Poly(ADP-ribose) Polymerase-1 (PARP-1) Contributes to the Barrier Function of a Vertebrate Chromatin Insulator*

The prototypic chromatin insulator cHS4 has proven effective in reducing silencing chromosomal position effects in a variety of settings. Most of this barrier insulator activity has been mapped to a 250-bp core region, as well as to several proteins that bind this region. However, recent studies from our laboratory demonstrated that an extended 400-bp core region of the cHS4 element is necessary to achieve full barrier insulator activity when used as a single copy in the context of recombinant gammaretroviral and lentiviral vectors. In this study, electrophoretic gel mobility shift assays revealed specific DNA-protein binding activities associated with the distal portion of this extended core region. Affinity purification and tandem mass spectrometry studies led to the identification of one of these proteins as poly(ADP-ribose) polymerase-1 (PARP-1). The identity of this binding activity as PARP-1 was subsequently verified by a variety of biochemical studies in vitro and by chromatin immunoprecipitation studies in vivo. Functional studies with gammaretroviral reporter vectors in cell lines and primary mouse bone marrow progenitor cultures showed that cHS4 barrier activity was abrogated upon mutation of the putative PARP-1-binding site or upon treatment with a PARP inhibitor, respectively. The barrier activity of the cHS4 element was also found to be abrogated in studies using bone marrow from Parp1-null mice. Taken together, this study demonstrates that binding of PARP-1 plays a key functional role in the barrier activity of the extended cHS4 insulator core element.

The prototypic chromatin insulator cHS4 has proven effective in reducing silencing chromosomal position effects in a variety of settings. Most of this barrier insulator activity has been mapped to a 250-bp core region, as well as to several proteins that bind this region. However, recent studies from our laboratory demonstrated that an extended 400-bp core region of the cHS4 element is necessary to achieve full barrier insulator activity when used as a single copy in the context of recombinant gammaretroviral and lentiviral vectors. In this study, electrophoretic gel mobility shift assays revealed specific DNA-protein binding activities associated with the distal portion of this extended core region. Affinity purification and tandem mass spectrometry studies led to the identification of one of these proteins as poly(ADP-ribose) polymerase-1 (PARP-1). The identity of this binding activity as PARP-1 was subsequently verified by a variety of biochemical studies in vitro and by chromatin immunoprecipitation studies in vivo. Functional studies with gammaretroviral reporter vectors in cell lines and primary mouse bone marrow progenitor cultures showed that cHS4 barrier activity was abrogated upon mutation of the putative PARP-1-binding site or upon treatment with a PARP inhibitor, respectively. The barrier activity of the cHS4 element was also found to be abrogated in studies using bone marrow from Parp1-null mice. Taken together, this study demonstrates that binding of PARP-1 plays a key functional role in the barrier activity of the extended cHS4 insulator core element.
Recent advances in the analysis of global patterns of gene regulation and genomic architecture have led in part to a greater appreciation for the role of a class of cis-regulatory elements known as chromatin insulators. Chromatin insulators are DNA elements that help form functional boundaries between adjacent chromatin domains (as reviewed Refs. [1][2][3][4]. They have been reported in species as diverse as yeast and man and have been shown to play a key role in global gene regulation. There are two basic classes of chromatin insulators: enhancer-blocking insulators and barrier insulators. Enhancerblocking insulators prevent enhancer-mediated transcriptional activation of adjacent promoters, whereas barrier insulators block the encroachment of silencing heterochromatin into adjoining regions of open chromatin that are otherwise transcriptionally permissive. The elements that mediate these activities are physically separable and mechanistically distinct, although they are sometimes found in close proximity. Of the insulator elements identified to date, most are of the enhancer-blocking variety. In vertebrates, the function of these elements is mediated through the zinc finger DNA-binding protein CCCTC-binding factor (CTCF) 2 (as reviewed Refs. 3 and 4). In general, these elements are thought to function through physical interactions between adjacent insulator elements or through CTCF-mediated tethering of the chromatin fiber to structural elements within the nucleus (5). Fewer barrier insulators have been reported in the literature, and less is known about the mechanisms underlying their function. In the context of higher eukaryotes, these elements appear to function through the formation of constitutionally "open" chromatin regions with histone modifications that block the self-perpetuating spread of heterochromatin (as reviewed in Ref. 3).
Much of what is known about the structure and function of vertebrate chromatin insulators comes from studies of cHS4 (chicken ␤-globin locus control region DNase-hypersensitive site 4). This prototypic chromatin insulator exhibits separable enhancer-blocking and barrier insulator activities in a variety of settings (6). Most of both activities were initially mapped to a 250-bp fragment containing the dominant DNase-hypersensitive site (6,7), although full activity requires the use of multiple copies of this core. Subsequent deletion, footprinting, biochemical, and functional studies identified five DNA-binding components of this fragment (as reviewed in Ref. 3). One site (FII) is bound by CTCF and is both necessary and sufficient for enhancer blocking (8). Another site (FIV) is bound by the proteins USF1 and USF2 and plays an important role in cHS4 barrier activity (9,10). These proteins serve to recruit histonemodifying enzymes that create a peak of hyperacetylated histone 3 Lys 9 . This competes in turn with methylation of this moiety, a modification that is involved in heterochromatin protein 1 recruitment and heterochromatin progression (as reviewed in Ref. 3). Three other sites of the 250-bp cHS4 core involved in barrier activity are bound by Vezf1/BGP1, a protein that regulates genomic DNA methylation through its effects on DNA methyltransferase Dnmt3b (11). * This work was supported, in whole or in part, by National Institutes of Health Unlike more conventional cis-regulatory elements, chromatin insulators do not exhibit inherent transcriptional enhancing or repressing activities on their own. This property, combined with the ability of insulators such as cHS4 to effectively block both enhancers and silencing heterochromatin, has led several groups to investigate the use of chromatin insulators for improving the expression and safety of recombinant retroviral vectors commonly used in gene therapy applications (see Refs. 12 and 13 and references therein). Much of this work has focused on the use of the cHS4 insulator to reduce the silencing of such vectors by chromosomal position effects. While seeking to define the minimum sequences necessary for this activity, we found that flanking a gammaretroviral vector with a single copy of the 250-bp cHS4 core is ineffective in preventing transgene silencing. In contrast, a 400-bp fragment containing the 250-bp core plus 3Ј-flanking sequence protects vector expression to the same degree as the 1.2-kb fragment used to initially characterize the cHS4 insulator in this setting (12). We report here studies that identify a DNA-binding activity associated with this extended core region as poly(ADP-ribose) polymerase-1 (PARP-1). We also provide functional evidence that PARP-1 plays a direct role in the barrier insulator activity of this extended core region. The protein PARP-1 has been physically associated with the activity of enhancer-blocking insulators in a variety of settings and has been shown to modify CTCF. However, this is the first functional demonstration of a role for PARP-1 in the insulating activity of cHS4 and the first demonstration of a role for PARP-1 in the barrier activity of any insulator.

EXPERIMENTAL PROCEDURES
Electrophoretic Gel Mobility Shift Assay-Nuclear extracts were prepared and analyzed by EMSAs based on previously described methods (14). Specifically, nuclear extracts were prepared by washing subconfluent cells in cold PBS and resuspending cells in hypotonic lysis buffer consisting of 20 mM HEPES (pH 7.9), 25% glycerol, 10 mM KCl, 1.5 mM MgCl 2 , 2 mM DTT, 1ϫ protease inhibitor mixture (Sigma P8340), 0.1% Nonidet P-40, and 1 mM PMSF. Nuclear pellets were collected by centrifugation and resuspended in low salt buffer (hypotonic lysis buffer except with 20 mM KCl and 0.2 mM EDTA) to onehalf of the packed nuclear volume. An equal volume of high salt buffer (same as low salt buffer except with 1.2 M KCl) was added dropwise, and after 30 min on ice, the preparation was again centrifuged to obtain cleared nuclear lysate, which was stored at Ϫ80°C. Protein concentrations were determined spectrophotometrically with Bradford reagent. Oligonucleotide probes (50 pmol) were 5Ј-labeled with [␥-32 P]ATP and T4 polynucleotide kinase and then purified over Sephadex G-50 columns. Probes were incubated with 10 g of nuclear extract at room temperature for 30 -40 min in EMSA buffer consisting of 8 mM HEPES (pH 7.9), 12% glycerol, 160 mM KCl, 7 mM MgCl 2 , 1 mM DTT, 2% polyvinyl alcohol, 0.1 mM EGTA, 1ϫ protease inhibitor mixture, 1 mM PMSF, and 0.05% IGEPAL CA-630 and then separated on 4% acrylamide and 2.5% glycerol nondenaturing gels in 0.5ϫ Tris borate/EDTA. Gels were imaged using a PhosphorImager. Complementary probes (50 M) were annealed by denaturing and slowly cooling in 10 mM Tris-HCl (pH 8), 1 mM EDTA, and 100 mM NaCl. Probes for FVI, FVII, and FVIII consisted of 56-bp oligonucleotides spanning the region between the HindIII site that defines the 3Ј-boarder of the 250-bp cHS4 core and the SbfI restriction site that defines the 3Ј-boarder of the extended 400-bp cHS4 core. Competitors included poly(dI-dC), salmon sperm genomic DNA, unlabeled cHS4 FVIII oligonucleotides, biotin-conjugated cHS4 FVIII oligonucleotides, unlabeled cHS4 FII oligonucleotides, and oligonucleotides spanning the PARP-1-binding site in the promoter of the EEF1A1 gene (15). Sequences for all probe and competitor oligonucleotides are provided in Table 1. Nuclear extract was either replaced with purified PARP-1 protein (Trevigen 4668-100-01) or supershifted by the addition of an anti-PARP-1 antibody (Santa Cruz Biotechnology sc-7150).
Affinity Capture and Tandem Mass Spectrometry-Biotinylated single-stranded DNA (ssDNA) oligonucleotide probes for FVIII and a nonspecific poly(T) oligonucleotide probe were bound to streptavidin-coated magnetic beads (Qiagen) following the manufacturer's directions. They were then incubated overnight at 4°C with nuclear extracts from K562 cells, along with poly(dI-dC) as nonspecific competitor. The samples were subsequently washed progressively with 100 -350 mM NaCl and boiled in SDS reducing buffer to release the remaining bound proteins. The samples eluted by boiling were size-fractionated on 4 -15% gradient denaturing gels (Bio-Rad). Specific bands were excised, washed with 100 mM ammonium bicarbonate (3ϫ), dehydrated in acetonitrile (3ϫ), dried under vacuum, and resuspended in sequencing-grade trypsin on ice for 45 min. Tryptic fragments were then extracted with 50% acetonitrile and 0.1% TFA and analyzed on a tandem mass spectrometer (Thermo linear trap quadrupole Orbitrap LTQ-OT MS/MS mass spectrometer, Thermo Electron Corp.). Peptide sequences were identified using ProteinProphet software.
Western Blotting-The gradient gel was soaked in transfer buffer consisting of 25 mM Tris base, 192 mM glycine, and 20% methanol and transferred to PVDF membrane overnight at 4°C. The membrane was blocked with Starting Block (Thermo Scientific) supplemented with 0.05% Tween 20 and hybridized with an anti-PARP-1 antibody (Chemicon AB3565), followed by HRP-conjugated anti-rabbit IgG (Pierce). The membrane was developed with the SuperSignal West detection kit (Pierce) and imaged on film.
Retroviral Vectors-The gammaretroviral reporter vectors MGPN2 and INS4(ϩ) have been described previously (16) and are diagrammed in Fig. 1A. The MGPN2-based vector containing the 400-bp cHS4 core in the 3Ј-long terminal repeat (LTR) has also been described previously (12). The PARP-1-binding site was removed by Bal31 digestion, generating a 92-bp deletion starting at 122 bp 3Ј of the HindIII site used to define the 3Ј-boundary of the 250-bp cHS4 core. A 32-bp segment starting at nucleotide 9 of the FVIII probe diagrammed in Fig. 2 and spanning the PARP-1-binding site was replaced with a "scrambled" sequence with similar GC content (see Table 1 for sequence details). Retroviral vector producer lines were generated using the amphotropic packaging line PA317 and the ecotropic packaging line GPϩE86 as described (12). All cell lines were maintained at 37°C and 7.5% CO 2 in Dulbecco's modified Eagle's medium with 10% heat-inactivated fetal bovine serum.
Gene Transfer-Human fibrosarcoma HT1080 cells were plasmid-transfected using FuGENE6 (Roche Applied Science) following the manufacturer's directions and plated at limiting dilution under G418 selection. After selection, individual colonies were picked under an inverted microscope and expanded for ChIP studies. Human erythroleukemia K562 cells were transduced by 24 h of culture with virus supernatant and 4 g/ml Polybrene at a limiting multiplicity of infection (Ͻ1 infectious unit/cell) to assure low vector copy numbers. The cells were then washed and plated at limiting dilution in 96-well dishes under G418 selection. After selection, individual colonies were isolated and expanded for expression analysis. Mouse bone marrow cells were transduced by co-cultivation on vector producer cells as described previously (16) and included the following strains: wild-type B6xD2 F1 and Parp1-deficient 129S-Parp1 tm1Zqw /J (Parp1 Ϫ/Ϫ ; Jackson Laboratory 002779) (17).
Flow Cytometry Analysis-Vector GFP expression was analyzed directly on a FACScan flow cytometer (BD Biosciences) using CellQuest software. The percentage of GFP-positive cells in the experimental samples was determined by subtracting the amount of background signal within the established gate (typically set at 1%) of mock-transduced cells. The GFP mean fluorescence in the experimental samples was determined by subtracting the background mean fluorescence of mock-transduced cells.
ChIP-Cells were cross-linked with formaldehyde, and chromatin was prepared and sheared as described (18). ChIP analysis was performed on two to six biological replicates using the previously described microplate-based Matrix ChIP method (19). Antibodies included anti-PARP-1 (Chemicon AB3565) and anti-histone H3 (Abcam ab1791). The level of target template in precipitated and input samples was determined in triplicate by real-time PCR as described (19). Target templates included FVIII, the exon 2-3 region of the gene encoding ␤-actin (ACTB), and promoters of the negative control SEMA4G and positive control HSP70.1. Primer sequences are listed in Table 1.
Progenitor Colony Assay-Transduced marrow cells were plated at 1-2 ϫ 10 4 cells/ml in medium containing 1% methylcellulose, 15% fetal bovine serum, and multiple growth factors, including interleukin-3, interleukin-6, stem cell factor, and erythropoietin (STEMCELL Technologies). G418 was used for selection at 0.9 mg/ml. The PARP inhibitor 3-aminobenzamide (3-AB) was used at 2 and 8 mM. Myeloid and erythroid colonies were scored after 6 -10 days of incubation at 37°C and 5% CO 2 . Colonies were picked individually under an inverted microscope, and washed with Hanks' buffered saline solution before further analysis. The presence of vector provirus in individual bone marrow progenitor colonies was determined as described (20). In short, DNA was prepared by incubating cells in lysis buffer and proteinase K and precipitating DNA with glycogen and EtOH. Vector template was amplified by PCR using the GFP-specific primers listed in Table 1, and visualization was performed by gel electrophoresis and staining by EtBr. The identity of amplified bands was confirmed by Southern blotting using a probe for GFP.

RESULTS
Initial Screen for Protein Binding to the Extended cHS4 Core Region-As diagrammed in Fig. 1A, our previous studies involved flanking the gammaretroviral vector MGPN2 with  fragments of the cHS4 region using a "double-copy" arrangement and then assessing the effects of these elements on expression of the vector GFP expression cassette (12,16). As reported recently (12), we found that both the full-length 1.2-kb cHS4 fragment and an extended 400-bp cHS4 fragment were equally effective in reducing silencing position effects and thereby improving the likelihood and level of vector GFP expression. In contrast, when used as a single unit, the originally described 250-bp cHS4 core was ineffective in reducing silencing position effects in this setting. To determine whether the additional activity associated with the extended 400-bp cHS4 core was associated with binding of proteins to the extended sequences, we carried out EMSA studies using nuclear extracts from human erythroleukemia K562 cells and 56-bp oligonucleotide probes equally spaced across the additional 3Ј-region (indicated as fragments VI, VII, and VIII in Fig. 1A). As shown in Fig. 1B, we found evidence for DNA-binding activities specifically associated with the most distal fragment, FVIII. This included a prominent faster moving doublet (bands d and e), an intermediate band (band c) that appears from later studies to be nonspecific, and a slower moving doublet (bands a and b). Similar results were seen with nuclear extracts from human fibrosarcoma HT1080 cells (data not shown).
During these initial studies, we noted that the doublestranded DNA (dsDNA) FVIII probe was prone to instability, resulting in the apparent generation of its ssDNA components or other complexed secondary structures. Further EMSA studies revealed that many of the FVIII-specific bands seen with the dsDNA FVIII probe were also generated with ssDNA versions of the FVIII probe (Fig. 1C). For the (ϩ)-strand, this included bands b and d, whereas with the (Ϫ)-strand, this included all four FVIII-specific bands. These results suggest a role for a ssDNA intermediate or unique secondary structure associated with the binding activity. To assess the specificity of this binding, we also carried out competition EMSA studies. As shown in Fig. 1D, essentially all of the binding activities associated with the dsDNA FVIII probe, except for band c, could be competed with both dsDNA and ssDNA versions of the FVIII probe, further demonstrating that both versions of this probe are bound by the same protein(s). The same results were seen with biotinylated versions of the dsDNA and ssDNA FVIII probes used in subsequent studies. Competition studies with probes for the FII site of the cHS4 element, which is known to be bound by the zinc finger protein CTCF, gave mixed results. When used in a dsDNA form, this probe appeared to provide some competition for bands a and b, but no competition for bands d and e. In contrast, competition with a ssDNA version of the FII probe failed to diminish any of the FVIII-specific bands.
Structural Analysis of the FVIII Segment-The studies described above suggest that the FVIII probes exhibited unique physical properties: the dsDNA version of the probe appeared to be unstable; the ssDNA probe appeared to be capable of forming two different structures (for example, see the unbound probes in Fig. 1D, fourth and seventh lanes); and both the dsDNA and ssDNA versions of the probe were apparently bound by the same FVIII-specific protein(s). Thermodynamic modeling (21) revealed the potential for stable secondary structures associated with both of the ssDNA FVIII probe sequences. As diagrammed in Fig. 2A, these structures involve stem-loop configurations consisting of stems formed through hybridization of complementary 8-base inverted repeats and 14 base loops. If these secondary structures were retained upon strand  (12,16). Vector MGPN2 includes a GFP cassette transcribed from the vector LTR and a neo expression cassette transcribed from an internal phosphoglycerate kinase (Pgk) promoter. Vector INS4(ϩ) is flanked with the 1.2-kb cHS4 fragment inserted in the double-copy position of the 3Ј-LTR, from which it is copied into the 5Ј-LTR during provirus formation. The subfragments composing the 250-bp core (7), as well as the subfragments composing the additional sequences in the 400-bp core (12), are indicated in Roman numerals. The relative frequency of vector GFP expression for the vectors carrying the different cHS4 fragments is presented to the right. B, EMSA analysis with the three 56-bp contiguous subfragments FVI, FVII, and FVIII incubated with nuclear extracts from the human erythroid cell line K562. Note that bands a, b, d, and e appear to be specific for FVIII. C, EMSA showing similar binding patterns for dsDNA and ssDNA versions of FVIII. (ϩ), sense strand; (Ϫ), antisense strand (based on orientation of the native locus). Shifted bands are labeled as in A. D, competition EMSA using a radiolabeled dsDNA FVIII probe and a 400-fold molar excess of the indicated unlabeled dsDNA and ssDNA probes. Note that unlabeled dsDNA and ssDNA FVIII probes, with and without a biotin (bio) label, competed for all FVIII-specific bands. In contrast, a ssDNA probe for the CTCF-binding FII segment failed to compete for any of the EMSA bands, whereas a dsDNA FII probe appeared to have a modest effect on the FVIII-specific bands a and b. Note that the last two lanes of B and the first two lanes of D are from the same experiment and that the same amount of probe was used throughout the studies.

PARP-1 Contributes to Barrier Insulator Function
annealing, they would result in the cruciform structure diagrammed in Fig. 2B. To determine whether such structures exist in vitro, we digested large DNA fragments containing the full-length cHS4 element with the ssDNA-specific nuclease S1.
As diagramed in Fig. 2C, this study provided direct evidence for a ssDNA structure specifically at the FVIII site. In this setting, the FVIII-specific nuclease S1 digestion appeared to be specific for the fragment containing the FVIII site and only involved ϳ20% of the DNA strands, suggesting a stochastic and presumably dynamic equilibrium between the ssDNA structure being digested and a conventional dsDNA structure at the FVIII site, providing a ready explanation for the apparent instability of the dsDNA probes for this site. The presence of such a unique secondary structure could also provide a ready explanation for the ability of both dsDNA and ssDNA versions of the FVIII probes to be bound by the same protein(s) because the same stem-loop structures would be present in these settings.
In an effort to confirm the presence of a unique secondary structure associated with the smaller FVIII probe, we again digested dsDNA probes for FVI, FVII, and FVIII with increasing amounts of nuclease S1. As shown in Fig. 2D, the dsDNA probe for FVIII was degraded at a 3-fold lower concentration of nuclease S1 compared with the equivalent FVI and FVII probes. Fig.  2D also demonstrates that the dsDNA probe for FVIII started out as a single product, again suggesting a stochastic and presumably dynamic equilibrium between ssDNA and dsDNA at the FVIII site.
Identification of PARP-1 as a Binding Factor for cHS4 FVIII-To identify the factor(s) that bind the cHS4 FVIII segment, we carried out affinity capture studies with biotinylated ssDNA FVIII probes and streptavidin-coated magnetic beads. As shown in Fig. 3A, we were able to identify a specific protein associated with both of the ssDNA FVIII probes at a molecular mass of just below 120 kDa. These bands, along with the equivalent region of the control lane, were excised and submitted for identification by tandem mass spectrometry. This analysis recovered a total of 197 peptides covering 57.6% of the amino acid sequence for the PARP-1 protein in association with both of the FVIII samples, but not for the negative control. As shown in Fig. 3B, this partial purification was first confirmed by Western blotting of the polyacrylamide gel containing the affinitypurified protein product using an anti-PARP-1 antibody. Subsequent EMSA studies demonstrated that purified PARP-1 protein generated one of the same shifted bands, band b, that was observed with K562 nuclear extracts (Fig. 3C). Furthermore, band b seen with the K562 extracts was supershifted upon the addition of an anti-PARP-1 antibody (Fig. 3D). Upon treatment of K562 cells with the apoptotic agent staurosporine, shown previously to induce cleavage of the PARP-1 protein (22), the intensity of band b was reduced (Fig. 3E). Finally, competition EMSA studies with a probe from the promoter of the EEF1A1 gene, shown previously to bind PARP-1 (15), also specifically reduced the intensity of band b (Fig. 3F). Together, these studies provide strong evidence that PARP-1 binds the FVIII probes in vitro and is responsible for band b in the EMSA studies. The identities of the proteins responsible for the other EMSA bands associated with the FVIII probe (bands a, d, and e) remain to be elucidated.
In Vivo Confirmation of PARP-1 Binding to cHS4 FVIII-To determine whether PARP-1 binds cHS4 FVIII in vivo, we transfected the human fibrosarcoma cell line HT1080 with plasmid forms of the gammaretroviral vector MGPN2 containing three FIGURE 2. Physical properties of FVIII probes. A, predicted secondary structure for ssDNA FVIII probes. The sequence of the ssDNA FVIII (ϩ)-strand probe was predicted to have the thermodynamically favorable structure diagrammed (Mfold Quickfold program (21); ⌬G ϭ Ϫ8.2, T m ϭ 70°C). Note that a similar structure was also predicted for the ssDNA FVIII (Ϫ)-strand probe (⌬G ϭ Ϫ12.2, T m ϭ 90°C; B). B, hypothesized cruciform secondary structure for the dsDNA FVIII probe. Note that the stem structures for both strands share identical primary sequences. C, confirmation of the cruciform for dsDNA FVIII. The bacterial plasmid encoding the gammaretroviral vector INS4(ϩ) was digested with KpnI, and the resulting mixture containing 4.7and 3.7-kb fragments was further digested with increasing amounts of nuclease S1. This included 3-fold dilutions starting at 4 units/l nuclease S1 and incubation at 37°C for 1 h. The samples were then separated on a 1% agarose gel, Southern-blotted, hybridized with a probe for the 3Ј-end of the cHS4 core as indicated, and imaged using a PhosphorImager. Left panel, gel stained with EtBr before blotting. Note that both bands were eventually degraded with increasing amount of nuclease S1 but that the top band appeared to generate a faint but specific smaller band. Right panel, Southern blot confirming the identity of the specific nuclease S1-digested band. Note the appearance of this band at the size expected for digestion at FVIII. Quantification indicated that this lower band represents ϳ20% of the signal from the higher intact band (two lower nuclease S1 concentrations). Pg, phosphoglycerate kinase promoter; K, sample digested with KpnI only. D, confirmation of complex secondary structure for the dsDNA FVIII probe. Nuclease S1 studies were also carried out using dsDNA probes for FVI, FVII, and FVIII, essentially as described for C with a 4% agarose gel and EtBr staining. Note that the FVIII probe was degraded at a 3-fold lower concentration of nuclease S1 than the other probes. Also notice that all three dsDNA probes, including the probe for FVIII, existed as a single product prior to further manipulation.

PARP-1 Contributes to Barrier Insulator Function
NOVEMBER 26, 2010 • VOLUME 285 • NUMBER 48 different versions of the 1.2-kb cHS4 fragment in the 3Ј-LTR and derived individual clones under G418 selection. This included the wild-type cHS4 sequence, a version in which most of FVIII was deleted, and a version in which most of the FVIII sequence was replaced with a scrambled sequence with the same CG content. We then carried out ChIP studies using antibodies to PARP-1 and histone H3 (as a control). We chose to normalize the PARP-1 signal to the H3 signal for two reasons: both proteins are highly abundant in chromatin, and this was found to be a useful comparison in the genome-wide assessment of PARP-1 binding (24). As shown in Fig. 4, analysis of the intragenic region of ACTB gene and the promoter of a gene found previously to bind only low levels of PARP-1 (SEMA4G) 3 revealed PARP-1/H3 ratios of 0.3 and 0.5, respectively. In contrast, the promoter of a gene shown previously to bind higher levels of PARP-1, HSP70.1 (23), exhibited a PARP-1/H3 ratio of 1.1.
Although the difference between the negative (SEMA4G) and positive (HSP70.1) control genes was only 2.2-fold, this difference was statistically significant (p Ͻ 0.01). In addition, this difference is consistent with the results reported recently in a benchmark genome-wide study of PARP-1 binding, which found that areas of high and low level PARP-1 binding typically differed by only 1.6-fold (24). Analysis of the FVIII segment containing the wild-type sequence also revealed a relatively high PARP-1/H3 ratio of 1.0, similar to that seen with the positive control HSP70.1 and statistically higher than that seen with the negative control SEMA4G (p Ͻ 0.01). In contrast, analysis of the constructs containing the deleted and scrambled versions of the FVIII segment revealed PARP-1/H3 ratios of 0.5 and 0.6, respectively, both of which are statistically indistinguishable from the SEMA4G negative control (p ϭ 0.45 and 0.28, respectively). Taken together, these results provide direct evidence that PARP-1 binds the cHS4 FVIII region in vivo.
Functional Confirmation of PARP-1 in cHS4 Barrier Insulator Activity-As a first step toward determining whether PARP-1 binding plays a functional role in the barrier insulator activity of the cHS4 element, we transduced K562 cells with the gammaretroviral vector MGPN2 flanked with four different versions of the cHS4 element. Panels of transduced clones were 3 R. Krishnakumar and W. L. Kraus, personal communication.  (as a control). Constructs included FVIII sequences that were wild-type (wt), deleted (del), or scrambled (scram). Control target sequences included exons 2-3 of the ␤-actin gene ACTB and the promoters of the SEMA4G (negative) and HSP70.1 (positive) genes. The percent input was determined by comparing the ratio of target in precipitated versus input samples by real-time PCR and was adjusted by subtracting the signal from a nonspecific polyclonal antibody control. Each histogram represents the mean Ϯ S.E. for data from two to six independent biological replicates, with PCR performed in triplicate.
G418-selected and then analyzed for vector GFP expression by flow cytometry. As shown in Fig. 5A, inclusion of the full-length 1.2-kb or 400-bp core versions of the cHS4 insulator increased the likelihood of vector GFP expression by nearly 2-fold, from an average of 3.8% GFP-positive cells to 6.9 and 6.0% GFPpositive cells, respectively. Although the overall level of vector GFP expression remained low due to the inefficiency of the LTR promoter/enhancer in the erythroid environment, this increase was significant (p Յ 0.001). In contrast, the likelihood of GFP expression for vectors flanked with the full-length cHS4 element in which the FVIII segment was either scrambled or deleted averaged only 3.6 and 3.5%, respectively. These levels are indistinguishable from those seen with the uninsulated vector and indicate a complete loss of barrier insulator function.
We next used our established mouse bone marrow culture model (12,16) to assess the effects of the general PARP inhibitor 3-AB (22) on cHS4 barrier activity. This entailed transducing bone marrow cells from wild-type B6xD2 F1 mice with the uninsulated and insulated vectors MGPN2 and INS4(ϩ) (diagrammed in Fig. 1A) and then analyzing vector GFP expression in panels of individual G418-selected progenitor colonies. In this case, we focused on the overall level of vector GFP expression, which is a function of both the frequency of GFP-positive cells and expression variegation in these positive cells. As shown in Fig. 5B, inclusion of the cHS4 insulator increased the level of vector GFP expression in three independent experiments from an average of 8 -45 mean fluorescence units (m.f.u.) to an average of 62-94 m.f.u. in the absence of 3-AB. However, this difference was completely abrogated upon the addition of 3-AB at two different concentrations, such that the average level of vector GFP expression ranged from 15-54 m.f.u. for the uninsulated vector compared with 16 -40 m.f.u. for the insulated vector. In each case, the difference between uninsulated and insulated vectors changed from a statistically significant 2-7-fold to statistically indistinguishable.
To rule out the possibility that the effects seen with the general PARP inhibitor 3-AB were due to other PARP family members, we turned to transduction studies using bone marrow from Parp1-null mice. This entailed transducing bone marrow cells from wild-type and Parp1 Ϫ/Ϫ donors with the uninsulated and insulated vectors MGPN2 and INS4(ϩ) and then analyzing vector GFP expression in individual progenitor colonies. Because the Parp1 knock-out was generated by insertion of a neo selection gene, we could not use G418 selection to ensure that all colonies contained vector provirus. Instead, we relied on PCR to identify individual colonies that contained vector provirus. This also allowed us to measure both the level of vector GFP expression and the frequency of vector GFP expression without the constraints introduced by G418 selection. Data from two independent experiments in wild-type progenitors indicated that inclusion of the cHS4 insulator increased the average level of vector GFP expression from 11-15 m.f.u. to 54 -107 m.f.u. (Fig. 5C), as well as the average frequency of vector GFP expression from 15-16 to 51-59% (Fig. 5D). However, these differences were abrogated in the Parp1 Ϫ/Ϫ progenitors. In this setting, the average level of vector GFP expression ranged from 23-30 m.f.u. for the uninsulated vector and 39 -87 m.f.u. for the insulated vector (Fig. 5C), whereas the average frequency of vector GFP expression ranged from 23-30% for the uninsulated vector and 30 -39% for the insulated vector (Fig. 5D). We believe that these results provide very strong evidence that PARP-1 specifically plays a key functional role in the barrier activity of the extended cHS4 insulator in the context of gammaretroviral vectors. That said, it is worth noting that the 2-kb cHS4 with FVIII sequences deleted. Each data point represents the percentage of GFP-positive cells for independent clones. The mean for each group of clones is indicated. p values were based on t test versus the uninsulated control. B, gammaretroviral vector GFP expression in transduced wild-type B6xD2 F1 mouse bone marrow progenitor cell colonies with or without the indicated amounts of the PARP inhibitor 3-AB. The level of vector GFP expression (m.f.u.) was determined for uninsulated and insulated vectors MGPN2 and INS4(ϩ), respectively, in three independent experiments as described (12,16). Histograms represent the mean Ϯ S.D. of each group of clones (at least 12 each). p values were based on t test for the indicated comparisons. C and D, gammaretroviral vector GFP expression in transduced bone marrow progenitor colonies from wild-type (WT) and Parp1-null (Parp) mice. The level (C) and frequency (D) of vector GFP expression were determined for uninsulated and insulated vectors MGPN2 and INS4(ϩ), respectively, in two independent experiments as described for the previous panels. Histograms represent the mean Ϯ S.D. of each group of clones. p values were based on t test.

PARP-1 Contributes to Barrier Insulator Function
NOVEMBER 26, 2010 • VOLUME 285 • NUMBER 48 loss of insulator activity shown in Fig. 5B appears to be greater than the loss of insulator activity shown in Fig. 5C. Although this difference is not statistically significant, it suggests that PARP variants other than PARP-1 may also play a partial role in cHS4 barrier insulator activity.

DISCUSSION
The cHS4-binding protein identified in our studies, PARP-1, is an abundant nuclear protein that has the capacity to bind DNA through zinc finger motifs and to catalyze the addition of poly(ADP-ribose) chains to itself and other proteins. It has been implicated in many biochemical pathways, including DNA repair, signaling in apoptosis, and gene regulation (25). As reviewed recently (26), PARP-1 has been shown to play a direct role in gene regulation through several modes, including binding to enhancers, serving as a transcriptional co-regulator, modulating chromatin structure, and helping to mediate the activity of enhancer-blocking insulators. Indeed, the main mediator of enhancer-blocking insulator function, CTCF, can itself be poly(ADP-ribosyl)ated (PARylated) by PARP-1 (27), and PARP-1 has been found to co-immunoprecipitate with CTCF (5). This may help explain why FII, which is known to bind CTCF, is able to partially compete for PARP-1 binding to FVIII (Fig. 1D). However, the studies presented here provide the first evidence to date that PARP-1 can also play a direct role in barrier insulator activity. Although the studies presented here do not directly address the underlying mechanism(s) of this action, others have identified three properties of PARP-1 activity that are likely candidates. First, PARP-1 has been shown recently to exclude a key component of compact chromatin, linker histone H1, from the promoters of most transcriptionally active genes either by directly competing with H1 binding to nucleosomes or by PARylating H1 (24). Second, it is possible that PARP-1 plays a less direct role in cHS4 barrier activity, functioning instead by the PARylation of other proteins bound to the cHS4 core region. Although CTCF is not thought to play a role in the barrier insulator activity of the cHS4 element (6), there is good evidence that PARP-1 modulates the enhancerblocking activity of cHS4 and other insulators through the PARylation of CTCF (5,28). Third, the genome-wide analysis of PARP-1 binding revealed an inverse correlation between PARP-1 and histone H3 levels (24), suggesting that high levels of PARP-1 binding may deplete or retard nucleosome formation. Our own data indicate that barrier insulator activity is also associated with an elevated ratio of PARP-1 to H3 levels. Taken together, these observations suggest that PARP-1 binding may prevent the spread of heterochromatin by creating a physical gap in the chain of nucleosomes.
It is unclear what is mediating the specific binding of PARP-1 to the cHS4 FVIII segment. Although PARP-1 binding is mediated by zinc finger motifs, only a weak functional consensus binding sequence has been identified to date (29). Instead, some studies suggest that PARP-1 recognizes secondary DNA structures (30,31). In particular, these studies suggest a specificity for stem-loop and cruciform structures similar to those suggested for the cHS4 FVIII site in Fig. 2.
We have demonstrated previously that the 400-bp cHS4 core exhibits all of the barrier activity associated with the 1.2-kb cHS4 fragment, whereas the 250-bp cHS4 core is ineffective as a barrier element when used as a single copy in the context of retroviral vectors (12). The studies presented here provide a functional explanation for this additional activity. These studies also provide new insight into the mechanisms underlying the barrier activity of the cHS4 element. Previous studies have identified two independent sets of DNA-binding proteins and related mechanisms associated with this element's barrier activity: USF1/2 and histone hyperacetylation (10) and Vezf1/ BGP1 and DNA methylation (11). Our studies provide a novel third protein and related mechanism involved in cHS4 barrier activity: PARP-1 and H1 exclusion, protein PARylation, or nucleosome depletion. Future studies will be needed to differentiate between these two mechanisms and to determine whether PARP-1 plays a role in the activity of other barrier insulators.