ABSTRACT
This report describes the isolation, sequencing and preliminary characterization
of the first 1 kb of the 5
'
-regulatory region of the human
QM
gene. This region and the 5
'
-half of the transcribed region of the
QM
gene are enriched for C and G nucleotides with no bias against CpG
dinucleotides-indicative of a CpG island. Several consensus GC boxes are present within
the sequence. Most are clustered at the distal end, with one site present in
the proximal 200 bp of the promoter. Electrophoretic mobility shift experiments
and luciferase assays done in insect cells transfected with an Sp1 expression
construct suggest that most of these sites can bind Sp1 or a closely related
factor. In addition, the promoter is shown to be responsive to cAMP via a
response element (CRE) in the proximal promoter. Studies with 5
'
-end and internal deletion mutants suggest that elements in the distal
promoter exert their positive effect through interactions with a proximal
element(s). Candidate proximal elements include the proximal GC box and a 43 bp region between a
Kpn
I site (at -182) and a
Sma
I site (at -139).
The
QM
gene was first identified by subtractive hybridization as a gene for which
increased expression correlated with the non-tumorigenic phenotype of a Wilms' tumor microcell hybrid (
1
). Southern analysis showed
QM
to be a member of a large, multigene family in mammals (
1
).
QM
is the only member of this family that is known to be expressed. Other family
members so far isolated appear to have arisen by retrotransposition events and
may be pseudogenes.
The QM protein is a highly basic, 25 kDa protein that shows no significant similarity to any other known human proteins. However, QM homologs have been isolated from a diverse array of other species,
encompassing not only the animal, but also the plant and fungal kingdoms (
2
,
3
). Amongst these homologs there has been striking conservation, such that even
yeast (
Saccharomyces cerevisiae
) and human QM are 70% identical at the amino acid level (
2
). Much of this conservation is within extensive domains of similarity that
stretch over the first 170 residues of the protein (
2
).
The function of QM remains obscure. It has been reported that QM can bind to c-Jun
in vitro
and repress c-Jun-mediated transcriptional activation, suggesting that QM may be a novel transcription factor (
4
). However, as yet, no
in vivo
data, such as co-immunoprecipitation of c-Jun and QM from cell extracts, has been presented in support of the
observed
in vitro
association. Moreover, subfractionation studies done in this laboratory suggest
that QM is located on the membrane of the endoplasmic reticulum and not in the
nucleus (T.M.Loftus and E.J.Stanbridge, unpublished results). The yeast homolog
of QM has also been localized to the cytosolic compartment by
immunocytochemistry (
5
). Thus, the apparent association of c-Jun and QM may not be physiological, although the formal possibility that
QM translocates to the nucleus under certain conditions, or in quantities
undetectable by the methods used, cannot be excluded.
Whatever the true role of QM, its normal function is critical to eukaryotes.
This is suggested both by its extreme conservation across the animal, plant and
fungal kingdoms (
2
) and by the finding that deletion of the yeast homolog of
QM
is lethal in yeast (
5
). Not surprisingly,
QM
appears to be expressed in all mammalian tissues examined (
1
,
6
). However, the level of
QM
expression shows considerable variation between different tissues, as well as
within tissues at different stages of development (
1
,
6
). In particular, there appears to be an inverse correlation between the level
of
QM
expression and developmental stage. For example, Northern analysis of normal
mouse tissues from different stages of development revealed a decrease in
QM
expression in heart, kidney, liver and skin between embryonic and adult stages
(
1
). Moreover, the mouse homolog of
QM
was isolated by subtractive hybridization as a gene whose expression decreases
70% upon differentiation of pre-adipocytes into mature adipocytes (
7
,
3
). A similar reduction in
QM
expression was also observed in differentiating rat adipocytes (
3
). In plants also, decreased expression of
QM
is associated with differentiation into adult tissues (
8
). Taken together, these findings suggest that
QM
may have a function in cell growth or differentiation. However, it remains to
be demonstrated whether the changes in
QM
expression are a cause or consequence of the changes in cell
growth/differentiation. As a first step toward a better understanding of the possible mechanisms for
regulating QM expression we report here on the isolation, sequencing and preliminary characterization of the 5'-regulatory region of human
QM
.
A luciferase reporter construct containing ~1 kb of sequence upstream of the
QM
transcription start site was generated in two steps. First, a fragment
extending from an
Xho
I site 995 bases upstream of the transcription start site to a
Dpn
I site in the second exon was cloned into pGL2basic (Promega) at the
Xho
I and blunted
Hin
dIII sites. This was done as a three-part ligation using an
Xho
I-
Aat
II fragment and an
Aat
II-
Dpn
I fragment. Ligation of the
Dpn
I half-site to the filled-in
Hin
dIII site recreates a
Hin
dIII site. Second, the construct generated in the first step (pGL2QM1) was
opened with
Hin
dIII and
Aat
II, and the excised fragment further digested to generate an
Aat
II-
Fsp
I fragment. This fragment was then religated into the opened pGL2QM1 together
with an oligomer encoding from the
Fsp
I site to the fifth base of the first exon of QM, followed by a
Hin
dIII site. The sequence of this oligomer is: sense strand, 5'-GCAGGCGGAGGAGCGCCTCTTA; antisense strand, 5'-AGCTTAAGAGGCGCTCCTCCGCCTGC. 5'-End and internal deletion mutants were
generated from this construct by digestion with various restriction enzymes
followed by reclosure of the vector. (see Fig.
3
).
Sequencing was done by the chain termination method (
9
) using Sequenase 1.0
®
(US Biochemical). The primers used were as follows:
GL1, 5'-TGTATCTTATGGTACTGTAACTG;
GL2, 5'-CTTTATGTTTTTGGCGTCTTCCA;
QM35, 5'-GTAAGAACCATAGAGTCCTGT;
QM2-53, 5'-AGCACAGTGGAGTGGGAA;
QM2-35, 5'-TTCCCACTCCACTGTGCT;
QM3-35, 5'-CGTTAACTGTGACAGACGTA;
QM3-53, 5'-CTCTCAGAAATATACGTCTG;
QM4-35, 5'-GAGAAATCTCCACGGAGA;
QM5-53, 5'-CGGGTTGACAAAGGAACG.
Primers GL1 and GL2 are specific for the flanking vector sequences (pGL2Basic).
The remaining primers are
QM
specific.
NIH3T3 cells were maintained in DMEM with 10% bovine calf serum (Hyclone). SL2
(Schneider line 2) insect cells were cultured in Shields and Sang M3 insect
medium (Sigma) supplemented with 10% inactivated bovine calf serum.
For transfection, 4 * 10
5
NIH3T3 cells or 5 * 10
6
SL2 cells were plated in a 60 mm diameter dish and allowed to grow overnight. For each experiment, equimolar quantities of the various promoter constructs
were used and the total mass of DNA was standardized between transfections
using pGL2basic DNA. For the insect cell experiments, promoter constructs were
transfected either in the presence or absence of the Sp1 expression vector
pPacSp1 (
10
). The pGL2promoter vector (Promega), which contains six Sp1 binding sites from
the SV40 promoter, served as a positive control for Sp1 activity. All
experiments were done in triplicate. The DNA was transfected by calcium phosphate precipitation, as
previously described (
11
). After two days, the cells were washed with PBS and lysed in 300 [mu]l of luciferase reporter lysis buffer (Promega). The lysate was clarified by
centrifugation, and the supernatant transferred to a new tube. The nuclear
pellet was extracted for DNA and used to standardize transfection efficiency.
Luciferase activity of the supernatant was assayed using Promega's luciferase
assay system and a Monolight 2010 luminometer (Analytical Luminescence
Laboratories).
NIH3T3 transfection efficiencies were standardized to the quantity of luciferase
DNA present in the nuclei of the transfected cells. To do this, the
concentration of DNA in one sample was determined and the volume containing 5 [mu]g DNA calculated. This volume was then taken from each sample and
transferred to a nylon membrane (Nytran; Schleicher and Schuell) using a slot
blot apparatus (Schleicher and Schuell). The membrane was then probed using a
2.5 kb (
Hin
dIII-
Cla
I) luciferase fragment from pGL2basic that had been random prime labeled (
12
). After overnight hybridization at 65oC, the blot was washed (once in 2* SSC, 0.2% SDS for 15 min at 25oC, once in 0.2* SSC, 0.2% SDS for 30 min at 25oC and once in 0.2* SSC, 0.2% SDS for 30 min at 65oC) and exposed to X-ray film (X-Omat; Kodak). The resulting
autoradiogram was scanned onto a Macintosh 660AV computer using a Hewlett
Packard flatbed color scanner and densitometry was done on the image using the
program NIH Image 1.5f.
For the insect cell experiments, luciferase activity was standardized to the
quantity of total cellular protein in the luciferase extracts.
Dibutyryl cyclic AMP (DBcAMP; Sigma) was solvated in 100% dimethyl sulfoxide
(DMSO) and stored at -20oC. The drug was added to the cells at doses of 0.1-1.0 mM 36 h post-transfection. Cells were harvested 12 h later and
assayed for luciferase activity as described above. Cells treated with 0.4%
DMSO (the final concentration of DMSO in the DBcAMP treatments) acted as a
control.
EMSAs were done using the following oligomers and restriction fragments of the
QM
promoter as probes:
GC box consensus oligomer, 5'-attcgatcGGGGCGGGGCgagc (Santa Cruz);
GC box mutant oligomer, 5'-attcgatcGG
CRE consensus oligomer, 5'-agagattgccTGACGTCAgagagctag (Santa Cruz);
CRE mutant oligomer, 5'-agagattgccTG
QM
wild-type CRE, 5'-tatggtcaTGACGTCTgacagagc;
QM
mutant CRE, 5'-tatggtcaTG
For these oligomers, only the sense strand is shown. Upper case nucleotides
represent the binding site and underlined bases depict mutations. The
QM
promoter restriction fragments containing potential GC boxes were
Pst
I (at -925)-
Bgl
I (at -844),
Fsp
I (at -866)-
Pvu
II (at -702) and
Pvu
II (at -203)-
Sma
I (at -139). The positions of these fragments are detailed in Figure
4
A. All probes were phosphorylated with [[gamma]-
32
P]ATP using T4 polynucleotide kinase (NEB) and purified over Sephadex G-50 columns. Restriction fragment probes were dephosphorylated using calf intestinal alkaline phosphatase prior to being labeled. For
each reaction, 6.5 pmol probe was labeled using 70 [mu]Ci [[gamma]-
32
P]ATP (>5000 Ci/mmol) in a reaction volume of 15 [mu]l. Binding reactions were done by mixing 5 [mu]g HeLa nuclear extract (Promega) with ~20 000-40 000 c.p.m. probe in a mixture containing 10 mM Tris, pH
7.5, 50 mM NaCl, 1 mM DTT, 1 mM Na
2
EDTA, 5% glycerol and 1 [mu]g poly(dI[middot]dC). The reaction volume was 20 [mu]l. Reactions were incubated at room temperature for 15 min and
then electrophoresed through a 6%, 29:1 acrylamide:bis polyacrylamide gel in 1* TGE (25 mM Tris base, 190 mM glycine, 1 mM Na
2
EDTA) at room temperature, with water cooling, for ~2 h at a constant 200 V. After electrophoresis, the gel was dried and
autoradiographed. For competition experiments the extracts were first incubated
with excess competitor DNA for 15 min. The probe was then added, and the
mixture incubated a further 15 min prior to electrophoresis. For super-shift experiments with anti-Sp1 antiserum (1C6; Santa Cruz), 1 [mu]g antiserum was mixed with the extract for 15 min at room
temperature, prior to adding the probe.
Analysis of the
QM
5'-flanking region and transcribed sequence for G/C content and
dinucleotide composition was done using the program COMPOSITION (
13
). Sequence comparisons were done using BLAST (
14
). Analysis of the 5'-regulatory region for potential transcription factor binding sites
was done using DNA Strider 1.2 for the Macintosh together with a database of transcription factor binding sites compiled by H.Mangalam (Department of
Microbiology and Molecular Genetics, University of California, Irvine).
Several groups have mapped the human
QM
gene to Xq28 (
6
,
15
,
16
). In particular, Bione
et al
. mapped
QM
to a single cosmid within a 450 kb contig stretching from the
G6PD
locus to the color vision genes (
16
). To clone the
QM
promoter from this cosmid, a probe specific for the 5'-end of genomic
QM
was generated by PCR using primers that lie within the first and fifth introns
of the gene (DC1, 5'-TAGGTCTGTTCTCGTCTTG and DC2, 5'-AATGTAGAGACTCCAACTGC). The amplified fragment was `TA'-cloned into pCRII (Invitrogen) and an ~500 bp fragment encoding from intron 1
through to the
Not
I site in the second intron was isolated by digestion with
Not
I and
Eco
RI. This fragment was used to probe restriction digests of the cosmid. In this
way, a
Not
I-
Eco
RI fragment that extended ~6 kb 5' of the
QM
transcription start site was identified and cloned into pBSK
+
(Stratagene). Subsequently, a 1 kb fragment encoding from nucleotides -995 to +5 relative to the transcription start site was cloned into the
firefly luciferase reporter vector pGL2Basic (Promega). This fragment was
sequenced in both directions, with the exception of the region 3' of the CRE (Fig.
1
A). The sequence of this segment has been published previously (
15
) and we found no differences between the published sequence and that described
here.
A striking feature of the sequence shown in Figure
1
A is its high G/C content, which averages 65% over its length (Fig.
2
A). Within the gene itself the percentage G/C falls, though at 52%, it continues
to remain above the 40% average G/C content for the human genome (
17
). In addition to the high G/C content, the usual bias against CpG dinucleotides
over GpC is not seen (Fig.
2
B). In higher vertebrates, the ratio CpG:GpC is typically ~0.25 (
18
). In contrast, this ratio is close to 0.8 within the QM promoter and rises further to 1.2 over exons 1-2 (Fig.
2
B). Although G/C content falls after exon 2, the CpG:GpC ratio remains high (~0.8) through exon 4, where it returns abruptly to the more common figure of
~0.2. The high CpG:GpC ratio of the
QM
promoter region was not unexpected, since a cluster of rare restriction enzyme
sites had been mapped to the 5'-end of the gene (
15
,
16
). Indeed, the cosmid contig that contains the
QM
sequence was developed by looking specifically for sequences on the X
chromosome that contain CpG islands, as defined by both restriction mapping and
the absence of methylation on CpG dinucleotides (
16
).
To assess the contribution of various regions of the promoter to its overall
activity, a series of 5'-end and internal deletion mutants was generated using appropriate
restriction enzymes. These constructs (shown in Figure
3
) were transfected into NIH3T3 cells, and promoter activity was measured as
described in Materials and Methods. The results of this analysis are tabulated
in Figure
3
. Fully 60% of the activity seen in the 1000 bp promoter fragment is lost on
deletion of the 5'-most 460 bases of the promoter up to the unique
Stu
I site (at -533). This region contains five putative consensus GC boxes, four
putative AP-2 sites and has the highest G/C content (Figs
1
A and
2
A). The region containing the 33 bp palindrome and the most distal AP-2 site at the very 5'-end of the sequence analyzed is not responsible for this
activity, since loss of this sequence alone (
Pst
I construct, Fig.
3
) does not result in any loss of promoter activity. In fact, loss of this region
gives a mild but significant increase in activity (>95% confidence by Student's
t
-test), suggesting that it may have a repressor function.
Promoter activity declines further (60%) to 14% activity with removal of
sequences up to the
Kpn
I site at position -182. This region contains two further putative AP-2 binding sites and the remaining GC box. Activity then drops
sharply to just 1-2% of maximal activity upon deletion up to the
Sma
I site (-139), just 43 bp less than the
Kpn
I fragment. This segment contains a putative glucocorticoid response element.
Further reductions, ultimately down to the region surrounding the TATA box have
little further effect. Interestingly, an internal deletion that removes ~200 bp between the
Avr
II site at -281 and the
Aat
II site at -99 generates a promoter with only 1% of the activity of the 1000 bp
construct. While it should be borne in mind that this deletion does not retain
the distal elements in the same positions relative to the TATA box as in the
wild-type construct, it does suggest that the distal elements alone are not
sufficient to provide promoter activity. Nonetheless, they do contribute
considerably to it, as witnessed by the effect of deleting them (
Stu
I construct).
As discussed above, there is considerable loss of promoter activity upon removal
of the distal half of the promoter. This region is very G/C rich and, as noted
in Figures
1
and
3
, contains all but one of the putative GC boxes. GC boxes bind members of the
Sp1 family of transcription factors. Thus, it seemed possible that Sp1, or a
related factor, acting through these sites might be responsible for the
transcriptional activity of this region. As a first step towards answering this
question, we tested the ability of these sites to bind to such a factor. To do
this, three restriction fragments were chosen that together covered all but one
of the putative GC boxes. These were
Pst
I (at -925)-
Bgl
I (at -844),
Fsp
I (at -866)-
Pvu
II (at -702) and
Pvu
II (at -203)-
Sma
I (at -139). The positions of these fragments are outlined in Figure
4
A. These fragments were gel purified and labeled with [[gamma]-
32
P]ATP as described in Materials and Methods and used in EMSAs to determine if
Sp1 or an Sp1-like factor could specifically bind to them. Two typical autoradiograms
are shown in Figure
4
B and C. This shows that both the
Fsp
I-
Pvu
II and the
Sma
I-
Pvu
II fragments bind a factor that can be specifically competed by excess unlabeled
consensus Sp1 oligomer, but not by an excess of a mutant Sp1 oligomer. No
similar binding was found using the
Pst
I-
Bgl
I fragment (Fig.
4
C). Moreover, under the conditions used, no similarly migrating species is seen
using a labeled mutant Sp1 oligomer as a probe (Fig.
4
B and C). These results suggest that at least some of the putative GC boxes
within the
QM
promoter can bind Sp1 or an Sp1-like factor that may contribute to the activity of the
QM
promoter.
To further investigate the potential functional role of Sp1 in regulating activity of the
QM
promoter, the luciferase assays were repeated using SL2 insect cells, which
lack endogenous Sp1. In these experiments, the various
QM
promoter constructs were transfected either in the presence or absence of an
Sp1 expression vector. The results are tabulated in Figure
3
, together with the results from the NIH3T3 experiments described above. In the absence of any exogenous Sp1, the activity of all the promoter constructs is
minimal. In contrast, addition of Sp1 results in a pattern of promoter activity
closely resembling that seen in the NIH3T3 assays. In particular, a large drop
in activity is seen with removal of the distal half of the promoter up to the
Apa
I site, and both the
Sma
I fragment (lacking all GC boxes) and the
Avr
II-
Aat
II internal deletion construct (lacking the proximal GC box) have little
activity either in the presence or absence of Sp1. Together with the EMSA data,
these data strongly suggest that the GC boxes identified are functional Sp1
binding sites and that synergy between the distal GC boxes and an element(s) in
the proximal promoter is important for promoter function.
Interestingly, the
Pst
I fragment (lacking the palindrome sequence) still shows significantly greater activity than the
Xho
I fragment, as was seen in the NIH3T3 experiments.
As mentioned above, there is a near-consensus CRE (5'-TGACGTCT-3') in the proximal region of the
QM
promoter (position -99 with respect to the transcription start site). We have obtained
evidence that
QM
transcription is responsive to cAMP through this site. As shown in Figure
5
, the 1000 bp promoter fragment is responsive to the cAMP analog DBcAMP at doses
of >= 0.5 mM. This effect is not seen with DMSO alone, used to solubilize the
DBcAMP. The effect of DBcAMP (0.5 mM) was also tested on the various deletion
mutants. As shown in Figure
6
A and B, deletion up to the
Sma
I site (position -139) does not block the response to DBcAMP. However, deletions that
remove the CRE [
Aat
II, [Delta](
Avr
II-
Aat
II) and [Delta]
Aat
II] result in a loss of responsiveness (Fig.
6
B and C). Significantly, the [Delta]
Aat
II construct (in which the CRE alone was destroyed by opening the promoter
construct at the overlapping
Aat
II site, blunting by fill-in with Klenow fragment and reclosing of the vector) has the same basal
activity as the wild-type promoter yet fails to respond to DBcAMP (Fig.
6
C).
The human
QM
gene has been mapped to a single cosmid within a 450 kb contig on Xq28 that was
generated by probing an Xq28-specific cosmid library with probes for CpG islands (
16
). In all cases, the CpG island probes mapped to the 5'-ends of genes contained within the contig (
16
). For
QM
, the presence of a CpG island was inferred by the presence of a cluster of rare
restriction enzyme sites 5' of the gene (
15
,
16
). The sequencing data presented here shows that the
QM
promoter, as well as the first 1 kb of transcribed sequence, is G/C rich
(averaging 65% G/C) and lacks the usual bias seen against CpG dinucleotides.
This confirms the presence of a CpG island at the 5'-end of the
QM
gene.
It has been estimated that up to 50% of all human genes are associated with CpG
islands (
20
). Most of these are housekeeping genes, all of which have CpG islands (
21
). However, 40% of all sequenced tissue-restricted genes (e.g. [alpha]-globin) are also associated with CpG islands (
17
,
22
). Whereas bulk vertebrate DNA is highly methylated, CpG islands are almost
never methylated at cytosines
in vivo
(except on the inactivated mammalian X chromosome and, in some instances, of
imprinting;
21
,
23
). Thus, CpG islands appear to be remnants of ancestral invertebrate DNA which,
unlike vertebrate DNA, is mostly unmethylated at cytosine residues and shows no
bias against CpG dinucleotides (
23
). It is the methylation of CpG within the vertebrate genome that accounts for
the observed paucity, since 5-methylcytosine is prone to mutation via deamination to thymidine (
24
). Thus, methylated CpG dinucleotides become replaced over time by TpG and CpA
dinucleotides. By remaining unmethylated, CpG islands escape this mutational
loss and so are not biased against CpG. Why vertebrates should have methylated
most of their DNA except for these short stretches that overlap the 5'-ends of many genes remains unclear. Their localization to the 5'-ends of genes is provocative. However, it is not clear whether
transcription at these sites is the result or the cause of the lack of
methylation. Thus, it remains to be explained why only some genes have islands
and why the islands of tissue-specific genes (e.g. [alpha]-globin) remain unmethylated even in tissues where they are
not expressed.
An interesting feature of many genes associated with CpG islands is the lack of
TATA and CCAAT box sequences. The lack of a TATA box in these promoters results
in a wide variation in the position of transcription initiation (
25
). The
QM
gene, however, retains a TATA box and initiates transcription from a single
site (
1
). Likewise, the promoter for triose phosphate isomerase (TPI) is also G/C rich,
but has both TATA and CCAAT boxes and initiates transcription from a single site (
26
). The mechanisms that enable some G/C-rich promoters to function in the absence of a TATA box remain to be
defined. The importance of GC content (
27
) and the presence of long palindromic sequences in the region of transcript
initiation (
28
) and polypurine/polypyrimidine tracts (
29
) have all been proposed. It would be interesting to determine if the TATA boxes
in G/C-rich promoters such as those for
QM
and
TPI
are functionally redundant or still required for efficient transcription of
these genes. If the TATA box is required, then comparison of the difference
between TATA-less and TATA-containing G/C-rich promoters may yield insights into the mechanism of TATA-independent transcription in G/C-rich promoters.
Consistent with its high G/C content, there are multiple putative GC boxes
within the
QM
promoter. These are binding sites for members of the Sp1 family of transcription
factors. In the
QM
promoter, most of these sites are clustered in the 5'-half of the sequence, which is the most G/C rich. The clustering of
distal GC boxes is quite common (
30
). In addition, a single GC box is often found in the proximal promoter (
30
), as is seen here for
QM
(Fig.
1
). Synergism between adjacent Sp1 sites, as well as between distal and proximal
sites, can occur through the formation of multimeric Sp1 complexes, dependent
on activation domains in Sp1 (
30
). This may be occurring in the
QM
promoter. Deletion analysis shows that the region containing the distal GC boxes accounts
for 60% of basal transcription activity. Moreover, an internal deletion [[Delta](
Avr
II-
Aat
II) construct] that removes proximal sequences, including the proximal GC box, results in almost no activity from
the promoter, even though the distal GC boxes remain. This was the case in both
NIH3T3 and SL2 cells transiently transfected with an Sp1 expression construct.
One caveat, however, is that this deletion alters the spacing between the
distal elements and the TATA box. It is possible that other elements may also
be involved in the apparent synergy between proximal and distal elements. In
particular, the 43 bp region between the
Kpn
I site at -182 and the
Sma
I site at -139 seems to be important. This region is also lost in the [Delta](
Avr
II-
Aat
II) construct. Moreover, while deletion up to the
Kpn
I site removes all GC boxes, the promoter still retains some 15% of its full
activity. Yet, this is all lost when the promoter is further deleted up to the
Sma
I site (see Fig.
3
). Finer deletion analysis, site-directed mutagenesis and EMSAs should further resolve the functions and
interactions of these elements.
Additional support for the functional activity of at least some of the GC boxes
has come from gel shift analysis using various restriction fragments of the
QM
promoter (presented in Fig.
4
) and from luciferase assays done on the promoter fragments transfected into
insect cells in the presence or absence of Sp1 (Fig.
3
). The EMSA studies revealed GC box-specific binding both to the proximal GC box (box 6, see Fig.
4
A), present on a
Pvu
II-
Sma
I fragment, and to an
Fsp
I-
Pvu
II fragment covering GC boxes 2-4. A
Pst
I-
Bgl
I fragment containing the most distal GC box (box 1, see Fig.
4
A) did not show GC box-specific binding. Thus, this site may not be functional. From the present
data, it is not possible to determine which GC box(es) in the
Fsp
I-
Pvu
II fragment is responsible for the binding seen with this fragment.
Interestingly, all the
QM
promoter GC boxes, except the non-functional distal-most element, share the same core sequence (5'-GGGCGG-3'). This sequence, recognized by the second
and third zinc fingers of Sp1 and related proteins, is the most critical
determinant of binding (
31
). Thus, it is possible that all of these sites (
2
-
6
) are functional. Work ongoing in the laboratory will address this question.
Although our gel shift competition experiments demonstrate specific binding to
the
QM
GC boxes, we were unable to identify the factor binding to these sequences in
the gel shifts. However, the results of the promoter activity experiments
obtained using insect cells not only confirm the functional role of the distal
and proximal GC boxes in regulating
QM
transcription, but also support the argument that Sp1 is able to transactivate
QM
promoter activity through these sites.
The
QM
promoter was found to be responsive to cAMP. Sequence analysis revealed the
presence of a near-consensus CRE in the proximal promoter (5'-TGACGTCT-3'), as well as five putative AP-2 sites. Responses to cAMP can occur
through both types of sites (
32
). In the case of the
QM
promoter, only the CRE appears to respond to cAMP, since deletion of this
element, alone, abolishes the cAMP response. In contrast, 5'-end deletions of the promoter that remove the AP-2 sites but which retain the CRE do not affect cAMP
responsiveness. Gel shift data further support the functionality of the
QM
CRE. The wild-type
QM
CRE generates an identical pattern of shifted complexes with nuclear extract as an oligomer encoding a consensus CRE, whereas an oligomer
encoding a mutant
QM
CRE does not show this pattern of binding. Moreover, the wild-type
QM
sequence can compete with the consensus CRE for binding of the CRE-specific factors, but the mutant sequence cannot. We have also found that
the wild-type, but not the mutant,
QM
CRE can bind to the recombinant DNA binding/dimerization domain of CREB (Santa
Cruz) (data not shown). It is interesting that the AP-2 sites do not appear to function in the cAMP response of the
QM
promoter. There are at least two other promoters that contain AP-2 sites and a CRE in which the cAMP response is found to occur only
through the CRE (
33
,
34
). In contrast, there are also reports of promoters that have consensus CRE
sequences that bind CREB in gel shift experiments but which do not respond to
cAMP (
34
). Sometimes, this is cell type dependent, e.g. the rat cytochrome c promoter
has a consensus CRE that functions in NIH3T3 but not COS cells (
35
). It is becoming clear for other sites too that context is important in
determining activity. Spacing as well as the presence of other neighboring
sites can be important in determining the activity of a given site (
36
).
In addition to previously identified binding sites, our analysis also revealed a
33 bp near-perfect palindrome ~900 bp upstream of the transcription start site. Removal of this
element resulted in a mild but significant increase (>95% confidence by
Student's
t
-test) in promoter activity, suggesting a potential repressor activity may
bind to this site. It will be of interest to determine the nature of the
factor(s) binding here. As discussed above, potentially novel positive elements may also be present in the 43 bp region between the
Kpn
I and
Sma
I sites.
In summary, this study has revealed several mechanisms for controlling activity
of the
QM
promoter, including both known and novel elements. The interactions between the
distal Sp1-responsive GC boxes and proximal promoter elements highlight the
considerable potential for synergy between different promoter elements and,
thus, support the growing understanding that transcription factor binding sites
do not necessarily function independently, but can have significant functional
interactions with each other, depending on their context within the promoter.
Presumably, such interactions are important in increasing the subtlety of
transcriptional regulation. The potential for regulation of the
QM
promoter is evident. How this potential for regulation determines the impact of
this essential eukaryotic gene on cell growth and differentiation remains to be
determined.
The authors thank Richard Schuppek for his excellent technical assistance and Dr
H.Mangalam for the use of his transcription factor database. The cosmid
containing the genomic
QM
sequence from Xq28 was a kind gift of D.Toniolo (Instituto di Genetica
Biochemica ad Evoluzionistica, Pavia, Italy). This work was funded by The
National Cancer Institute, National Institutes of Health grant GM 20379 and a
Post-doctoral Fellowship (KPS), The Norwegian Cancer Society (JIJ), a National
Science Foundation Graduate Fellowship (TML) and a University of California
Cancer Research Coordinating Committee Post-doctoral Associate Award (16137; AAF/EJS).
REFERENCES
Return