ZNF143 protein is an important regulator of the myeloid transcription factor C/EBPα

The transcription factor C/EBPα is essential for myeloid differentiation and is frequently dysregulated in acute myeloid leukemia. Although studied extensively, the precise regulation of its gene by upstream factors has remained largely elusive. Here, we investigated its transcriptional activation during myeloid differentiation. We identified an evolutionarily conserved octameric sequence, CCCAGCAG, ∼100 bases upstream of the CEBPA transcription start site, and demonstrated through mutational analysis that this sequence is crucial for C/EBPα expression. This sequence is present in the genes encoding C/EBPα in humans, rodents, chickens, and frogs and is also present in the promoters of other C/EBP family members. We identified that ZNF143, the human homolog of the Xenopus transcriptional activator STAF, specifically binds to this 8-bp sequence to activate C/EBPα expression in myeloid cells through a mechanism that is distinct from that observed in liver cells and adipocytes. Altogether, our data suggest that ZNF143 plays an important role in the expression of C/EBPα in myeloid cells.

Pluripotent hematopoietic stem cells are capable of differentiating along myeloid, erythroid, or lymphoid pathways (1), via intermediate, more lineage-restricted progenitor cells. Once a cell has committed to the myeloid pathway, it can further develop into either monocytic or granulocytic lineages. This stepwise differentiation process requires the concerted action of several TFs, 5 and aberrant regulation by these TFs has been shown to play a role in leukemia (2)(3)(4). One of the critical TFs during myelopoiesis is C/EBP␣, which is a member of the bZIP family of TFs. This family of proteins contains a leucine zipper domain that is important in protein dimerization and DNA binding (5). Six C/EBP family members have been identified thus far. These proteins share homology in the DNA-binding C terminus while diverging from each other within their N-terminal domain (6). C/EBP␣ is indispensable for inducing the granulocytic differentiation of myeloid progenitors (7)(8)(9) and for the acquisition and maintenance of adult hematopoietic stem cell quiescence (10). It does so by binding to and repressing genes involved in stem cell renewal and by activating the promoters of a myriad of genes involved in myeloid differentiation (11,12). In line with these observations, defects in C/EBP␣ have been reported in ϳ10% of patients with acute myeloid leukemia (13)(14)(15).
Although much is known about the general transcriptional machinery in eukaryotic cells, the events that are required for the transcription of specific genes that encode master controllers of differentiation, such as CEBPA, are not sufficiently understood. In human liver cell lines, C/EBP␣ was reported to autoregulate its promoter indirectly through an upstream stimulatory factor-binding site located at Ϫ268 bp relative to the transcription start site (TSS) (16). In addition to autoactivation, the murine Cebpa promoter also contains two repressive AP2 sites, one at Ϫ218 bp and the other in the 5Ј-untranslated region, and both sites must be occupied to fully suppress the transcription of C/EBP␣ (17).
Whereas most of the known regulatory regions within the CEBPA promoter have been described in hepatocytes or adipocytes, less is known about its regulation in hematopoietic cells. However, Runx1 was recently reported to activate the expression of C/EBP␣ by binding to a conserved site in the murine promoter and to a distal enhancer in hematopoietic cells (18). To further extend our knowledge with regard to myelopoiesis, we examined the proximal promoter of the human CEBPA gene in search of previously undescribed cisregulatory elements. This analysis revealed a highly evolutionarily conserved 8-bp sequence in the CEBPA promoter that is crucial for its transcription. Using affinity chromatography, we were able to demonstrate that zinc finger protein 143 (ZNF143), a human homolog of Xenopus STAF (19,20), bound this element. ZNF143 was previously shown to bind to Ͼ3000 non-coding RNAs and protein-coding genes genome-wide (21). Binding of ZNF143 to this conserved 8-bp regulatory sequence is crucial for the transcriptional activation of the CEBPA gene promoter in myeloid cells.

Identification of a regulatory region essential for expression of C/EBP␣ in myeloid cells
To identify regulatory elements required for myeloid expression of human C/EBP␣, we mapped the minimal promoter region necessary for gene expression in the immature myeloid cell line, U937. We performed transient transfections of luciferase reporter constructs to test the activity of a series of five deletion constructs extending from Ϫ2.2 kb to Ϫ78 bp relative to the TSS (Fig. 1A). We observed that most of the transactivation was directed by a region from Ϫ438 to Ϫ78 (Fig. 1B). Next, we tested a series of internal deletions within the P-438 promoter construct (Fig. 1C). High levels of activity were maintained upon deleting the sequences between positions Ϫ329 and Ϫ121 and between positions Ϫ250 and Ϫ121. However, a significant drop in reporter activity was observed upon deleting the sequence between positions Ϫ309 and Ϫ78 (Fig. 1D). Collectively, these data suggest the presence of one or more impor- Figure 1. Identification of the promoter region essential for C/EBP␣ expression in myeloid cells. A, C, and E, schematic representation of the human CEBPA promoter and luciferase reporter constructs used to study CEBPA transcriptional activation. The numbers indicate the position of the 5Ј-end of each construct relative to the TSS at position ϩ1. The rectangle labeled LUC represents the luciferase gene. A, luciferase map of CEBPA promoter fragments within Ϫ2200 bp relative to the TSS. C, deletion constructs within the 438-bp proximal promoter. ⌬ followed by a number designates the location of internal deletions shown in the schematics. E, deletion constructs within the 438-bp proximal promoter. B, D, and F, graphs depicting luciferase activity in U937 cells in relative luciferase units (RLU). Note that stepwise deletions demonstrate that the region between Ϫ110 and Ϫ95 is crucial for transcriptional activation. PXP1 is the empty vector. Error bars, S.D.

ZNF143 regulates C/EBP␣ expression
tant regulatory elements that are located between Ϫ329 and Ϫ309 and/or between Ϫ121 and Ϫ78 of the human CEBPA promoter.
Next, we generated a series of deletions between Ϫ438 and Ϫ78 (Fig. 1E). Whereas a consistently high level of luciferase activity was observed from the P-438, P-121, and P-110 promoter constructs, of which the two latter lack the region between Ϫ328 and Ϫ306, a significant drop in activity was observed with deletion of the sequence between Ϫ110 and Ϫ95 (Fig. 1F). These findings indicate that a 116-bp region is sufficient to drive proximal promoter activity in U937 cells and also that the region between Ϫ110 and Ϫ78 is required for maximal gene transcription.

A specific DNA-protein interaction occurs between ؊112 and ؊88 of the CEBPA promoter
Based on our deletion studies, we hypothesized that a TF bound to the proximal promoter between Ϫ121 and Ϫ78. To further narrow down the binding region, we generated a series of overlapping double-stranded oligonucleotide probes ( Fig.  2A) and performed EMSAs using nuclear extracts from U937 cells, which express C/EBP␣. Probe II (Ϫ112 to Ϫ88; Fig. 2B, lanes 5-8) generated two complexes (lane 6) that could be outcompeted by a 100-fold excess of non-labeled ("cold") specific self-probe (lane 7). These complexes could not be outcompeted by a cold nonspecific probe containing an SP1 site from the PU.1 promoter (lane 8). Probe I (Ϫ124 to Ϫ101) did not generate a specific shifted band on the gel (Fig. 2B, lane 2). Probe III (Ϫ96 to Ϫ71; lanes 9 -12) generated a complex that could be partially outcompeted with excess unlabeled self-probe (lane 11) but also by the non-self-probe used as a negative control (lane 12), indicative of a nonspecific interaction. Probe II overlaps with portions of both probes I and III but has a unique core region. Given that this probe generated a complex of greater intensity and specificity, we concluded that a specific DNAprotein interaction does indeed occur within this region of the CEBPA promoter.
EMSA on eight other cell lines demonstrated a similar banding pattern (Fig. 2C), indicating that this factor is widely expressed. Two bands were observed in some cell lines. Given that the assay is performed under non-denaturing conditions, it is difficult to assess whether the probe is binding two different factors, the same factor with different protein modifications, or a single factor that is partially degraded.

A highly conserved 8-bp element within the proximal promoter of different C/EBPs is essential for promoter activation
The results obtained in the luciferase assays and EMSA experiments strongly suggested the presence of an important regulatory region between Ϫ112 and Ϫ88 upstream of the human CEBPA TSS. Therefore, we examined its sequence conservation between different species. We identified an octameric sequence motif (CCCAGCAG) that was fully conserved among human, mouse, rat, and chicken CEBPA proximal promoters (Fig. 2D). In humans, it is located from position Ϫ106 to Ϫ99. This octameric sequence is also present in human CEBPD and CEBPE promoters, but not in CEBPB or CEBPG promoters. Interestingly, we also observed a similar motif in the Xenopus CEBPA promoter, with all but the seventh base conserved Figure 2. Protein binding to a conserved motif centered around ؊100 bp. A, schematic representation of the probes designed to pinpoint the region of the CEBPA promoter in which binding occurs. B, EMSAs of the region between positions Ϫ124 and Ϫ71 using the three overlapping probes (I, II, and III). Note that only probe II (lanes 5-8) generated two complexes that could be outcompeted by a 100-fold excess of unlabeled double-stranded self-oligonucleotide (lane 7). S, self-competitor (non-labeled probe); N, non-self-competitor (unrelated DNA probe). C, EMSA demonstrates binding in nine different cell lines when using probe II. Probe II alone and with self-competitor (S) are loaded as controls in lanes 1 and 3, respectively. D, presence of the conserved 8-base sequence "CCCAGCAG" (gray) shared with several C/EBP family members across different species (h, human; m, mouse; r, rat; c, chicken). The locations of the sequences relative to the TSS are indicated. (CCCAGCCG), at Ϫ108 to Ϫ101 bp relative to the TSS, suggesting a spatial and sequential conservation of this motif.

ZNF143 regulates C/EBP␣ expression
To identify which residues are crucial for protein binding, methylation interference assays were performed using the antisense strand of probe II (Fig. 3A). Comparing the patterns produced by the free probe and the retarded band, which represents the DNA-protein complex, we observed three guanosine residues at positions Ϫ105, Ϫ104, and Ϫ101 (Fig. 3A, lane 3) with diminished intensity relative to the free probe (Fig. 3A, lane 4), indicating their close contact with a DNA-binding protein. No other protected regions were observed, showing that this specific site is critical for the formation of the observed complex. We then introduced mutations into these three guanosine residues (G 3 T at positions Ϫ101, Ϫ104, and Ϫ105) and used this mutated probe to perform EMSA. We assessed the effects on DNA binding but only observed partial loss (data not shown). Next, we performed EMSAs using probes that harbored different numbers of mutated nucleotides. These mutational analyses demonstrated that binding could be abolished only upon mutating at least six of the eight bases in the core sequence CCCAGCAG (Fig. 3B) (data not shown). Fig. 3B demonstrates that alteration of the ACCCAGCAG sequence in probe II to TTCACCAAA abolishes the interaction (lanes [1][2][3][4]. This unlabeled mutant probe was unable to compete with the observed complex when using the wild-type labeled oligonucleotide, probe II (Fig. 3B, lane 8).
These mutant sequences were then cloned into various luciferase constructs. As shown in Fig. 3C, the basic pattern among the wild-type constructs is consistent with that observed previously (Fig. 1C). The luciferase activity of the Ϫ443/ϩ124, Ϫ181/ϩ124, and Ϫ125/ϩ124 constructs was consistently higher than that of the Ϫ97/ϩ124 and Ϫ78/ϩ124 constructs, leading us to further focus on the region surrounding position Ϫ100. Additionally, mutation of the CCCAGCAG site significantly decreased luciferase activity in each of the three mutant constructs to a level comparable with that of the Ϫ78 bp minimal sequence (Fig. 3C). These results support our hypothesis that the sequence between Ϫ106 and Ϫ99 is important for promoter activity of these constructs in luciferase assays.

The zinc finger, ZNF143, binds to the CCCAGCAG site in the CEBPA promoter
Next, we determined the molecular mass of the DNA-binding protein that interacts with this region by UV cross-linking (22). Two BrdU-labeled bases were incorporated in the sequence of probe II at the putative DNA-binding site located between positions Ϫ106 and Ϫ99. After performing EMSA, excising the bands, and exposing them to UV light, we loaded the gel slices on an SDS-polyacrylamide gel. As shown in Fig.  3D, the lower band migrates with apparent molecular mass of 95 kDa, and the upper band migrates similar to a 105-kDa protein. Attempts to identify the protein using a variety of screen- Figure 3. The identified protein binds to the conserved CCCAGCAG site. A, methylation interference analysis of DNA contact points of the DNA-binding factor to the C/EBP␣ promoter demonstrates that a protein makes specific contact with this region. Lanes G and GϩA are sequence standards prepared from unmethylated probe by DNA sequencing. Lane F is free DNA, and lane X is DNA from the observed DNA-protein complex. Both samples were subjected to piperidine-mediated cleavage of methylated guanosine residues. The asterisks indicate a direct interaction between the DNA and protein at bases Ϫ101, Ϫ104, and possibly Ϫ105. B, mutation of the conserved CCCAGCAG core region abolishes binding. U937 nuclear extract was incubated with radiolabeled probe II or a mutant derivative (probe M). FP, free probe; *, specific band. C, schematic representation of the WT and mutant luciferase constructs used to test the effect of the full CCCAGCAG sequence on promoter activity. Mutations were introduced in three constructs containing various segments of the CEBPA promoter region. The associated graph reports luciferase activity of each construct in U937 cells after transient transfection. D, UV cross-linking determines the molecular weight of the DNA binding factor at ϳ80 -95 kDa. A double-stranded oligonucleotide containing BrdU was synthesized and radiolabeled. Samples were electrophoresed on a 6% non-denaturing gel, and two complexes were excised and further separated on a 10% SDS-polyacrylamide gel. Lane 1, upper complex, estimated to be 95 kDa; lane 2, lower band, estimated to be 80 kDa. Error bars, S.D.

ZNF143 regulates C/EBP␣ expression
ing techniques, including FROGS (23) and yeast one-hybrid (24), were unsuccessful. In addition, EMSAs using control sequences demonstrated that the protein was not Sp1 or a member of the STAT family. Finally, we identified the protein binding to the CCCAGCAG site by partially purifying a HeLa nuclear extract using column chromatography in combination with SDS-PAGE and EMSA. HeLa cells were used, as they can be grown in large quantities in spinner flasks and therefore often used to isolate TFs. We first partially purified the DNAbinding activity of the crude nuclear extract by anion-exchange chromatography using a Mono Q column. Approximately 90% of the DNA-binding activity was eluted from the column, leading to a 100-fold purification of the factor compared with crude nuclear extract (Fig. 4A). This fraction was subsequently separated on an 8% SDS-polyacrylamide gel, 12 slices of different sizes in the 100-kDa range were excised (Fig. 4B), and the polypeptides within were renatured and tested for binding in an EMSA using probe II. Fraction 5 contained the highest binding activity (Fig. 4C), so we submitted the gel slice corresponding to this fraction for mass spectrometry analyses.
Simultaneously, motif analysis on 566 bp surrounding the TSS of the human CEBPA gene was performed with the Mat-Inspector tool of Genomatix. This is a software tool that utilizes a large library of matrix description for TF binding sites to locate matches in DNA sequences. MatInspector suggested that the Xenopus protein STAF could bind to the CCCAGCAG sequence (Table 1). This finding was in close agreement with our mass spectrometric findings, as the human homolog of STAF, ZNF143 (19,20), was identified in fraction 5 (Table 2). Subsequently, using EMSA, we demonstrated that a protein within fraction 5 bound to a synthetic ZNF/STAF consensus binding site (Fig. 4D) (25) as well as to two other published ZNF143 probes (26) (Fig. 4E). EMSA using HeLa extracts with ZNF143 antibodies demonstrates that ZNF143 binds to the CCCAGCAG motif in the CEBPA promoter (Fig. 4F).
To assess whether ZNF143 binds the CEBPA proximal promoter in myeloid cells, we performed ChIP-exo experiments (27) using ZNF143 antibodies and U937 cells. ChIP-exo is a modification of chromatin immunoprecipitation followed by deep sequencing (ChIP-seq), in which the binding site is narrowed down by employing an exonuclease to trim the immunoprecipitated DNA to a precise distance from the cross-linking site to which the protein binds (27). As expected, we observed binding of ZNF143 to the CEBPA promoter (Fig. 5, A  and B). By performing de novo motif discovery on the ChIP-exo data, we identified a motif containing the conserved CCCAG-CAG sequence (Fig. 5C), which corresponds to the previously published ZNF143-binding motif (21,28). Twelve fragments were selected for further analysis. C, EMSA of size-fractionated extract using the probe from the Ϫ112 to Ϫ86 region of the CEBPA promoter demonstrates binding to fractions 4, 5, and 6. D, EMSA demonstrating binding of the protein in fraction 5 to a probe containing the STAF consensus binding site, CCCAGCAG. E, EMSA demonstrating binding to three different ZNF143 probes in U937 cells and that all three probes undergo a similar supershift as probe II (Fig. 2, A and B) by adding the anti-ZNF143 antibody. F, EMSA demonstrates that ZNF143 binds to the CCCAGCAG motif in the CEBPA promoter in vitro. Note that the negative control, normal rabbit serum, did not supershift the complex (lanes 4 and 8).

ZNF143 regulates C/EBP␣ expression ZNF143 transactivates the human CEBPA promoter in Schneider cells
To test whether ZNF143 is able to transactivate the human CEBPA promoter, we performed transient transfection studies in Drosophila S2 Schneider cells. This cell line was utilized because the presence of binding activity for ZNF143 or the related family member ZNF76 (19) could not be detected by EMSA (Fig. 6A, lane 1). Overexpression of ZNF143 in Schneider cells led to the ϳ10-fold transactivation of the Ϫ179 bp promoter, whereas overexpression of ZNF76, as well as the empty vector, generated only low luciferase activity (Fig. 6B). In addition, mutating the ZNF143 site reduced transactivation 5-fold. Altogether, we show that ZNF143 can transactivate the CEBPA promoter by binding directly to a site within it.

ZNF143 is required for C/EBP␣ expression in myeloid cells
Finally, we investigated whether ZNF143 could regulate C/EBP␣ expression. We identified two shRNA sequences that reduced ZNF143 transcription. Transduction of these shRNA constructs into U937 myeloid cells reduced ZNF143 expression at both the mRNA and protein level by 80% in comparison with a non-silencing control (NSC) (Fig. 7, A-C). In addition, ZNF143 silencing resulted in a strong reduction of CEBPA transcript levels (Fig. 7, A-C). Because ZNF143 binds a motif conserved in CEBPA, CEBPD, and CEBPE (Fig. 2D), we examined the consequences of ZNF143 knockdown on CEBPE expression (as CEBPD is not expressed in U937 cells (29), the effects of CEBPD knockdown could not be analyzed). A significant reduction of CEBPE transcripts was observed in U937 cells transduced with either of two ZNF143 shRNA constructs (Fig. 7D). Further, we analyzed the effect of ZNF143 silencing on other CEBP members and observed that CEBPB and CEBPG transcription was similarly reduced (Fig. 7D). C/EBP␤ reduction was also detectable at the protein level (Fig. 7E). Although CEBPB and CEBPG do not present the binding motif identified Table 2 Mass spectrometric data of identified proteins present within fraction 5 Bold indicates that identified protein is human ZNF143.

ZNF143 regulates C/EBP␣ expression
in the proximal promoter of CEBPA, CEBPD, and CEBPE (Fig.  2D), analysis of the 5Ј-untranslated region demonstrated the presence of a conserved sequence with high similarity to the consensus octamer (CEBPB, CCgAGCAG at position Ϫ318; CEBPG, CCCAGCga at position Ϫ769), which might explain the observed down-regulation. In contrast, there was no reduc-tion in transcript levels of FOS (Fig. 7D), a transcription factor that does not present the ZNF143 conserved motif. Next, a rabbit polyclonal ZNF143-specific antibody, the specificity of which is shown in Fig. 7E, was generated. This antibody was used to perform quantitative ChIP-PCR on shRNA-transduced U937 cells; the observed reduction of ZNF143 protein levels resulted in reduced enrichment of ZNF143 at the CEBPA promoter region (Fig. 7F, left). A similar effect was seen in the U6 promoter, known to be bound by ZNF143 (30) (Fig. 7F, right), whereas no significant changes were observed within an intergenic region that is not bound by ZNF143 (Fig. 7F, middle). Taken together, this study provides evidence that ZNF143 can positively regulate C/EBP␣ expression in myeloid cells by binding to the CCCAGCAG site ϳ100 bp upstream of the CEBPA TSS.

Discussion
In this study, to gain insights into the transcriptional regulation of myeloid development, we investigated the regulatory control elements of the promoter of one of the master regulators of this process, C/EBP␣. Because the programs driving hematopoietic development are largely driven by alterations in transcriptional activity, studying potential TF binding sites within such genes may prove to be key in defining the mechanisms that drive lineage development. Indeed, we demonstrate here that ZNF143 binds to the proximal promoter of CEBPA in vitro and that down-regulation of ZNF143 results in a marked decrease in C/EBP␣ mRNA and protein levels in myeloid cells.

ZNF143 regulates C/EBP␣ expression
Interestingly, mutational analyses showed that there is some flexibility in the 8-bp sequence for formation of a DNA-protein complex. Although, methylation interference assays defined three guanine residues as being in close proximity to a DNAbinding protein (Fig. 3A), it seems that other bases within this sequence compensate and allow the DNA-protein interaction to occur, as we were unable to abolish binding by just mutating one or two bases of this 8-bp motif. We could only completely abolish binding by mutating at least 6 bases in the core sequence CCCAGCAG (Fig. 3B). This demonstrates that, as for most transcription factors (31), there is flexibility in the DNAbinding motif. A study of ϳ100 mouse transcription factors showed that virtually all of them have a unique preference, but almost half of them can recognize several different sequences in addition to the known consensus sequences (32). This is also suggested in Fig. 5B, in which the sixth base C and the eighth base G in the 8-bp motif CCCAGCAG have a very low consensus score. Previously, ZNF76 was reported to repress TBP activity via direct protein-protein interaction (33), but we observed no effect on C/EBP␣ promoter activity (Fig. 6B). We conclude that ZNF143 is probably responsible for most of the observed C/EBP␣ transcriptional activity on this conserved cis-element. On the other hand, it is also possible that other factors necessary for ZNF76 to exert its effect on C/EBP␣ promoter activity are absent in S2 cells. Further, we observed that ZNF143 downregulation in myeloid cells resulted in a strong reduction in CEBPA, CEBPB, CEBPE, and CEBPG transcript levels. CEBPB and CEBPG do not retain the conserved ZNF143 motif; however, sequences with a 1-or 2-nucleotide difference from the core sequence, CCCAGCAG, were noted elsewhere in the 5Ј-untranslated region. As already mentioned, our data suggest that there might be some flexibility in the 8-bp sequence for formation of a DNA-protein complex, and we hypothesize that, via the alternative core sequences, ZNF143 can also regulate CEBPB and CEBPG expression. These observations challenge our ability to understand how transcription factors and their DNA-binding motifs exactly interact to regulate subsequent gene expression, as this opens a wider array of potential TF-binding sites, further increasing the complexity of the protein-DNA binding landscape.
ZNF143 is a ubiquitously expressed transcriptional activator that belongs to the Kruppel family of zinc finger proteins. It has been shown to associate with CHD8 and contribute to the efficient transcription of the U6 RNA polymerase III (30). Furthermore, ZNF143 has been implicated in the transcriptional regulation of genes associated with cell cycle and DNA replication (34) and DNA repair genes following exposure to cisplatin (35). ZNF143 is also a key component of the three-dimensional chromatin structure (36) and is required for human embryonic stem cell pluripotency (37). However, until this study, ZNF143 had not been implicated in hematopoietic development,  (sh#1 and sh#2). B, bar graphs quantifying the C/EBP␣ protein levels from A demonstrate that knockdown of ZNF143 in U937 cells leads to down-regulation of C/EBP␣ protein levels using two independent shRNA-targeting constructs. C, bar graphs demonstrating levels of ZNF143 mRNA in cells expressing NSC, sh#1, or sh#2. D, CEBPA, CEBPB, CEBPE, CEBPG, and FOS mRNA levels in cells expressing NSC, sh#1, or sh#2. E, Western blots demonstrating that ZNF143 knockdown in U937 cells decreases C/EBP␣ and C/EBP␤ levels. A ZNF143-specific antibody was employed. Actin was used as a loading control. Lanes 1-3, S2 cells overexpressing empty vector control (EV), ZNF76, and ZNF143, respectively. Lanes 4 and 5, U937 cells transduced with NSC, ZNF143 sh#1, and ZNF143 sh#2, respectively. F, quantitative ChIP-PCR using the U937 cells transduced with NSC (black bars), ZNF143 sh#1 (white bars), and ZNF143 sh#2 (gray bars). Immunoprecipitation was performed with antibodies against IgG control and ZNF143, and enrichment was determined in a CEBPA promoter region, intergenic region, and U6 promoter. y axes indicate enrichment as a percentage of input. Error bars, S.D.

ZNF143 regulates C/EBP␣ expression
although its knockdown in Zebrafish causes reduced blood cell formation (38).
Whereas the expression of ZNF143 is ubiquitous (Fig. 2C), at the same time, it appears to be important for expression of C/EBP␣ in myeloid cells. Furthermore, the regions of the CEBPA promoter required for activity in myeloid cells appear to be distinct from those that drive its expression in hepatic cells (16). We hypothesize that ZNF143, rather than being a sole regulator of CEBPA expression in myeloid cells, might interact with other cell type-specific regulatory factor(s) that are yet to be identified. Therefore, it will be important in future studies to identify the component(s) that might further contribute to myeloid transcriptional activity of the CEBPA promoter. A number of ubiquitously expressed transcription factors can contribute to myeloid specificity of promoters, including Sp1 (39 -42). One possibility is that other transcription factors cooperate with ZNF143 to provide specificity. Among various candidates, the transcriptional activator Notch1 and its associated DNA-binding repressor, RBPJ, strongly overlap with ZNF143-binding sites in other cell types (21,28). Interestingly, the Notch target, Hes-1, represses the Cebpa promoter in myeloid cells, with a Hes-1 site at Ϫ170 bp (43). In addition, a number of binding sites for the transcriptional repressor THAP11 were observed to coincide with ZNF143 sites (21,44). Notch1 signaling is aberrant in many human cancers and in myeloid leukemia (45) and has been implicated in regulating self-renewal of tumor stem cells (46). Similarly, THAP11 is involved in self-renewal of embryonic stem cells (47). C/EBP␣ has also been implicated in regulating self-renewal of fetal liver and adult hematopoietic stem cells (10), and aberrant C/EBP␣ regulation plays a key role in myeloid leukemia (15) and solid tumors (48). Taken together, it appears that ZNF143 may modulate genes involved in biological processes related to cell growth in rapidly dividing cells, such as stem cells and tumors, via potential cofactors THAP11 and Notch1.
A second possibility is the recent observation that ZNF143 can contribute to specificity by contributing to interactions between promoters and distal regulatory elements (49). Recently, two groups have identified an important CEBPA enhancer at ϩ42 kb relative to the transcription start site that interacts with the promoter to mediate myeloid-specific expression of C/EBP␣ (50,51). The effect of competitive or combinatorial binding of ZNF143 and other potential partners on the transcriptional activity of target genes, such as C/EBP␣, as well as the potential interaction of ZNF143 binding to the CEBPA promoter and distal regulatory elements (18,50,51) merit further investigation.

Constructs
We used the GRCh37/hg19 reference genome to determine the sequence of the CEBPA proximal promoter region. All CEBPA constructs used in this study are numbered here relative to the TSS determined previously (16), as illustrated in supplemental Fig. S1. Luciferase constructs were kindly provided by Dr. Gretchen Darlington and were previously used in a study involving Hep3B2 cells (16). These constructs were gen-erated by first subcloning a 2.2-kb EcoRI/NruI fragment into PXP1. Further deletions were generated using BamHI, KpnI, StuI, EagI, AvrII, and SmaI sites. Deletion constructs starting from position Ϫ110 or Ϫ95 were generated by PCR, subcloned into the pGEM-T vector, and subsequently cloned into the PXP1 vector (52). These constructs were sequenced using the dideoxy method. We then used PCR to generate another series of deletions within 443 bp of the TSS and extending to the initiating ATG (ϩ124). Mutant constructs were generated, using the method of Zaret et al. (53). Two internal primers containing a 9-base mutation, from Ϫ117 to Ϫ108, were synthesized, primers C (5Ј-CTAGGTTCACCAAAGCGCCG-CGCCG-3Ј) and B (5Ј-CGGCGCGGCGCTTTGGTGAACCT-AG-3Ј). Two external BamHI-containing primers, A (5Ј-CGG-GATCCCGCAGGCCTGGTTCTGGCT-3Ј) and D (5Ј-GGG-GTACCCCGGGGGAGTTAGAGTTCT-3Ј), were synthesized. 40 PCR cycles (94°C, 1 min; 55°C, 1 min; 72°C, 1 min) were performed using primers A and B in one reaction and primers C and D in the other. PCR products were gel-purified. 1 l of each purified product was added to a 48-l PCR for 3 cycles of the following: 94°C, 1 min; 55°C, 1 min; 72°C 1.5 min. 1 l of primers A and D (10 M) was added, and 40 PCR cycles were performed: 94°C, 1 min; 55°C, 1 min; 72°C, 1.5 min. The PCR fragment was size-fractionated on a 1.2% agarose gel and purified using Gene-Clean (BIO 101). The fragment was digested with BamHI and KpnI (New England Biolabs) and subcloned into PXPI. Subsequent deletions were made using internal primers. shRNA sequences specifically targeting and consequently down-regulating ZNF143 were cloned into the retroviral vector pRSeG (54). sh#1 (5Ј-GGCAGATGGTGACAACT-TA-3Ј) targets a sequence from exon 3 that was the human version of an RNAi sequence used previously to target murine ZNF143 in stem cells (55). sh#2 (5Ј-GGACATGCTACAAGA-GTAA-3Ј) targets a sequence located in the 3Ј-UTR and was found using Dharmacon's design program for shRNA. ZNF143 and ZNF76 expression constructs were generously donated by Dr. Carbon (Université de Strasbourg, France) and Dr. Kunkel (Texas A&M University, College Station, TX).

Cell culture, transient transfections, and luciferase assays
U937 cells were maintained in RPMI 1640 medium supplemented with 10% FBS and transiently transfected with the indicated constructs (see above, and see Pahl et al. (56)). Briefly, 10 7 cells were electroporated at 220 V at 960 microfarads with a Bio-Rad Genepulser with 20 g of supercoiled luciferase reporter construct and 1 g of CMV-hGH, a plasmid expressing human growth hormone from a CMV promoter. Cells were incubated at 37°C for 6 h, harvested by centrifugation at 1000 rpm for 5 min, and lysed. Luciferase activity was measured on a Monolight 2010 Luminometer (Analytical Luminescence Laboratories). Transfection efficiencies were controlled by measuring human growth hormone levels by radioimmunoassay (Nichols Institute). Schneider (S2) cells were obtained from ATCC and cultured in insect medium (Sigma) supplemented with 10% FBS at 25°C and transfected with Transfectin (Bio-Rad).

ZNF143 regulates C/EBP␣ expression Preparation of nuclear extracts
Nuclear extracts were prepared as followed; cells were spun down and washed with cold PBS and resuspended in isotonic buffer A (10 mm Hepes, 1.5 mM MgCl 2 , pH 7.9) for 30 min. Cells were lysed with a 1-ml tuberculin syringe (18-gauge) by rapidly pumping the sample through at least five times. Nuclei were then lysed and resuspended in buffer B (10 mM Hepes, 1.5 mM MgCl 2 , 400 mM NaCl, pH 7.9). Protein concentration was determined using the method of Bradford (57) (Bio-Rad), and extracts were aliquoted and frozen at Ϫ80°C.

EMSA
Gel shifts assays were performed by incubating 20 g of nuclear extract with 5 l of 5ϫ buffer (50 mM Hepes, pH 7.9, 250 mM KCl, 10 mM MgCl 2 , 1 mM DTT), 2 g of poly(dI-dC), and 50,000 cpm of 32 P-labeled double-stranded probe. Samples were loaded on a 6% polyacrylamide non-denaturing gel and run at 200 V for 3 h in 1ϫ TBE (89 mM Tris, pH 7.6, 89 mM boric acid, 2 mM EDTA). Gels were dried and exposed to film for 12 h or less. ZNF143 antibodies were generously donated by Dr. Philippe Carbon (58). ZNF76 antibodies were generously donated by Dr. Yu-Chung Yang (Case Western University) (33). The sequences of the probes used in EMSA are provided in Table 3.

ZNF143 knockdown experiments
Two shRNA constructs were used to generate virus in Phoenix A cells. A mixture of plasmid DNA and Lipofectamine 2000 was added to 150-mm plates of 80% confluent Phoenix A cells containing DMEM without FBS. 4 h later, 10% FBS was added. 60 h later, the medium was collected, concentrated with a Centricon Plus-70 concentration device (Millipore), and stored at Ϫ80°C. 5 ϫ 10 5 U937 cells were transduced with these viral stocks by co-incubation for 48 h and then grown for 3 days. GFP-positive subpopulations were purified by FACS, resuspended in medium, and grown for an additional week. Wholecell lysates were tested on Western blots using antibodies against ZNF143 (58)

Generation of antibodies against human ZNF143
Rabbit polyclonal ZNF143 antibody was raised against a C-terminal epitope of human full-length ZNF143 (from amino acid 572 to 594, AFHTASSEMGHQQHSHHLVTTET). Briefly, preimmunized blood was taken 2 weeks before first antigen injection. Antigen was injected into rabbits on days 0, 21, and 42. Blood samples were taken on days 28 and 49. Rabbits were euthanized, and blood was collected for further antibody purification on day 56.

UV cross-linking
EMSA was performed with a double-stranded oligonucleotide containing a halogenated analog of thymidine (BrdU). The sequence of the probe was 5Ј-CTAGGACCCAGCAGGCGCC-GCGCCG-3Ј. 20 g of U937 nuclear extract was incubated with 50,000 cpm of double-stranded probe. Samples were then electrophoresed on a 6% non-reducing 1ϫ acrylamide gel. The gel was exposed to X-ray film without drying. The X-ray film was superimposed on the gel, and the area of interest containing the shifted complex was marked. Two bands were carefully excised from the gel using a razor blade. Gel slices were irradiated with UV light at 254 nm for 5 min to cross-link the labeled probe to the DNA-binding protein of interest. Gel slices were then boiled in 200 l of 2ϫ SDS sample buffer containing both DTT and ␤-mercaptoethanol for 5 min and loaded into the wells of a 10% SDS-polyacrylamide gel. 1% agarose was loaded into the well to seal it into the well. The gel was run for 3 h at 50 mA, dried on a gel dryer, and then exposed to X-ray film overnight. The apparent molecular weights of the bands of interest were determined by comparison with known size markers.

ZNF143 regulates C/EBP␣ expression
␤-mercaptoethanol) was added, and the partially methylated fragment was ethanol-precipitated. 50,000 cpm of this methylated probe was incubated with 20 g of U937 nuclear extract. This sample was then run on a 6% acrylamide gel in 1ϫ TBE. The gel was exposed to film, and the complexes of interest were located by superimposing the X-ray film over the gel. The fragments of interest were isolated and extracted from the gel using 0.3 M ammonium acetate. The pellet was resuspended in 100 l of 1 M piperidine and heated to 90°C to cleave the DNAprotein complex. Samples were then lyophilized and run on a 4% denaturing urea gel. Chemical sequencing of unmethylated oligonucleotides was performed by the Maxim-Gilbert method (59), and samples were run along with test samples for comparison.

Protein purification
A 10-liter preparation of HeLa cells was obtained from the National Culture Facility. After centrifugation of this culture, the 10-ml pellet of cells was resuspended in 10 ml of isotonic buffer A (10 mM Hepes, 1.5 mM MgCl 2 , pH 7.9) for 30 min. Cells were lysed with a 1-ml tuberculin syringe (18-gauge) by rapidly pumping the sample through at least five times. Nuclei were then lysed and resuspended in buffer B (10 mM Hepes, 1.5 mM MgCl 2 , 400 mM NaCl, pH 7.9). Protein concentration was determined using the method of Bradford (57) (Bio-Rad). Ultimately, we obtained 7.5 ml of protein at a concentration of 10 g/l. This sample was then desalted into 10 M Tris, pH 8.0, using a 5-ml HiTrap desalting column, containing Sephadex-G25 (GE Healthcare). The sample was then passed over a 5-ml HiTrap-Q HP column (GE Healthcare). The salt concentration was increased in a step gradient from 100 mM salt to 1 M salt in 100 mM increments. 5 ml/step at a flow rate of 1 ml/min were collected. Fractions were collected and tested for the presence of the DNA-binding activity of interest using EMSAs. As 95% of the activity was found in the 200 mM salt fraction, we concentrated this fraction using a Viva Spin protein concentrator (GE Healthcare), resuspended it in 4ϫ SDS sample buffer (40% glycerol, 240 mM Tris, pH 6.8, 8% SDS, 0.04% bromphenol blue, 5% ␤-mercaptoethanol), and boiled it for 10 min. We subsequently ran this sample on a 10% polyacrylamide SDS gel and excised 20 different fractions around the 100-kDa marker. Each sample was ground up, resuspended in 300 l of renaturation buffer (100 mM NaCl, 10 mM Hepes, pH 7.5, 1 mM EDTA), and stored at 4°C overnight. We performed EMSA on the fractions with probe II to determine which samples contained proteins that were capable of binding to 32 P-labeled probe II. Those samples, as well as a sample that did not show any DNA binding activity, were sent to the Biopolymer facility at Harvard Medical School.

ChIP-exo and motif analyses
8.5 ϫ 10 6 U937 cells were harvested, washed with ice-cold PBS, and cross-linked with 1% formaldehyde (Sigma) for 10 min at room temperature with gentle agitation. 2 M glycine was added to quench the reaction. Cell pellets were obtained by 10 min of cold centrifugation at 2000 rpm and lysed (50 mM Hepes, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS ϩ protease inhibitors (Complete Mini, catalogue no. 04693139001, Roche Applied Sci-ence). The solution was incubated at 4°C for 30 min. Cell pellets were collected by 10 min of centrifugation at 2000 rpm at 4°C, resuspended in lysis buffer, and sonicated in a Bioruptor Next Gen water bath sonicator (Diagenode) to 200 -500 bp and sent to Peconic LLC (State College, PA) for ChIP-exo analysis (60). For both ChIP-exo and ChIP-seq input, reads were mapped onto the human genome (hg19, without mitochondrial chromosome and unassembled contigs) using Bowtie2 version 2.1.0 (option: very sensitive). Reads with mapping quality score under 10 were removed (using Samtools view Ϫq10) to remove unmapped and multimapped reads. Identification of enriched regions (peak calling) was performed with Peakzilla (default options), identifying 10,413 peaks. For visualization, reads were elongated to 123 bp, corresponding to the fragment found by Peakzilla (GSE65130). De novo motif discovery was performed using RSAT peak motifs (algorithm: oligo-analysis, dyads, and position analyses; word length: 6 -8; other options as default).

ChIP-seq input library preparation
8.5 ϫ 10 6 U937 cells were cross-linked in 1% formaldehyde (Sigma) for 10 min at room temperature. After cross-linking, cells were washed once with ice-cold PBS. Cell lysis was carried out for 10 min on ice using 350 l of lysis buffer containing 1% SDS (1st Base), 10 mM EDTA (1st Base), 50 mM Tris, pH 8 (1st Base), and Complete protease inhibitors (Roche Applied Science). Genomic DNA was sheared in a Bioruptor TM Next Gen water bath sonicator (Diagenode) to ϳ200 -500 bp by 10 cycles of 30 s ON, 30 s OFF, on High. DNA fragment size was verified via agarose gel electrophoresis. Input chromatin was reversecross-linked in 125 l of 1% SDS and 0.1 M NaHCO 3 (Sigma) for 6 -7 h at 65°C. The chromatin was purified using the QIAquick PCR spin kit (Qiagen), and at least 5 ng of input DNA was processed for deep sequencing using ThruPLEX-FD kits (Rubicon Genomics) according to the manufacturer's instructions. The input library was sequenced using Illumina HiSeq2000 at the Duke-NUS Genome Biology Facility, and tags were mapped to human genome build hg19.

Quantitative ChIP followed by PCR
60 ϫ 10 6 U937 cells transduced with ZNF143 targeting shRNA constructs (sh#1 and sh#2) or NSC shRNA were used for quantitative ChIP-PCR as described previously (61). Briefly, cells were fixed with 1% formaldehyde and then neutralized with glycine at a final concentration of 0.125 M. After washing with ice-cold PBS three times, cells were lysed in lysis buffer and sonicated to obtain 500-bp genomic DNA fragments. Samples were divided into three equal aliquots, and each was incubated overnight with IgG or antibodies to ZNF143 or CEBP␣. The next day, magnetic beads were added for 2 h and subsequently washed (61). After reverse-cross-linking and purifying chromatin, quantitative PCR was performed using the following oligonucleotides: CEBPA promoter region, 5Ј-GTTAGAGT-TCTCCCGGCATG-3Ј and 5Ј-GTCGAAAATATGGGCCG-TCC-3Ј; U6 promoter region, 5Ј-GCAAAACGCACCACGTG-ACG-3Ј and 5Ј-CTATGTTCCGACAATCTCTC-3Ј; and intergenic region, 5Ј-CAGGCGGTCATTGTCACTG-3Ј and 5Ј-GTGGCTGAAGAACCGGAAC-3Ј.