ABSTRACT
Metaxin (
Mtx
) is an essential nuclear gene which is expressed ubiquitously in mice and
encodes a mitochondrial protein. The gene is located upstream and is
transcribed divergently from the thrombospondin 3 (
Thbs3
) gene; 1352 nucleotides separate the putative translation start sites. Although
the
Mtx
and
Thbs3
genes share a common intergenic region, transient transfection experiments in
rat chrondrosarcoma cells and in NIH-3T3 fibroblasts demonstrated that the elements required for expression of
the
Mtx
gene are situated within a short proximal promoter and have no major effect on
the transcription of
Thbs3
. The metaxin -377 bp promoter contains four clustered GC boxes between nucleotides -146 and -58 and an inverted GT box between nucleotides -152 and -161, but does not contain TATA or CCAAT
boxes. Like many genes regulated by a TATA-less promoter, the transcription start site of metaxin is heterogeneous.
The major start site is only 13 bp upstream from the putative translation start
site. Electrophoretic mobility shift, competition and supershift assays showed
that the ubiquitous transcription factor, Sp1, and, to a lesser extent, the Sp1-related protein, Sp3, bind to four of these Sp1-binding motifs. Co-transfection of metaxin promoter-luciferase constructs and an Sp1 expression vector into
Schneider
Drosophila
cells, which do not synthesize Sp1, demonstrated that the metaxin gene is
activated by Sp1. Deletion of the four upstream Sp1-binding elements, on the other hand, demonstrated that these motifs are
superfluous in context of the larger
Mtx
promoter. Thus, despite the potential for common regulatory mechanisms, the
available evidence indicates that the
Mtx
minimal promoter does not significantly affect
Thbs3
gene expression.
The mouse metaxin (
Mtx
) gene has been mapped to chromosome 3E3-F1 and is closely linked to the episialin (
Muc1
), glucocere- brosidase (
Gba
) and thrombospondin 3 (
Thbs3
) genes (
1
,
2
). The transcriptional start site of Muc1, which encodes a tumor- associated polymorphic epithelial mucin (
3
,
4
), is located 2.3 kb downstream from the polyadenylation signal of
Thbs3
(
5
), while the
Gba
polyadenylation signal is situated only 6 kb upstream from the 5'-end of
Thbs3
(
2
). Mutations within the
GBA
gene are responsible for Gaucher disease, the most common lysosomal storage disorder in humans
(reviewed in
6
), while thrombospondin 3 is a secreted extracellular matrix glycoprotein of unknown function (
7
). The fourth transcription unit in this region, the metaxin gene, is situated
within the 6 kb region separating the glucocerebrosidase and thrombospondin 3
genes. Thus,
Mtx
and
Gba
are transcribed convergently and
Mtx
and
Thbs3
are transcribed divergently (
2
). The major polyadenylation sites of
Mtx
and
Gba
are only 431 bp apart, while the putative translation start sites of the
Mtx
and
Thbs3
genes are separated by 1.4 kb of DNA. It is, therefore, possible that the
primary transcripts of the
Mtx
and
Gba
genes overlap and that
Mtx
and
Thbs3
are regulated by common promoter elements.
In an attempt to establish a mouse model for the mild form of Gaucher disease,
an A to G substitution was introduced into exon 9 of the mouse
Gba
gene by homologous recombination in embryonic stem cells. The selectable
marker, phosphoglycerate kinase-neomycin, was inserted downstream from the
Gba
gene, within, as it turned out, the terminal exon of metaxin (
2
). Homozygous mice that contain the disrupted
Mtx
gene die early during gestation (
2
). These findings strongly suggested that the metaxin gene was required for
embryonic development and that it has an essential function. Unlike the
Thbs3
gene, which has a more limited distribution (
5
,
8
,
9
), the metaxin gene is expressed ubiquitously in murine tissues with the highest
levels detected in the kidney (
2
). The 1.3 kb metaxin mRNA encodes a protein of 317 amino acids. The 35 kDa
protein has been localized to the mitochondria and cytosol (
10
), and preliminary data suggest that metaxin is part of the mitochondrial import
machinery (L.C. Armstrong, T. Komiya, K. Mihara and P. Bornstein, manuscript in
preparation). Furthermore, metaxin has significant sequence similarity with the
yeast mitochondrial import protein, Mas37p, (
10
) and a mitochondrial localization sequence has also been identified within the
C-terminus of the protein (L.C. Armstrong, T. Komiya, K. Mihara, and P. Bornstein, manuscript in preparation). This finding strengthens the hypothesis that metaxin is involved in
mitochondrial protein import.
An increasing number of eukaryotic genes, that are in a head-to-head orientation and are transcribed in opposite directions, have
been characterized. These gene pairs include, among others, the
COL4A1
and
COL4A2
genes (
11
,
12
), the human dihydrofolate reductase and mismatch protein 1 genes (
13
), the chicken glycinamide ribonucleotide transformylase and 5-aminoimidazole ribonucleotide carboxylase genes for
de novo
purine nucleotide synthesis (
14
), and the human histidyl-tRNA synthetase (HRS) and HRS homologous genes (
15
). The transcription start sites of these divergent genes are separated by fairly short intergenic regions,
ranging from 114 to 229 bp. The transcriptionally divergent yeast histone H2A and H2B genes (
16
),
Gal1
and
Gal10
(
17
), and Drosophila yolk protein genes,
yp1
and
yp2
(
18
), on the other hand, are separated by much larger intergenic regions of between
606 and 1225 bp in length. The majority of these genes have related functions and are coordinately
regulated by bi-directional promoters. However, there are examples of divergently
transcribed genes which are regulated by a bi-directional promoter and encode proteins that have no apparent
relationship, like the human dihydrofolate reductase and mismatch protein 1
genes (
13
).
Since the divergently transcribed metaxin and thrombospondin 3 genes could share
common
cis
-acting elements, the minimal promoter region in metaxin was identified and
partially characterized in this study. Promoter elements in metaxin were mapped
to the 377 bases upstream from its translation start site and were shown to
contain four functional Sp1-binding motifs. Three of the distal functional Sp1-binding elements were nevertheless shown to be superfluous when
deleted from the -1.7 kb metaxin promoter, since additional upstream elements appeared to
compensate for their absence. The influence of this minimal metaxin promoter on
thrombospondin 3 gene expression was also studied, and no major effects were
detected.
A 2.3 kb
Hin
dIII-
Bam
HI fragment, containing the mouse metaxin-thrombospondin 3 intergenic region, the first exons of both genes and the
5'-portions of their first introns (
1
), was digested with the appropriate restriction enzymes (see Figs
3
and
7
) and subcloned into Bluescript. Double-stranded plasmid DNA was sequenced using the dideoxy chain-terminating method and the Taq DyeDeoxytm Terminator Cycle Sequencing Kit (Applied Biosystems), as
described by the manufacturer. The samples were analyzed with an Applied
Biosystems model 373A automated sequencer and the resulting sequences were
analyzed with the GENEPRO (Riverside Scientific, Seattle) or the Genetics
Computer Group (GCG) programs.
Poly A+ RNA was extracted from 4-week-old female BALB/c mouse liver and kidney by established methods, and
the transcription start site of the metaxin gene was determined using 5'-rapid amplification of cDNA ends (5'-RACE). Briefly, cDNA was synthesized from 2 [mu]g mouse liver or kidney poly A+ RNA using an
oligonucleotide primer (5'-GGAAAGGGCATAGCCTCTGCATATC-3') corresponding to sequences within exon 5 of the
metaxin gene (
2
). The cDNA was purified and ligated to the anchor oligonucleotide using the
AmpliFINDERtm RACE Kit (Clontech), as described by the manufacturer. The anchor-ligated metaxin cDNA or mouse liver 5'-RACE-Readytm cDNA (Clontech) was PCR amplified using
20 pmol of anchor primer, 20 pmol of primer (5'-GATGGGTGATCTTGTCTGGCAC-3') corresponding to sequences within exon 3, and the
Stoffel fragment of AmpliTaq DNA polymerase (Perkin Elmer) in a final reaction
volume of 100 [mu]l containing 10 mM Tris-HCl (pH 8.3); 10 mM KCl; 2.5 mM MgCl
2
and 0.2 mM each dATP, dCTP, dGTP and dTTP. Two [mu]l of the primary PCR reaction mixture was re-amplified as above with 20 pmol each of the anchor primer and a nested
metaxin primer (5'-CTAGGATCCAGCACGGCCAGACTATCCAG-3') corresponding to sequences within exon 1. Thirty
five cycles, denaturing at 95oC for 1 min, annealing at 55oC for 1 min and, extending at 72oC for 2 min, with a final extension time of 7 min were used for
both the primary and secondary PCR reactions. The secondary PCR products were
cloned into pUC19 and 96 positive clones were analyzed by restriction
digestion. Selected clones were sequenced as described above.
Rat chondrosarcoma cells (RCS) were a gift from Dr J. Kimura and were originally
obtained from a rat Swarm chrondrosarcoma as described in Choi
et al
. (
19
) and Mukhopadhyay
et al
. (
20
). RCS cells were cultured as monolayers in high glucose Dulbecco's modified
Eagle's medium (DMEM), supplemented with 20% heat-inactivated fetal calf serum, 2 mM l-glutamine, 100 U/ml penicillin, 100 [mu]g/ml streptomycin and 0.25 [mu]g/ml amphotericin B. NIH-3T3 cells (ATCC CRL-1658) were cultured in low glucose DMEM
containing the above supplements, except that 10% heat-inactivated fetal calf serum was added to the medium. Schneider Drosophila
Line 2 (SL2) cells (ATCC CRL-1963) were cultured in Schneider Drosophila Medium (Gibco) containing 10% heat-inactivated fetal bovine serum (Sigma, F-3018), 2 mM l-glutamine, 100 U/ml penicillin, 100 [mu]g/ml streptomycin and 0.25 [mu]g/ml amphotericin B at 25oC.
The
Mtx
and
Thsb3
promoter-luciferase reporter gene constructs were prepared by digesting the 2.3 kb
Hin
dIII-
Bam
HI fragment of clone 11 (
1
) with the appropriate restriction enzymes (see Figs
3
,
6
and
7
) and ligating the fragments into the pGL2-Basic vector (Promega).
Rsa
I and
Cel
II sites situated immediately upstream from the
Mtx
and
Thbs3
translation start sites, respectively, were used to define the 3'-ends of the promoters.
Sub-confluent NIH-3T3 and RCS cells were transiently transfected with 10 [mu]g of the promoter-luciferase plasmid constructs in 60 mm Petri dishes using
the calcium phosphate-DNA precipitation method (
21
). To control for variations in transfection efficiency, the cells were co-transfected with 2 [mu]g of a [beta]-galactosidase construct in which the gene is driven by the
SV40 promoter and enhancer (Promega). SL2 cells were transfected with 5 [mu]g of the luciferase construct in 60 mm dishes using the calcium phosphate-DNA precipitation method (
22
). These cells were co-transfected with 4 [mu]g of the plasmid [delta]F-gal containing the [beta]-galactosidase gene driven by the Drosophila
hsp70 core promoter and the transposable element F enhancer-like sequences to control for transfection efficiency and with 1 [mu]g of either the human Sp1 expression vector, pPacSp1, or the parent vector pPac0. Plasmids pPacSp1 and pPacO were a gift from Dr R. Tjian (
23
), while plasmid [delta]F-gal was a gift from Dr P. De Nocera (
24
). Cell lysates were prepared from the transfected eukaryotic and insect cells
using the freeze-thaw method (
21
). The luciferase activity in the extracts were measured using the Luciferase
Assay System (Promega) as described by the manufacturer, while the [beta]-galactosidase activity was assayed using
O
-nitophenyl-[beta]-d-galactopyranoside as a substrate (
21
), or the Galacto-Light Chemiluminescent Reporter Assay (Tropix) as described by the
manufacturer.
The various DNA fragments used as probes in EMSA (Fig.
4
A) were prepared from the mouse metaxin proximal promoter, radiolabeled with the
Klenow fragment of DNA polymerase. Nuclear proteins were isolated from NIH-3T3 and RCS cells using the method of Lee and Green (
25
) and their concentrations were determined by the Bradford method (
26
) using chymotrypsin as a standard. Crude nuclear extract (4 [mu]g) was incubated in 20 mM HEPES (pH 7.9), 50 mM KCl, 0.5 mM dithiothreitol, 0.2 mM EDTA, 1 mM MgCl
2
, 4% Ficoll 400 and 4 [mu]g poly dI-dC.poly dI-dC in a final volume of 20 [mu]l for 10 min at room temperature, prior to the addition of
10
4
c.p.m.
32
P-labeled probe. The incubation was continued for an additional 30 min at 4oC and the DNA-protein complexes were analyzed on 5% non-denaturing polyacrylamide gels in 0.5* TBE at 4oC. The gels were dried and exposed to X-ray film for 16 h.
In the competition assays, double-stranded oligonucleotides were prepared by denaturing complementary single-stranded oligonucleotides (synthesized by IDT®) at 90oC for 5 min and allowing them to anneal at 37oC for 60 min prior to cooling to room temperature. The unlabeled double-stranded GC box consensus (5'-ATTCGATCGGGGGGGGGCGAGC-3') or GC box mutant (5'-ATTCGATCGGTTGGGGGCGAGC-3') oligonucleotides were incubated with the nuclear proteins prior to the addition of the
32
P-labeled probe.
In the supershift assays, 1.0 [mu]l anti-Sp1 (PEP2), Sp2 (K-20), Sp3 (D-20) or Sp4 (V-20) Trans Cruztm affinity-purified rabbit polyclonal IgG (Santa
Cruz Biotechnology) was added to the reaction mixtures after the probe and
incubated at 4oC for 60 min. To test for the specificity of the antibodies, 1.0 [mu]l rabbit IgG was incubated with the appropriate blocking peptide prior
to its addition to the reaction mixture.
In a previous study we identified a novel murine gene, metaxin, which is located
within the 6 kb region separating the glucocerebrosidase (
Gba
) and thrombospondin 3 (
Thbs3
) genes on chromosome 3E3-F1 (
2
). Metaxin is situated upstream from the thrombospondin 3 gene, in a head-to-head orientation, so that the putative translation start sites of
the two genes are separated by 1352 base pairs of intergenic sequence (Fig.
1
). Analysis of the intergenic region revealed that it is G+C rich (61%), that it
contains 86 CpG dinucleotides, and that both the metaxin and thrombospondin 3
promoters lack a CCAAT and a TATA box. Comparison of the mouse and human (
27
) intergenic regions reveals an identity of 59%. However, there is a sequence of
184 bp situated between nucleotides -592 and -409 of the mouse intergenic region which is 88% identical between
the two species (Fig.
1
). The sequence of the mouse intergenic region has been deposited in the
DDBJ/EMBL/GenBank Data Libraries under Accession no. U66257.
The transcription start site of the metaxin gene was determined using 5'-rapid amplification of cDNA ends (5'-RACE) as described in Materials and Methods. The
transcription start site was shown to be heterogeneous (Fig.
2
) with the major start site being only 13 bp upstream from the putative
translation start site (seven out of 17 clones sequenced). The longest and
shortest 5'-UTR were determined to be 26 and 7 bp, respectively. As shown in
Figure
2
, four overlapping initiator (INR)-like elements were identified within the short untranslated region. All four sequences
(5'-GTGCTTC-3'; 5'-TGCTTCC-3'; 5'-CCGGGTC-3'
and 5'-TCAGAGA-3'), contained 2-base mismatches from the consensus sequence, 5'-YYANWYY-3' (
28
). The majority of the transcription start sites were mapped to the downstream
INR-like element, suggesting that this element was functionally the most
important. RT-PCR of liver or kidney poly A+ RNA using 5'-primers which flanked the most upstream and downstream
transcription start sites, or included some of these sites, confirmed the 5'-RACE results (data not shown).
To determine the minimal promoter elements required to stimulate transcription
of the metaxin gene, several 5'-deletion fragments from the thrombospondin 3-metaxin intergenic region were cloned upstream from the
luciferase reporter gene in the metaxin direction (Fig.
3
). In addition to the intergenic region, the largest construct (-1661) also contained exon A and the 5'-region of intron A of the thrombospondin 3 gene. The various
metaxin promoter-luciferase constructs were co-transfected, together with the [beta]-galactosidase reporter gene, into a rat chondrosarcoma
cell line (RCS) and into NIH-3T3 fibroblasts. Northern blot analysis and/or RNase protection assays
showed that the metaxin and thrombospondin 3 genes were transcribed at
relatively high levels in RCS cells, while only trace amounts of the metaxin
message were detected in NIH-3T3 cells (data not shown). Similar luciferase activity was obtained when
all of the 5'-deletions, from the
Bam
HI site, 1661 bases upstream from the metaxin translation start site, to the
Sma
I site at -377, were assayed in both cell lines (Fig.
3
). There was, however, a slight decrease in luciferase activity when the -635 (
Hph
I) promoter construct was assayed in both RCS (
P
= 0.04) and NIH-3T3 (
P
= 0.002) cells, suggesting that factors which bind upstream from the
Sma
I site play a minor role in the regulation of the metaxin gene. Further analysis
of the -377 and -138 bp proximal promoters showed that there was a significant
difference in expression of luciferase when the activities of these constructs
were compared with that of the -83 bp promoter construct. Thus, the major elements responsible for
activating the metaxin gene are primarily situated within the proximal promoter
between -138 and -83 bp upstream from the translation start site.
A comparison of luciferase activity in extracts prepared from cells transfected
with the largest metaxin promoter construct, with those transfected with the
luciferase gene driven by the strong SV40 promoter, showed that the relative
activity of the metaxin promoter was 64.3 +- 5.9% (n = 4) and 38.9 +- 6.6% (n = 4) in RCS and NIH-3T3, cells respectively. These findings suggest that the
metaxin promoter functions as a fairly strong promoter in both cell lines.
However, the constructs that were tested did not appear to be dramatically less
active in NIH-3T3 fibroblasts, which only produce trace amounts of metaxin mRNA, than in
RCS cells.
Cis
-acting elements situated upstream or downstream from the -1661 bp promoter and/or post-transcriptional mechanisms may be responsible for down-regulating metaxin in NIH-3T3 fibroblasts.
Since the -377 bp proximal promoter contained all the major
cis
-acting elements required for directing expression of the metaxin gene in transfected cells, candidate regulatory elements were
identified within the minimal promoter by searching the transcription factor
databases, TFD sites (release 7.4) (
29
) and Transfac (release 2.2) (
30
). Four putative GC boxes, the proximal two in the positive orientation and the
distal two inverted, were identified within the promoter between -146 and -58 nucleotides upstream from the translation start site (Fig.
1
). The three downstream GC boxes all contained two base mismatches when compared
with the Sp1 decanucleotide consensus sequence, 5'-KGGGCGGRRY-3' (
31
). The fourth upstream GC box, on the other hand, contains only a single base
mismatch, making it more similar to the Sp1 consensus sequence. A fifth
potential Sp1-binding motif, an inverted GT box (
32
), was also identified between nucleotides -152 and -161. This element also contains two base mismatches from the
published sequence. A potential inverted metal responsive element (MRE) was
identified downstream from the Sp1 elements between nucleotides -54 and -46 (
33
). Furthermore, a C/EBP (-354 to -346), an inverted NF-[kappa]B (-252 to -244) and an AP-2 (-191 to -184) consensus
sequence (
34
-
36
) were identified upstream from the five Sp1-binding motifs.
Since transient transfection experiments showed that the -138 metaxin promoter-luciferase construct, which contained three out of the four GC
boxes, had over 70% of the activity of the larger promoter constructs (Fig.
3
), we decided to test whether GC box-binding proteins were the main factors responsible for activating the
metaxin gene. Accordingly, the 192 bp
Bsu
36I-
Rsa
I, 133 bp
Sac
II-
Rsa
I and 74 bp
Bgl
I-
Rsa
I fragments of the metaxin proximal promoter were radiolabeled, as described in
Materials and Methods, and used as probes A, B and C, respectively, in
electrophoretic mobility shift assays (EMSA) (Fig.
4
). Probes A, B and C contained four, three and one intact GC boxes,
respectively. Probe A also contained the GT box and the AP-2 consensus sequence. All three probes contained the MRE. Similar DNA-protein complexes were observed when RCS or NIH-3T3 nuclear extracts were assayed with these probes (data not
shown). Two complexes, a major complex and a faster migrating minor complex,
were identified (indicated by arrowheads) when RCS nuclear extracts were
assayed with probes B or C (Fig.
4
B). In addition to these two complexes, a second more slowly migrating minor
complex was identified when nuclear extracts were assayed with probe A. The
additional faster migrating minor complexes shown in the figures were not
reproducible.
We next tested the ability of increasing molar excesses of unlabeled double-stranded GC box consensus and mutant oligonucleotides to compete with
32
P-labeled probes A, B or C for complex formation with RCS nuclear extracts
in EMSA (Fig.
4
B). Figure
4
B shows that the GC box consensus sequence specifically inhibited the formation
of the major and two minor complexes when RCS nuclear extracts were assayed
with probe A. Similarly, the GC box consensus oligonucleotide inhibited the
formation of both the major and minor complexes when nuclear extracts were
assayed with probes B or C. No detectable competition was observed when the mutant GC box oligonucleotide was used in the assays. The GC box consensus oligonucleotide also specifically
inhibited the formation of the major and minor complexes when NIH-3T3 nuclear extracts were assayed with probes A and B (data not shown).
These findings suggest that
trans
-acting factors present in RCS and NIH-3T3 nuclear extracts bind to one or more of the GC boxes within the
metaxin proximal promoter.
To date, at least three factors, Sp1 and two related proteins Sp3 and Sp4, have
been shown to bind GC boxes with similar specificities and affinities (
37
,
38
). All three factors are also able to bind to the GT box. A fourth member of the
Sp transcription factor family, Sp2, is only able to interact with the GT box (
38
). To distinguish which member(s) of this transcription factor family binds to
the metaxin promoter, anti-Sp1, Sp2, Sp3 and Sp4 polyclonal antibodies were used in supershift
assays. As shown in Figure
5
, anti-Sp1 supershifted the major complex when RCS nuclear extracts were assayed
with probes A and B. Prebinding of the antibody with a control peptide
containing the anti-Sp1 epitope prevented the major complex from supershifting in EMSAs (data
not shown). The anti-Sp3 antibody, on the other hand, resulted in the supershift of the minor
band(s) when nuclear extracts were assayed with probe A or B. However, there
was no detectable shift in any of the complexes when anti-Sp2 or Sp4 was assayed, suggesting that mainly Sp1 and, to a lesser extent
the transcriptional repressor, Sp3, are responsible for the formation of the
observed DNA-protein complexes. Similar results were obtained when NIH-3T3 nuclear extracts were assayed with the antibodies (data not
shown).
The functional studies in Figure
3
clearly suggest that more than one of the Sp1-binding motifs are involved in regulating the metaxin gene, and binding
experiments with purified protein showed that four out of the five Sp1-binding motifs were capable of binding Sp1
in vitro
(data not shown). Schneider
Drosophila
line 2 (SL2) cells, which do not contain endogenous Sp1, were therefore
transfected with metaxin-promoter luciferase constructs in the presence or absence of an insect
cell Sp1 expression vector. As shown in Figure
6
A, there was a marked increase in luciferase activity when SL2 cells were co-transfected with metaxin-promoter constructs and the Sp1 expression vector, as compared with
the activity in cells transfected with the metaxin constructs alone. The fold
increase in activities from the -83 to the -138, from the -138 to the -377, and from the -377 to the -1661 promoter fragments were, however,
not similar to that seen in RCS and NIH-3T3 cells. The increase in luciferase activity in the insect cells
appeared to be proportional to the number of Sp1-binding elements present in the
Mtx
promoter constructs (Fig.
1
). In addition, an insect-specific transcriptional activator(s) could activate the
Mtx
promoter constructs in insect cells, or the cells could lack a transcriptional
repressor(s). An insect-specific DNA-protein complex was observed when SL2 nuclear extracts were assayed
with probe A (data not shown). In any event, the data strongly suggest that Sp1
is a transcriptional activator of the metaxin gene.
To test whether the Sp1-binding elements were critical for
Mtx
promoter activity, internal deletion mutants of the -1661 (
Bam
HI) promoter were generated and assayed (Fig.
6
B). As shown in Figure
6
B, there was no significant decrease in luciferase activity when the -1661 ([Delta]
Bsu
36I-
Sac
II) promoter construct, in which the distal GC box and GT box had been deleted,
was assayed in RCS and NIH-3T3 cells. There was however, only a slight decrease in luciferase
activity when the -1661 ([Delta]
Bsu
36I-
Ava
II) promoter construct, in which the four upstream Sp1-binding motifs were deleted, was assayed in both RCS and NIH-3T3 cells, suggesting that factors which bind upstream from the Sp1- binding elements are able to activate the
Mtx
gene by compensating for the deletion of the GC and GT boxes. To test this hypothesis, -377 ([Delta]
Bsu
36I-
Sac
II) and -377 ([Delta]
Bsu
36I-
Ava
II) metaxin promoter constructs were assayed. Except for a C/EBP and an NF-[kappa]B consensus sequence (Fig.
1
), all of the other potential upstream
cis
-acting elements were deleted in these constructs. As shown in Figure
6
B, there was no significant decrease in luciferase activity when the -377 and the -377 ([Delta]
Bsu
36I-
Sac
II)
Mtx
promoter constructs were assayed in both RCS and NIH-3T3 cells, suggesting that the distal GT and GC boxes are not important
for
Mtx
gene regulation. This finding is consistent with the data shown in Figure
3
, where there was also no significant difference in activity between the -377 and the -138
Mtx
promoter constructs, which contained five and three Sp1-binding elements, respectively. There was however a significant decrease
in luciferase activity when the -377 ([Delta]
Bsu
36I-
Ava
II) promoter construct, containing only one proximal GC box and the C/EBP
consensus sequence, was assayed in RCS (
P
< 0.001) and NIH-3T3 (
P
= 0.013) cells. The activity of the -83 bp promoter construct, which contains only one intact GC box, was
about half that of the -377 ([Delta]
Bsu
36I-
Ava
II) construct. This slight increase in activity with the -377 ([Delta]
Bsu
36I-
Ava
II) promoter construct could be due to a cooperative interaction between Sp1 and
C/EBP (
39
). Taken together these data suggest that the proximal three GC boxes are
involved, but are not critical, in regulating the metaxin gene. Factors which
bind upstream from the minimal
Mtx
promoter, but which normally may not have a major effect on
Mtx
gene expression (Fig.
3
), can compensate for deletions of any or all of the four upstream Sp1 motifs.
In order to test whether the Sp1-binding elements within the minimal metaxin promoter were also able to
regulate the thrombospondin 3 gene, various thrombospondin 3 promoter-luciferase reporter gene constructs were tested in RCS and NIH-3T3 cells (Fig.
7
). The largest construct, -2034
Thbs3
promoter-luciferase, contains the entire intergenic region, exon 1 of metaxin and
the 5'-part of the metaxin first intron. The second construct, -1348
Thbs3
promoter-luciferase, contains the entire metaxin thrombospondin 3 intergenic region, which includes the minimal metaxin promoter. In the -1150 bp
Thbs3
promoter-luciferase construct, the Sp1-binding elements in the metaxin minimal promoter were deleted. As shown in
Figure
7
, there was a minor, but significant, decrease in luciferase activity when the -1348 and -1150
Thbs3
promoter-luciferase constructs were assayed in both RCS (
P
= 0.035) and NIH-3T3 (
P
= 0.002) cells, suggesting that the minimal metaxin promoter does not, at least
in these cell lines, play a major role in regulating the expression of the
thrombospondin 3 gene. Further deletions of the -1150
Thbs3
promoter-luciferase construct, resulted in a gradual decrease in luciferase
activity in both cell lines. These experiments also showed that the -2034 thrombospondin 3 promoter construct was only about a sixtieth or a
tenth as active as the -1661 metaxin promoter in RCS and NIH-3T3 cells, respectively. It is possible that an as yet unidentified
enhancer element(s) situated further upstream, within the body of the metaxin gene, or downstream within the body of the thrombospondin 3 gene, contributes
to the expression of the
Thbs3
gene. A potential enhancer element for the human
THBS3
gene has recently been identified within the body of the human
MTX
gene (J. Silver, M. Collins, and P. Bornstein, unpublished data).
The mouse metaxin and thrombospondin 3 genes are arranged in a head-to-head orientation on chromosome 3E3-F1 so that their putative translation start sites are
separated by 1352 nucleotides (
2
). Since these divergently transcribed genes are regulated by a common promoter
region, the metaxin minimal promoter was characterized and its effect on
Thbs3
gene expression was studied. Transient transfection experiments in
Mtx
-expressing rat chondrosarcoma cells and NIH-3T3 fibroblasts demonstrated that the elements required for
expression of the
Mtx
gene are situated within the short G+C rich proximal promoter region containing
five clustered Sp1-binding motifs, but lacking a CCAAT or a TATA box. Electrophoretic
mobility shift, competition and supershift assays demonstrated that Sp1, and to
a lesser extent the Sp1-related protein, Sp3, bind to these elements
in vitro
. Functional studies in SL2 insect cells, co-transfected with metaxin promoter constructs in the presence or absence of
an Sp1 expression vector, confirmed that Sp1 activates the
Mtx
gene. The four upstream Sp1-binding motifs were also shown to be superfluous when deleted from the -1.7 kb promoter. Additional elements probably compensate for these
deletions.
Sp1 is a ubiquitous transcription factor which is responsible for activating
many cellular and viral genes (
31
). Like metaxin, many eukaryotic genes with TATA-less promoters contain multiple Sp1-binding motifs arranged in tandem and in close proximity within the
proximal promoter (
40
). Sp1 can activate many of these promoters synergistically by Sp1-Sp1 interactions and these interactions are an important mechanism for modulating the expression of this class of genes (
40
,
41
). These genes are further regulated by Sp1-related factors, such as Sp3. Hagen
et al
. (
32
) have demonstrated that the ubiquitous transcription factor, Sp3, represses Sp1-mediated transcriptional activation in a linear dose-dependent manner.
Besides the GC and GT boxes, an inverted NF-[kappa]B and an AP-2 consensus sequence were identified upstream from the Sp1-binding motifs within the -377 bp metaxin minimal promoter, between
nucleotides -252 and -244 and between -191 and -184, respectively. Both NF-[kappa]B and AP-2 are inducible transcription
factors (
35
,
36
), suggesting that metaxin gene expression could be regulated in response to
external stimuli. A highly specific cooperative interaction between NF-[kappa]B and Sp1 is responsible for the induction of the HIV-1 long terminal repeat expression (
42
,
43
). Since the NF-[kappa]B element is located upstream from the cluster of Sp1-binding elements within the metaxin proximal promoter, it is
possible that a similar mechanism may serve to regulate the metaxin gene in
response to external stimuli. A cooperative interaction between Sp1 and C/EBP
has also been previously described (
39
). Since, the
Mtx
minimal promoter also contains a C/EBP element, it is possible that C/EBP may
also play a role in regulating
Mtx
gene expression. Interestingly, the clustered Sp1-binding motifs as well as the NF-kB, AP-2 and C/EBP elements are all conserved within the human
MTX
proximal promoter (
27
), strengthening the hypothesis that these elements modulate metaxin gene
expression.
The transcription start site of the metaxin gene was shown to be heterogenous.
The major start site is only 13 bp upstream from the putative translation start
site, and the longest and shortest 5'-UTR is 26 and 7 bp, respectively. The 5'-UTR of metaxin is therefore shorter than the average
length of between 20 and 100 nucleotides for most eukaryotic mRNAs (
44
). Investigators have shown that the accuracy of translation from an initiation
AUG codon can decrease when it is very close to the transcription start site.
Initiation at a downstream AUG can also occur with increased efficiency under
these circumstances, even when the upstream AUG is in a better sequence context
(
45
,
46
). As shown in Figure
2
, there is an in-frame AUG situated 12 bp downstream from the putative metaxin translation
start site. This downstream AUG is also conserved within the human gene (
27
). In both species the upstream AUG is situated within an ideal sequence
context, 5'-A
-3
AG
Previous studies have demonstrated that metaxin is a ubiquitously expressed gene which encodes a mitochondrial protein and is essential for embryogenesis (
2
,
10
). Unlike metaxin, thrombospondin 3 has a much limited tissue distribution. The
Thbs3
gene is expressed predominately in the lung, in the hippocampus of the brain,
in cartilage, and in the gastrointestinal tract, with lower levels of
expression in other extracellular matrix-producing tissues (
5
,
8
,
9
). Thrombospondin 3, like other members of the thrombospondin protein family, probably plays a role in modulating the function of the extracellular matrix. Although these two divergent genes
encode proteins with apparently different functions, it was still possible that
common elements regulated
Mtx
and
Thbs3
. Thus, the divergent human dihydrofolate reductase and mismatch protein 1 genes
encode proteins which have no apparent relationship, but are nevertheless
regulated by a bi-directional promoter (
13
). The -2 kb
Thbs3
promoter was shown to be relatively weak in both rat chondrosarcoma (RCS) and
NIH-3T3 cells. The promoter was about a sixtieth or a tenth as active as the
metaxin promoter in RCS and NIH-3T3 cells, respectively. There was a minor, but significant, difference in
promoter activity when the -1.3 kb and the -1.1 kb
Thbs3
promoter, in which the metaxin minimal promoter had been deleted, was assayed
in both cell lines. These findings suggest that the minimal metaxin promoter
only plays a minor role in modulating the expression of the thrombospondin 3
gene in RCS or NIH-3T3 cells. Since the thrombospondin 3 promoter was weak in both cell
lines, it is possible that an as-yet-unidentified enhancer element(s) may contribute to the expression of
the
Thbs3
gene in cells, such as RCS cells, in which this gene is expressed. If this is
the case, then the effects of the
Mtx
minimal promoter on
Thbs3
gene expression would probably be even less significant.
This work was supported in part by grant DE 08229 from the National Institutes
of Health. M.C. was supported by a Fogarty International Research Fellowship (1
F05 TW05218-01), and a Foundation for Research Development Postdoctoral Fellowship
from South Africa. We thank Lynn Law for assistance with the insect cell
expression system. We also thank members of our laboratory for helpful comments
on the manuscript.
REFERENCES
Return