Repression of the Human Adenine Nucleotide Translocase-2 Gene in Growth-arrested Human Diploid Cells THE ROLE OF NUCLEAR FACTOR-1*

Adenine nucleotide translocase-2 (ANT2) catalyzes the exchange of ATP for ADP across the mitochondrial membrane, thus playing an important role in maintaining the cytosolic phosphorylation potential required for cell growth. Expression of ANT2 is activated by growth stimulation of quiescent cells and is down-regulated when cells become growth-arrested. In this study, we address the mechanism of growth arrest repression. Using a combination of transfection, in vivo dimethyl sulfate mapping, and in vitro DNase I mapping experiments, we identified two protein-binding elements (Go-1 and Go-2) that are responsible for growth arrest of ANT2 expression in human diploid fibroblasts. Proteins that bound the Go elements were purified and identified by matrix-assisted laser desorption ionization time-of-flight mass spectrometry as members of the NF1 family of transcription factors. Chromatin immunoprecipitation analysis showed that NF1 was bound to both Go-1 and Go-2 in quiescent human diploid cells in vivo , but not in the same cells stimulated to growth by serum. NF1 binding correlated with the disappearance of ANT2 transcripts in quiescent cells. Furthermore, overexpres-sion of NF1-A, -C, and -X in NIH3T3 cells repressed expression of an ANT2-driven

The adenine nucleotide translocase (ANT) 1 proteins catalyze the exchange of mitochondria ATP for cytosolic ADP. In doing so, they play an important role in maintaining the cytosolic phosphorylation potential and therefore normal cell growth and function. In addition, the ANTs have been implicated in early events in initiation of mitochondrion-dependent apoptosis (1).
Three ANT isoforms are encoded in separate genes in mammals (2)(3)(4)(5) and yeast (6,7). Two of these isoforms (ANT1 and ANT2) are differentially expressed in mammalian tissues (8 -10) and in differentiating cells (8,(11)(12)(13). ANT2 expression is down-regulated in the latter case (8,(11)(12)(13). Expression of the ANT2 isoform is also growth-dependent (11). Rapid expression of ANT2 mRNA has been demonstrated in a variety of growtharrested mammalian cell types activated to enter the G 1 phase of cell growth (11, 14 -17). ANT2 mRNA expression occurs together with the immediate-early genes required for activation of cell cycle progression and is accounted for solely by the activation of transcription (18). However, unlike the other immediate-early genes, ANT2 expression is maintained throughout the cell cycle. Expression is down-regulated only as cells become growth-arrested at confluence (18).
The mechanism(s) by which gene expression is repressed in cells entering G 0 is poorly understood. Growth arrest-specific genes have been identified in growth-inhibited NIH3T3 cells (19), but these do not participate directly in transcription initiation (20). Transforming growth factor-␤ induces growth arrest of many cell types, leading to repression of many individual genes via the Smad proteins (21,22). However, microarray analysis of growth-stimulated human primary fibroblasts (23,24) revealed up-regulation of relatively few transcription factors, suggesting that modulation of transcription factor expression may not be not a commonly used mechanism for regulating G 0 -specific gene expression.
The mechanism by which ANT2 expression is repressed during growth arrest is not known. We demonstrated previously that removal of a 700-bp upstream region of the human ANT2 promoter prevents growth arrest repression (18). In the present study, we show that growth arrest of ANT2 is mediated by members of the nuclear factor-1 (NF1) family of transcription factors via two DNA elements (Go-1 and Go-2) in the upstream repressor region. The NF1 family consists of four genes, NF1-A, -B, -C, and -X, and a large number of splice variants that can act either as transcriptional activators or repressors (see Ref. 25 for review) depending on the cell context. However, to our knowledge, the repression of ANT2 reported here is the first example of NF1 acting as a growth arrest repressor.

EXPERIMENTAL PROCEDURES
Cell Culture-Human primary diploid foreskin fibroblasts were used in passages 7-17. Diploid fibroblasts and NIH3T3 cells were grown as described (18). For serum starvation, cells were washed twice with phosphate-buffered saline; serum-free medium was added; and incuba-tion was continued for 48 h. Confluent NIH3T3 cells were produced as described (18).
Plasmids-ANT2-Luc reporter plasmids used in stable transfection experiments were prepared using unique restriction enzyme sites (BglII, XbaI, BamHI, and SmaI) in the human ANT2 promoter PstI/PstI fragment (26). These restriction fragments were inserted into the Hin-dIII/NheI sites of pGL3-basic (Promega). ANT2 promoter fragments bearing mutations in the C box were prepared as described (26) using the Mut-2 sequence. All clones were checked for fragment size and orientation. An oligonucleotide containing the mutated ANT2 Go-2 element (nucleotides (nt) Ϫ822 to Ϫ794) was prepared by PCR using a mutated 5Ј-primer (5Ј-CCA ATT CCT TAA AAG ATC TTT GTC GAA C-3Ј, where the underlined nucleotides represent mutation of the wildtype GGC sequence) and the GL2 primer (5Ј-CTT TAT GTT TTT GGC GTC TTC CA-3Ј) from pGL3-basic. The PstI/PstI ANT2-Luc reporter plasmid (26) was used as the template DNA. An oligonucleotide with both the Go-2 (nt Ϫ822 to Ϫ794) and Go-1 (nt Ϫ726 to Ϫ701) elements mutated was prepared by PCR amplification of a short fragment using a set of primers in which the core GGC sequence was changed to TAA. The 5Ј-primer contained the mutated Go-2 element (5Ј-CCA ATT CCT TAA AAG ATC TTT GTC GAA C-3Ј), and the 3Ј-primer contained the mutated Go-1 element (5Ј-GTG TGC TGT CCT GGA TTA AGT GAA ACC-3Ј). The amplified fragment was then used as the 5Ј-primer for another round of amplification using the PstI/PstI ANT2-Luc reporter plasmid (26) as the template and the GL2 primer as the 3Ј-primer. This amplified fragment of ANT2 was used in turn as the template for amplification with the wild-type Go-2 element primer (5Ј-CCA ATT CCT GGC AAG ATC TTT GTC GAA C-3Ј) to obtain a combination of mutations. PCR was performed with Vent DNA polymerase (New England Biolabs Inc.) according to the manufacturer's recommendations. All clones were verified by sequencing.
Stable transfections were performed as described (18). Resistant colonies (100 -200) for each of the luciferase constructs were pooled and grown in the presence of Geneticin (0.4 mg/ml). Luciferase activity measurements were performed as described (18). Protein concentration was measured by the Bio-Rad protein assay.
In Vitro DNase I Footprinting-Nuclear extracts were prepared from human diploid fibroblasts and NIH3T3 cells by the method of Dignam et al. (27). The DNase I protection assay was performed as described by Promega (28). Radioactive probes were prepared by PCR using 5Ј-32 Plabeled chloramphenicol acetyltransferase primer, the M13 primer, and pCAT-ANT2(Ϫ917/Ϫ654) as the template.
In Vivo Dimethyl Sulfate (DMS) Footprinting-DMS footprinting of human diploid cells was performed as described (29) with minor modifications. Cells were treated with 0.1% DMS in 2 ml of phosphatebuffered saline for 2 min at room temperature. After washing with phosphate-buffered saline, cells were lysed with 1 ml of lysis solution (50 mM Tris-Cl (pH 8), 300 mM NaCl, 25 mM EDTA, 0.2% SDS, and 200 g/ml proteinase K). Modified genomic DNA was isolated according to established procedures (29). As a control, unmodified genomic DNA isolated from cells was exposed to 0.125% DMS in vitro for 2 min at room temperature. DMS-modified DNA was cleaved with piperidine (29) and subjected to ligation-mediated PCR with primers directed against the upper strand. Thus, the radiolabeled bands on the gels represent C residues on the coding strand.
SDS-PAGE and Protein Identification by MALDI-TOF/MS-Samples from the DNA affinity column were precipitated for 20 min on ice in 10% trichloroacetic acid, followed by a 15-min centrifugation at 10,000 ϫ g and two washes with ice-cold acetone. Samples were airdried, dissolved in sample buffer, and separated by 10% SDS-PAGE (34). Proteins were visualized by silver staining (35), and bands of interest were cut out. In-gel tryptic digestion and sample preparation were done as described (35). MALDI-TOF analysis was performed in reflector mode using a Voyager-DE STR MALDI-TOF mass spectrometer from Applied Biosystems (Foster City, CA). Internal calibration was done with autodigested trypsin. Data were analyzed using Moverz software (Proteometrics LLC, Winnipeg, Canada), and data base searches were done with Mascot (36). 2 A search of all NCBInr Database entries was performed allowing one missed cleavage for trypsin, carbamidomethylated cysteine residues, and variable modification of oxidized methionine. Peptide tolerance for monoisotopic values was set to 50 ppm. The different forms of NF1 were identified with significant scores.
Western Blot Analysis-Trichloroacetic acid-precipitated samples from the DNA affinity column were subjected to 10% SDS-PAGE and electroblotted onto polyvinylidene difluoride membrane. Membranes 2 Available at www.matrixscience.com. Total RNA (10 g) was loaded on each lane, and Northern analysis was performed with 32 P-labeled ANT2 cDNA. rRNA was used as a measure of the total RNA loaded.
were incubated with antibodies against the N terminus of human NF1 (Santa Cruz Biotechnology) and developed with alkaline phosphataseconjugated secondary antibodies as described (29).
Chromatin Immunoprecipitation-Chromatin immunoprecipitation of NF1 from growth-arrested and growth-induced human diploid cells was performed as described (38), except that 100 l of protein A-Sepharose was used instead of Staphylococcus aureus cells. Immunoprecipitation was performed with 2 l of antiserum 8199 prepared against a central domain of the NF1 C-protein (kindly provided by Dr. Tanese). Amplification of immunoprecipitated DNA fragments (2 l) was performed using primers Ϫ982 and Ϫ795 (5Ј-GTTCGACAAA-GATCTTGCCAGGAATTGG-3Ј) for the Go-2 element and primers Ϫ726 (5Ј-GGTTTCACTGGCTCCAGGACAGCACAC-3Ј) and Ϫ499 (5Ј-GGGT-GAGGCAAGCGAGACAAGGTCATG-3Ј) for the Go-1 element. PCR was performed for 32 cycles, with 30 s of denaturation at 94°C, followed by 30 s of annealing at 60°C and 30 s of extension at 72°C. The last step included extension for 10 min at 72°C.

Growth Arrest of Human Diploid Cells Down-regulates ANT2
Transcripts-Growth arrest repression of the ANT2 gene (18) was studied in vivo using human primary foreskin fibroblasts. Exponentially growing cells expressed ANT2 transcripts at high levels ( Fig. 1, lanes 1 and 5), but transcripts were barely detectable in growth-arrested cells after 48 h of serum starvation (lane 2). However, ANT2 transcript levels were restored to ϳ30% of the control levels after 6 h of serum induction (lane 3) and to 100% after 24 h of induction (lane 4).
Mapping of Proteins Bound to the ANT2 Promoter in Growtharrested Diploid Cells in Vivo-Proteins bound to the ANT2 promoter during growth modulation were mapped in vivo in growing and growth-arrested diploid fibroblasts by DMS modification ( Fig. 2; summarized in Fig. 3). Transcription of ANT2 is maintained by two adjacent, synergistically acting Sp1 elements in the proximal promoter (26,39). However, growth arrest repression appears to be located in a distal 700-bp fragment of the promoter (18). In agreement with these findings, a region of protein contact was detected in vivo within the 700-bp repressor region. In this region, protection from in vivo DMS modification was observed on C residues in growth-arrested cells ( Fig. 2A, S lane) that extended over a stretch of ϳ60 bp (nt Ϫ830 to Ϫ767) (Figs. 2A and 3A). More importantly, none of these C residues was protected in vivo in serum-activated ( Fig Protein contact is also detected in vivo within or near the Sp1 activation elements (AB boxes) and the Sp1 repressor element (C box) in the proximal promoter (26,39). Strongly protected nucleotides were found within the A and B boxes under all conditions of growth (i.e. exponentially growing, serumstarved, and serum-induced cells) ( nently occupied in vivo. Similarly, C residues within and 5Ј of the C box were protected, but less strongly than those in the A and B boxes. In addition, there appears to be at least one C residue between the B and TATA boxes that was protected in all cells, as well as a major hypersensitive site (Figs. 2B and 3B, asterisks). The significance of the latter is not clear.
Identification of Growth Arrest DNA Elements in the ANT2 Promoter-To further define the upstream growth arrest repressor in the human ANT2 promoter, stable transfectants were made with 5Ј-deletion fragments of the promoter (Fig. 4). Since stable transfection of diploid fibroblasts is hampered by their limited life spans, NIH3T3 cells were used. As shown in Fig. 4 (open circles), deletion of a 112-bp fragment between nt Ϫ804 and Ϫ692 abolished repression of luciferase activity in confluent cells. This is the region in which protein binding was observed in growth-arrested diploid cells in vivo (see above). For convenience, we refer to this extended 112-bp region as the Go repressor (GoR) region.
To identify proteins that bind the 112-bp GoR region, an overlapping fragment of the promoter (nt Ϫ917 to Ϫ654) was mapped in vitro with DNase I using nuclear extracts from human diploid fibroblasts and NIH3T3 cells. A 28-bp region (nt Ϫ822 to Ϫ794) was strongly protected by nuclear extracts from both cell types in the growth-arrested state (Fig. 5). This footprint, referred to as a Go-2 element, overlaps the 5Ј-end of the GoR region defined by deletion constructs (Fig. 4) and is part of an extended region that includes a second DNA element, Go-1 (see Figs. 7 and 9 below). DNase I protection of the Go-2 element was interrupted in both cell types by a hypersensitive site (Fig. 5, asterisks), suggesting that the same or a similar protein is bound. In agreement with the in vivo DMS mapping experiments (see above), nuclear extracts from serum-activated human diploid cells did not footprint the Ϫ822/Ϫ794 Go-2 element (Fig. 5, left panel, compare S and I lanes). Thus, either the amount or the binding ability of the DNA-binding protein is modulated by the growth state of the diploid cells. By contrast, nuclear extracts from serum-activated NIH3T3 cells protected the Ϫ822/Ϫ794 element from DNase I (Fig. 5, right panel, compare S and I lanes). Although it is not clear why DNA binding is retained in nuclear extracts from growing 3T3 cells, the result may provide an explanation for the weaker serum induction of ANT2 expression observed in 3T3 cells (31) compared with human diploid cells.
Additional Putative Repressor Elements in the ANT2 Promoter Are Not Involved in Growth Arrest Repression-The above data show that the GoR region plays a major role in repressing ANT2 expression in growth-arrested diploid cells. However, Sp1 bound to the C box also acts as a repressor of ANT2 promoter activity (26,40), raising the possibility that it may also contribute to growth arrest repression. To test this, stable transfectants of NIH3T3 cells were prepared using human ANT2 reporter gene constructs bearing various 5Ј-deletions with and without a mutated C box. Reporter gene activity was measured in cells approaching confluence. In agreement with data in Fig. 4, repression of luciferase activity was lost in growth-arrested 3T3 cells only when the GoR region was deleted (Fig. 6C). However, mutating the C box (open symbols) either in the presence (Fig. 6, A and B) or absence (Fig. 6C) of the GoR region did not significantly alter growth arrest repression, showing that the Sp1 C box plays little or no role in this process. However, the C box contributes to constitutive repression of the gene since mutations that prevent Sp1 binding increased the absolute activity of the reporter gene by 2-5-fold (data not shown), as described previously (26,39).
We have also identified a putative silencer centered at nt Ϫ332 in the ANT2 promoter (41). To test the contribution of this region to growth arrest repression, promoter-driven reporter gene constructs were prepared with deletions between the Go element and the AB boxes, some of which (clones ⌬Ϫ692/Ϫ235 and ⌬Ϫ546/Ϫ235) (Fig. 6D) delete the Ϫ332 silencer. Furthermore, since ANT2 expression is maintained via Sp1 on the AB boxes and the spatial arrangement of the Sp1 elements in the proximal promoter is critical for optimal gene expression (40), these clones also provide a test of the importance of the spatial relationship between Sp1 and the Go element. Stable transfections of these constructs (Fig. 6D) were made in NIH3T3 cells. None of the deletions prevented repression in cells approaching confluence (Fig. 6D). Thus, neither the silencer centered at nt Ϫ332 nor the distance between the Go element and the AB box Sp1 activation sites appears to be influence ANT2 growth repression.
Purification of the Growth Arrest Repressor Element-binding Protein-The above data indicate that repression of ANT2 in growth-arrested cells is achieved through protein(s) binding to the GoR region. Since the rat liver and HeLa nuclear extracts contain a DNA-binding activity that is indistinguishable from the Go element-binding activity in human diploid fibroblast nuclear extract, 3 both extracts were used for protein purification. The eluted fractions from individual purification steps (see "Experimental Procedures") were tested for DNA-binding activity by in vitro DNase I protection assay of the Ϫ917/Ϫ654 GoR region of the ANT2 promoter (see above). Fractions exhibiting protection were pooled and used in the subsequent purification step. Fig. 7 shows the DNase I protection pattern using fractions eluted from the last DNA affinity purification step. The first two fractions from the 500 mM NaCl elution step contained Go element-binding activity. However, binding to DNA was not limited to the Go-2 element (which was used as bait on the DNA affinity column), but was also observed on a sequence located between nt Ϫ726 and Ϫ701 (summarized in Fig. 3A), which we termed the Go-1 element. Both Go-1 and Go-2 are within the GoR region defined by transfection/deletion studies ( Fig. 6; see above). Go-1 was not be detected by DNase I protection assay with whole nuclear extracts. The reasons for this are not clear at present.
The SDS-PAGE profiles of the DNA affinity-purified fractions from rat liver (Fig. 8A) and HeLa cells (Fig. 8B) are very similar. Polypeptides from both cell types were eluted and analyzed by MALDI-TOF/MS. MS analysis identified several of these polypeptides as members of the NF1 family. These are indicated by small black boxes in Fig. 8 (A and B). The presence of NF1 in the active fractions was also confirmed by Western blot analysis (Fig. 8C). NF1 proteins were present only in those fractions from the DNA affinity column that footprinted the Go-1 and Go-2 elements.
Identical DNase I footprints were obtained with recombinant NF1 and with affinity-purified fractions from rat liver and HeLa nuclear extracts (Fig. 9A). All three preparations protected both the Go-1 (nt Ϫ726 to Ϫ701) and Go-2 (nt Ϫ822 to Ϫ794) elements. In addition, all three samples introduced identical hypersensitive sites within the two protected regions (Fig.  9A). This result indicates that NF1 and the Go-1-and Go-2binding proteins in HeLa and liver nuclear extracts are one and the same.
Electrophoretic mobility shift assays (EMSAs) were also conducted to confirm the specific binding of NF1 to Go-1 and Go-2 (Fig. 9B). Oligonucleotides carrying the wild-type Go-1 and Go-2 elements (Fig. 9B, WT) strongly competed for binding of HeLa nuclear extracts to a labeled probe containing the NF1 bipartite consensus binding sequence. However, competition was lost when the core NF1-binding sequence in the Go-1 and Go-2 elements was mutated (Fig. 9B, Mut). Similar results were obtained in experiments with purified recombinant NF1 (Fig. 9C). Finally, antibodies against NF1 prevented formation of the major electrophoretic mobility shifted band formed with HeLa cell nuclear extracts (Fig. 9B, last lane).
The above experiments identified members of the NF1 tran-3 P. Barath scription factor family as the rat liver and HeLa cell proteins that protect Go-1 and Go-2 in the ANT2 promoter. To determine whether NF1 is also the protein that binds to the GoR region in human diploid fibroblasts, we carried out in vitro DNase I protection analysis with nuclear extracts from quiescent diploid cells and competition with wild-type and mutated NF1 consensus oligonucleotides (Fig. 10). In agreement with the above results, nuclear extracts from quiescent diploid cells protected the Go-2 element, and protection was eliminated by a competitor oligonucleotide bearing a wild-type NF1 consensus element, but not by one bearing a mutated element (Fig. 10, WT versus Mut lanes). Furthermore, no protection of Go-2 was found using nuclear extracts from serum-induced diploid cells (Fig. 10, right panel), in agreement with the data presented in Figs. 2 and 5 indicating occupation of this site only in the quiescent state. By contrast, Go-1 (Fig. 9A) was not protected by nuclear extracts from quiescent diploid cells (data not shown), even though this element was occupied by NF1 in vivo (see below).

Mutation of the NF1-binding Element Prevents Growth
Arrest Repression-To further test the functionality of Go-1 and Go-2, stable transfections of NIH3T3 cells were done with constructs in which the putative NF1 half-site (TGGC) in each element was mutated. Reporter gene activity was measured in cells approaching confluence. We found that luciferase activity was repressed upon growth to confluence if either the Go-1 or Go-2 element was mutated (Fig. 11, closed circles, closed   FIG. 9. Purified recombinant NF1 and affinity-purified fractions from rat liver and HeLa cell nuclear extracts exhibit identical properties in the DNase I and EMSAs. A, DNase I assay; B and C, EMSA. A, DNA affinity-purified fractions from rat and HeLa cells (fraction 4) (see Fig. 8) and purified recombinant NF1 protein were used for DNase I protection analysis of the ANT2 GoR promoter region. Digested probe (P lanes) is shown. Only the coding strand is shown. Asterisks denote hypersensitive nucleotides. Nucleotide numbering is relative to the transcription start site. B, EMSA was done with HeLa nuclear extracts and an oligonucleotide probe containing an NF1 bipartite consensus binding sequence (probe NF1 WT; Santa Cruz Biotechnology). Wild-type (WT) or mutated (Mut) competitor oligonucleotides containing the NF1 bipartite consensus binding element, the Go-1 element (nt Ϫ726 to Ϫ701), or the Go-2 element (nt Ϫ822 to Ϫ794) were added in 50-fold excess. In all mutations, the core NF1-binding sequence (TGGCA) was changed to TTAAA. The sample in the Ab NF1 lane was preincubated with antibody (Ab) 8199. The major shifted complex is marked with an arrow. The asterisk denotes a nonspecifically bound probe. C, EMSA was done with purified recombinant NF1 and the Go-2 element oligonucleotide (nt Ϫ822 to Ϫ794) as the probe. Competitor oligonucleotides containing the NF1 consensus sequence (NF1 WT; left panel) or the Go-2 element (nt Ϫ822 to Ϫ794; right panel) were added in 50-fold excess.
FIG. 10. Identification of NF1 as the protein in quiescent human diploid fibroblasts that binds to the ANT2 G 0 repressor element. DNase I protection assay of the ANT2 GoR region was carried out with nuclear extracts from quiescent diploid fibroblasts (left and middle panels). Extracts were incubated prior to DNase I digestion with a 10-, 20, or 50-fold excess of competitor oligonucleotide (oligo) containing either a wild-type (WT) or mutated (Mut) NF1 bipartite consensus binding element. The P lanes contain a digest of the probe in the absence of protein. DNase I protection assay of the ANT2 GoR region was carried out with nuclear extracts from growth-arrested (starved) and serum-activated (induced) diploid cells (right panel). The asterisk denotes a hypersensitive site. Nucleotide numbering is relative to the transcription start site. squares, and open circles), but not when both were mutated (open squares). Thus, both NF1-binding sites play a role in repression of ANT2 expression in cells entering the G 0 state. Moreover, these data suggest that both sites most probably function independently of each other.
NF1 Binds to the Go Elements Only in Growth-arrested Diploid Cells-To test the binding of NF1 to the ANT2 promoter in vivo during growth modulation of diploid cells, we performed chromatin immunoprecipitation experiments (Fig. 12). In agreement with in vivo DMS mapping ( Fig. 2A) and in vitro DNase I protection (Fig. 5) experiments, NF1 was bound to Go-2 (Fig. 12A) in growth-arrested cells, but not in seruminduced cells. Similarly, NF1 was also bound in vivo to Go-1 (Fig. 12B) in quiescent cells, but not in serum-activated cells. Thus, repression of the human ANT2 gene in growth-arrested human diploid cells is associated with NF1 binding to both Go elements in vivo, and binding is released upon growth activation.
Expression of NF1 Isoforms Represses ANT2 Promoter Activity in NIH3T3 Cells-To assess the effect of NF1 isoforms on ANT2 gene expression, we transfected HeLa and NIH3T3 cells with plasmids expressing mouse NF1-A, -B, -C, and -X proteins (Fig. 13). Expression of all four NF1 isoforms activated the ANT2 promoter from 5-to 10-fold in HeLa cells. In contrast, expression of NF1-A, -C, and -X in NIH3T3 cells repressed the ANT2 promoter by 60 -75%. In these cells, NF1-B slightly activated the promoter. This observation extends existing data on the dual role of NF1 in activating and/or repressing gene expression.

DISCUSSION
The ANT2 gene was identified as one of the immediate-early genes expressed in cells stimulated to enter growth and division (11). However, unlike most immediate-early genes, expression of ANT2 is not cell cycle-regulated, but rather is downregulated when cells become growth-arrested (18). Repression of ANT2 in growth-arrested cells is eliminated by deleting an upstream region of the promoter (18), suggesting the presence of an active growth-related repressor. In this study, we identified two protein-binding DNA elements (Go-1 and Go-2) that are necessary for growth arrest repression of ANT2. The proteins binding these elements were purified by DNA affinity chromatography and found to be members of the NF1 family of transcription factors. We further showed, by chromatin immunoprecipitation analysis, that both elements were occupied by NF1 in quiescent human diploid cells in vivo, but not in the same cells stimulated to growth. Our data support the view that NF1 binding to specific ANT2 repressor elements is a necessary event for growth arrest repression of the gene.
That the Go-1 and Go-2 elements described here are responsible for growth arrest repression of ANT2 is strengthened by experiments that exclude the participation of additional repressor or silencer regions reported to be present in the ANT2 promoter (26,41,42). One of these is an Sp1-binding element FIG. 12. NF1 is bound to the GoR region only in growth-arrested human diploid fibroblasts. Chromatin immunoprecipitation was carried out on formaldehyde-cross-linked human diploid cells. 48-h serum-starved cells (Starved), cells serum-starved and serum-activated for 24 h (Induced), or untreated cells were cross-linked with 0.25% formaldehyde. Chromatin was isolated, and specific DNA-protein complexes were immunoprecipitated with anti-NF1 antibody (Ab) 8199. After reversal of the cross-link, DNA was amplified by PCR using primers Ϫ982 and Ϫ795, covering the Go-2 element (A), and primers Ϫ726 and Ϫ499, covering the Go-1 element (B). Total, total chromatin; ϪAb, a negative control without antibodies. PCR also included the negative control without the template (ϪDNA). As a marker (Marker lanes), a 100-bp gene ruler (MBI Fermentas) was used. The expected sizes of the amplified immunoprecipitated fragments are given on the right.
FIG. 13. NF1 isoforms repress the ANT2 promoter in transfected NIH3T3 cells. NIH3T3 and HeLa cells (0.25 ϫ 10 6 cells) were transfected with 5 g of PstI/PstI ANT2-Luc plasmid carrying the full-length human ANT2 promoter (19); 2.5 g of cytomegalovirus vector expressing murine NF1-A, -B, -C, or -X protein (64); and 0.5 g of SV40-␤-galactosidase vector (Promega). Cells were collected 48 h after transfection, and luciferase activity was normalized to ␤-galactosidase activity. The data represent the means Ϯ S.D. of two independent experiments with triplicate samples and are expressed relative to the activity of the ANT2 promoter reporter gene in the absence of the NF1 expression vector.
(C box) that is juxtaposed to the transcription start site (26). Mutating this element increased promoter activity many-fold in several cell types. However, mutating the C box had no influence on growth arrest repression. Furthermore, a silencer element between the Go repressor elements and the Sp1 activation boxes (AB boxes) (41) and the Ϫ1196/Ϫ1184 region of the ANT2 promoter, which is reported to contain a glycolysisregulated box (GR Box) (42), could be deleted with no effect on growth arrest repression. Thus, the GoR region (and specifically, the Go-1 and Go-2 NF1-binding elements) appears to be responsible for growth arrest repression of the gene.
The NF1 isoform(s) that repress ANT2 remain to be established. The NF1 family is composed of four expressed genes (NF1-A, -B, -C, and -X) (see Ref. 25 for review) and a large number of splice variants (43)(44)(45) that form homodimers and heterodimers (46,47), thus creating an extensive network of possible functional dimers. The SDS-PAGE profiles of our DNA affinity-purified preparations are similar to those reported for rat and HeLa cells (48,49) and are consistent with the presence of multiple expressed isoforms. However, NF1 proteins isolated by DNA affinity chromatography most likely contain the highly conserved N-terminal DNA-binding/transactivation domain (45,47,50), the inclusion of which would contribute to MS peptide mass identification of different isoforms whether they are present or not. In attempts to obtain information on the repressor isoform(s), we investigated the effects of expressed NF1-A, -B, -C, and -X on the human ANT2 promoter transfected into NIH3T3 and HeLa cells. As reported elsewhere (51), the response is cell-type dependent. All four isoforms activated the promoter in HeLa cells, whereas in 3T3 cells, NF1-A, -C, and -X repressed expression of the reporter gene. The results are consistent with the idea that NF1 represses ANT2 in cells that are susceptible to growth arrest.
Regardless of the NF1 isoforms involved, our data show that repression of ANT2 requires occupation of the Go-1 and Go-2 elements by NF1 and that binding is lost in growth-activated diploid cells. Such reversible occupation implies that the NF1 protein is altered by different growth conditions or that availability of the Go elements is altered. Relatively small changes in nucleosome positioning on the glucocorticoid receptor promoter are sufficient to hide the NF1-binding site and to prevent NF1 binding (52), and direct interactions of NF1 with coactivator p300/CBP (53) and with histone-3 (54) have been observed, showing the importance of chromatin organization in NF1 function. That nucleosome reorganization might play a role in the growth-regulated expression of ANT2 is supported by experiments from this laboratory showing that the histone deacetylase inhibitor trichostatin A activates ANT2 expression in quiescent diploid fibroblasts in vivo, 4 as well as in ANT2 reporter gene constructs stably transfected into chromatin of NIH3T3 cells (55).
Growth arrest repression of ANT2 might also be explained by changes in NF1 binding affinities or by changes in NF1 nuclear concentration. NF1 has been reported to undergo phosphorylation (56 -59), but in those cases where tested, phosphorylation did not appear to alter NF1 binding properties (56 -59). Glycosylation of NF1 has also been reported (60), but no regulatory function has been attached to this modification. To our knowledge, no other post-translational modifications of NF1 are known. Early studies have reported changes in the levels or type of NF1 protein expressed in different growth states, including studies on quiescent and growth-activated NIH3T3 cells (61) and 3T3-L1 cells (56). In the latter studies, however, it is not clear if the changes observed reflect expression of different isoforms or different phosphorylation states. To date, we have detected no differences in the content of NF1 in quiescent and growth-activated diploid foreskin fibroblasts as judged by Western analysis (data not shown). This result does not, however, exclude the possibility that expression of low abundant isoforms or splice variants is altered.
To our knowledge, this is the first study reporting a direct growth arrest repressor function of NF1. NF1 can act as a repressor (Refs. 51, 62, 63, and 65; see Ref. 25 for review); and in several genes, repression is mediated by NF1 binding directly to a repressor or silencer element in the promoter (63, 66 -70), similar to what has been described here for the ANT2 promoter. However, none of these repressed genes is associated with growth arrest. Despite this, NF1 does appear to play a role in growth-related processes. For example, NF1 activates expression of p53 (71) and gadd153 (68), both of which have key roles in growth arrest of damaged or environmentally stressed cells. Paradoxically, NF1 has also been shown to prevent growth arrest. NF1-X1 is one of three genes (together with c-myc and MDM2) that prevent transforming growth factor-␤ induced growth arrest (37). Finally, studies with transfected 3T3-L1 cells show that reporter gene constructs driven by NF1-dependent promoters are more active in quiescent cells than in growing cells (56), in contrast to our findings with ANT2. However, the same promoters are strongly repressed in quiescent cells by overexpressing Myc. Since NF1 can act as an activator or a repressor of the same gene depending on the cell type or promoter context, the above study does not eliminate the possibility that Myc promotes a repressor function of NF1 in quiescent cells, as observed for ANT2 in this study. Thus, even though ANT2 is currently the only clear example of an NF1-dependent growth-repressed gene, we cannot exclude a broader role for NF1 in the growth arrest process.