An Analysis of CAF-1-interacting Proteins Reveals Dynamic and Direct Interactions with the KU Complex and 14-3-3 Proteins*

CAF-1 is essential in human cells for the de novo deposition of histones H3 and H4 at the DNA replication fork. Depletion of CAF-1 from various cell lines causes replication fork arrest, activation of the intra-S phase checkpoint, and global defects in chromatin structure. CAF-1 is also involved in coordinating inheritance of states of gene expression and in chromatin assembly following DNA repair. In this study, we generated cell lines expressing RNAi-resistant versions of CAF-1 and showed that the N-terminal 296 amino acids are dispensable for essential CAF-1 function in vivo. N-terminally truncated CAF-1 p150 was deficient in proliferating cell nuclear antigen (PCNA) binding, reinforcing the existence of two PCNA binding sites in human CAF-1, but the defect in PCNA binding had no effect on the recruitment of CAF-1 to chromatin after DNA damage or to resistance to DNA-damaging agents. Tandem affinity purification of CAF-1-interacting proteins under mild conditions revealed that CAF-1 was directly associated with the KU70/80 complex, part of the DNA-dependent protein kinase, and the phosphoserine/threonine-binding protein 14-3-3 ζ. CAF-1 was a substrate for DNA-dependent protein kinase, and the 14-3-3 interaction in vitro is dependent on DNA-dependent protein kinase phosphorylation. These results highlight that CAF-1 has prominent interactions with the DNA repair machinery but that the N terminus is dispensable for the role of CAF-1 in DNA replication- and repair-coupled chromatin assembly.

CAF-1 is essential in human cells for the de novo deposition of histones H3 and H4 at the DNA replication fork. Depletion of CAF-1 from various cell lines causes replication fork arrest, activation of the intra-S phase checkpoint, and global defects in chromatin structure. CAF-1 is also involved in coordinating inheritance of states of gene expression and in chromatin assembly following DNA repair. In this study, we generated cell lines expressing RNAi-resistant versions of CAF-1 and showed that the N-terminal 296 amino acids are dispensable for essential CAF-1 function in vivo. N-terminally truncated CAF-1 p150 was deficient in proliferating cell nuclear antigen (PCNA) binding, reinforcing the existence of two PCNA binding sites in human CAF-1, but the defect in PCNA binding had no effect on the recruitment of CAF-1 to chromatin after DNA damage or to resistance to DNA-damaging agents. Tandem affinity purification of CAF-1-interacting proteins under mild conditions revealed that CAF-1 was directly associated with the KU70/80 complex, part of the DNA-dependent protein kinase, and the phosphoserine/threonine-binding protein 14-3-3 . CAF-1 was a substrate for DNA-dependent protein kinase, and the 14-3-3 interaction in vitro is dependent on DNA-dependent protein kinase phosphorylation. These results highlight that CAF-1 has prominent interactions with the DNA repair machinery but that the N terminus is dispensable for the role of CAF-1 in DNA replication-and repair-coupled chromatin assembly.
DNA in eukaryotic cells is packaged into nucleosomes that form higher order chromatin. During DNA replication, histones H3 and H4 that form the core of the nucleosome are randomly partitioned to the daughter strands, and new histones are deposited de novo to maintain histone density in the wake of the replication fork (1). Although many histone chaperones exist in human cells, the primary DNA replication-coupled histone deposition factor is CAF-1 as determined from biochemical purification experiments (2). Depletion of CAF-1 from human cells has been shown to lead to defects in DNA replication fork progression, activation of the intra-S phase checkpoint, and global defects in chromatin structure (3)(4)(5).
CAF-1 is a heterotrimeric complex consisting of 150-, 60-, and 48-kDa subunits that binds to histones H3 and H4 (6,7). The p60 and p48 subunits consist largely of WD40 repeats that play a role in histone binding, and p48 is known to be a member of numerous H3/H4 binding complexes such as HAT1 and HDAC1 (8 -10). CAF-1 p60 also interacts with the histone chaperones ASF1A and ASF1B through their B-domain, which resides outside of the WD40 repeat region (11). The p150 subunit has been better characterized and contains the regions that mediate interactions with proliferating cell nuclear antigen (PCNA), 4 HP1, MBD1, BLM, and the CAF-1 p60 subunit. The PCNA mediates the connection between CAF-1 and the DNA replication fork (12)(13)(14). There are two PCNA binding sites that have been mapped in the p150 subunit, but CAF-1 is localized to replication forks through a conserved PCNA binding region motif called PIP2 in the middle of the p150 open reading frame (5,12,15). The N terminus of p150, containing a second PCNA binding region as well as the HP1-interacting region, is dispensable for DNA replication-coupled CAF-1 chromatin assembly activity in vitro and for localization to replication foci in vivo (12,16). This implies that the N-terminal region of CAF-1 might be specialized as a regulatory domain or in mediating other functions of CAF-1 in DNA damage or heterochromatin maintenance.
CAF-1 is recruited to chromatin after DNA damage in human cells, and in vitro, CAF-1 can direct chromatin assembly on many types of damaged DNA templates (16,17). Moreover, yeast cells lacking CAF-1 are sensitive to a variety of DNAdamaging agents, including bleomycin, methane methylsulfonate (MMS), and UV radiation (18,19). The BLM helicase also interacts with CAF-1 and inhibits its activity in vitro, and the WRN helicase has been shown to also interact with CAF-1 in a DNA damage-dependent manner (20,21). CAF-1 has been shown to be responsible for the deposition of new H3.1 at localized sites of UV or oxidative DNA damage, directly demonstrating that CAF-1 is involved in chromatin assembly at these loci (22). CAF-1 also functions in epigenetic inheritance, most likely through maintenance of histone density in the genome during DNA replication (24). 5 In yeast, loss of CAF-1 causes defects in the expression of genes near a telomere and at the mating type loci (18,25,26). These effects are due to aberrant gene expression in cells with low nucleosome density. 5 Yeast CAF-1 binds the histone acetyltransferase SAS-1, and human and mouse CAF-1 binds to HP1 proteins and to the methyl-DNA-binding protein MBD1, all of which are involved in the transcriptional silencing of DNA (27,28). The data from yeast and mammalian systems suggest that one way in which CAF-1 contributes to epigenetic inheritance is by the recruitment of chromatin modifiers to ensure copying of histone marks.
We are interested in defining the functions of CAF-1 in vivo and have previously shown that the p150 subunit was essential for cell viability and efficient DNA replication (3). CAF-1 requires other trans factors present in cell extracts to assemble chromatin in addition to PCNA loaded onto the replicated DNA in cis (13). We therefore sought to identify factors interacting with CAF-1 under the hypothesis that these proteins may be directly involved in replication-coupled chromatin assembly. CAF-1-interacting factors may also play a role in processes other than DNA replication-coupled chromatin assembly in which CAF-1 functions. In this work, a set of proteins interacting with full-length and a functional, N-terminally truncated CAF-1 p150 were defined by immunoprecipitationcoupled mass spectrometry analysis. Direct interactions between the CAF-1 with the KU complex and with 14-3-3 proteins, both of which are involved in DNA damage responses, were confirmed, suggesting that CAF-1 is centrally involved in histone deposition at multiple types of DNA lesions. This analysis also showed that the previously mapped N-terminal PCNA-interacting region contributes to stable CAF-1-PCNA interaction even though this region is dispensable for chromatin assembly during S phase and DNA damage.

EXPERIMENTAL PROCEDURES
Cell Culture, Generation of Transgenic Cell Lines, and siRNA Transfections-RKO cells were maintained in McCoy's medium ϩ 10% FBS. Transgenic cell lines expressing tagged siRNA-resistant CAF-1 variants were generated by standard methods (see supplemental methods). Transfection of cell lines with the GL3 (control) or 150-1 siRNA was done using a 200 nM siRNA concentration exactly as detailed previously (3). For initial preparation of nuclear extract for large scale protein purification, transgenic cells and parental RKO cells were adapted to grow in suspension medium by transfer to Joklik's medium ϩ 10% FBS followed by expansion in Joklik's medium ϩ 5% FBS and 5% calf serum. Human 293-HEK cells and HeLa cells were grown in Joklik's medium ϩ 5% calf serum.
Affinity Purification, Mass Spectrometry Analysis, and Validation Screens-400 mM NaCl nuclear extracts were prepared from cells as described (2). The nuclear extract was dialyzed against Buffer A (25 mM Tris-Cl, pH7.5, 10% glycerol, 0.01% Nonidet P-40) ϩ 25 or 100 mM NaCl and protease inhibitors, spun at 100,000 ϫ g for 30 min, and snap frozen in aliquots for storage at Ϫ70°C. For large scale tandem affinity purification, 4 ml of extract (ϳ12 mg of total protein) was thawed on ice. ATP was added to a final concentration of 1 mM, and MgCl 2 was added to 2.5 mM. After a clarification spin, the supernatant was incubated with 750 l of anti-HA 12CA5 beads with rotation for 3 h at 4°C. Beads were washed two to three times with Buffer A 25 or 100 ϩ Mg/ATP ϩ protease inhibitors, and protein was eluted with three successive 15-min incubations with 1 mg/ml HA peptide in Buffer A 25 or 100 ϩ Mg/ATP ϩ protease inhibitors at 30°C with shaking. Eluates were pooled and incubated with 200 l of anti-FLAG beads, rotated for 2 h at 4°C, and eluted as above with 1 mg/ml FLAG peptide in Buffer A 25 or 100 ϩ Mg/ATP. Eluates were pooled, and proteins were precipitated overnight from the final eluate with 20% TCA followed by shotgun mass spectrometry sequencing of the entire sample, or the eluate was run on a 5-20% Tris-glycine gel followed by staining with GelCode Blue stain. In the latter case, 10% of the eluate was run in parallel for silver staining.
Visibly stained bands were excised and processed for in-gel tryptic digestion following standard protocols. The resulting peptides were extracted and purified on C 18 ZipTips (Millipore) according to the manufacturer's protocol and resuspended in 5 l of 10% methanol, 0.1% acetic acid. A fraction of the purified peptides was analyzed on an ABI 4700 MALDI-TOF/TOF mass spectrometer using 3-indole acrylic acid as the matrix. The data were analyzed using the MASCOT search engine (Matrix Science) or the PROFOUND search engine (29). The remaining sample was analyzed by LC-MS/MS using an LTQ mass spectrometer (Thermo Electron) using standard protocols. Briefly, samples were separated using 0.075 ϫ 50mm Discovery C 18 (Supelco) picofrit columns (New Objective) developed at 0.5 l/min using a 60-min linear gradient of 10 -90% methanol in 0.1% acetic acid. The resulting spectra were analyzed with the SONARS software package (Genomic Solutions) or the Global Proteome Machine (30). Complex mixtures were digested with trypsin for 6 -18 h in the presence of 10% acetonitrile. The resulting digests were either directly analyzed by LC-MS/MS as described above or prefractionated using a 0.200 ϫ 150-mm Discovery strong cation exchange (Supelco) column. The resulting data were analyzed using the Global Proteome Machine (30).
Validation of Mass Spectrometry Hits-Proteins that were co-precipitated with CAF-1 and identified by mass spectrometry (hits) were screened for direct interaction with recombinant CAF-1 by co-immunoprecipitation of in vitro translation products produced in the presence of [ 35 S]methionine (Amersham Biosciences) using a coupled reticulocyte lysate in vitro transcription-translation system (Promega) that was programmed with individual cDNAs encoding hits cloned into the pDNR-1r backbone.
Protein Purification-Recombinant CAF-1 complex was purified from insect cells that had been infected with recombinant baculoviruses as described previously (13). To purify dephosphorylated protein, antibody bead-bound CAF-1 was treated with 20,000 units of -phosphatase (New England Biolabs) prior to elution with 5 M LiCl and further purification over a hydroxyapatite column. Protein was dialyzed against Buffer H25 (25 mM HEPES, pH 8.0, 25 mM NaCl, 10% glycerol, 0.01% Nonidet P-40, 1 mM DTT) and stored in small aliquots.
Recombinant His-tagged KU complex was purified in two steps based on a modification of two published protocols (31,32). GST and GST-14-3-3 was purified from 1 liter of freshly transformed BL21 Codon Plus cells (Stratagene) induced with 400 M isopropyl 1-thio-␤-D-galactopyranoside at 24°C for 4 h. Details of the purification can be found in the supplemental methods.
Immunoblots and Immunofluorescence-Total cell extracts or pulldown samples were separated by SDS-PAGE and transferred to nitrocellulose for immunoblot analysis. CAF-1 p150 was detected with the SS48 antibody (33), which recognizes both full-length and truncated CAF-1, at 1:1000 dilution. CAF-1 p60 was detected by a SS53 (33) or by a mixture of the SS53 and SS96 monoclonal antibodies at 1:5000. Anti-PCNA (PC10; Cold Spring Harbor Laboratory) and anti-p53 (CM1, NovoCastra) were used at 1:4000 and 1:2000, respectively. Members of the KU complex were detected with rabbit anti-KU80 and rabbit anti-KU70 at 1:2000 and rabbit anti-DNA-PKCS (all from Serotec) at 1:1000, and members of the inhibitor of histone acetyl transferases complex were detected with anti-SET, anti-ANP32A, and anti-ANP32B all at 1:1000.
Immunofluorescence analysis was performed essentially as described (3). For DNA damage recruitment, cells were pretreated with 0.02% MMS (Fluka) diluted from a freshly prepared 10% aqueous stock for 2 h prior to Triton X-100 pre-extraction, fixation, and staining for HA with the 12CA5 antibody at a 1:500 dilution.
Immunoprecipitation and in Vitro Pulldown Assays-Immunoprecipitation from nuclear extract of 293 cells or full-length and ⌬296 CAF-1 p150-expressing RKO cells was done by incubating extracts with SS24 anti-p60 beads or anti-HA 12CA5 beads with rotation for 3 h at 4°C followed by washes in Buffer A 100. Where necessary, CAF-1 was predepleted with SS1 anti-p150 beads. For dithiobis(succinimidyl propionate) cross-linking and immunoprecipitation from whole cell extracts, 2 ϫ 10 7 tagged CAF-1-expressing cells were trypsinized and washed in 1 ml PBS. Cells were resuspended into 500 l of cold PBS, dithiobis(succinimidyl propionate) (Pierce) was added to 2 mM from a 25 mM stock freshly prepared in DMSO, and cells were incubated for 15 min on ice. Tris, pH 7.5 was added to 50 mM to quench the cross-linking reaction, and cells were incubated another 15 min. Cell extracts were prepared by resuspension of the pellet in 500 l of radioimmune precipitation assay buffer ϩ protease inhibitors followed by immunoprecipitation with 25 l of washed anti-HA 3F10 antibody beads (Roche Applied Science) for 3 h at 4°C followed by washes in radioimmune precipitation assay buffer. Cross-links were reversed by boiling the beads in SDS-PAGE sample buffer (containing 0.1 M DTT) prior to loading on SDS-polyacrylamide gels for immunoblotting.
To map CAF-1-KU interactions, CAF-1 subunits were in vitro translated and diluted with 200 l of imidazole wash buffer (Buffer A 100 ϩ 2.5 mM MgCl 2 , 1 mM ATP, 10 mM imidazole, 2 mM ␤-mercaptoethanol, protease inhibitors), and 100 l of diluted reactions was added to 25 l of washed Talon resin (Clontech) with or without 3 g of recombinant His-KU70/ KU80 complex. Beads were washed with wash buffer, resus-pended in SDS sample buffer, and separated by SDS-PAGE, and gels were dried and exposed to film.
For GST 14-3-3 pulldown experiments, 5 g of GST or GST 14-3-3 was incubated for 2 h with 1 g of recombinant dephosphorylated CAF-1 (pretreated for 1 h with different kinases) in 100 l of Buffer A 100 ϩ protease inhibitor mixture and 20 l of glutathione-Sepharose beads (Amersham Biosciences). Beads were washed three times in Buffer A 100, resuspended in sample buffer, and processed for immunoblotting. R18 peptide (Biomol) was added from an aqueous stock to reactions as indicated.
Immunodepletion/Chromatin Assembly Assays-The KU complex was depleted from 293 S100 extract by successive incubation with Gammabind G beads (Amersham Biosciences) to which the anti-KU80 monoclonal antibody KU15 (Sigma) or the control anti-T-antigen monoclonal antibody PAB419 (Cold Spring Harbor Laboratory antibody facility) was bound. Replication-coupled chromatin assembly assays have been described previously (13).
Analysis of Global Nucleosome Structure-Cells were collected after siRNA treatment by trypsinization, and washed once in PBS and four successive times with 500 l of extraction buffer (1% Triton X-100, 0.25 M sucrose, 10 mM MgCl 2 , 10 mM HEPES, pH 8.0), spinning cells for 5 min at 2700 ϫ g in between washes. Cells were washed another two times in extraction buffer ϩ 0.02% Triton X-100 instead of 1% Triton X-100 and resuspended in a normalized volume of this buffer. 250 l of cell suspension was transferred to a new tube on ice to which CaCl 2 was added to a final concentration of 1 mM. Micrococcal nuclease was added, the mixture was transferred to a 37°C shaker, and 75-l samples were collected at 2, 3, and 4 min and transferred to tubes containing 37.5 l of stop solution (1.25% SDS, 25 mM EGTA). Samples were treated with RNase and Pronase, ethanol-precipitated, run on a 1.5% agarose 1ϫ Tris acetate-EDTA gel, and stained with ethidium bromide to visualize the DNA.
DNA Damage Dose-Response Assays-For dose-response assays, cells were treated with two rounds of control or 150-1 siRNA transfection, and at the 48-h time point, cells were trypsinized and plated in triplicate in 96-well plates at 1.9 ϫ 10 5 cells/well in the presence of DNA-damaging agents. All drugs were diluted in 10% DMSO, MMS from a freshly prepared 10% aqueous stock, and etoposide (Sigma) and 4-nitroquinoline 1-oxide (Sigma) from 100 and 10 mM stocks, respectively, in 100% DMSO. 10 l of diluted stock was added per well. Cells were grown for 72 h and stained with sulforhodamine B (Fluka) exactly as published (34). Graphs and IC 50 values were generated with Prizm software (GraphPad).

Generation and Analysis of CAF-1-expressing Cell Lines-
Cell lines expressing N-terminally FLAG-HA-tagged fulllength and truncated (⌬1-296) CAF-1 p150 were generated to facilitate the biochemical analysis of CAF-1-containing protein complexes in human cells. The full-length transgene was engineered to be RNAi-resistant to the previously used 150-1 siRNA with six silent point mutations, and the ⌬296 version of CAF-1 was not targeted by this siRNA. Single cell clones expressing a wild-type level of protein were identified. Interestingly, cell lines expressing ⌬296 CAF-1 expressed reduced amounts of full-length CAF-1 in immunoblots, suggesting that there is tight control over the level of the CAF-1 p150 subunit in cells (Fig. 1A).
Both the full-length and truncated epitope-tagged transgenes complemented depletion of endogenous CAF-1 p150 by the 150-1 siRNA. The transgenes were functional in vivo as measured by their ability to rescue the replication defect seen when endogenous CAF-1 p150 was depleted from cells (Fig.  1B). Unlike CAF-1 p150-depleted cells, cells containing the complementing transgene did not activate p53 and showed normal global nucleosome spacing after RNAi treatment (Fig.  1, A, C, and D). Also, both full-length and truncated CAF-1 showed the normal localization to DNA replication foci in S phase cells (Figs. 1B and 7D). This result demonstrates conclusively that the N terminus of CAF-1 is completely dispensable for the essential functions of CAF-1 and is consistent with the notion that the key function of CAF-1 in human cells resides in its ability to deposit histones. These experiments also showed that the epitope tag did not interfere with CAF-1 function in vivo.
Tandem Affinity Purification of CAF-1-CAF-1 was purified from nuclear extracts from cells expressing the full-length or ⌬296 large subunit using the parental RKO cell line as a negative control (Fig. 2). In our initial experiments, very few CAF-1-interacting proteins were found when using buffers at standard ionic strength with the exception of histones H3 and H4. However, when salt levels were dropped to 25 mM, many more specifically co-purifying proteins were seen (Fig. 2B). This suggests that many proteins are dynamically associated with the core CAF-1-histone complex. It is important to note, however, that CAF-1-containing complexes were previously extracted from chromatin using high salt because all of the p150 subunit in human cells is tightly associated with chromatin.
Proteins co-purifying with CAF-1 (Fig. 2, A and B) were identified by sequencing individual bands or by one-and two-dimensional LC-MS/MS sequencing of the entire immunoprecipitate, and the results were collated from the two approaches. To be considered a hit, greater than three different peptides from a protein had to be identified and judged significant by the software used in the different experiments. Proteins found at high levels in the control RKO purification were discarded as probable background. Interacting proteins for which subsequent rounds of verification were done are displayed in Table 1, and the complete data are presented in supplemental Table S1. Beyond the well known core CAF-1-interacting proteins consisting of histones H3 and H4, PCNA, and the HP1 proteins, there was little overlap in proteins identified with the different approaches. Note that the interaction between CAF-1 and PCNA was very salt-sensitive, whereas the interactions between CAF-1 and histones and HP1 were more salt-resistant. HP1 proteins were not found in the immunoprecipitates from N-terminally truncated CAF-1 as expected, but a significantly reduced amount of PCNA was found to associate with the ⌬296 version of CAF-1 even though the CAF-1 subunits were present at equal or higher levels in this purification. This result implies that the N terminus of the p150 subunit contributes to PCNA binding by CAF-1 and provides support for the relevance of the N-terminal PCNA binding region (12,16). We also found components of the DNA-dependent protein kinase (DNA-PK) in multiple experiments, although there was some variability in recovery. Using the shotgun approach, we found the catalytic DNA-PKCS subunit, and the KU80 subunit of the KU complex was also detected in a band recovered from the truncated CAF-1 immunoprecipitate.
A number of other proteins were found in only one of the analysis approaches used. HMGB2, RBM3, NONO, and a number of other proteins were found only using the two-dimensional shotgun strategies and not in bands isolated from a gradient gel. By contrast, DDX41, SET, ANP32A, and YWHAB/Z were only detected in bands isolated from the gel. DDX41, a putative RNA helicase, and YWHAB/Z, the ␤ and 14-3-3 isoforms, were only detected in bands from the full-length CAF-1 immunoprecipitate. 14-3-3 ␤ and are very similar, and several peptides common to both isoforms were identified. However, we also identified some -specific peptides, implying that this isoform was the primary 14-3-3 protein in the precipitate. The components of the histone-binding inhibitor of histone acetyl transferases complex, SET, ANP32A, and ANP32B, were found in individual bands in both full-length and truncated immunoprecipitates (35). Although the inhibitor of histone acetyl transferases proteins were one of the strongest hits when we sequenced individual bands, we failed to detect significant amounts of all but ANP32B in the shotgun sequencing approaches. This might have been due to the acidic nature of the proteins or to their smaller size.
Two extra bands were clearly visible in the ⌬296 CAF-1 immunoprecipitate (Fig. 2B). These corresponded to ribosomal proteins RPS3 and RPS9. RPS3 has been reported to play a role

CAF-1 Interacts with KU and 14-3-3
in DNA repair as part of an apurinic/apyrimidinic endonuclease activity (36). Rather than discarding these proteins as probable background, they were therefore included in subsequent validation experiments.
Validation of Mass Spectrometry Hits-Because different sets of CAF-1-interacting proteins were identified depending on the analysis approach used, verification of these putative interactions was essential. Therefore, cDNAs encoding a subset of these hits were cloned into the pDNR-1r series shuttle vector (Clontech) that is amenable to in vitro transcription and translation and can be transposed in-frame into plasmids encoding different epitope tags. Some cDNAs were also obtained in a similar vector backbone from the FLEXGene consortium at the Harvard Institute of Proteomics. Individual proteins were screened for direct interactions with recombinant CAF-1 by incubating in vitro translated protein with baculovirus-produced CAF-1 immobilized on antibody beads. These experiments demonstrated that 14-3-3 (YWHAB encoding 14-3-3 ␣/␤ and YWHAZ encoding 14-3-3 /␦) and the KU70/80 complex interact directly with CAF-1 ( Fig. 3A and supplemental Fig. S1). Note that the subunits of the KU complex did not interact with CAF-1 when translated individually, indicating that KU complex formation was required for the interaction. Because KU has DNA and RNA binding activity (37), the interaction was reproduced in the presence of DNase or RNase, suggesting that it is nucleic acid-independent (supplemental Fig.  S2). We were unable to detect any direct interaction between CAF-1 and SET, ANP32A, or ANP32B even when co-translated or co-incubated with CAF-1 in the presence of histones H3 and H4. CAF-1 also did not interact when SET and ANP32B were co-translated either in the presence or absence of histones H3 and H4 (data not shown); however, SET was seen to interact with CAF-1 in transient transfection experiments (supplemental Fig. S1).
As a final test, CAF-1 was immunoprecipitated with anti-CAF-1 monoclonal antibodies from 293 nuclear extract. To control for nonspecific interactions with the antibody resin, immunoprecipitations were also done from CAF-1-depleted extracts. In these experiments, all the components of the DNA-PK complex were present in the CAF-1 immunoprecipitate, whereas ANP32A and ANP32B were not identified (Fig.  3B). Collectively, these data further confirm that CAF-1 interacts with DNA-PK, most likely through the KU complex, and reinforce the idea that CAF-1 interacts with some 14-3-3 proteins. Interactions between CAF-1 and SET, HMGB2, and DDX41 were also seen but were most likely due to indirect binding.
Characterization of KU-CAF-1 Interaction-The KU complex is intimately involved in the non-homologous end joining  of double strand breaks and interacts with DNA-PK (38). It is also involved in tethering silent genes to the nuclear periphery and recruitment of silencing regulators (39,40). We were unable to detect an interaction between KU and CAF-1 in a reverse immunoprecipitation reaction with anti-KU antibodies, which can be attributed to KU being far more abundant than CAF-1 in human cell extracts. His-tagged KU complex was therefore purified from baculovirus-infected insect cells, incubated with recombinant CAF-1, and precipitated with Talon resin, which binds the His tag. CAF-1 was purified in the presence but not in the absence of His-KU70/80, indicating that this interaction was specific although not stoichiometric (Fig.  4A). These results also suggest that the interaction between CAF-1 and KU is direct and not dependent on nucleic acid. Talon resin pull-down experiments with His-KU and in vitro translated CAF-1 were used to map the CAF-1 subunit interaction. Only CAF-1 p150 interacted robustly with His-KU70/80. Further mapping experiments using an overlapping deletion set of CAF-1 p150 fragments demonstrated that the interacting region resided within the C terminus and most likely between residues 469 and 703 of CAF-1 (Fig. 4B). Attempts to delineate this interacting region further were hampered by the nonspecific stickiness of smaller deletion fragments.
The 469 -703 region of CAF-1 p150 lies within the region essential for chromatin assembly. KU has also been reported to interact with NASP, a histone chaperone found in complex with H3 that is also essential for cell proliferation in vivo, sug-gesting a widespread association between KU and chromatin assembly factors (41)(42)(43). To test whether KU played a role in CAF-1-mediated chromatin assembly, KU-depleted extracts were tested for chromatin assembly activity. These extracts were still able to support SV40 T-antigen-mediated DNA replication as well as chromatin assembly, indicating that KU does not play a role in chromatin assembly per se (Fig. 5).
Although a DNA-KU-CAF-1 complex could not easily be detected by gel shift analysis (data not shown), it remains possible that KU might recruit CAF-1 to sites of double-stranded DNA breaks in vivo. This hypothesis is attractive because sites of non-homologous end joining generally have limited DNA synthesis and may thus be deficient in PCNA that would otherwise recruit CAF-1. If true, this hypothesis would predict that an increase in the CAF-1 and KU interaction should be seen after DNA damage. However, we were unable to detect such increases in the KU-CAF-1 interaction after multiple types of DNA damage, including ionizing radiation, MMS, Zeocin, and extended hydroxyurea treatment. Nevertheless, the p150 and p60 subunits of CAF-1 proved to be very good in vitro substrates for DNA-PK in a DNA-dependent manner (supplemental Fig. S3A and data not shown). Thus, we cannot rule out a role for KU in the recruitment of CAF-1 to sites of double-stranded DNA breaks in the cell.
Phosphorylation-dependent Interaction of 14-3-3 with CAF-1-14-3-3 proteins are phospho-Ser/Thr-binding proteins that bind targets primarily harboring one of two types of recognition motifs (44). Although no strong 14-3-3 consensus binding sites were detected in CAF-1 subunits, we did find that GST-tagged 14-3-3 was able to bind recombinant CAF-1 in vitro (Fig. 6A). Because CAF-1 prepared from insect cells is phosphorylated,  . KU is not required for chromatin assembly in vitro. A, the KU complex was depleted from cytosolic 293 S100 extracts by five successive incubations with anti-KU80 antibody or mock-depleted (depl) with the T-antigen monoclonal antibody PAB419. Aliquots of each round were immunoblotted with anti-KU antibody. B, replication-coupled chromatin assembly assays were performed with increasing amounts of mock-depleted or KUdepleted extract in the presence of CAF-1, T-antigen, and [␣-32 P]dATP to visualize replicated DNA. Assembled replicated DNA is negatively supercoiled and migrates more quickly in the gel. inp, input.
large amounts of phosphatase-treated recombinant CAF-1 were purified and combined with various protein kinases and ATP. CAF-1 was a good in vitro substrate for the cyclin-dependent kinases (CDKs) in addition to DNA-PK, although they phosphorylated the p150 and p60 subunits with different specific activities (supplemental Fig. S3A). In contrast to phosphorylated recombinant CAF-1, phosphatase-treated CAF-1 did not interact with GST-14-3-3 (Fig. 6A, compare lanes 1 and 2  (dephosphorylated) with lanes 9 and 10 (phosphorylated)). When kinase-treated CAF-1 was pulled down with GST-14-3-3 , only DNA-PK-treated protein was efficiently precipitated.
These data indicate that 14-3-3 interacted directly with phosphorylated CAF-1 and that this interaction was supported most efficiently by DNA-PK treatment but not treatment with any of the CDKs tested. ATP was also required for the CAF-1-14-3-3 interaction, further demonstrating that the interaction depends on the kinase activity of DNA-PK (data not shown).
CAF-1 prepared from baculovirus can contain varying amounts of a co-purifying protein that we identified as the baculovirus protein NMAP. Baculovirus NMAP was also phosphorylated by DNA-PK (supplemental Fig. S3A) but is unlikely to contribute to the interaction with 14-3-3 because its interaction was not salt-resistant, whereas p150 and p60 were still bound under high salt conditions (supplemental Fig. S3B). Moreover, GST-14-3-3 but not GST precipitated CAF-1 p150 and p60, but not PCNA, from human cell extracts, indicating that baculovirus NMAP was not required for the CAF-1 interaction (supplemental Fig. S3C). Because the interaction between CAF-1 and GST-14-3-3 depends on DNA-PK phosphorylation in vitro, we tested whether this interaction was enhanced in extracts from damaged cells, but no change in the amount of interacting protein was seen after treatment with a variety of DNA-damaging agents (data not shown).
14-3-3 proteins are dimeric and can act as scaffolds to facilitate interactions between two phosphorylated proteins (45). CAF-1 requires both kinase and phosphatase activities for chromatin assembly and has been reported to dimerize in a phosphorylation-dependent manner (46,47). It is also becoming clear that histone chaperones bind soluble histone H3 and H4 as dimers, not tetramers, suggesting that the regulated dimerization of CAF-1 might contribute to the deposition of two histone H3/H4 dimers onto DNA to form the H3/H4 core tetramer (43,48). To test whether 14-3-3 might play a role in chromatin assembly, we interfered with the CAF-1-14-3-3 interaction using a peptide known to disrupt interactions between 14-3-3 and other targets (49). The R18 peptide was able to inhibit the 14-3-3-CAF-1 interaction even at the lowest concentration used (Fig. 6B). In contrast, titration of the R18 peptide into DNA replication-coupled chromatin assembly assays had no effect on chromatin assembly even at very high levels, suggesting that the 14-3-3 interaction plays no role in CAF-1-mediated assembly at DNA replication forks (Fig. 6C). This conclusion was also supported by the fact that 14-3-3 was not identified in the immunoprecipitate of the truncated CAF-1 p150 subunit.
Contributions of N Terminus of CAF-1 p150 Subunit to PCNA Binding-The interaction between CAF-1 and KU and the DNA-PK-dependent interaction between 14-3-3 proteins and CAF-1 provide further support for the idea that there is an intimate association between the DNA repair machinery and the replication-dependent nucleosome deposition machinery. PCNA is recruited to the site of many types of DNA damage and may play a role in the recruitment of CAF-1 onto damaged DNA. Because the N terminus of CAF-1 appears to contribute to PCNA binding even though it is not essential for normal chromatin assembly in vivo and in vitro, we examined whether this region contributed to the DNA damage response.
The decreased interaction between N-terminally truncated CAF-1 and PCNA compared with the interaction between full-FIGURE 6. CAF-1 interacts directly with 14-3-3 in DNA-PK-dependent manner. A, dephosphorylated CAF-1 was untreated or incubated with different cyclin/CDKs (Cyc) as well as DNA-PK and assayed for interaction with GST-14-3-3 (labeled "14") in pulldown assays. GST was used as a control (labeled "C"). CAF-1 purified without phosphatase treatment from insect cells was used as a positive control (last two lanes). 1% of input and 10% of pulldown reactions were immunoblotted with antibody against CAF-1 p150 or p60. B, the R18 peptide blocks the CAF-1-14-3-3 interaction. Dephosphorylated CAF-1 was treated with DNA-PK and assayed for interaction with GST-14-3-3 in the absence or presence of 2 or 20 M R18 peptide. Some nonspecific interaction between CAF-1 p150 and GST was induced at 20 M R18. 1% of input (inp) and 10% of pulldown reactions are shown. C, R18 does not block CAF-1 mediated chromatin assembly. Replication-coupled chromatin assembly assays were performed with increasing amounts of R18 peptide (2-200 M) at low and high CAF-1 concentrations (1:1 and 2:1 CAF-1:DNA ratios, respectively). unt, untreated; Cont, control.

CAF-1 Interacts with KU and 14-3-3
MARCH 25, 2011 • VOLUME 286 • NUMBER 12 length CAF-1 p150 and PCNA was first reproduced in two different assays. First, epitope-tagged full-length and truncated CAF-1 was purified from nuclear extracts by anti-HA precipitation under low and moderate stringency. Under these conditions, PCNA was only detected in the full-length immunoprecipitate but not in the truncated precipitate by immunoblot analysis, and this interaction was salt-sensitive (supplemental Fig. S4). Because the interaction between PCNA and CAF-1 was labile, we then cross-linked proteins in situ in tagged CAF-1-expressing cell lines using the membrane-permeable, DTT-cleavable cross-linking agent dithiobis(succinimidyl propionate) followed by immunoprecipitation under denaturing conditions. With this protocol, PCNA was detected in the truncated CAF-1 immunoprecipitate albeit at much lower levels than with the full-length CAF-1 (Fig. 7A). These results demonstrate that the N terminus of CAF-1 indeed contributes to PCNA binding in vivo even though this binding is dispensable for the essential functions FIGURE 7. N terminus of CAF-1 contributes to PCNA binding but is not required for survival after DNA damage. A, RKO, full-length CAF-1-, and truncated CAF-1-expressing cells were cross-linked with 2 mM dithiobis(succinimidyl propionate), quenched, and extracted with denaturing radioimmune precipitation assay buffer. Tagged CAF-1 was immunoprecipitated (IP) with anti-HA antibody beads, and pulldowns were immunoblotted with antibodies against CAF-1 p150 and PCNA. 1% of input (inp) and 40% of pulldown reactions are shown. B, full-length and truncated CAF-1 p150-expressing cells were treated with control or p150-targeting siRNAs (oligo 1) and plated onto 96-well plates with increasing amounts of etoposide, MMS, and 4-nitroquinoline 1-oxide (4NQO). Cell growth after 3 days was quantified with sulforhodamine B staining. Results plotted are from two independent experiments assayed in triplicate. Error bars are ϮS.E. C, triton-resistant full-length or truncated CAF-1 was visualized after incubation with or without 0.02% MMS for 2 h. Cells were pre-extracted with 0.5% Triton X-100 followed by indirect immunofluorescence with anti-HA antibodies to detect tagged CAF-1.
To examine whether the N terminus of CAF-1 plays a role in the DNA damage response, endogenous CAF-1 was depleted by RNAi from cells expressing the siRNA-resistant full-length or truncated p150 subunit, and the cells were then plated with increasing concentrations of the DNA-damaging agent MMS, 4-nitroquinoline 1-oxide, or etoposide. Growth was measured after 3 days using sulforhodamine staining. We observed no significant difference in the sensitivity of cells expressing either full-length or truncated CAF-1 to the different DNA-damaging agents, suggesting that the N terminus of CAF-1 does not contribute to the survival of cells after DNA damage (Fig. 7B).
CAF-1 is redistributed to a Triton-resistant chromatin fraction after many types of DNA damage, including MMS treatment (4,17). Therefore, even though the CAF-1 p150 N terminus does not contribute to survival after DNA damage, it might still contribute to the recruitment to chromatin after DNA damage. However, when redistribution and Triton resistance was monitored after MMS treatment, truncated CAF-1 p150 was recruited to chromatin as efficiently as full-length protein (Fig. 7C). Induction and recovery from DNA damage as monitored by checkpoint responses in truncated CAF-1-expressing cells also were completely normal, indicating that DNA repair pathways were intact (data not shown). Collectively, the data strongly suggest that the PCNA binding to the CAF-1 p150 N terminus plays no detectable role in histone deposition after DNA damage or during DNA replication.

DISCUSSION
CAF-1 in yeast has long been linked to the DNA damage response because cells lacking the complex are more sensitive to a variety of DNA-damaging agents. In human cells, CAF-1 is known to interact with the BLM and WRN helicases at sites of prolonged replication fork arrest and is recruited to sites of UV radiation-induced damage to deposit histones de novo (20 -22). We now show that in human cells CAF-1 binds directly to the KU complex that is part of DNA-PK, that CAF-1 is phosphorylated in vitro on its p150 and p60 subunits by this kinase, and that the phosphorylation of CAF-1 by DNA-PK supports an in vitro interaction with 14-3-3 . None of these interactions are important for chromatin assembly as such, and thus the most parsimonious explanation is that they reflect a role of CAF-1 in the DNA damage response.
Despite extensive efforts, however, an increased interaction between CAF-1 and either the KU complex or 14-3-3 proteins after DNA damage was not detected in vivo, and no CAF-1-KU complex was seen on DNA in vitro. One possibility is that these interactions are enhanced in the cell after DNA damage but lost after extraction: CAF-1 is tightly bound to chromatin, and only 50% of the protein is salt-extractable and available for study using our approach. Alternatively, the interaction between KU, 14-3-3, and CAF-1 may be related to DNA damage indirectly or to some other cellular process. An indirect relationship between DNA-PK and CAF-1 is supported by a report that the re-expression of CAF-1 in quiescent normal human fibroblasts after bleomycin treatment is dependent upon DNA-PK (4). This study did not determine whether the induction of CAF-1 was at the post-transcriptional level, but it is conceivable that direct phosphorylation of CAF-1 by DNA-PK stabilizes the protein, perhaps in a 14-3-3-dependent manner.
14-3-3 proteins have long been linked to the DNA damage response. 14-3-3 expression is induced by p53, and it mediates G 2 arrest by contributing to the sequestration of CDK1/ cyclin B in the cytoplasm (51). 14-3-3 proteins in fission yeast bind the checkpoint kinase Chk1 and are important in mediating the DNA damage response (52). In budding yeast, 14-3-3 proteins interact with the DNA damage response regulator RAD53 and positively regulate its function, and specific mutants in 14-3-3 proteins have been identified that increase the sensitivity to DNA-damaging agents, perhaps through an effect on global chromatin structure (53)(54)(55). Several large scale proteomics analyses have also identified interactions between 14-3-3 proteins and the DNA damage response proteins CHK1, RAD50, and HUS1 in human cells (56,57). It is important to note that the 14-3-3-CAF-1 interaction may have nothing to do with DNA damage, however, and 14-3-3 proteins have many functions in the cell outside of the DNA damage response. The DNA-PK-dependent interaction reflects a requirement for phosphorylation but not necessarily for DNA damage because in vitro treatment with DNA-PK may lead to some nonspecific phosphorylation events.
The KU complex is also involved in other cellular functions, including telomere maintenance and potentially the initiation of DNA replication (58 -60). In yeast, the KU complex has recently been shown to be involved in heterochromatin maintenance at the silent mating type loci in part by the recruitment of the silencing regulator Sir4 to epigenetically silenced genes (40,61). Thus, KU can promote silencing of gene expression.
The KU complex is also involved in recovery from a cell cycle checkpoint arrest caused by double-stranded DNA breaks (62). Recently, overlapping roles for CAF-1 and ASF1 have been described in turning off the double-stranded DNA break-induced cell cycle checkpoint arrest after repair has occurred (63). Therefore, the CAF-1-KU interaction may play a role in recruiting CAF-1 to sites of double-stranded DNA break damage to reassemble the appropriate chromatin structure at that site, and once this is achieved, cell cycle progression could occur. Because the global replication defects caused by CAF-1 deficiency globally distort chromatin structure, a definitive answer to this question will require the generation of CAF-1 mutants unable to bind KU.
Function of CAF-1 p150 N Terminus-CAF-1 p150 has long been reported to have two PCNA binding sites, one at the N terminus (PIP1), which is very robust when attached to GST for in vitro binding assays, and one that is internal (PIP2; Refs. 5, 12, 15, and 16). The PIP2 PCNA binding region is essential because it is conserved from yeast to human and is necessary to CAF-1 function in vivo for recruiting CAF-1 to sites of DNA replication and in vitro chromatin assembly assays (12). The role of the PIP1 binding site is not known. We now conclusively show that the N-terminal PCNA binding site is involved in PCNA binding in the cell but is dispensable for normal CAF-1 functions in DNA replication-and damage-coupled assembly. Similar results with N-terminally truncated CAF-1 p150 have also been reported using a conditional knock-out approach in chicken DT40 cells, although PCNA binding to CAF-1 was reportedly unaffected in this cell line (64).
Other CAF-1-interacting Proteins-Although we identified numerous novel CAF-1-interacting proteins in this analysis, we did not identify some previously identified interactors such as ASF1, MDB1, BLM, and WRN. MBD1 and BLM were identified as CAF-1-interacting proteins on the basis of two-hybrid screens and may thus interact with only a small minority of CAF-1 in the cell and be below the threshold of detection. Direct interactions between human ASF1 and CAF-1 have been demonstrated, and ASF1 is a component of the soluble H3.1 and H3.3 complexes (23,43). It may be that the ASF1-H3/H4 complex is largely separate from the CAF-1-H3/H4 complex in the cell and functions to deliver histones to CAF-1. Indeed, there are several lines of biochemical evidence that suggest that this may be one function of ASF1 in human cells (50). If this were true, only a small amount of CAF-1 might bind ASF1 at any particular time, perhaps below the detection limit of our affinity purification approach that involved extraction of CAF-1 off chromatin with high salt. Seen in this light, the proteins that were identified in this screen might represent factors that are bound to a significant percentage of CAF-1 albeit dynamically. It will thus be interesting to further investigate their function in the context of CAF-1 by identifying mutant versions of CAF-1 that have specific defects in binding one or more of these proteins.