interactions are regulated differentially after UV-induced damage and also potentially by apoptotic signaling, since the associations also increased at the time of PARP1 cleavage (120 min). We propose that BAF and emerin have dynamic and potentially regulated roles after DNA damage. Whether BAF and emerin function independently or concertedly, and whether their roles include the actual repair of damaged DNA, are open questions.
Our major conclusion is that this BAF proteome, validated for seven of 11 tested candidates, is a rich source of candidate proteins that might reveal how DNA damage responses, genome replication, epigenetic control and chromatin organization might be supported by BAF, and might be dynamically coupled to the nuclear envelope and nuclear lamina networks.

Materials and methods
Cell culture, nuclear protein lysates and affinity purification using BAF-beads HeLa cells were obtained from the American Type Culture Collection and cultured in Dulbecco's modified eagle's medium containing 10% fetal bovine serum. To prepare protein lysates, HeLa cells were resuspended in hypotonic buffer (10 mM HEPES pH 8, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM DTT, 0.5% NP-40), incubated on ice 10 min and then lysed by 8-10 passes through a 26G3/8 needle and syringe. Lysates were centrifuged (5000 rpm, 10 min, 4uC) to obtain the nuclear pellet, and nuclear proteins were solubilized by gentle rotation at 4uC in lysis buffer (20 mM HEPES pH 8, 150 mM NaCl, 0.1% Triton X-100, 0.2 mM EDTA, 0.2 mM EGTA, 1.5 mM MgCl 2 , 2 mM DTT, benzonase to digest DNA, and protease inhibitors) for 30 min; samples were then centrifuged (5,000 rpm, 5 min, 4uC) to remove insoluble material, and solubilized nuclear lysate proteins were either used fresh or stored at 280uC after being flash frozen with liquid nitrogen.
BAF-conjugated beads were made by incubating 30 mg purified isolated H 6 BAF dimers prepared by affinity purification and size exclusion chromatography as described [23,28], with 30 ml Ni-NTA beads (Qiagen, Valencia, CA) overnight at 4uC, at a final protein concentration of 1 mg/ml. Control beads were incubated with lysis buffer only. Ni-alone or Ni-BAF beads were then each washed twice with lysis buffer and incubated with 500 mg precleared nuclear lysate proteins, rotating, for 3 hrs at 4uC. Unbound material was removed by four washes with lysis buffer lacking detergent. Bound proteins were eluted sequentially, first with lysis buffer containing 0.3 M NaCl (no detergent), then with lysis buffer containing 1 M NaCl (no detergent), both compatible with in-solution trypsin digestion for LC-MS/MS, and finally with 0.5% SDS. The remaining beads were resuspended in 2X-SDS sample buffer (2X-SB). About 10% of each eluate was resolved by SDS-PAGE and silver stained for qualitative analysis. Two pairs of samples were analyzed in solution by LC-MS/MS: Ni-alone (0.3 M and 1 M NaCl eluates) and Ni-BAF (0.3 M and 1 M NaCl eluates). For samples analyzed by iTRAQ (below), we used 100 mg Ni-BAF beads (final BAF concentration 1 mg/ml) and 5 mg nuclear lysate proteins.

Liquid Chromatography-Coupled Tandem Mass Spectrometry
Eluted proteins were reduced and alkylated in solution using iodoacetamide, and proteolyzed with sequencing grade modified porcine trypsin (Promega, Madison, WI) as described [95]. Protein identification by LC-MS/MS analysis of peptides was performed with an LTQ ion trap MS (Thermo Fisher Scientific, San Jose, CA) interfaced with a 2D nanoLC system (Eksigent, Dublin, CA). Peptides were fractionated by reverse-phase HPLC on a 75 mm6100 mm C18 column with a 10 mm emitter using a 0-60% acetonitrile/0.5% formic acid gradient over 30 min at 300 nl/min. Peptide sequences were identified using Mascot software (version 2.2, Matrix Science, Boston, MA) to search the NCBI non-redundant database with the acquired fragmentation data. Identified sequences were confirmed manually by inspecting the fragmentation spectra. Scaffold (version Scaffold_2_04_00, Proteome Software Inc., Portland, OR) was used to validate MS/ MS based peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 0% probability as specified by the Peptide Prophet algorithm [96] embedded within Scaffold. Low probability enabled potentially significant protein identifications based on many moderateprobability peptides, but no such proteins emerged; instead nearly all identifications were based on peptides with individual confidences .80% (Table S1). Protein identifications were accepted if they could be established at greater than 89% probability based on at least 1 identified peptide. Two ''singlehit'' candidates were subsequently validated as BAF-associated in vivo, as described in Results. Protein probabilities were assigned by the embedded Protein Prophet algorithm [97]. Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony.

Isobaric Tagging for Relative and Absolute Quantitation
All four eluates (,10 mg protein, each) were dissolved in 20 ml 0.1% SDS in 0.5 M triethylammonium bicarbonate, reduced, alkylated and digested at 37uC overnight with trypsin at a proteinto-enzyme ratio of 20:1. The iTRAQ reagents (4-plex, Applied Biosystems -MDX Sciex, Foster City, CA) were dissolved in 70 ml ethanol, added to each digest, and incubated 1 hr at 22-25uC. The four samples, designated Ni-control-0.3 M eluate, BAF-0.3 M eluate, Ni-control-1 M eluate and BAF-1 M eluate, were labeled with iTRAQ reagents 114, 115, 116 and 117, respectively. After labeling, all samples were mixed and dried down to a volume of 50 ml. The combined peptide mixture was fractionated by Strong Cation Exchange chromatography (SCX) on a 1100 HPLC system (Agilent, Santa Clara, CA) using a PolySulfoethyl A column (2.16100 mm, 5 mm, 300 Å , PolyLC, Columbia, MD), by first dissolving the sample in 4 mls SCX loading buffer (25% v/v acetonitrile, 10 mM KH 2 PO 4 pH 2.8, adjusted with 1 N phosphoric acid), then loading the sample onto the column and washing isocratically for 30 min at 250 ml/min. Peptides were eluted by a gradient of 0-350 mM KCl (25% v/v acetonitrile, 10 mM KH 2 PO 4 pH 2.8) over 40 min at a flow rate of 250 ml/ min. The 214 nm absorbance was monitored and 10-SCX fractions were collected along the gradient. Each SCX fraction was dried down, dissolved in 40 ml 0.1% formic acid, and analyzed on a QSTAR Pulsar TM (Applied Biosystems -MDS Sciex, Foster City, CA) interfaced with an Eksigent nano-LC system (Dublin, CA). Peptides were separated on a reverse-phase column packed with 10 cm of C18 beads (360675 mm, 5 mm, 120 Å , YMC ODS-AQ; Waters Milford, MA) with a 10 mm emitter tip (New Objective, Woburn, MA) attached. Peptides eluting from an HPLC gradient of 5-36% acetonitrile in 0.1% formic acid, over 60 min at a 300 nl/min flow rate, were sprayed into the QSTAR. Survey scans were acquired from m/z 350-1200 with up to three precursors selected for MS/MS using a dynamic exclusion of 45 sec. A rolling collision energy was used to promote fragmentation and the collision energy range was ,20% higher than that used for unlabeled peptides due to iTRAQ tags.
MS/MS spectra were extracted and searched against the SwissProt database using ProteinPilot TM software (Applied Biosystems, Foster City, CA) with Paragon TM method and the following parameters: all species, trypsin as enzyme (one missed cleavage allowed), cysteine static modification with methylmethanethiosulfate and iTRAQ (peptide labeled at N-terminal and lysine) as sample type. Mass tolerance was set to 0.2 Da for precursor and 0.15 Da for fragment ions. The raw peptide identification results from Paragon TM Algorithm (Applied Biosystems, Foster City, CA) searches were further processed by the Pro Group TM Algorithm (Applied Biosystems, Foster City, CA) within the ProteinPilot TM software before they were displayed. The Pro Group TM Algorithm uses the peptide identification results to determine the minimal set of proteins that can be reported for a given protein confidence threshold. The peptide confidence threshold cutoff for the iTRAQ study was at least 95% (see Table 3).

UV irradiation of transiently transfected HeLa cells
About 22-24 hrs after transfection cells were UV-irradiated (20 J/m 2 ) using a Stratagene UV Stratalinker 2400 (Stratagene, La Jolla, CA), given fresh media and allowed to recover for the indicated time, then harvested with a rubber policeman, washed once with ice-cold PBS, pelleted (1,500 rpm, 5 min, 4uC) and processed for immunoprecipitation as described above.