ABSTRACT
Glutamine:fructose-6-phosphate amidotransferase (GFAT) is the enzyme that is rate
limiting in the synthesis of glucosamine and hexosamines. Glucosamine has been
proposed to contribute to the glucotoxicity of diabetes. Evidence that the gene
encoding GFAT is transcriptionally regulated prompted us to clone and
characterize its promoter. The position of the mouse GFAT promoter relative to
the translational start site was located by primer extension and found to be
149 bp upstream of the translational start site. A 1.9 kb
Sac
I fragment of the GFAT gene was found to contain the promoter and 88 bp of
sequence downstream of the transcriptional start site. This promoter segment
could drive expression of a luciferase reporter gene, could confer correct
transcriptional initiation to the reporter and could confer the EGF-responsiveness previously observed in the native gene. The mouse GFAT
promoter lacks a canonical TATA box and has several GC boxes within a highly GC-rich region. Deletional analysis of the promoter indicated that a proximal
element extending to -120 relative to the transcriptional start site could confer reporter
expression at a level of 57% of the 1.9 kb construct. Detailed analysis of this
proximal region by DNase I footprinting, electrophoretic mobility shift assays
and site-directed mutagenesis indicated that Sp1 binds to three elements in this
proximal promoter segment and plays a vital role in regulation of transcription
from this gene.
Glutamine:fructose-6-phosphate amidotransferase (GFAT) is the rate limiting enzyme in the
hexosamine biosynthetic pathway (
1
). This enzyme diverts 2-5% of the fructose-6-phosphate derived from glucose to glucosamine-6-phosphate, using glutamine as the nitrogen donor
(
2
). Subsequently, glucosamine-6-phosphate is metabolized to uridine diphosphate
N
-acetylglucosamine (UDP-GlcNAc), which serves as a substrate for protein glycosylation.
Recent studies have shown that the hexosamine biosynthetic pathway regulates a
diverse set of cellular events, including glucose-induced insulin desensitization in adipocytes (
3
), glycogen synthase activity (
4
), pyruvate kinase activity (
5
) and glucose-induced growth factor expression in vascular smooth muscle cells (
6
-
8
). These studies have shown that glucosamine, the product of GFAT activity, can
mimic the effects of glucose. In some of these studies (
7
,
8
) it has also been shown that the metabolism of glucose to glucosamine through
GFAT is necessary for the effect of glucose on the respective metabolic
function. Thus, it appears that it is the intracellular concentration of
glucosamine, or one of its products, that mediates these metabolic changes.
The intracellular concentration of glucosamine is likely to be tightly
regulated. One point of regulation of glucosamine synthesis is at the level of
GFAT activity itself. This enzyme is negatively regulated in eukaryotes by its
downstream product, UDP-GlcNAc, perhaps through an allosteric mechanism (
9
,
10
). However, the mass amount of this enzyme within the cell also appears to
affect the intracellular glucosamine content. Recently, we have shown that the
intracellular content of GlcNAc covalently bound to protein by
O
-linkage (
O
-GlcNAc) is an indirect measure of GFAT activity (
7
,
8
). This covalent modification can be measured with a monoclonal antibody, RL-2. It was observed that pharmacological blockage of GFAT activity or an
antisense block of GFAT expression resulted in a decrease in the intracellular
content of
O
-GlcNAc-modified proteins (
7
,
8
). In another study, overexpression of GFAT under the control of a heterologous
promoter in cultured cells resulted in increased intracellular glucosamine (
11
). The endogenous GFAT promoter also appears to regulate expression of this
enzyme. In yeast, GFAT gene transcription can be regulated by [alpha]-pheromone (
12
), and in human cells, GFAT gene transcription can be regulated by EGF, glucose
and glucosamine (
13
). This regulation of the GFAT promoter could potentially result in changes in
glucosamine synthesis with subsequent consequences on protein glycosylation and
cellular metabolism. These studies were therefore designed to molecularly clone
and identify the mouse GFAT promoter and perform an initial characterization of
this promoter. We found that the GFAT promoter has no TATA box, typical of most
genes encoding `housekeeping' enzymes, and that its basal transcription is mainly regulated by the transcription factor Sp1. The availability of this promoter will allow more detailed analysis of transcriptional
regulation of the GFAT gene.
A once amplified 129SV mouse genomic library in [lambda]FixII (Stratagene, La Jolla, CA) was used to isolate the clones
containing the 5'-region of the GFAT gene. Plaque lifts of 1 * 10
6
recombinants were hybridized with a
32
P-labeled 356 bp fragment of the mouse GFAT cDNA spanning from the ATG
translation initiation codon to the internal
Eco
RI site (
14
). After several low stringency washes, membranes were washed in 0.2* SSC, 1.0% SDS at 60oC for 60 min. Individual plaques were identified by autoradiography
and purified to homogeneity through subsequent rounds of screening.
5'-RACE of the GFAT transcripts was performed using the Gibco-BRL 5'-RACE kit (Grand Island, NY). Briefly, first
strand cDNA was synthesized using 3 [mu]g poly(A
+
) RNA isolated from NIH 3T3 cells and a GFAT gene-specific primer corresponding to nucleotides +91 to +48 relative to the
ATG translation initiation codon. After incubation with 8 U SuperScript reverse
transcriptase for 30 min at 42oC, the RNA template was degraded with RNase H and the cDNA was purified
using the GlassMax DNA Isolation Spin Cartridge System. The 3'-ends of the cDNA were tailed using 10 U terminal deoxynucleotide
transferase and 2 mM dCTP. The cDNA pool was then amplified by PCR using the
Gibco-BRL poly(dGTP) anchor primer and a nested GFAT gene-specific primer corresponding to cDNA nucleotides +54 to +30
relative to the ATG translation initiation codon. The primer contained an
Eco
RI restriction site. PCR amplification was carried out for 40 cycles with an
anealing temperature of 50oC and this gave rise to a single band of ~250 bp. The product was digested with
Eco
RI and
Sal
I, ligated into correspondingly cut pT7T3 (Pharmacia, Piscataway, NJ) and
independent clones were sequenced on both strands. The GFAT promoter (-470/+88) was placed upstream of the firefly luciferase cDNA (pGL-2 Basic; Promega, Madison, WI) and this plasmid was subsequently
transfected into NIH 3T3 cells. 5'-RACE of the luciferase transcript as directed by the GFAT promoter
was performed to determine the conferred transcriptional start site. At 40 h
post transfection, poly(A
+
) RNA was isolated to serve as a template for primer extension using nested
luciferase gene-specific primers as follows: cDNA synthesis primer, 5'-AATGGCGCCGGGCCTTTCTTTATGTTTTTGGCGTCTTCCATT-3'; PCR primer, 5'-CCATTTTACCAACAGTACCGGAATG-3'. After PCR
amplification, the unique product was digested with
Xho
I and
Sal
I, cloned into pT7T3 and independent clones were sequenced on both strands.
DNA sequencing was performed by the dideoxy chain termination method (
15
) employing the Sequenase Version 2.0 sequencing kit (USB, Cleveland, OH). The
Sac
I-
Sac
I 1.9 kb genomic DNA fragment was cloned into pT7T3 and sequenced using a series
of overlapping nested primers. Computer sequence analysis was done with the GCG
Version 7.2 Fragment Assembly System (FAS) and overlap was identified between
the genomic sequence and the 5'-RACE products.
The
Sac
I-
Sac
I 1.9 kb genomic DNA fragment (corresponding to -1822/+88 relative to the transcription initiation site) was cloned into
pGL-2 Basic. The -470/+88 construct was generated by cutting the vector with
Kpn
I and closing with ligase. The remaining deletant constructs were prepared by
PCR amplification using the following oligonucleotides: 5'-CGGGGTACCTACAGCTTTTCTCTCTGTC-3' (for -275/+88); 5'-CATGGTACCCGCAGCTCTGCGTCTG-3' (for -120/+88); 5'-CATGGTACCCAATGG-
GAGAGCCG-3' (for -55/+88); with the bottom strand primer being the luciferase
PCR primer listed above. The 5'-untranslated region (UTR) variants were also generated by PCR, with
the bottom strand oligonucleotides being 5'-CATGGATCCAGCGCTCGCTTCGCTCTC-3' (for -470/+19) and 5'-CATGGATCCGATGTTGGTCACGGGCGAG-3' (for -470/+149), with
the top strand primer being 5'-GTGCTGGGACTACTGAC-3'. For site-directed mutagenesis, constructs were generated
as previously described (
16
). Briefly, the GFAT-luciferase vector (-120/+88), which contains an F1 origin, was used to generate ssDNA
with R408 helper phage. Oligonucleotides bearing different restriction sites
for each of the seven mutant constructs were used to prime second strand
synthesis. After plasmid purification, all mutations were confirmed by both
restriction digestion and sequencing.
NIH 3T3 and MDA468 cells were passed weekly in Dulbecco's modified Eagle's
medium (DMEM) + 10% neonatal calf serum (NCS) supplemented with 100 [mu]g/ml penicillin and 50 [mu]g/ml gentamicin. For transfections, 6 * 10
6
cells were trypsinized, washed with cold phosphate-buffered saline and resuspended in DMEM + 10% NCS and 12 [mu]g appropriate luciferase plasmid were added in addition to 8 [mu]g pCMV-[beta]Gal reporter plasmid to normalize for transfection
efficiency. Cells were electroporated at 400 (NIH 3T3) or 350 V (MDA468) and
500 [mu]F in a Gene Pulser (BioRad, Richmond, CA) and plated at a density of 5 * 10
5
cells/well using 6-well plates (Fisher Scientific, Atlanta, GA).
Luciferase activity was measured in detergent extracts of cells in the presence
of ATP and luciferin as previously described (
17
) using a Monolight 2010 luminometer (Analytical Luminescence Laboratory, San
Diego, CA). Light output was integrated over a 10 s period and displayed as
relative light units (RLU). Background activity was found to be <4% of all readings. The same extracts were assayed for [beta]-galactosidase activity as previously described (
18
). All [beta]-galactosidase activities were within the linear range of the
absorbance curve.
The full-length human Sp1 cDNA was cloned downstream of the cDNA encoding the
glutathione S-transferase (GST) protein. The GST-Sp1 sequence was then cloned into the pTM3 vector and a recombinant
vaccinia virus (vGST-Sp1) was generated using selection methods previously described (
19
,
20
). After co-infecting BSC-40 cells with vGST-Sp1 and vTF7-3 viruses, protein extracts were prepared as described (
20
). The fusion protein was affinity purified by incubating with glutathione
covalently linked to Sepharose beads (Pharmacia, Piscataway, NJ). After several
washes, the beads were resuspended and the Sp1 protein was released by cleaving
with thrombin. The Sp1-containing supernatant was cleared of thrombin by incubation with
Sepharose beads covalently linked to benzamidine (Pharmacia, Piscataway, NJ).
Concentration was determined by the BioRad D
c
Protein Assay (Richmond, CA). SpN protein is encoded by the same cDNA but lacks
the N-terminal 83 amino acids, resulting in a protein of ~80 kDa. The SpN cDNA was cloned directly into pTM3, generating the recombinant virus vSpN. After co-infection with vTF7-3, protein extracts were prepared and SpN protein was
partially purified using wheatgerm agglutinin chromatography (
21
). After elution from the column, protein concentrations were determined using
the BioRad D
c
Protein Assay (Richmond, CA).
Nuclear extracts were prepared from MDA468 cells as previously described (
16
,
22
). The GFAT oligonucleotide were made by annealing two complementary sequences
spanning the indicated sequences in the GFAT promoter. 5'-Overlaps were provided to allow labeling with
32
P using Klenow enzyme. Approximately 20 fmol DNA were added to a nuclear extract
containing 2.5 [mu]g protein in a final volume of 25 [mu]l 20 mM HEPES, pH 7.9, 50 mM KCl, 1 mM dithiothreitol (DTT), 10%
glycerol and 1 [mu]g poly(dI[middot]dC). Following a 20 min incubation at room temperature, DNA-protein complexes were resolved by electrophoresis on a 5.5% non-denaturing polyacrylamide gel with 90 mM Tris-borate, 2 mM EDTA buffer. For antibody experiments, the
radiolabeled probe was added subsequent to a 25 min pre-incubation of nuclear extract with antibody at 4oC. The human AP-2 and Sp1 antibodies were purchased from Santa Cruz
Biotechnology (Santa Cruz, CA) and another human Sp1 antibody is described
elsewhere (
23
).
The vector containing the GFAT promoter (-470/+88) driving firefly luciferase was linearized with
Hin
dIII and end-labeled with Klenow enzyme. After secondary digestion with
Kpn
I, the purified labeled 450 bp fragment was incubated with proteins in a final
volume of 50 [mu]l with buffer containing 25 mM Tris, pH 8.0, 6.25 mM MgCl
2
, 0.5 mM EDTA, 0.5 mM DTT, 7.5% glycerol, 10 [mu]g bovine serum albumin and 200 ng poly(dI[middot]dC) for 25 min at 4oC. Varying concentrations of DNase I were added and allowed to
react for 30-120 s. The reactions were stopped with EDTA, phenol/chloroform extracted
and ethanol precipitated. The digested DNA was electrophoresed on a 6%
acrylamide-urea DNA sequencing gel prior to autoradiography. The exact locations of
the DNase I footprints were determined by comparison with sequences of known
length electrophoresed concurrently.
Although the sequence of the entire open reading frame for the mouse GFAT cDNA
was previously determined (
14
), the GFAT mRNA was found to be as much as 5 kb larger than the open reading
frame. In order to orient the GFAT promoter relative to the coding sequence and
identify the GFAT transcription start site(s), we used the 5'-RACE method to perform primer extension on the GFAT mRNA. The
single 5'-RACE product of ~200 bp was cloned and four individual clones were sequenced on
both strands. The sequence was identical for all clones and each contained the
50 bp sequence 3' of the ATG translation initiation sequence, allowing identification of
these sequences as GFAT. Furthermore, all clones were found to initiate from
the same guanidine nucleotide 149 bases upstream of the ATG translation initiation codon (DDBJ/EMBL/GenBank accession no. U39442).
A mouse genomic library was screened using a probe derived from the 5'-end of the GFAT cDNA (
14
). Four clones were isolated after screening ~2 * 10
6
[lambda] phage plaques. In order to identify which of these clones contained the
5'-end of the GFAT gene, phage plaques from these four clones were
probed with an oligonucleotide representing a sequence from the 5'-UTR (primer 642). Two of these clones were positive when probed
with this oligonucleotide. Restriction digests of DNA from the two positive
clones resulted in identification of an identical 1.9 kb
Sac
I fragment from both clones which could be hybridized to primer 642. Southern
blotting of mouse genomic DNA cut with
Sac
I and probed at high stringency with this segment of the GFAT gene revealed the
same unique 1.9 kb fragment (data not shown). This 1.9 kb
Sac
I fragment was subcloned into pT7T3 and sequenced using a series of overlapping
primers (DDBJ/EMBL/GenBank accession no. U39442). The sequence of the 3'-end of this
Sac
I genomic fragment corresponded exactly to 88 bp of the 5'-RACE cDNA product, beginning with the transcription start site and
continuing to the endogenous
Sac
I site. To determine whether an intron was present in the segment of the gene
encoding the 5'-UTR, the original [lambda] phage genomic DNA was sequenced using primer 642. The
resulting sequence corresponded exactly to the 5'-RACE product from the priming site to the ATG translation
initiation sequence, thus establishing the lack of an intron in this segment of
the GFAT gene.
The sequence immediately upstream of the transcription start site is GC rich and
lacks a TATA box, a feature typical of housekeeping genes. In a transient
transfection assay using a construct in which the 1.9 kb
Sac
I-
Sac
I fragment was placed upstream of firefly luciferase in the native orientation,
reporter activity was readily detected (see below). However, placement of this
gene segment in reverse orientation in this reporter plasmid resulted in a
total loss of transcriptional reporter activity (data not shown). To confirm
that this fragment of the GFAT gene confers proper transcriptional initiation
to a heterologous reporter gene, the transcriptional start site of the
luciferase reporter mRNA, whose expression was driven by this segment of the
gene, was determined by 5'-RACE. Five such clones were sequenced on both strands. Relative to
the +1 start site determined for the native gene, these reporter transcripts
were found to initiate from -4, -3, -3, +1 and +3. This finding indicates that the 5'-flanking sequence of the GFAT gene confers a
transcriptional start site either identical or very close to the start site of
the native gene to a reporter gene. The small variance observed between the
native and reporter start sites is consistent with the usual finding that
housekeeping genes lacking a TATA box often initiate at a cluster of sites.
However, it remains possible that the portion of the promoter used in the
reporter plasmid is incomplete or that the number of 5'-RACE products that were sequenced was too small to detect this
variability in the native gene.
In order to identify the proximal GFAT promoter, deletional analysis of the 5'-end of the promoter was undertaken and activity of the resulting
truncated promoter was determined by transient transfection assays into the
mouse fibroblast cell line NIH 3T3. The cells were co-transfected with a CMV-[beta]-galactosidase construct to control for transfection
efficiency (Fig.
1
). Deletion of sequences distal to -470 resulted in some augmentation of reporter function relative to the
1.9 kb
Sac
I promoter fragment, while deletion of the segment between -470 and -275 resulted in no loss of promoter activity. The segment of the
promoter between -275 and -120 contributed significantly to promoter function, but residual
activity following deletion of this segment left a construct with 57% of the
activity of the full-length promoter. Further deletion to -55 resulted in a loss of most of the promoter activity. This
finding indicated that the segment of the gene between -120 and +88 contained the core promoter element. Computer analysis of the
segment of the gene downstream (+1 to +88) of the transcriptional initiation
site(s) predicted an RNA transcript with secondary structure. The predicted
stem-loop structure has a calculated free energy of -46.1 kcal/mol and such potentially stable structures have been
shown to have an impact on activity of the adjacent promoter (
24
). Nevertheless, deletion of this segment of the gene had no detectable impact
on reporter activity when tested in the context of the -470 promoter. This delineation of the core promoter directed our further
studies on the DNA-protein interactions that control transcription from this segment of the
gene.
The sequence of the core promoter was computer searched against a transcription
factor database. In the region of the promoter near to the transcriptional
start site were overlapping putative binding sites for Sp1 (-17 to -5) and AP-2 (-28 to -16). More distally there was an additional
putative Sp1 site (-89 to -83). To determine if Sp1 indeed binds to these consensus sites,
DNase I footprinting analysis of this proximal portion of the GFAT promoter was
performed on the top strand (Fig.
2
A). Footprinting with purified Sp1 made using a recombinant vaccinia virus
system produced three distinct footprints. Two of the footprints were localized
to the consensus Sp1 binding sites at -17/-5 and -89/-83. The other site corresponded to a GC-rich sequence between -60 and -50. This latter sequence does
not correspond to a consensus Sp1 binding site, but did permit binding of
purified Sp1 at the concentration used in this experiment. The same segment of
DNA was subjected to footprint analysis with nuclear extract from MDA468 cells.
The extract from MDA468 cells proved to be of higher quality than that derived
from NIH 3T3 cells. A nearly complete footprint was observed at the proximal
Sp1 binding site (-17/-5), with only minor differences from the footprint created by pure
Sp1. No footprinting was observed at the -60/-50 site, while a partial footprint was observed at the more distal
site (-89/-83). The partial footprint at the more distal site appeared to be
shifted in the 5'-direction relative to the footprint made by Sp1 and was associated
with a DNase I-hypersensitive site upstream of this footprint (Fig.
2
B). No footprint was observed over the putative AP-2 binding site (-24/-17). The differences in the footprints generated by pure Sp1
versus nuclear extract probably relate both to the concentration of Sp1 in
these two preparations and to the presence of other DNA binding proteins in the
nuclear extract.
A
To confirm the findings of DNA-protein interactions from footprinting, we used EMSA to assess the
interaction of these elements on the GFAT promoter with proteins in nuclear
extracts from MDA468 cells. The three regions of the promoter that were
footprinted with pure Sp1 were investigated more thoroughly. Figure
3
shows EMSA using a probe spanning from -30 to +13. The major shifted bands, indicated by the arrows, were
specific, since incubation with the same unlabeled oligonucleotide derived from
the GFAT promoter resulted in a dose-dependent competition of binding to the probe. A consensus Sp1
oligonucleotide competed away the major upper band entirely, but was
ineffective at competing the less intense lower band. A consensus AP-2 oligonucleotide was less effective than the Sp1 oligonucleotide at
competing these bands. The AP-2 competitor oligonucleotide contained the sequence GGGCGG, which is at
the core of the Sp1 consensus site, and this similarity might explain the weak
competition observed with the AP-2 oligonucleotide. Further specificity testing also indicated that an Oct-1 binding site and a mutant Sp1 binding site could not compete for
binding to the GFAT-derived oligonucleotide in the EMSA (data not shown).
The functional significance of the Sp1 binding sites to the GFAT promoter was
investigated by site-directed mutagenesis of the promoter. Using the -120/+88 segment of the promoter as the parent construct, seven mutations were placed along the promoter. After co-transfection of these constructs into either NIH 3T3 or
MDA468 cells with CMV-[beta]-galactosidase, the cell extracts were assayed for [beta]-galactosidase and luciferase activities (Fig.
7
). Mutations at the three sites footprinted by purified Sp1 resulted in a
significant loss of promoter activity, whereas mutations at a site not
identified by footprinting (L/S -50) or at the potential AP-2 binding site (L/S -23) had no significant affect on promoter activity. Of note,
a mutation at the major transcriptional initiation site also had no effect on
transcription, suggesting that this region of the gene does not support binding
of transcription factors. These results strongly indicate that the Sp1 binding
sites identified in this gene play an important role in regulation of basal
expression of the GFAT gene. These results confirm that the potential AP-2 site identified in this segment of the promoter does not play a role in
regulation of this gene.
Prior studies have shown that in MDA468 cells EGF induces a late transcriptional
response from the GFAT promoter, as indicated by Northern blotting and nuclear
run-on experiments (
13
). To determine if the GFAT promoter can confer an EGF response, various
proximal segments of the promoter were transfected into MDA468 cells. The cells
were treated for 18 h with 5 nM EGF prior to assay of reporter function. Table
1
indicates that the promoter, even when 5'-deleted to -36, was still capable of confering the EGF response to the
luciferase reporter. When deleted to -7, no promoter activity could be detected. Since all of the tested
deletion constructs contained at least one Sp1 binding site, it is possible
that the EGF response might be confered by this element.
GFAT appears to be the unique enzyme that allows
de novo
synthesis of glucosamine from glucose (
1
,
8
). Glucosamine can then be utilized in the synthesis of other hexosamines and
these hexosamines are vital for post-translational modification of proteins by glycosylation. Thus, the rate of
glucosamine synthesis must be coupled to the rate of glycoprotein synthesis. In
addition, glucose metabolism to glucosamine appears to play a role in other
metabolic pathways, including glucose-induced insulin desensitization in adipocytes (
3
), glycogen synthase activity (
4
), pyruvate kinase activity (
5
) and glucose-induced growth factor expression in vascular smooth muscle cells (
6
-
8
). These metabolic consequences of glucosamine require that the synthesis of
this sugar be under strict control. The major known points of control appear to
be as follows: first, the concentration of the GFAT substrate, fructose-6-phosphate, is highly regulated at the level of its glycolytic
metabolism to fructose 1,6-bisphosphate (
28
,
29
); second, GFAT activity is regulated allosterically by its downstream product,
UDP-GlcNAc (
9
,
10
); third, GFAT activity can be influenced by the quantity of GFAT protein in the
cell (
7
,
8
,
11
). The amount of GFAT protein in the cell reflects the balance between synthesis
and degradation. While little is known about the half-life of the GFAT protein, our prior studies have shown that transcription
of the GFAT gene appears to be coupled to the cell cycle and can also be
regulated by glucose and glucosamine (
13
). To further characterize the transcriptional regulation of GFAT, we cloned the
5'-flanking region of the GFAT gene.
These studies show that the 5'-flanking region of the GFAT gene contains the promoter. This
segment of the gene directs transcriptional initiation of a luciferase reporter
gene from a tight cluster of nucleotides (-4 to +3) including and surrounding the transcriptional initiation site
observed in the native GFAT gene. We attribute this imprecise initiation to the
lack of a TATA box. Furthermore, restriction analysis of DNA from the four
independently cloned [lambda] phages containing the GFAT gene yielded fragment sizes that matched
those observed on Southern blots of mouse genomic DNA cut with the same
restriction enzymes. These observations suggest that the GFAT gene is a single
copy gene in the mouse genome and that the cloned gene did not contain a
rearrangement.
Table 1
Deletional analysis of the GFAT promoter indicated that most of the
transcriptional activity resided between bases -275 and +19 relative to the transcription start site. The sequence distal
to -470 appeared to repress promoter activity slightly and the basal promoter
appeared to reside between -120 and +19. The region between -120 and -275 enhanced activity almost 6-fold above basal, suggesting that this region
contains important elements that contribute to activity of the GFAT gene. The
identity of these elements remain to be determined. This paper describes a
detailed analysis of this proximal segment of the GFAT promoter between -120 and +19.
DNase I footprinting analysis of the proximal GFAT promoter indicated that
recombinant Sp1 binds to three sites in this segment of the GFAT gene. Gel
shift analysis confirmed that native Sp1 in nuclear extracts also specifically
bound these sites, whereas binding of other transcription factors to these
sites was not detectable. Mutagenesis at these Sp1 binding sites markedly
reduced promoter function, whereas mutations at two other sites had no
significant effect on function of the promoter. Taken together, these studies
indicate that transcription from the proximal GFAT promoter is principally
under control of the transcription factor Sp1. Sp1 is a ubiquitous
transcription factor that appears to be involved in the transcription of many
viral and mammalian genes. Sp1 is modified post-translationally both by phosphorylation (
30
) and glycosylation with
O
-GlcNAc (
31
). The role of these modifications has not been determined, although the major
phosphorylation appears to require binding of Sp1 to DNA (
30
) and the glycosylation may be involved in transcriptional activation but not
DNA binding by Sp1 (
31
). Sp1 has been considered to confer constitutive expression of housekeeping
genes. However, transcription of several genes considered to be housekeeping
genes has been shown to be regulated under certain circumstances. For example,
those genes encoding enzymes required for DNA synthesis are regulated in a cell
cycle-dependent manner (see for example
32
). The dihydrofolate reductase gene has been shown to be regulated in a manner
that depends on the transcriptional activity of the transcription factor E2F,
which in turn is regulated by retinoblastoma (Rb) protein (
33
-
37
). Rb phosphorylation is regulated in a cell cycle-dependent manner and the phosphorylation state of Rb controls its ability
to bind and neutralize E2F transcriptional activity. At the G1/S boundary, Rb
becomes hyperphosphorylated and thereby incapable of binding E2F. E2F is then
liberated to activate its target genes. Recently it has been proposed, but not
proven, that Sp1 activity may also be indirectly controlled by Rb. An inhibitor
of Sp1 (Sp1-I) has been described that blocks the ability of Sp1 to bind to DNA (
38
). Sp1-I also appears to be bound by Rb, raising the possibility that Rb might
control Sp1 activity through sequestration of Sp1-I. Sp1 activity has also been shown to be regulated in response to DNA
methylation (
23
) during the differentiation of a leukemic cell line in response to phorbol
esters (
26
) and cell cycle regulation of gene expression has been attributed in part to
Sp1 (
35
). Since GFAT activity is required for post-translational modification of proteins and protein synthesis varies with
the growth state of cells, transcriptional regulation of GFAT may contribute to
the control of GFAT activity under various growth conditions. Indeed,
activation of GFAT gene transcription appears to be a late response to EGF
stimulation (
13
).
The GFAT promoter conferred a response to EGF in transfection studies. This
response was seen even after normalization to the activity of [beta]-galactosidase, whose expression was under the control of the co-transfected CMV-[beta]Gal plasmid. The CMV construct has been shown to
be non-responsive to EGF in MDA468 cells (
7
). While more distal elements in the GFAT promoter may have contributed to the
EGF response, the proximal 120 bp of the promoter also displayed EGF
responsiveness. Deletional analysis of this 120 bp segment of the promoter
indicated that even the most proximal 36 bp of the promoter exhibited the EGF
response. This segment of the promoter contains one of the three Sp1 binding
sites characterized in the 120 bp segment of the gene. While it is tempting to
attribute the EGF response to these Sp1 binding sites, it is not possible to do
so definitively, because removal of all the Sp1 binding sites results in a
totally inactive promoter. The lack of a response from an otherwise inactive
promoter cannot be interpreted as resulting from removal of the EGF response
element. In the presence of the one Sp1 binding site in the -36 promoter, it remains possible that the EGF response resulted from a
cryptic enhancer element present elsewhere in the plasmid (
39
). Whether the proximal GFAT promoter and Sp1 that binds to this segment of the
gene are responsible for this transcriptional response to EGF or other stimuli
remains to be elucidated by direct studies on Sp1 protein and inducible
modifications of this protein. Such studies are underway in this laboratory.
This work was supported by grants from the National Institutes of Health,
DK43652, and the Juvenile Diabetes Foundation, International.
GFAT promoter
-EGF
+EGF
Ratio +/-EGF
-470 to +88
68.9
201.2
2.92
-120 to +88
11.7
39.3
3.35
-55 to +88
1.72
6.01
3.50
-36 to +88
0.84
2.73
3.23
-7 to +88
0.11 (background)
0.13 (background)
NA
REFERENCES
Return