The androgen-induced protein AIbZIP facilitates proliferation of prostate cancer cells through downregulation of p21 expression

Androgen-Induced bZIP (AIbZIP) is structurally a bZIP transmembrane transcription factor belonging to the CREB/ATF family. This molecule is highly expressed in androgen-sensitive prostate cancer cells and is transcriptionally upregulated by androgen treatment. Here, we investigated molecular mechanism of androgen-dependent expression of AIbZIP and its physiological function in prostate cancer cells. Our data showed that SAM pointed domain-containing ETS transcription factor (SPDEF), which is upregulated by androgen treatment, directly activates transcription of AIbZIP. Knockdown of AIbZIP caused a significant reduction in the proliferation of androgen-sensitive prostate cancer cells with robust expression of p21. Mechanistically, we demonstrated that AIbZIP interacts with old astrocyte specifically induced substance (OASIS), which is a CREB/ATF family transcription factor, and prevents OASIS from promoting transcription of its target gene p21. These findings showed that AIbZIP induced by the androgen receptor (AR) axis plays a crucial role in the proliferation of androgen-sensitive prostate cancer cells, and could be a novel target of therapy for prostate cancer.


Results
Expression of AIbZIP in androgen-sensitive prostate cancer cell line LNCaP. AIbZIP has been reported to be highly expressed in androgen-treated prostate cancer cell lines 11 . To confirm the upregulation of AIbZIP in prostate cancer, we examined the expression levels of AIbZIP in various tumor types using the ONCOMINE Cancer Profiling Database. AIbZIP was highly expressed in sex hormone-related cancers including prostate, breast, endometrium, and uterus cancers (Fig. 1a). Notably, the expression levels of AIbZIP were much higher in prostate cancer than in other cancers (Fig. 1a), implying that AIbZIP expression could correlate with progression of prostate cancer. An experimental investigation using various cancer cell lines showed the mRNA levels of AIbZIP were extremely high in LNCaP (androgen-sensitive prostate cancer) cells, and moderate in MCF-7 (breast cancer) and HeLa (cervical cancer) cells (Fig. 1b). In contrast, AIbZIP was hardly detected in PC-3 (androgen-insensitive prostate cancer) and Caco-2 (colon carcinoma) cells (Fig. 1b). Interestingly, cells expressing AIbZIP have a tendency to express AR (Fig. 1b). These findings suggested a link between AIbZIP expression and AR signaling.
Next, we checked the induction of AIbZIP in response to androgen stimulation. Treatment of LNCaP cells with the synthetic AR agonist R1881 resulted in a time-dependent increase of AIbZIP expression at both the mRNA and protein levels (Fig. 1c,d). Interestingly, western blot (WB) analysis showed that two bands of AIbZIP protein at approximately 50 and 43 kDa were increased after treatment with R1881 (Fig. 1d). AIbZIP contains a putative N-glycosylation site within the luminal domain (Fig. 1e). Treatment of cells with tunicamycin (Tm), which blocks N-linked glycosylation, decreased the 50 kDa band and increased the 43 kDa one, indicating that the 50 kDa AIbZIP is a glycosylated form, while the 43 kDa one is not (Supplementary Figure S1).
AIbZIP belongs structurally to the CREB/ATF transcription factor family with a similarity to the ER stress transducers ATF6 and OASIS (Fig. 1e). To determine if AIbZIP is upregulated or activated in response to ER stress, we treated LNCaP cells with ER stressors Tm, thapsigargin (Tg), and brefeldin A (BFA), respectively. The expression of AIbZIP mRNA was never changed although ER stress markers BiP and spliced XBP1 (sXBP1) were upregulated (Fig. 1f). AIbZIP has potential cleavage sites for both S1P and S2P (Fig. 1e). Indeed, treatment with BFA, which is a compound that causes Golgi tubules to fuse with the ER 22 , induced cleavage of AIbZIP (Fig. 1g). However, AIbZIP was never cleaved by ER stressors Tm, Tg, or R1881 (Fig. 1g), indicating that AIbZIP is not activated in response to ER stress and could function as a full-length form in prostate cancer cells. The reason why AIbZIP transported from the ER to the Golgi apparatus is not cleaved by S1P and S2P remains unknown, but it is possible that some molecules inhibit cleavage by masking the recognition sites for S1P and S2P at the Golgi apparatus.
To investigate the subcellular localization of AIbZIP, we performed immunofluorescence staining of HeLa cells transfected with a vector expressing FLAG-tagged AIbZIP (FLAG-AIbZIP). FLAG-AIbZIP immunoreactivity was detected in the perinuclear region (Fig. 1h). Co-staining for FLAG and the ER marker calnexin showed that AIbZIP partially co-localized with calnexin ( Fig. 1h). Furthermore, FLAG-AIbZIP signals were also detected from GM130-positive Golgi apparatus (Fig. 1h). These findings implied that AIbZIP could localize and function not only in the ER, but also in the Golgi apparatus as a full-length form.
AIbZIP is induced by SPDEF acting downstream of AR. Next, we analyzed the mechanisms responsible for the upregulation of AIbZIP expression. The AR activated by androgen binds to androgen response elements (AREs) in the promoter regions of target genes 23,24 . Knockdown of AR suppressed the transcriptional induction of its target gene PSA by R1881 (Fig. 2a,b). Similarly, AIbZIP expression induced by R1881 was significantly reduced in AR-knockdown cells (Fig. 2a,b). In addition, the AR inhibitor bicalutamide also suppressed the transcriptional induction of AIbZIP and PSA in LNCaP cells (Fig. 2c,d), indicating that the AR is indispensable for the upregulation of AIbZIP by androgen stimulation. However, no exact sequence of AREs exists in the promoter region of AIbZIP, suggesting that the AR does not directly activate transcription of AIbZIP. To disclose the leading molecule for transcription of AIbZIP, we focused on SAM pointed domain-containing ETS transcription factor (SPDEF). It was shown that SPDEF promotes transcription of PSA through direct binding to the PSA promoter region, and interacts with AR to cooperatively enhance PSA promoter activity 25 . SPDEF has also been reported to be upregulated by R1881 stimuli in LNCaP cells 26 . Indeed, knockdown of AR or treatment with bicalutamide suppressed the transcriptional induction of SPDEF by R1881 treatment (Fig. 2a-d). Moreover, SPDEF mRNA levels were increased at an earlier time point compared with AIbZIP mRNA in LNCaP cells treated with R1881 ( Fig. 2e,f). Therefore, we hypothesized that SPDEF mediates the androgen-induced transcriptional regulation of AIbZIP. Overexpression of AIbZIP had no effect on the expression levels of SPDEF, while LNCaP cells expressing SPDEF showed a 2.5-fold increase in AIbZIP expression (Fig. 2g,h). Alternatively, knockdown of SPDEF in LNCaP cells dramatically decreased the transcriptional induction of AIbZIP by R1881 treatment (Fig. 2i,j). These results supported our hypothesis that AIbZIP is upregulated by SPDEF, which acts downstream of AR.  AIbZIP, we conducted reporter assays using a luciferase reporter gene driven by a 5-kb fragment of the promoter region of AIbZIP. Reporter activity was markedly enhanced (33-fold) by SPDEF expression (Fig. 3a). A serial deletion analysis showed that the region between − 1042 and − 19 confers the maximal promoter activity of AIbZIP (Fig. 3a). Moreover, deletion of the DNA-binding domains of SPDEF resulted in a decrease of the promoter activity induced by SPDEF (Fig. 3b). The ETS family proteins contain an evolutionarily conserved DNA-binding domain, which mediates binding to conserved purine-rich DNA sequences with a GGA(A/T) core consensus sequence 25,27 . We found that this core sequence is distributed in the promoter region of AIbZIP ( Fig. 3c and Supplementary Figure S2). To narrow down the binding region of SPDEF, we designed four primer sets (P1-4) (Fig. 3c) and performed chromatin immunoprecipitation (ChIP) assays. Specific amplification was only detected by PCR analysis using primer set P2 (Fig. 3d). Collectively, these data indicated that SPDEF induced by androgen treatment regulates AIbZIP transcription through direct binding to the promoter region (− 686 to − 570) of endogenous AIbZIP, which contains two GGA(A/T) core sequences.
AIbZIP is involved in prostate cancer cell proliferation. Next, we investigated whether AIbZIP is involved in the proliferation of prostate cancer cells. We performed knockdown of AIbZIP in LNCaP cells using a mixture of three small interfering RNAs (siRNAs) (Supplementary Figure S3), and then counted the number of cells for 5 days. Without R1881 treatment, the number of cells was significantly reduced by 39.6% in AIbZIP-knockdown cells compared with control cells on day 5 (Fig. 4a). Incorporation of 5-Bromo-2′ -deoxyuridine (BrdU), which determines the frequency of cells undergoing DNA synthesis and division, was dramatically decreased in AIbZIP-knockdown cells (Fig. 4b). Furthermore, the percentage of cells positive for the cancer proliferation marker Ki67 was also significantly decreased by AIbZIP-knockdown (Fig. 4c), indicating that knockdown of AIbZIP suppresses the proliferation of LNCaP cells.
To address the mechanism responsible for the antiproliferative effects of AIbZIP silencing, we examined the effects of AIbZIP-knockdown on the expression of cell cycle-related genes. Following AIbZIP knockdown, the mRNA levels of cyclin A2 (CCNA2), cyclin E1 (CCNE1) and cyclin-dependent kinase 2 (CDK2) were reduced compared with control sample (Fig. 4d). In contrast, both the mRNA and protein expression levels of cyclin dependent kinase inhibitor 1 (p21) were highly upregulated in AIbZIP-knockdown cells, although the phosphorylation of tumor protein p53 (p53) was unaffected (Fig. 4d,e), suggesting that silencing of AIbZIP upregulates the expression of p21 independently of p53. p21 is a major molecule to inhibit the activities of cyclin-CDK complexes and negatively modulates cell cycle progression 28 . We therefore focused on p21 especially among the cell cycle-related genes whose expressions were affected by AIbZIP-knockdown.
To confirm that AIbZIP regulates the expression of p21, we examined the changes in p21 expression patterns after transfection with siRNA targeting AIbZIP. The expression of AIbZIP was effectively suppressed for 3 days (Fig. 5a-d), while the expression pattern of p21 was inversely correlated with that of AIbZIP ( Fig. 5a-d). We also performed an additional proliferation assay using AIbZIP-knockdown or control cells with or without R1881 treatment. R1881 stimuli induced a reliable increase in the proliferation of control cells (Fig. 5e), but failed to rescue the suppression of cell proliferation by AIbZIP-knockdown (Fig. 5e). Consistently, the increase of p21 expression in AIbZIP-knockdown cells was not affected by R1881 treatment, although R1881 stimuli decreased p21 expression in control cells ( Fig. 5f-i). In contrast, overexpression of AIbZIP dramatically increased the proliferation of LNCaP cells without R1881 treatment (Supplementary Figure S4). These results indicated that AIbZIP may regulate LNCaP cell proliferation induced by AR signaling via inhibition of p21 expression.

AIbZIP represses p21 expression via prevention of OASIS activation. The next issue is how AIbZIP
promotes cell proliferation through regulation of p21 expression. Interestingly, the bZIP transcription factor OASIS, which is structurally similar to AIbZIP (Fig. 1e), has been reported to directly activate p21 transcription 29 . We investigated the expression levels of OASIS in several cell lines. Unexpectedly, OASIS was strongly expressed in PC-3 and Caco-2 cells, in which AIbZIP was hardly expressed (Fig. 1b and Supplementary Figure S5). These data suggested that the pattern of OASIS expression in various cancer cell lines is not correlated with that of AIbZIP. However, OASIS demonstrated moderate expression in LNCaP cells (Supplementary Figure S5), and knockdown of OASIS significantly decreased p21 expression in LNCaP cells (Fig. 6a-d). Moreover, knockdown of AIbZIP alone increased p21 mRNA and protein, and this increase was blocked by simultaneous knockdown of OASIS (Fig. 6a-d), suggesting that AIbZIP modulates the function of OASIS to regulate p21 expression, at least in LNCaP cells. The reason why AIbZIP knockdown induced p21 protein more than its mRNA remains unknown, but it is well known that the stability of p21 protein is increased when cell cycle progression is negatively regulated 30,31 . Therefore, it is conceivable that the apparent induction of p21 protein was much higher than that of its mRNA. To identify how AIbZIP affects the function of OASIS, we expressed a constant amount of FLAG-OASIS and various amounts of Myc-tagged full-length AIbZIP (Myc-AIbZIP) in HEK293T cells and analyzed the effects on the activation of OASIS. RT-PCR showed the mRNA levels of OASIS were not affected by any amounts of AIbZIP (Fig. 6e). OASIS is activated by sequential cleavage by S1P and S2P at the Golgi apparatus and then completely cleaved N-terminal fragments translocate to the nucleus to promote gene transcription 15 (Supplementary Figure S6). Interestingly, WB analysis showed that the amounts of S1P-cleaved OASIS were strongly increased by AIbZIP in a dose-dependent manner (Fig. 6e,f). However, the amounts of full-length and S2P-cleaved OASIS remained relatively unchanged (Fig. 6e,f), suggesting that AIbZIP does not affect S1P-mediated cleavage of OASIS at the Golgi apparatus, but could prevent S2P-mediated cleavage, thereby preventing release of the N-terminus of OASIS from the Golgi apparatus, resulting in inhibition of p21 transcription by OASIS.
AIbZIP prevents S2P-mediated cleavage of OASIS via direct interaction with OASIS at the Golgi apparatus. bZIP family proteins are known to form homo-or heterodimers to switch their target genes 16,32 . Hence, we performed co-immunoprecipitation assays using HEK293T cells co-transfected with Myc-AIbZIP and FLAG-OASIS to confirm whether these two molecules interact with each other. Full-length and N-terminal OASIS co-precipitated with full-length AIbZIP (Fig. 7a, lane 1), and alternatively full-length AIbZIP co-precipitated with full-length and N-terminal OASIS (Fig. 7a, lane 4). To determine the interaction sites of both AIbZIP and OASIS, we constructed mutant vectors expressing full-length AIbZIP or OASIS lacking the bZIP domain (Myc-AIbZIP Δ bZIP or FLAG-OASIS Δ bZIP). Co-immunoprecipitation showed that FLAG-OASIS was not co-precipitated with Myc-AIbZIP Δ bZIP (Fig. 7a, lanes 3 and 5), and Myc-AIbZIP was not co-precipitated  with FLAG-OASIS Δ bZIP (Fig. 7a, lanes 2 and 6), indicating that full-length AIbZIP interacts with full-length and N-terminal OASIS through their respective bZIP domains.
To assess the impact of the interaction of AIbZIP with full-length and N-terminal OASIS on their functions, we analyzed their subcellular localizations. When FLAG-OASIS was introduced alone, it was localized mainly in the ER, but also partly in the Golgi apparatus and nucleus (Fig. 7b,c). Interestingly, when AIbZIP and FLAG-OASIS were co-transfected, both OASIS and AIbZIP accumulated at a juxtanuclear, Golgi apparatus-like region (Fig. 7b). Immunostaining for OASIS and GM130 showed the majority of OASIS signals overlapped with GM130 (Fig. 7b,c). Furthermore, FLAG-OASIS signals in the nucleus were reduced by co-transfection with AIbZIP (Fig. 7b,d). When FLAG-OASIS was co-transfected with Myc-AIbZIP Δ bZIP, the signals of Myc-AIbZIP Δ bZIP were also detected from ER-and Golgi apparatus-like regions (Fig. 7b). However, the signals of OASIS hardly overlapped with Myc-AIbZIP Δ bZIP and GM130 (Fig. 7b,c), and the nuclear signals of OASIS were   (Fig. 7b,d). This observation indicated AIbZIP could cause the accumulation of both full-length and N-terminal OASIS at the Golgi apparatus by direct binding of each other, resulting in the inhibition of N-terminal OASIS translocation to the nucleus. To examine whether the interaction of AIbZIP and OASIS affects cleavage of OASIS, we co-transfected HEK293T cells with FLAG-OASIS and Myc-AIbZIP Δ bZIP. Interestingly, in contrast to the results in Fig. 6e,f, the amounts of full-length, S1P-, and S2P-cleaved N-terminal OASIS were not affected by any amounts of Myc-AIbZIP Δ bZIP, which could not interact with OASIS ( Fig. 7e and Supplementary Figure S7). These results suggested that AIbZIP prevents OASIS activation by inhibiting S2P-cleavage through direct binding at the Golgi apparatus.

Discussion
AR signaling plays critical roles in the development, proliferation, survival, and progression of prostate cancer 4 . Understanding the physiological functions of downstream molecules of AR is crucial for the development of effective therapeutic strategies targeting prostate cancer. In this study, we demonstrated that AIbZIP is a key molecule of the AR axis regulating the proliferation of prostate cancer by the following evidence: (1) AIbZIP was highly expressed in androgen-sensitive LNCaP cells but not in androgen-insensitive PC-3 cells; (2) AIbZIP was upregulated by SPDEF, which acts downstream of AR signaling; (3) knockdown of AIbZIP decreased the proliferation of LNCaP cells while overexpression of AIbZIP increased it; (4) AIbZIP prevented OASIS from promoting transcription of its target gene p21; (5) AIbZIP directly bound with OASIS via the bZIP domain, and inhibited S2P-mediated cleavage of OASIS (Fig. 8). To date, treatments for the prostate cancer are focused on the reduction of serum androgen levels and inhibition of AR 33,34 . However, androgen deprivation therapy (ADT) increases the sensitivity of AR to androgen, meaning that the AR signal becomes hyperactivated in response to the trace amounts of androgen 33,34 . Therefore, our findings provide a great possibility that targeting AIbZIP could be a more effective therapy for androgen-sensitive prostate cancer than ADT because AIbZIP is an essential factor for the proliferation of these cancer cells. Notably, we demonstrated that overexpression of AIbZIP significantly increased the proliferation of LNCaP cells without androgen stimulation, implying that the increase of AIbZIP expression could contribute to androgen-independent growth of prostate cancer cells and/or acquired resistance to ADT. Indeed, AIbZIP expression has been reported in all grades of prostate cancer 35 , and the expression levels of AIbZIP are increased in malignant tissue compared with benign tissue 17,18 . It is conceivable that interfering with AIbZIP expression or its function could effectively control the proliferation of prostate cancer cells and reduce the side effects of ADT.
Our study demonstrated that the AR axis upregulates the expression of SPDEF, which in turn SPDEF directly promotes transcription of AIbZIP. It is well known that SPDEF is highly expressed in androgen-sensitive prostate cancer cells LNCaP and LNCaP C4-2, but depleted in androgen-insensitive cells PC-3 and DU-145 [36][37][38] . These expression patterns are consistent with those of AIbZIP. Furthermore, it was reported that knockdown of SPDEF decreased the proliferation rates of LNCaP and LNCaP C4-2 cells 38,39 , the data of which support our present results. The expression of genes relevant for cell proliferation including insulin-like growth factors (IGF-I, IGF-II) and the IGF-receptor (IGFR) are considered to be regulated downstream of SPDEF 38,39 , but the significance of these genes in the proliferation of prostate cancer cells remain unclear. Knockdown of AIbZIP, which is a direct target of SPDEF, suppressed the proliferation of prostate cancer cells, indicating that AIbZIP functions downstream of SPDEF as a major regulator of proliferation in these cells. However, several groups reported that SPDEF is a negative regulator of cell proliferation in some cancer types including prostate cancer 40,41 . It was shown that overexpression of SPDEF in PC-3 cells inhibited proliferation and induced apoptosis via negative regulation of stathmin and survivin expression 36,41 . Taken together with our findings, SPDEF has different roles in the regulation of androgen-sensitive and -insensitive prostate cancer cell proliferation. Further investigation of the roles of SPDEF in several cancers including androgen-insensitive prostate cancer is required to understand the mechanism leading to its contradictory effects on cell proliferation.
AIbZIP is structurally a transcription factor that belongs to the OASIS family 11 . However, it displays some features that differ from other OASIS family members. First, AIbZIP was localized to the Golgi apparatus as well as the ER under normal conditions. Second, AIbZIP was never cleaved in response to various stimuli including ER stress although it has the potential to be cleaved by S1P and S2P. These findings indicated that AIbZIP may function as the full-length form at the Golgi apparatus. A previous study of the interaction between bZIP transcription factors using protein arrays showed that AIbZIP could interact with other OASIS family members 42 . However, the physiological significance of those complexes has not been elucidated. Our data showed that AIbZIP may prevent S2P-mediated cleavage of OASIS to inhibit its nuclear translocation via interaction with OASIS. It provides a new insight that AIbZIP could negatively modulate the activation of other bZIP transmembrane proteins by forming a heterodimer at the Golgi apparatus. However, the mechanism by which AIbZIP inhibits protease processing remains unclear. Given that AIbZIP can bind to OASIS via their bZIP domains, one possibility is that this binding may cause a conformational change in the transmembrane domain of OASIS that conceals the S2P recognition site. Further studies are needed to clarify the detailed mechanism involved in the prevention of S2P-mediated cleavage by AIbZIP.
In conclusion, we demonstrated that AIbZIP is upregulated by SPDEF acting downstream of AR in prostate cancer cells. AIbZIP tightly suppresses p21 expression by inhibiting OASIS activation to promote LNCaP cell proliferation. These findings provide a new possibility that AIbZIP could be a potential cancer therapeutic target of prostate cancer.
Plasmids, transfection and electroporation. AIbZIP and SPDEF cDNAs were cloned from LNCaP mRNA by using PCR, and inserted into the pMX retroviral vector, the pTRE2hyg tet-off expression vector, or the pcDNA3.1(+ ) expression vector. The expression vectors for FLAG/Myc-AIbZIP, and Myc-AIbZIP Δ bZIP were amplified by PCR using full-length AIbZIP as a template. The expression vectors for FLAG-SPDEF and truncated mutants of FLAG-SPDEF were constructed from full-length SPDEF by PCR. The expression vector for FLAG-OASIS was described in previous our report 43 . The expression vector for FLAG-OASIS Δ bZIP was constructed using FLAG-OASIS plasmid. The transfections were performed using ScreenFect A (Wako) according to the manufacturer's protocol. The transfections of LNCaP cells in ChIP assays were performed using electroporation (BEX CUY21EX). The reporter plasmids, driven by the human AIbZIP promoter regions, were cloned from LNCaP genome by using PCR and inserted into pGL3-basic reporter plasmids. All primer sets for cloning are indicated in Supplementary Table 1.
Western blot analysis and Antibodies. Western blot analysis was performed as described previously 44 . The quantification of western blot analysis was performed by Quantity one software (Bio-Rad). Antibodies used in western blot analysis, immunoprecipitation, and immunofluorescence were summarized in Supplementary Table 2. RNA extraction and RT-PCR. Total RNA was extracted from cells using ISOGEN (Nippongene) according to the manufacturer's protocol. RT-PCR assays were performed according to our published procedures 44 . Primer sequences are summarized in Supplementary Table 3. The quantification of RT-PCR was performed by Quantity one software (Bio-Rad).
Immunofluorescence staining. Immunofluorescence staining was performed as described previously 43 .
Staining was visualized under a confocal microscope (Olympus FV1000D). The quantification of co-localization between OASIS and GM130 signals was calculated on a pixel-by-pixel basis WCIF version of ImageJ (http:// www.uhnresearch.ca/facilities/wcif/imagej/), which generates a scatter-plot of the pixel intensities to calculate the threshold for each channel 45 . The scatter-plot is then used to calculate the number of co-localized pixels and their intensities. To calculate the proportion of co-localization, we used the sum of intensities greater than the threshold that showed co-localization divided by the sum of intensities greater than the threshold of the respective channel that did not co-localize. Fluorescence intensity in nuclear area was measured by Image_analysis_software CS Analyzer 4 (ATTO CORPORATION).

RNA interference and Cell proliferation assay. LNCaP cells were transfected using Lipofectamine
RNAiMAX (Invitrogen) according to the manufacturer's protocols. The siRNA sequences were summarized in Supplementary Table 4. For cell proliferation assay, after treatment with doxycycline or transfection with siRNAs into LNCaP cells (1 × 10 5 cells/well in 6-well plates), the numbers of cells were counted at the indicated time.
Luciferase assay. Reporter plasmids containing the AIbZIP promoter region were co-transfected into HEK293T cells along with Renilla luciferase (internal control) and vectors expressing a series of truncated SPDEF mutant using Screenfect A (Wako). 24 hours after transfection, cells were harvested into passive lysis buffer (Promega), and dual luciferase activity was assayed with a GloMax Multi + Detection System (Promega). Reporter luciferase activity was normalized to the internal Renilla control activity.
Chromatin Immunoprecipitation assay and Immunoprecipitation. The chromatin immunoprecipitation assay was performed as previously described 15 . Briefly, LNCaP cells were infected with FLAG-SPDEF and cross-linked using formaldehyde for 10 min at 37 °C. Then cells were lysed using SDS lysis buffer and sonicated (25 × 5-s sonication pulses at 10-s intervals). Equal amounts of chromatin from each sample were incubated overnight at 4 °C with anti-FLAG M2 or anti-Histone H3 antibody. Cross-linking was reversed (> 6 h at 65 °C) and the DNA was purified by phenol-chloroform extraction and ethanol precipitation. The purified DNAs were subjected to PCR analysis using primer sets summarized in Supplementary Table 5. For immunoprecipitation, cell lysates were incubated with anti-FLAG or anti-Myc antibodies overnight, and then rotated with protein G-agarose (Invitrogen) at 4 °C for 3 hours. The agarose beads were washed with TNE buffer (10 mM Tris, 1 mM EDTA, and 150 mM NaCl). Statistical Analysis. Statistical comparisons were made using the unpaired Student's t-test. Statistical significance between two samples was determined by a p-value of less than 0.05. p-values of less than 0.05, 0.01 or 0.001 are described as *p < 0.05; **p < 0.01; or ***p < 0.001, respectively.