Characterization of GTP cyclohydrolase I gene expression in the human neuroblastoma SKN-BE(2)M17: enhanced transcription in response to cAMP is conferred by the proximal promoter

Abstract

GTP cyclohydrolase I (GTPCH) gene expression was investigated in the human monoamine-containing neuroblastoma cell line SK-N-BE(2)M17. Northern blot analysis revealed a single GTPCH mRNA transcript that was confirmed by RNase protection assay to encode for Type 1 GTPCH; no alternatively spliced forms of GTPCH mRNA were detected with this assay. Incubation with 8Br-cAMP, but not nerve growth factor or leukemia inhibitory factor, produced a rapid increase in GTPCH mRNA and protein levels; protein levels remained elevated during the entire treatment period while mRNA content declined rapidly between 10 and 24 h. Treatment with 8Br-cAMP did not significantly modify the stability of GTPCH mRNA but did increase GTPCH transcription as determined by transient transfection assays of a luciferase reporter construct containing 1171 bp of human GTPCH 5′-flanking sequence. Cis-acting elements required for maximal basal and cAMP-dependent transcription were localized by deletion analysis to the 146 bp proximal promoter. DNase I footprint analysis of the proximal promoter using SK-N-BE(2)M17 nuclear extracts identified two protein binding domains: one an upstream Sp1-like site and the other a combined CRE-Sp1-CCAAT-box element. EMSA and supershift assays demonstrated that the combined CRE-Sp1-CCAAT-box element recruits ATF-2 and NF-Y but not Sp1–4 or Egr-1–3. NF-Y binding was confirmed using pure recombinant human NF-Y protein. Transcription of the human GTPCH gene in human SK-N-BE(2)M17 cells is thus enhanced by cAMP acting through regulatory elements located in the proximal promoter and may involve the transcription factors NF-Y and ATF-2.

Abbreviations used

ATF

activating transcription factor

BH4

5,6,7,8-tetrahydrobiopterin

bp

base pair(s)

CBP

CREB-binding protein

C/EBP

CCAAT enhancer-binding protein

CRE

cAMP response element

CREB

CRE binding protein

EMSA

electrophoretic mobility shift assay

GTPCH

GTP cyclohydrolase I

kb

kilobase(s)

LIF

leukemia inhibitory factor

MA

monoamine

NF-Y

nuclear factor Y

NGF

nerve growth factor

NO

nitric oxide.

The first and rate-limiting step in the biosynthesis of the pteridine co-factor 5,6,7,8-tetrahydrobiopterin (BH4) is catalyzed by the enzyme GTP cyclohydrolase I (GTPCH; EC 3.5.4.16) (Nichol et al. 1985). GTPCH expression within the human (Nagatsu et al. 1999), rat (Hirayama et al. 1993; Lentz and Kapatos 1996; Dassesse et al. 1997; Hirayama and Kapatos 1998) and mouse (Nagatsu et al. 1995; Shimoji et al. 1999) brain has been localized at the cellular level to monoamine (MA)-secreting neurons, where BH4 serves as the essential co-factor for the enzymes tyrosine and tryptophan hydroxylase (Kaufman 1974). The routine detection of GTPCH mRNA in extracts from regions of the rat brain that do not contain intrinsic MA neurons (Hirayama et al. 1993; Kapatos et al. 1999) strongly suggests that low levels of GTPCH are also associated with nitric oxide (NO)-secreting neurons and/or glial cells, where BH4 would serve as an essential co-factor for nitric oxide synthase activity (Kwon et al. 1989; Tayeh and Marletta 1989). Intracellular concentrations of BH4 generally appear to be subsaturating for BH4-dependent enzymes (Kettler et al. 1974; Miwa et al. 1985; Werner et al. 1990). Alterations in BH4 levels resulting from changes in GTPCH gene expression are thus predicted to be a common point of control for both MA and NO synthesis.

GTPCH gene expression within rat MA cell types is highly dynamic and regulated by numerous signal transduction pathways, including those that utilize the second messenger cAMP (Zhu et al. 1994; Anastasiadis et al. 1998). Cyclic-AMP-dependent enhancement of GTPCH transcription is observed in only a subset of rat cell types the normally express GTPCH (Kapatos et al. 1981). Located within the proximal promoter of the rat GTPCH gene is a non-canonical cAMP response element (CRE) and an adjacent perfect CCAAT-box that are both required for maximal and cell type-specific cAMP-dependent transcription (Kapatos et al. 2000). This combined CRE and CCAAT-box element acts as an enhancer and can confer sensitivity to cAMP on a heterologous minimal promoter. Moreover, in vitro assays have revealed that this CRE recruits the transcription factors CCAAT enhancer-binding protein β (C/EBPβ) and activating transcription factor 4 (ATF-4), whereas the CCAAT-box is bound by nuclear factor Y (NF-Y). The rat and human GTPCH proximal promoters exhibit almost 80% sequence homology (Kapatos et al. 2000) and the location and spacing of this CRE and CCAAT-box cassette are conserved in the human sequence. This suggests that in the appropriate cellular milieu the human GTPCH gene might also be enhanced by cAMP although, to our knowledge, this has never been demonstrated.

Despite their high level of homology, however, there are also differences between the rat and human promoter sequences that may impart variation in the control of this gene. These include a lack of homology in the 3′ end of the non-canonical CRE, the presence in the human promoter of a GC-rich putative Sp1-like site overlapping the 3′ end of the CRE and an E-box in the rat promoter that is not found in the human sequence (Kapatos et al. 2000). Like the rat and mouse genes (Shimoji et al. 1999; Kapatos et al. 2000) the 5′-flanking region of the human GTPCH gene has recently been shown to support transcription in transient transfection assays of gene reporter constructs (Witter et al. 1996). Unlike the rat gene, deletion analysis of the human promoter did not provide evidence for inhibitory elements located upstream of the proximal promoter. This same study also showed that transcription driven by the human GTPCH promoter was not enhanced in response to a stimulus that is known to induce expression of the endogenous GTPCH gene. There is thus reason to believe that what may appear to be minor variations between human and rat GTPCH promoter sequences may actually be responsible for significant differences in promoter function. Moreover, unlike in the rat, the product of the human GTPCH gene is alternatively spliced to generate three (Togari et al. 1992; Ichinose et al. 1995) or more (Golderer et al. 2001) transcripts with different coding sequences which introduces an additional level of complexity to the study of human GTPCH gene expression.

The major goals of the present study were therefore to use the human MA (Biedler et al. 1978) and BH4 (Albrecht et al. 1978) containing neuroblastoma cell line SK-N-BE(2)M17 to: (1) characterize human GTPCH mRNA expression within a neuronal cell type with regard to alternative splicing; (2) discover a stimulus capable of increasing human GTPCH gene expression at the transcriptional level; and (3) characterize the human GTPCH 5′ flanking sequence with respect to cis-acting elements and trans-acting factors involved in basal and enhanced transcription.

Materials and methods

Cell cultures and transfection

The human neuroblastoma cell line SK-N-BE(2)M17 (a gift from Dr John Haycock, Louisiana State University, LA, USA) was maintained in Dulbecco's modified Eagle's medium supplemented with 10% heat inactivated fetal calf serum, 100 µg/mL penicillin and 100 µg/mL streptomycin. Cells were cultured in a humidified atmosphere of 5% CO₂ at 37°C and prior to transient transfection were plated onto poly d-lysine-coated 24-well plates and grown to approximately 70% confluency. Cultures were transfected using lipofectamine and 0.4 µg of human GTPCHluc DNA construct along with 0.1 µg of pCMVβgal DNA to correct for transfection efficiency. Twenty-four hours after transfection, cells were treated for 8 h with either fresh media or media containing 2.5 mm 8Br-cAMP. All cultures were then harvested and assayed for luciferase (Promega, Madison, WI, USA) and β-galactosidase (β-gal; Clontech, Palo Alto, CA, USA) activities. Luciferase activity was divided by β-gal activity and is reported as relative luciferase activity.

Pteridine analysis

BH4 levels were determined by the method of Fukushima and Nixon (1980).

Western blot analysis

Human GTPCH Type 1 cDNA (a gift from Dr Hiroshi Ichinose, Fujita Health University, Aichi, Japan) was cloned in frame into pGEX4T at EcoR1 sites and GTPCH was isolated as a GST-fusion protein by chromatography on GSH-Sepharose beads. Polyclonal antisera to this fusion protein were raised in rabbits (HTI Bioproducts, Ramona, CA, USA). The specificity of the antiserum was demonstrated by comparing control HeLa cell extracts, which do not contain GTPCH, with extracts from HeLa cells transfected to express human GTPCH protein in frame with an N-terminal X-press epitope. Sodium dodecyl sulftate polyacrylamide gel electrophoresis (SDS–PAGE) and sequential western blot analysis using the GTPCH antiserum, and the X-press antibody showed that only extracts from transfected cells contained the GTPCH/X-press fusion protein. In subsequent analyses, 30 µg of protein from each sample were analyzed using 12% SDS–PAGE. After transfer to nitrocellulose, the membrane was incubated with rabbit anti-human GTPCH antibody (1 : 12 000 dilution) for 2 h at room temperature and then with peroxidase-conjugated goat anti-rabbit IgG (1 : 70 000 dilution) for 1 h. A chemiluminescence detection reaction was performed and detected by exposing the blot to X-ray film. Care was taken to insure that chemiluminescence signals were within the linear response range of the film. Films were scanned and the blot was stripped and probed again using rabbit antihuman β-actin as the primary antibody. The GTPCH protein content of each sample is expressed relative to that of β-actin.

Northern and nuclease protection analyses

Northern blot analysis was performed as described using poly(A)⁺-selected SKNBE(2)M17 cell RNA (Hirayama et al. 1993). A ribonuclease protection assay to simultaneously determine human GTPCH and β-actin mRNA levels in samples of total RNA was also performed as described (Stegenga et al. 1996). The [³²P]-labeled GTPCH antisense RNA used for both the northern blot and nuclease protection analyses was transcribed from the plasmid hGTPCH/T3T7, which contains human GTPCH Type 1 cDNA ligated into the T3T7BM vector at SacI and BamHI sites. The [³²P]-labeled β-actin antisense RNA probe was transcribed from a BlueScript plasmid containing full-length human β-actin. The GTPCH mRNA content of each sample is expressed relative to that of β-actin.

Isolation of human GTPCH genomic clone

A human genomic library (Clontech) was screened by PCR using primers anchored in GTPCH exon 1 and 5′-flanking DNA (Ichinose et al. 1995; Witter et al. 1996). A PCR amplicon of predicted size (1620 bp) was obtained, digested at MluI and NcoI sites and the 1171 bp product ligated into the pGL3 luciferase reporter vector to produce 1171GTPCHluc. As confirmed by sequence analysis, 1171GTPCHluc contains 167 bp of the human GTPCH 5′ UTR upstream from the translation start site and 1004 bp upstream from the reported transcription start site (− 1004 to + 167). The plasmid 1171GTPCHLuc was digested with NcoI and either HindIII or AccIII and re-ligated into pGL3 basic to generate the deletion constructs 613GTPCHluc (− 446 to +167) or 313GTPCHluc (−146 to +167), respectively (Fig. 5a).

Figure 5. Human GTPCH promoter activity is stimulated by 8Br-cAMP treatment. (a) GTPCH 5′-flanking sequence showing restriction sites used for generating the luciferase reporter gene deletion constructs 1171GTPCHluc (DraI), 613GTPCHluc (HindIII), and 313GTPCHluc (AccIII). Also shown are putative Sp1, CRE, CCAAT-box and TATA-box like sequences located within the 313GTPCHluc construct. (b) SK-N-BE(2)M17 cells were transfected with GTPCHluc constructs and pCMV-β-gal DNA. Then, 5 mm 8Br-cAMP were added 18 h later and the cultures continued for another 8 h, when cells were lysed and assayed for luciferase and β-galactosidase activities. Luciferase activity was divided by β-galactosidase activity to correct for transfection efficiency and was expressed as relative luciferase activity. Data are the mean ± SE of three independent experiments each determined in triplicate.

Nuclear extracts and DNase I footprint analysis

Nuclear proteins were prepared from SK-N-BE(2)M17 cells (Dignam et al. 1983) except that in addition to phenylmethylsulfonyl fluoride (PMSF; 1 mm) the protease inhibitors leupeptin (1 µg/mL) and pepstatin (1 µg/mL) were added to all buffers. Protein content was determined by the method of Bradford (1976). Single-stranded probes for DNase I footprinting of the human GTPCH promoter were produced as follows. 1171GTPCHluc DNA digested with AccIII or Bsu36I was labeled using Klenow and [α-³²P]dCTP (3000 Ci/mmol). [³²P]-labeled AccIII- and Bsu36I-DNA were then digested with either NcoI or HindIII, respectively, and gel-purified to generate probes labeled on the coding and non-coding strand, respectively. DNase I footprint analysis was performed as previously described using 45 µg of nuclear extract protein (Kapatos et al. 2000). The exact location of each footprint on each strand was ascertained by establishing the position of G + A or C + T nucleotides using chemical cleavage sequencing reactions (Maxam and Gilbert 1977).

Electrophoretic mobility shift assay

EMSA was performed as previously described (Kapatos et al. 2000). Single-stranded complementary oligonucleotides corresponding to positions − 100/−42 in the human GTPCH promoter were annealed and end-labeled with polynucleotide kinase and [α-³²P]dGTP (3000 Ci/mmol; PerkinElmer Life Sciences, Boston, MA, USA). Assays typically contained 1 µg of nuclear extract protein and in some cases up to 200 ng of recombinant human NF-Y. For supershift or competition assays, antibodies (4 µg: Santa Cruz Biotechnology, Santa Cruz, CA, USA) or competing double-stranded oligonucleotides were added to the reaction buffer at 4°C for 2 or 1 h, respectively, before addition of the probe. Protein-DNA complexes were resolved from free DNA on 6% polyacrylamide gels in TBE (100 mm Tris pH 8.4, 90 mm boric acid and 1 mm EDTA) with 1% glycerol.

Production of recombinant human NF-Y

Human NF-YA and NF-YB both in pET3b and His-tagged thioredoxin-NF-YC in pET32b were obtained from Dr R. Mantovani (Universita di Modena e Reggio, Modena, Italy). Recombinant NF-Y subunits were produced in bacterial strain BL21(DE3)pLysS following induction by isopropyl-1-thio-β-D-galatoside (IPTG). Frozen bacterial pellets were resuspended in 20 mm Tris–HCl (pH 7.5), 10 mm EDTA, 1.0% Triton X-100 containing 0.2 mm PMSF and 10 µg/mL of aprotonin, leupeptin and pepstatin A and were disrupted by sonication on ice. Homogenates were centrifuged and inclusion bodies were washed twice in the disruption buffer. Inclusion bodies were solubilized in 20 mm Tris–HCl (pH 7.8), 6 m guanidine HCl (GnHCl), 0.1 m KCl, centrifuged and the protein content of the supernatant was determined. Subunits were mixed and dialyzed at 4°C against decreasing concentrations of GnHCl in 20 mm Tris–HCl (pH 7.8), 0.1 m KCl, 10% glycerol and 5 mmβ-mercaptoethanol (β-Me). The final dialyzate in buffer only was cleared by centrifugation and the supernatant passed over a Ni⁺-agarose column. The column was washed with 10 bed volumes of 20 mm Tris–HCl (pH 7.8), 0.1 m KCl, 10% glycerol, 5 mmβ-Me and the assembled NF-Y heterotrimer was eluted with the same buffer containing 300 mm imidazole. Fractions were pooled, dialyzed and analyzed for purity by SDS–PAGE.

Results

Northern blot and nuclease protection assays detect only Type 1 GTPCH mRNA in SK-N-BE(2)M17 cells

The coding regions of human Type 1, 2, and 3 GTPCH mRNA differ only at their 3′ ends (Fig. 1a). We designed an in vitro-transcribed 503 nt single-stranded antisense RNA probe containing GTPCH coding sequence to hybridize to all three of these forms on northern blots yet to distinguish between Type 1 and Types 2 and 3 in nuclease protection assays. When a northern blot of poly(A)⁺-selected SK-N-BE(2)M17 mRNA was hybridized with this probe, only a single transcript of approximately 3.6 kb was detected (Fig. 1b), which is within the range of the reported size of Type 1 GTPCH mRNA (Gutlich et al. 1992; Nomura et al. 1993). No smaller bands corresponding to the sizes predicted for Type 2 or 3 mRNA (Ichinose et al. 1995) were observed, even upon over-exposure of the autoradiogram. Nuclease protection analysis of total RNA extracted from SK-N-BE(2)M17 cells revealed only a single band at 470 nt which corresponds to Type 1 GTPCH mRNA (Fig. 1c). A protected fragment of 318 nt indicating Type 2 or 3 mRNA was never present, even upon over-exposure of the autoradiogram or when Type 1 mRNA levels were increased more than four-fold following treatment with 8Br-cAMP (Fig. 1c, lane 2 and see below). These data unequivocally identify the major transcript expressed in SK-N-BE(2)M17 cells as Type 1 GTPCH mRNA.

Figure 1. Northern blot and nuclease protection assays detect only Type 1 GTPCH mRNA in SK-N-BE(2)M17 cells. (a) Schematic representation of the alternatively spliced forms of human GTPCH mRNA known as Types 1, 2 and 3 and the in vitro transcribed 503 nt GTPCH probe located between SacI and BamHI sites of Type 1 cDNA that was used to detect them. This probe hybridizes with all three forms of GTPCH mRNA by northern blot but is able to distinguish between Type 1 (468 nt) and 2 or 3 (318 nt) GTPCH mRNA by nuclease protection analysis. (b) Northern blot analysis of 2.9 µg poly(A⁺) selected RNA from SK-N-BE(2)M17 cells shows a single band of 3.6 kb GTPCH mRNA. (c) Nuclease protection analysis of total RNA from SK-N-BE(2)M17 showing a single band of 468 nt corresponding to Type 1 GTPCH mRNA. No Type 2 or 3 GTPCH mRNA was detected, even after GTPCH induction by treatment with 8Br-cAMP. Lane 1: 40 µg of total RNA from control cells. Lane 2: 40 µg of total RNA from cells treated with 5 mm 8Br-cAMP for 8 h. Lane 3: RNA size marker. In this experiment, 8Br-cAMP induced GTPCH Type 1 mRNA approximately five-fold.

BH4, GTPCH protein and GTPCH mRNA levels are increased by treatment with 8Br-cAMP

To discover signaling pathways that are capable of modifying GTPCH gene expression in SK-N-BE(2)M17 cells, we next performed a screening assay using BH4 levels as a response indicator. Cultures were treated for 24 h with agents that are known to modify GTPCH expression in rat neurons; 400 ng/mL of nerve growth factor or 5 mm 8Br-cAMP to increase BH4 levels (Zhu et al. 1994; Hirayama and Kapatos 1995) or 2 ng/mL of leukemia inhibitory factor to decrease BH4 levels (Stegenga et al. 1996). The results of this screen (data not shown) revealed that only 8Br-cAMP was able modify BH4 content and it did so by increasing levels more than two-fold.

In order to determine whether this cAMP-dependent increase in BH4 results from an increase in GTPCH protein and to establish the time course of this response, western blot analysis was performed on protein extracts from SK-N-BE(2)M17 cells treated for 0–48 h with 5 mm 8Br-cAMP. As shown in the top of Fig. 2, basal levels of GTPCH protein are quite low in these cells but were increased throughout the entire treatment period. In contrast, β-actin protein content was unaltered by 8Br-cAMP. When expressed relative to the amount of β-actin protein present, GTPCH protein levels increased linearly for the first 8 h of treatment, with a 2.5-fold increase observed by 4 h and a five-fold increase after 8 h that remained stable for at least 48 h (Fig. 2, bottom).

Figure 2. A time course of the induction of GTPCH protein following stimulation with 8Br-cAMP. Top: GTPCH and β-actin protein signals detected by sequential western blots of total cellular extracts from control and 8Br-cAMP-treated SK-N-BE(2)M17 cells. Bottom: Data from a single experiment plotted as percent control of GTPCH protein normalized to β-actin protein content. Essentially identical data were obtained from two separate experiments.

Whether 8Br-cAMP alters GTPCH mRNA content and/or the pattern of gene splicing was determined next using a nuclease protection assay that simultaneously measures human GTPCH and β-actin mRNA (Fig. 3a). Following normalization to β-actin mRNA content, which was not altered by 8Br-cAMP, this analysis revealed that the relative abundance of Type 1 GTPCH mRNA more than doubled within 1 h and continued to increase until reaching a maximum five-fold stimulation by 10 h (Fig. 3a). Fourteen hours later and 24 h after treatment began, GTPCH mRNA content had declined to almost basal levels. This rapid decline is in stark contrast to the stable elevated levels of GTPCH protein that remain 48 h after 8Br-cAMP treatment. At no time point in any of these experiments was Type 2 or 3 GTPCH mRNA observed, suggesting that treatment with 8Br-cAMP does not make the alternatively spliced forms of GTPCH detectable by nuclease protection assay (see Fig. 1c).

Figure 3. GTPCH mRNA levels are increased by treatment with 8Br-cAMP without a change in GTPCH mRNA stability. (a) Time course of the induction of GTPCH mRNA following stimulation with 8Br-cAMP. SK-N-BE(2)M17 cells were treated with or without 2.5 mm 8Br-cAMP and were harvested 1, 3, 5, 10 and 24 h later. GTPCH and β-actin mRNA was measured by simultaneous nuclease protection analysis. Data from three experiments were plotted as a percent of control GTPCH mRNA normalized to β-actin mRNA content. Insert shows a dose–response curve for cultures treated for 5 h with varying concentrations of 8Br-cAMP. The maximum response to 8Br-cAMP was observed at 2.5 mm. (b) Estimation of the half-life of GTPCH mRNA following induction by 8Br-cAMP and inhibition of transcription with actinomycin D. SK-N-BE(2)M17 cells were treated with or without 2.5 mm 8Br-cAMP and after 3 h received 10 µg/mL of actinomycin D. Cultures were harvested 1, 2, 4, 7 and 10 h later and GTPCH and β-actin mRNA were measured by simultaneous nuclease protection analysis. The half-life estimated of GTPCH mRNA in control cells was 10.1 h and in 8Br-cAMP-treated cells 8.8 h.

8Br-cAMP does not modify GTPCH mRNA turnover

The half-life of GTPCH mRNA in rat renal mesangial cells is increased two-fold by cAMP (Pluss et al. 1996). The rapid increase in GTPCH mRNA levels following 8Br-cAMP treatment could thereby result from the stabilization of existing molecules of GTPCH mRNA. We investigated the effect of 8Br-cAMP on GTPCH mRNA turnover by treating cells continuously with 8Br-cAMP and, after 3 h of this treatment, inhibiting gene transcription with actinomycin D (10 µg/mL). Cultures were harvested from 1 to 10 h after addition of actinomycin D and GTPCH and β-actin mRNA analyzed by simultaneous nuclease protection assay. This 13-h treatment paradigm was chosen based upon preliminary experiments to maximize the decline in GTPCH mRNA produced by actinomycin D and to minimize the contribution to this decline made by the decrease in GTPCH mRNA observed following continuous treatment with 8Br-cAMP. These same experiments showed that inhibition of transcription over this 10-h period did not result in any measurable decline in β-actin mRNA abundance, which permitted us to continue to normalize GTPCH mRNA to β-actin mRNA content. As can be seen in Fig. 3(b), inhibition of transcription resulted in a first-order decline in GTPCH mRNA with an estimated half-life of 10.1 h. This rate of decline was marginally increased to a half-life of 8.8 h by continuous treatment with 8Br-cAMP. These data demonstrate that the increase in SK-N-BE(2)M17 Type 1 GTPCH mRNA in response to 8Br-cAMP does not result from an increase in mRNA stability.

Human GTPCH promoter activity is stimulated by 8Br-cAMP treatment

In order to study the effect of 8Br-cAMP on human GTPCH gene transcription, 1171 bp of the human GTPCH 5′ flanking region including 167 bp of GTPCH mRNA 5′ UTR were isolated, sequenced and cloned upstream of a luciferase reporter gene to produce 1171GTPCHluc (Fig. 4a). Sequence analysis identified the amplicon as being human GTPCH flanking sequence. Transient transfection of SK-N-BE(2)M17 cells with this construct revealed that transcription was up to 10-fold greater than from the parental vector (Fig. 4b). Progressive deletion analysis showed that the minimal promoter sequence necessary to initiate and sustain transcription is located within the 313GTPCHluc construct, which contains 146 bp upstream from the reported transcription start site (Witter et al. 1996) (Fig. 5). Deletion analysis also showed a step-wise increase in transcription as 5′ flanking sequence was eliminated, which suggests that repressor elements may be located immediately upstream from the proximal promoter.

Figure 4. Nucleotide sequence and organization of the proximal 5′-flanking region of the human and rat GTPCH genes. This sequence corresponds to the 146 bp at the 5′ end of the 313GTPCHluc construct. Circled letters indicate transcription start sites. Numbering is based upon the distance from the human transcription start site at A (+1). Putative protein binding sites determined by computer analysis are identified by boxes and named below or on top of the sequence. Actual protein binding sites determined by DNase I footprint and EMSA analysis are labeled Domains 1 and 2 and are drawn as lines above the sequence. The oligonucleotide used as the probe for EMSA spanned from −100 to −42. Note that the location and spacing between the CRE and CCAAT-box and the purine-rich ATAAAAA sequence are conserved but that the Sp1-like element located between the CRE and CCAAT-box is found only in the human gene.

SK-N-BE(2)M17 cells transfected with 1171GTPCHluc responded to an 8-h incubation with 8Br-cAMP with a three- to five-fold increase in relative luciferase activity, supporting the premise the cAMP enhances human GTPCH transcription (Fig. 4b). Deletion analysis showed that elimination of promoter sequence up to the AccIII site did not modify the response to 8Br-cAMP. The cis-acting elements required for cAMP-dependent enhancement of human GTPCH gene transcription are thus localized within the 146 bp proximal promoter found within the 313GTPCHluc construct. A computerized search (TFMATRIX) (Heinemeyer et al. 1998) for potential regulatory elements in this region yielded (5′-3′) two GC-rich Sp1-like binding sites, a non-canonical CRE (TGACGCGA) and 3′ overlapping GC-rich Sp1-like site (GAGGCGGGGC), a perfect CCAAT-box and a purine-rich sequence (ATAAAAA) that may serve as a TATA-box (Fig. 5).

DNase I footprint analysis of the human GTPCH core promoter reveals two protein binding domains

[³²P]-end labeled probes AccIII–NcoI (−146/+167 coding) and Bsu36I–HindIII (+ 54/−446 non-coding) were incubated with nuclear extract prepared from SK-N-BE(2)M17 cells and the non-specific polynucleotide competitor Poly (dI/dC), and then digested with DNase I. Analysis of regions within the proximal promoter that are protected from nuclease digestion revealed a protein binding domain located on both strands which spans from −97 to −48 bp and covers the entire combined CRE-Sp1-CCAAT-box element (Fig. 6). In addition, a second domain spanning from −122 to −113 bp and encompassing the most 5′ of the two predicted upstream Sp1-like binding sites was observed. No footprint was associated with the Sp1-like site located between −107 and −97. At least three hypersensitive bases were observed within the combined CRE-Sp1-CCAAT-box domain; on both strands a G and a C at position − 50 that lie 16 bp 3′ to the CCAAT-box and on the coding strand a C at position −79 that is located in the center of the Sp1-like site. These two protein-binding domains are shown mapped onto the sequence of the proximal promoter in Fig. 5.

Figure 6. DNase I footprint analysis of the human GTPCH proximal promoter reveals two protein binding domains. [³²P]-end-labeled probe AccIII–NcoI (− 156/+167 coding) and Bsu36I–HindIII (+ 58/−446 non-coding) fragments were incubated with nuclear extract prepared from SK-N-BE(2)M17 cells and the non-specific polynucleotide competitor Poly (dI/dC) and then digested with DNase I. Maxam–Gilbert sequencing reactions were performed in parallel. Proximal promoter sequence and cis-elements are presented for each strand. Two protein binding domains were detected: domain 1 spanning from −97 to −48 and associated with the entire CRE-Sp1-CCAAT-box element and Domain 2 spanning from −125 to −111 and associated with the most upstream Sp1-like element. Asterisks represent the positions of hypersensitive bases.

Electrophoretic mobility shift analysis reveals protein binding to the combined CRE-SP1-CCAAT-box element

Mutagenesis of both the CRE and CCAAT-box in the rat proximal promoter decreases basal transcription by more than 70% and completely eliminates the promoter response to cAMP (Kapatos et al. 2000). We therefore focused our attention on the human CRE-Sp1-CCAAT-box element in an attempt to discover the protein factors that bind to it and presumably mediate the response of the human promoter to cAMP. EMSA using nuclear protein extracts from SK-N-BE(2)M17 cells and a [³²P]-labeled double-stranded oligonucleotide spanning from −100 to − 42 (see Fig. 5), and centered upon the CRE-Sp1-CCAAT-box element, revealed one major and occasionally a larger but very minor band which were observed only in the presence of nuclear protein and increased in intensity as the amount of nuclear protein in the assay was increased (Fig. 7a). Formation of the single major complex was competed away in a concentration-dependent manner by unlabeled −100/−42 oligonucleotide (Fig. 7a). Supershift analysis using antibodies directed against numerous transcription factors that can bind to CRE-like sequences indicated that ATF-2 is recruited by this oligonucleotide but not the ATF-2 binding partners ATF-3 and c-Jun or the related factors ATF-1, ATF-4, Jun B, Jun D, c-fos, Fos B, Fra-1, or Fra-2 (Figs 7b and c). A small signal associated with CREB binding could also be detected following overexposure of the autoradiogram (see insert, Fig. 7b). The co-activators CBP and P300 were not part of this binding complex (Fig. 7c). Supershift assays were used next to determine whether proteins occupy the Sp1-like site located within the CRE-Sp1-CCAAT-box element. Incubation with antibodies directed against Sp1, Sp2, Sp3, or Sp4 did not modify the formation or size of the binding complex (Fig. 7f). Similarly, antibodies directed against members of the early growth response family (Egr a.k.a. zif268), which also recognize GC-rich elements and include Egr-1, Egr-2 and Egr-3, did not interact with the binding complex (Fig. 7g). The CCAAT-box binding factors NF-1, C/EBPα and C/EBPβ were also not recruited by this oligonucleotide (Figs 7b,d). In contrast, antibodies directed against the C terminus of the A subunit of NF-Y shifted the entire binding complex without altering complex formation while antibodies directed against the C termini of the NF-YB or NF-YC subunits were able to both shift the complex and inhibit complex formation (Fig. 7d). These results strongly suggest that NF-Y is the cognate binding protein of the human GTPCH CCAAT-box and is the major binding activity found in this complex. In order to corroborate this finding EMSA was performed using 0.2–200 ng of pure recombinant human NF-Y protein. As predicted, the GTPCH CCAAT-box recruited human NF-Y in a protein concentration-dependent manner (Fig. 7e). SK-N-BE(2)M17 nuclear proteins recruited by the combined CRE-Sp1-CCAAT-box element thus include the transcription factors NF-Y and ATF-2.

Figure 7. Electrophoretic mobility shift analysis identifies proteins binding to the combined CRE-SP1-CCAAT-box element. The double-stranded oligonucleotide probe used in these assays corresponds to positions − 100 to − 42 in Fig. 5. Arrows represent either specific binding or supershifted bands. In (e) the concentration of recombinant human NF-Y protein was varied from 0.2 to 200 ng in 10-fold increasing steps.

Discussion

Five main conclusions can be reached from the present work. First, only Type 1 GTPCH mRNA can be detected by nuclease protection assay in the human neuroblastoma cell line SK-N-BE(2)M17. Second, the second messenger cAMP increases levels of BH4 in these cells by rapidly stimulating GTPCH gene transcription and increasing GTPCH mRNA and protein content. Third, cis-acting elements required for maximal basal and cAMP-dependent transcription are found within the 146 bp proximal promoter. Fourth, two protein binding domains are located within the proximal promoter, one associated with an upstream Sp1-like site and the other encompassing the entire combined CRE-Sp1-CCAAT-box element. Fifth, the CRE-Sp1-CCAAT-box binding domain recruits the transcription factors NF-Y and ATF-2 in vitro.

The major form of GTPCH mRNA found in human tissues and cell lines (Gutlich et al. 1992; Nomura et al. 1993) is reported to be approximately 2.9–3.6 kb long and encodes for the Type 1 transcript. Only Type 1 GTPCH mRNA was detectable by nuclease protection assay in SK-N-BE(2)M17 cells under conditions of basal and cAMP-enhanced transcription and this transcript was approximately 3.6 kb in size. This is a significant observation because unlike the rat GTPCH gene, which generates two forms of GTPCH mRNA (Gutlich et al. 1992; Hirayama et al. 1993) with identical coding regions (S. I. Lentz and G. Kapatos, unpublished data), the human GTPCH gene is reported to encode for three forms of mRNA (Togari et al. 1992) that differ in their coding regions and are produced by alternative splicing of exon 6 (Ichinose et al. 1995). Moreover, Golderer et al. (2001) have recently reported additional alternatively spliced forms of human GTPCH. Neither Type 2 nor 3 transcripts have ever been observed by northern blot but are predicted to contain at least 1000 fewer bases than Type 1. While we could not detect Type 2 and 3 mRNA by nuclease protection assay, these transcripts can be amplified by RT-PCR using RNA isolated from human liver RNA (Togari et al. 1992) and the human neuroblastoma cell line SK-N-SH (Golderer et al. 2001) and are reported to make up as much as 20% of the GTPCH mRNA found in some human tissues (Hirano et al. 1997). Although Type 2 and 3 and less abundant alternatively spliced forms of GTPCH mRNA may be present in some human cells it should be kept in mind that the proteins translated from these transcripts have never been observed in vivo. Indeed, the physiological significance of Type 2 and 3 GTPCH remains enigmatic because, when expressed in recombinant form in Escherichia coli, neither protein is reported to have enzyme activity (Gutlich et al. 1992).

Incubation of SK-N-BE(2)M17 cells with 8Br-cAMP resulted in a rapid and coordinated increase in GTPCH mRNA and protein levels that was detectable as early as 1 h after treatment. The rapidity of these responses indicates that new protein synthesis is not required for the induction of GTPCH transcription by cAMP. Following 10 h of 8Br-cAMP treatment, however, GTPCH mRNA levels declined while protein levels remained stable. A similar rapid increase and equally rapid decline in GTPCH mRNA in response to continuous stimulation by cAMP has previously been shown to occur within cultured mesencephalic and diencephalic dopamine neurons (Zhu et al. 1994). These data suggest a cellular process that functions to limit the magnitude of the increase in GTPCH protein and thus BH4 synthesis after induction of GTPCH gene transcription by cAMP. At the nucleic acid level this would presumably involve either inhibition of GTPCH transcription or a destabilization of existing GTPCH mRNA molecules. A cAMP-dependent increase of this magnitude in GTPCH mRNA turnover is not supported by the data presented here or by studies showing that cAMP can actually prolong the half-life of GTPCH mRNA (Pluss et al. 1996). On the other hand, the decline in GTPCH mRNA that occurs between 10 and 24 h of continuous 8Br-cAMP treatment is similar to that observed here following inhibition of transcription with actinomycin D. This implies that at some time beyond 10 h of treatment GTPCH transcription is no longer stimulated and/or may actually be inhibited by cAMP. The loss of a transcriptional response to continuous cAMP stimulation is referred to as attenuation and most often results from a cAMP-dependent activation of protein phosphatase 1 and the dephosphorylation of CREB and/or some other phosphorylated transcription factor (Montminy 1997). Alternatively, a switch from an increase to a decrease in transcription in response to the same stimulus has been shown to result from a change in the mix of available trans-acting factors. For example, the Gadd153 (a.k.a. Chop) gene expressed in PC12 cells exhibits a biphasic response to continued presentation of the inducing stimulus that involves the displacement on the Gadd153 promoter of the activator ATF-4 by the newly synthesized repressor protein ATF-3 (Fawcett et al. 1999). With respect to GTPCH gene transcription and cAMP either mechanism may be operating in SK-N-BE(2)M17 cells.

This is not to say the post-transcriptional regulation of GTPCH mRNA stability does not take place. Indeed, the 3′ UTR of human (Nomura et al. 1993) and rat (Hatakeyama et al. 1992) GTPCH mRNA both contain cis-acting elements that are often involved in determining mRNA stability. Furthermore, while cAMP is reported to increase GTPCH mRNA stability (Pluss et al. 1996) glucocorticoids appear to destabilize this transcript (Simmons et al. 1996). Even under basal conditions the stability of GTPCH mRNA differs across cell types, with the estimated half-life reported here for GTPCH mRNA in SK-N-BE(2)M17 cells being two- to seven-fold greater than that observed previously for either rat cardiac endothelial cells (4.8 h) (Simmons et al. 1996) or rat renal mesangial cells (1.5 h) (Pluss et al. 1996). Post-transcriptional control of GTPCH mRNA stability may thus be an important process regulated in a cell type-specific manner.

The results of the present study show that in transient transfection assays the human GTPCH promoter performs very much like that of the rat. This analysis differs from that reported previously by Witter et al. (1996), possibly because in that study a non-neuronal cell type that does not constitutively express GTPCH was used to examine promoter function. While the human and rat proximal promoters demonstrate similarity of form and function there are still a number of interesting differences that should be noted here.

First, in vitro footprinting of the human proximal promoter identified an upstream Sp1-like site and the entire combined CRE-Sp1-CCAAT-box element as the only two protein-binding domains. This is in contrast to the rat promoter, where an identical analysis using PC12 nuclear extracts identified three protein-binding domains including an upstream Sp1-like site, the CCAAT-box and an E-box but not the CRE (Kapatos et al. 2000). In the human sequence the rat E-box is disrupted by a single base change (CAGCTT for human versus CAGCTG for the rat) that presumably eliminates binding of what we believe to be a ubiquitous factor. Located 17 bp upstream from the rat cap site (Kapatos et al. 2000) and imbedded within the E-box is the transcription start site for the human gene (Witter et al. 1996). As protein binding to the E-box and binding of TFIID to the transcription start site are presumably mutually exclusive this evolutionary change in the E-box sequence may have been responsible for the upstream shift in the human start site. The ability of nuclear extracts to footprint the human but not the rat CRE may also be related to a single base change (TGACGCGA for the human vs. TGACGCAA for the rat) and thus the proteins recruited by this DNA element. Alternatively, the proteins available for binding to this site may differ in nuclear extracts prepared from PC12 and SKN-BE(2)M17 cells.

Second, the footprint of the human upstream Sp1-like element was found to span 10 bp and corresponds to the most 5′ Sp1-like site located 20 bp upstream from the CRE. In contrast, the footprint of the Sp1-like element in the rat promoter spans over 20 bp and corresponds to the more 3′ Sp1-like site located within 5 bp of the CRE. The upstream Sp1-like element in the rat GTPCH proximal promoter appears to inhibit transcription (Kapatos et al. 2000).

Third, the human CRE-Sp1-CCAAT-box domain spans almost 50 bp and located within it are three DNase I hypersensitive sites. This differs from the rat promoter, where the CCAAT-box footprint covers only 20 bp and no evidence for protein-dependent DNA deformation was apparent. The CRE-Sp1-CCAAT-box element also binds the transcription factors NF-Y and ATF-2 while the rat CRE-CCAAT-box cassette binds NF-Y, ATF-4 and C/EBPβ. Although these differences could simply be due to the mix of proteins available for binding there is also no doubt that the human CRE-Sp1-CCAAT-box element attracts a protein complex that is larger and able to distort DNA in a way that proteins recruited to the rat CRE-CCAAT-box cassette cannot.

Finally, while the spacing between the CRE and CCAAT-box is identical in the human and rat promoters, the intervening sequence in the human element is GC-rich (GAGGCGGGGC for human versus AAGAGGCTCG for rat). While we predicted that this overlapping Sp1-like site might have a role in human transcription, no members of the Sp1 or Egr-1 families of transcription factors known to bind GC-rich sequences were found associated with it. These negative results do not necessarily mean that the putative element is unoccupied and the hypersensitive C located in the center of this site may indicate otherwise.

The studies reported here identify NF-Y and ATF-2 as proteins recruited to the CRE-Sp1-CCAAT-box cassette. Based upon our previous study of the rat GTPCH promoter it is highly likely that this combined element is involved in the response of the human GTPCH promoter to cAMP and that ATF-2 is recruited to the CRE, while NF-Y is bound by the CCAAT-box. ATF-2 is an activator protein, a member of the family of bZip transcription factors and is able to heterodimerize with c-Jun and thus to bind to non-canonical CRE response elements such as that found in the GTPCH promoter (Hai and Curran 1991). However, c-Jun was not a part of the ATF-2 binding complex reported here. ATF-2 is phosphorylated by MAPK, p38 and JNK on two threonine residues located within the transactivation domain but does not appear to be a substrate for PKA (Kageyama et al. 1991; Gupta et al. 1995). NF-Y is a ubiquitous heterotrimeric transcription factor composed of A, B and C subunits all of which must be present before binding DNA (Chodosh et al. 1988; Li et al. 1992; Sinha et al. 1995). NF-Y is not reported to be a substrate for any protein kinase. While it is likely that ATF-2 is recruited to the human GTPCH CRE, a recent report has shown that NF-Y and ATF-2 can form a stable complex in the absence of a CRE and thus DNA binding by ATF-2 (Fritz and Kaina 2001). Perhaps this direct interaction between ATF-2 and NF-Y also forms the basis for the ability of ATF-2 bound to a CRE to facilitate the binding of NF-Y to an adjacent CCAAT-box (Srebrow et al. 1993). ATF-2 has recently been shown to have intrinsic histone acetyltransferase activity that is controlled by phosphorylation (Kawasaki et al. 2000). Similarly, NF-Y is able to mobilize histone acetyltransferase activity by interacting directly with CBP and P300/CBP-associated factor (Bevilacqua et al. 1997; Currie 1997) as well as now with ATF-2 (Fritz and Kaina 2001). These observations suggest that the combination of NF-Y and ATF-2 binding may activate transcription of the endogenous human GTPCH promoter by direct effects on chromatin structure.

The human and rat GTPCH promoters have much in common with the promoters of the fibronectin (Muro et al. 1992), hexokinase II (Osawa et al. 1996) and arylalkylamine N-acetyltransferase (Baler et al. 1997) genes in that each responds to cAMP with an increase in transcription that is dependent upon a CCAAT-box and a CRE separated by approximately one or two turns of the DNA helix and that bind NF-Y along with the bZip family members CREB, ATF-1, ATF-2, ATF-4 or C/EBPβ. The combination of a CRE and an adjacent CCAAT-box is therefore a highly conserved and evolutionarily successful cis-acting element for the cAMP-dependent enhancement of gene transcription and presumably is capable of stimulating transcription in a manner that a CRE or a CCAAT-box acting alone cannot.

Acknowledgements

We would like to thank Drs Hiroshi Ichinose and Roberto Mantovani for cDNAs and Dr John Haycock for the SK-N-BE(2)M17 cell line. This work was supported by a grant from the NINDS (NS26081).