|
|
|
|||||||
|
|||||||||
(Received for publication, August 10,
1994; and in revised form, December 1, 1994) From the
Human carcinoembryonic antigen (CEA) belongs to a family of
membrane glycoproteins that are overexpressed in many carcinomas; CEA
functions in vitro as a homotypic intercellular adhesion
molecule and can inhibit differentiation when expressed ectopically in
myoblasts. The regulation of expression of CEA is therefore of
considerable interest. The CEA gene promoter region between -403
and -124 base pairs upstream of the translation initiation site
directed high levels of expression in CEA-expressing SW403 cells and
was 3 times more active in differentiated than in undifferentiated
Caco-2 cells, correlating exactly with the 3-fold increase in CEA mRNA
seen in differentiated Caco-2 cells. Inclusion of additional upstream
sequences between -1098 and -403 base pairs repressed all
activity. By in vitro footprinting and deletion analyses, four cis-acting elements were mapped within the positive regulatory
region, and one element within the silencing region. Several nuclear
factors binding to these domains were identified: USF, Sp1, and an
Sp1-like factor. By co-transfection, USF directly activated the CEA
gene promoter in vivo in both SW403 and Caco-2 cells. In
addition, the levels of factors binding to each positively acting
element increased dramatically with differentiation in Caco-2 cells.
Thus the transcriptional control of the CEA gene depends on the
interaction of several regulatory elements that bind multiple specific
factors.
CEA, ( CEA is a member of the
immunoglobulin supergene family and is the prototype for its own
subfamily of closely related molecules that vary in domain composition
and tissue distribution (for a review, see (6) ). This
subfamily consists of 29 closely linked genes on chromosome 19,
including those coding for CEA itself, nonspecific cross-reacting
antigen (NCA), biliary glycoprotein (BGP), CEA gene family member 6
(CGM6), and a number of other genes with yet undetermined products
(hsCGMs)(6, 7) . As with CEA, NCA and BGP have been
shown to function in vitro as intercellular adhesion
molecules(2, 3, 6) . Cloned cDNAs for CEA,
NCA, and BGP have been used as probes to study their expression in
normal and tumor tissue(6) . Whereas CEA mRNA is present at low
levels in normal adult colon and is usually overexpressed in malignant
colon and other cancers of epithelial cell origin, NCA mRNA is found in
normal colon, lung, and granulocytes and is elevated by a greater
factor in tumors of the colon, breast, and lung(6) . Several
forms of BGP have been isolated from bile ducts, gallbladder mucosa,
and various tumors(6) . In contrast to CEA and NCA mRNAs, the
expression of two BGP mRNAs have been shown to be down-regulated in
colorectal carcinomas in comparison to normal adjacent
mucosa(8) . The increased levels of CEA have been shown not to
be due to gene rearrangements or amplification (9) but,
instead, to hypomethylation of upstream regions (9, 10, 11) and/or factor changes leading to
altered rates of transcription; post-transcriptional changes have also
been implicated(9, 12) . CEA and NCA mRNA levels
have also been investigated in the differentiating Caco-2 cell
system(12) . When cultured in vitro, between 4 and 11
days after reaching confluence, this human colon adenocarcinoma cell
line differentiates and becomes highly polarized, with tight junctions
between individual cells and a brush border membrane containing enzymes
characteristic of a fully differentiated intestinal
epithelium(13, 14) . We found that CEA transcript
levels were 3-fold higher in fully differentiated Caco-2 cells than in
undifferentiated monolayers(12) . In the present study, we have
used the differentiation control of the CEA gene in Caco-2
cells to check the biological validity of the CEA promoter
analysis. To carry out this analysis, we characterized the upstream
noncoding region of the CEA gene. A 424-bp 5` flanking sequence has
been reported to confer cell-type specific expression on a reporter
gene(15) . Numerous purine-rich sites were postulated to play a
role in transcriptional control(11) , but the exact regulatory
elements involved remained unknown. We now demonstrate that both
positive and negative elements reside within 1098 bp upstream of the
translational start site and that a 403-bp upstream sequence confers
cell type-specific and differentiation-dependent expression on the
luciferase reporter gene. We identify five nuclear factor binding sites
and three of the multiple trans-acting factors (USF, Sp1, and
Sp1-like) interacting with the stimulatory domain. In addition, we
present direct evidence that USF activates CEA gene transcription in vivo.
5` deletion mutants of the CEA gene promoter lacked the CEA
translation initiation codon and were fused immediately 5` to the
firefly luciferase (LUC) reporter gene (17) at the SmaI site of the pXP2 vector(18) . The resulting
plasmids were named according to the sizes of their respective CEA
promoter restriction fragments as shown in Fig. 1. Internal
deletion mutants of the CEA gene promoter, p1098
Figure 1:
Localization of the elements
determining CEA gene promoter activity. CEA promoter activity in
various cell lines is shown. The activity of each construct in each of
the cell lines is presented relative to the activity of the
promoterless vector, pXP2. In all cases, CEA sequences from -2 to
-124, including the 5` untranslated region and transcription
initiation site, are present in plasmid constructs. The translational
start site is at position +1. Ovals represent regions
protected in DNase I footprinting assays. Names of constructs
correspond to the sizes of the fragments tested. Sites used to create
constructs are shown on the restriction map, with precise positions
shown in Table 1. The data represent the mean ± S.D. of
three to four independent experiments, each performed in duplicate, and
corrected for protein concentration and for transfection efficiency by
the activity of the internal RSVZ
Cells were
plated at a density of 1 To assess the effect of USF on
CEA promoter activity in vivo, 10 µg of p403LUC and either
5 or 15 µg of pRSV.USF (as indicated in Table 2) were
coprecipitated with calcium phosphate and transiently transfected into
the SW403 or Caco-2 cell line. pRSV.USF was constructed by inserting
human USF cDNA (24, 25) into a pUC18-derived
expression vector driven by the RSV long terminal repeat (18) .
Rat HNF-4 cDNA in the pSG5 expression vector (26) was
generously provided by Dr. F. M. Sladek (University of California,
Riverside, CA).
DNase I footprinting assays were performed
essentially as described by LeFèvre et al.(28) with modifications as by Howell et
al.(29) . Briefly, 1-2 fmol (10,000 cpm) of DNA
probes, labeled at only one end, were incubated with 0-160 µg
of nuclear extracts for 15 min on ice; restriction enzyme-grade bovine
serum albumin (Boehringer Mannheim) was added so that equal total
amounts (160 µg) of protein were present in each reaction. Freshly
diluted DNase I (Life Technologies, Inc.) at 20 ng/30 µl reaction
volume was added for 3 min on ice. Reactions were stopped, and the DNA
was digested with proteinase K, phenolextracted, and precipitated. The
dried DNAs were suspended in formamide loading dye, and equal
counts/min were loaded on an 8% polyacrylamide, 8 M urea
sequencing gel. Sequencing reactions (23) were run alongside as
size markers. The dried gels were autoradiographed at -75 °C
with Cronex Quanta III intensifying screens (E. I. du Pont de Nemours
& Co.).
To determine equivalent amounts of
nuclear extracts from Caco-2 cells in their undifferentiated and
differentiated states, oligonucleotides binding to ubiquitous factors
such as Sp1 (31) or AP1 (32) could not be used since
the levels of these factors have been observed to change with
differentiation ( (33) and data not shown). Thus equal
concentrations of isolated nuclei were used to normalize the extracts.
Using this approach, with the preparations shown in Fig. 6, 20
µg of undifferentiated Caco-2 nuclear extract was equivalent to 18
µg of differentiated Caco-2 nuclear extract. The data shown in Fig. 6are representative of several independent experiments
using independent nuclear extracts.
Figure 6:
Analysis of Caco-2 nuclear proteins
binding to CEA regulatory elements versus differentiation.
Oligonucleotides containing the DNA sequence of three of the CEA
regulatory elements were used as probes to reveal Caco-2 nuclear
proteins which specifically recognized them and to indicate the
relative abundance of these factors in equivalent amounts of nuclear
extracts prepared from undifferentiated (U) and differentiated (D) Caco-2 cells. By using equal concentrations of isolated
nuclei before the elution of nuclear proteins, 20 µg of
undifferentiated Caco-2 nuclear extract was found to be equivalent to
18 µg of differentiated Caco-2 nuclear extract in the experiment
shown here. The indicated unlabeled oligonucleotides used as competing
DNAs assessed the specificity of the complexes formed. Arrows point to the specific complexes.
Deletion of the
-403 to -124 sequence, leaving only the transcription
initiation site and 5` untranslated region intact (p124LUC construct),
abolished all promoter activity (Fig. 1). Thus regulatory
elements responsible for the control of cell-specific expression of the CEA gene reside within this 279-bp upstream region. Inversion
of the 403-bp minimal promoter in the p403aLUC antisense construct
strongly reduced but did not abolish activity, suggesting that the
elements within the minimal promoter may use cryptic signals on either
strand for transcriptional initiation (no TATA box can be found in the
upstream sequence of the CEA gene). Inclusion of further
upstream sequences (-1098 to -403) in conjunction with the
minimal promoter, as seen with the p1098LUC construct, repressed
promoter activity markedly in SW403 cells (Fig. 1). The
-1098 to -403 region alone, placed in either orientation
before the fragment containing the transcription initiation site,
showed no activity whatsoever (constructs p1098
As seen with the SW403 cell line, the inclusion of additional
upstream (-1098 to -403) sequence also repressed promoter
activity in Caco-2 cells, and further deletions of the minimal promoter
past position 403 generally decreased activity. Only in Caco-2 cells
does the p300LUC construct give higher activity than the p403LUC
construct, however. It is thus possible that a second silencing
element, recognized by factors only in Caco-2 nuclear extracts, is
present in the -403 to -300 region. Since
Figure 2:
Localization of binding sites for nuclear
proteins within the CEA promoter using DNase I footprinting. A, restriction enzyme map of the CEA promoter showing sites
used to generate the probes used in the footprinting experiments.
Probes I-III were labeled as indicated by asterisks at
either the 5` or 3` end. The circlednumbers show the
positions of the footprints detected. B, localization of sites
within the -2 to -403 region. 5`-end-labeled probe I was
incubated as described under ``Materials and Methods,'' with
nuclear extracts from LR-73, SW403, and Caco-2 (undifferentiated
monolayers) as indicated. G refers to the G sequencing track; lanes0 contained 160 µg of bovine serum albumin
only, and lanes 10-160 show results for 10-160
µg of the indicated nuclear protein extracts. The positions of the
various footprints are indicated on the right by bars and labels; their precise sequences are listed in Table 3.
The positions of some guanine residues are shown on the left for orientation purposes. Footprints were numbered according to
their proximity to the transcription initiation site. C,
localization of binding sites with a 3`-end-labeled -403 to
-2 probe (Probe II). Footprints FP2 and FP3 are revealed using
HepG2 nuclear extract. This probe is also shown using SW403 nuclear
extract for comparison. DNase I hypersensitive sites (
Two additional footprints,
labeled FP2 and FP3, were revealed using nuclear extract from HepG2
cells and were also present, although less apparent, with extracts from
SW403 cells (Fig. 2C). Two DNase I-hypersensitive sites
are visible between FP2 and FP3. In Fig. 2D, the
-835 to -403 silencing region was used as an end-labeled
probe in DNase I footprinting experiments; only one protected region
could be detected in repeated experiments. This footprint, labeled FP5,
was more apparent in HT29 nuclear extract and was flanked by two DNase
I-hypersensitive sites. Deletion of the -1098 to -403
region containing this element led to an increase in promoter activity
in all cell lines tested (Fig. 1, p1098LUC versus p403LUC constructs).
Figure 3:
Gel mobility shift assay reveals that the
USF transcription factor binds to the FP1 regulatory element. A, synthetic, double-stranded oligonucleotide representing FP1 (Table 1) was end-labeled, incubated with 60 µg of SW403
nuclear extract and electrophoresed, showing two complexes (C1 and C2).
5-, 10-, or 100-fold molar excesses of competing oligonucleotides were
added as indicated. B, USF synthesized in vitro (see
``Materials and Methods'') was added to the FP1 probe in lanes 9-11; a 5-fold molar excess of unlabeled FP1 DNA
was added in lane10; lanes11 and 12, USF specific antibody was first added to USF protein or
SW403 nuclear extract, respectively, for 15 min at 0 °C, after
which the FP1 probe was added and incubated for another 15 min on ice.
The supershifted (s.s.) complex is indicated by an arrow on the right. C, for comparison, 40 µg of
HepG2 nuclear extract was analyzed for binding under the same
conditions as that with SW403 extract. 5- and 50-fold molar excesses of
unlabeled, double-stranded FP1 oligonucleotide were used as competitor
DNAs.
Since the GAL2 USF oligonucleotide competed effectively for
complex formation, we tested whether purified USF transcription factor
would recognize the FP1 element. The complex formed (Fig. 3B, lanes9 and 10)
comigrated with C1 from SW403 extract (Fig. 3B, lane8). As expected, specific anti-USF antibody
supershifted the DNA/USF complex (Fig. 3B, lane11) but also supershifted the C1 complex from SW403
nuclear extract (Fig. 3B, lane12).
Thus colon carcinoma cell lines clearly contain USF, which can bind to
the FP1 element in the CEA gene promoter. The presence of two complexes
in HepG2 nuclear extracts comigrating with those obtained with SW403
extract (Fig. 3C) suggests that USF is also present in
this hepatoma-derived cell line.
Figure 4:
Element FP2 is recognized by Sp1 and
another novel transcription factor. A, end-labeled,
double-stranded FP2 oligonucleotide (Table 1) was incubated with
40 µg of SW403 nuclear extract and subjected to electrophoretic
analysis. Three complexes (C1-C3) were revealed. Unlabeled,
double-stranded FP2, FP3, Sp1, and AP1 oligonucleotides (Table 1;
10- and 100-fold molar excesses) were used as competitor DNAs. B, 3.5 ng of purified Sp1 protein was incubated with FP2 probe
in lanes 12 and 13 under the same binding conditions.
Antibody specific to Sp1 was preincubated with 14 ng of Sp1 protein (lane 14) and 40 µg of SW403 nuclear extract (lane15). The Sp1 supershifted (s.s.) complex is
indicated by an arrow on the right.
Figure 5:
Element FP3 is bound by Sp1. A,
electrophoretic analysis of end-labeled, double-stranded FP3
oligonucleotide (Table 1) incubated with 40 µg SW403 nuclear
extract shows two specific complexes (C1 and C2). Unlabeled,
double-stranded FP3, FP2, Sp1, AP1, and PY oligonucleotides (Table 1; molar excesses as shown) were used as competitor DNAs. B, in lanes 15 and 16, FP3 probe and 3.5 ng
of purified Sp1 protein were incubated as described in Fig. 4.
Sp1-specific antibody was preincubated with 14 ng of Sp1 protein (lane17) and 40 µg of SW403 nuclear extract (lane18). The Sp1 supershifted (s.s.)
complex is indicated by an arrow on the right. C, 20 µg of HepG2 nuclear extract was incubated with FP3
probe under the same binding conditions for comparison
purposes.
Probes FP1,
FP2, and FP3 each produced several complexes (Fig. 6, lanes
1-12), similar to those seen with SW403 extracts (Fig. 3Fig. 4Fig. 5). The levels of the complexes
obtained were dramatically higher using differentiated (D)
than undifferentiated (U) Caco-2 nuclear extracts (Fig. 6). Thus, the levels of USF, Sp1, the Sp1-like, and the
unknown factor responsible for the C3 complex with the CEA FP2 element
all appear to increase with differentiation in Caco-2 cells. The higher
levels support our contention that an increase in the abundance of
positive factors interacting with the regulatory elements in the
minimal promoter are partially responsible for the rise in CEA
transcription observed in this differentiating system. In this study, we have delineated the basic organization of
the CEA gene promoter (summarized in Fig. 7and Table 3),
which is the first to be determined for genes of the CEA subgroup of
this human tumor marker family. Functional assays using various 5`
flanking sequences of the CEA promoter linked to the luciferase
reporter gene transfected into CEA-producing cells coupled with DNase
footprint assays revealed that the 5` upstream region contains four
positive (FP1-FP4) and one negative regulatory element (FP5). The
upstream stimulatory factor (USF)(34) , also known as the
adenovirus major late transcription factor (MLTF), was shown to bind to
the FP1 element; the positive control of CEA transcription by USF was
confirmed directly by the demonstration of specific stimulation of the
CEA promoter in vivo by a co-transfected USF-producing
plasmid. The Sp1 (38) and an Sp1-like transcription factor were
found to bind to both the FP2 and the FP3 element. Through computer
sequence analyses, FP4 was found to resemble an AP-2 transcription
factor site (30, 39) , and preliminary experiments
(data not shown) confirmed this possibility; oligonucleotides
containing AP-2 binding sites competed with the FP4 site for nuclear
factor binding and purified AP-2 protein was capable of forming a
complex with the FP4 element. Other factors binding to the FP4 site, a
third factor binding to the FP2 site and factors recognizing the FP5
silencer element remain to be identified.
Figure 7:
Schematic representation of the various
nuclear proteins binding the regulatory elements of the CEA promoter.
Those factors that have been identified are indicated. Factors binding
to the BGP promoter are also shown to allow comparison of transcription
factor complexes specific to the CEA gene promoter. The arrow signifies the major transcriptional start site. Sequences and
positions are compared in Table 3.
The biological
significance of these element and factor assignments was further tested
using the Caco-2 colonocyte system, which shows an increase in CEA mRNA
with differentiation into polarized epithelium. Since the positively
acting factor levels increased dramatically, it is thus possible that
the transcription factors identified here could also control the
expression of CEA in normal colonic epithelial cells, which show an
increase in CEA mRNA (40) and protein (41) production
during their differentiation in transit from the bottom to the top of a
crypt. The basis for transcriptional changes seen for CEA and other CEA
family members in tumors remains to be investigated. Sequence
comparisons of the CEA gene regulatory elements with the upstream
noncoding sequences of other CEA gene family members are shown in Table 3. The control of expression of this family is of
particular interest because of the unusually close alignment of the
nucleotide sequences of its members (often over 90%). PSG5 and PSG11
are members of the pregnancy-specific glycoprotein (PSG) subgroup of
the CEA gene family whose upstream sequences showed homology to the CEA
FP2 and FP3 elements only. We have also analyzed the control of
transcription of a second CEA family member, BGP (42) which,
unlike CEA(6) , can show decreased transcript levels in colon
carcinomas relative to adjacent normal tissue(8) . Comparison
of the CEA and BGP gene promoters (see Fig. 7for summary)
revealed differences that could explain their differential regulation.
Only two of the corresponding BGP gene elements could be shown to bind
nuclear factors: thus the Sp1, the Sp1-like (recognizing FP2 and FP3 in
CEA), and the silencer factors (recognizing FP5 in CEA) do not bind to
the BGP promoter, while a second factor, HNF-4(26) , as well as
USF, binds to the USF site(42) . Experiments are in progress to
determine whether the different changes in transcription of the CEA and
BGP genes seen with colon carcinogenesis can be rationalized by changes
in these factors. Although the Sp1 transcription factor has long
been characterized(31) , only recently have dramatically
increased levels of this ubiquitously expressed factor been correlated
with differentiation(33) . Genes regulated by Sp1 include the
following: fibronectin, which shows greatly inhibited expression upon
neoplastic transformation(43) ; E-cadherin, which is generally
down-regulated in tumors(44) ; and the human papillomavirus
type 18 E6-E7 oncogene, which also has an unusual Sp1
site(45) . Specific recognition of the DNA sequence is provided
for by the three zinc finger domains of Sp1(46) . Although the
core sequence is a typical GC box, substitutions are tolerated as long
as certain G residues are present for contact with the
``fingers''(46) . These contact points exist in the
FP2 and FP3 elements of the CEA gene, within the aligned homologous
sequences of the NCA gene, and within the CGM1 sequence aligned to the
CEA FP2 element (see Table 3). The USF gene has recently been
cloned and is now known to code for a ubiquitous factor with a
helix-loop-helix repeat domain and a leucine zipper(34) . Its
protein-binding interface is similar to that of Myc and
Max(47) . All Myc family members recognize an identical core
DNA target sequence of CACGTG and appear to bind to DNA as homo- or
heterodimers, dependent on specificities contained within the leucine
zipper(34, 48, 49) . Since we have directly
demonstrated that USF activates the CEA promoter in vivo, any
factors interacting with and modulating this nuclear factor should also
modulate CEA gene expression. Although purified Myc protein did not
bind to the CEA FP1 site (data not shown), it remains possible that the
heterodimer c-Myc/Max could bind to this element or to USF itself. Both
Myc/Max and USF have also been shown to bend DNA toward the minor
groove to the same angle and orientation(50) . Pognonec et
al.(51) have also demonstrated that the DNA-binding
activity of USF is regulated via a redox dependent mechanism. The
activity of the CEA promoter could therefore be partially controlled by
a complex balance between the binding of various other b-HLH proteins
to form heterodimers with USF, the redox mechanism, and competition for
DNA binding by other factors, such as HNF-4 as shown for the BGP
promoter(42) . As previously mentioned, CEA mRNA levels are
up-regulated in colon carcinomas. About 70% of colon carcinomas
overexpress the c-Myc gene as well as other members of the Myc gene
family(52, 53) , although the status of USF expression
is presently unknown. This study has identified many of the cis-acting elements involved in the transcriptional control of
the CEA gene and some of the trans-acting factors which
interact with them. USF, in particular, is involved in both CEA and BGP
control and may play an important part in the overall control system of
the CEA gene family. These assignments, coupled with further studies on
other family members, should lead to the rationalization of the
observed tissue-specific, differentiation-dependent expression of the
family. The possible deregulation of these trans-acting
transcriptional factors in colon carcinogenesis could represent the
basis for changes in the expression of CEA and other family members
seen in tumors, changes that could be instrumental in the
carcinogenetic process(2) . Ectopic expression of CEA, for
example, has been shown recently to block myogenic differentiation and
leave cells with division potential(5) . It will now be of
interest to determine whether the CEA regulatory elements identified
here are targets for the action of oncogenes and tumor suppressor
genes.
Volume 270,
Number 8,
Issue of February 24, 1995 pp. 3602-3610
©1995 by The American Society for Biochemistry and Molecular Biology, Inc.
IDENTIFICATION OF REGULATORY ELEMENTS AND MULTIPLE NUCLEAR FACTORS (*)
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
FOOTNOTES
ACKNOWLEDGEMENTS
REFERENCES
)a membrane glycoprotein first observed in
human fetal colon and colorectal cancer(1) , is a widely used
clinical tumor marker. CEA has been shown to function in vitro as a homotypic intercellular adhesion molecule (2, 3) and could thus play an important role during
development. A model for a possible carcinogenic role of CEA
overproduction in the colon has been suggested(2, 4) .
In support of this model, we have recently shown that the ectopic
expression of CEA on the surface of rat L6 myoblasts can completely
block terminal differentiation and the normal loss of proliferative
capacity(5) . These attributes make CEA an important candidate
for studies on control of gene expression.
Plasmid Constructions
Upon probing a human
epithelial genomic EMBL3 phage library (gift from N. Laskin, Cold
Spring Harbor Laboratory, Cold Spring Harbor, NY) with the 5`
untranslated and leader region of human CEA cDNA(16) , a
17.7-kilobase pair genomic clone was found to contain sequence
identical to that of the CEA coding region. In addition, 2.3 kilobase
pairs of upstream sequence were found to be virtually identical to the CEA gene (COSCEA01) sequence already published(15) ;
only five dispersed single base differences were found in 1098 bp of
sequenced DNA immediately upstream of the initiation codon. The
-3500 to +1 (translational start site of the CEA gene)
upstream region of this genomic clone was used for promoter analyses. 
279LUC and
p1098a279LUC, were constructed as follows: the AvrII
to AvaI
fragment was blunt-ended and inserted in the sense and antisense
(a) orientation into the blunt-ended AvrII site in p124LUC,
thus preserving the transcriptional initiation site at its proper
position. p1098+279LUC contains the 974-bp AvrII
fragment, spanning from -124 to -1098, inserted into the AvrII site of p403LUC. pRSVLUC(18) , containing the
RSV long terminal repeat fused to the LUC gene in pXP2, was obtained
from Dr. M. Featherstone (McGill Cancer Centre, Montreal);
pRSVZ
-gal was obtained from Dr. E. Shoubridge (Montreal
Neurological Institute, Montreal) and contains the -galactosidase
gene driven by the RSV long terminal repeat.
-gal control plasmid. N.D., not determined.
Cells and DNA Transfections
Four human cultured
cell lines, SW403 (CCL 230, human colon adenocarcinoma), HT29 (HTB 38,
human colon adenocarcinoma), Caco-2 (HTB 37, human colon
adenocarcinoma), and HepG2 (HB 8065, human hepatocellular carcinoma)
were obtained from ATCC (Rockville, MD) and maintained in
Dulbecco's modified Eagle's medium supplemented with 10%
fetal bovine serum and penicillin/streptomycin (Life Technologies,
Inc.). The LR-73 (Chinese hamster ovary cell-derived) (19) and
HeLa R19 (human epithelial cervical carcinoma) (20) cell lines
were grown in 
-minimal essential medium(21) , supplemented
with 10% fetal bovine serum and penicillin/streptomycin.
10
cells/100-mm plastic
Petri dish 24 h before transfection. 10 µg of pRSVLUC or equimolar
amounts of the promoter-LUC gene constructs were cotransfected with 5
µg of pRSVZ-gal and l0 µg of calf thymus carrier DNA by
calcium phosphate-DNA coprecipitation as described
previously(22) . The precipitate was removed after 15 h, and
the cultures were incubated for another 72 h in normal growth medium.
Undifferentiated Caco-2 monolayers were transfected at 2 days before
confluence; fully differentiated Caco-2 cells were transfected at 11
days after confluence. Their state of differentiation was confirmed by
the presence of domes and the development of a brush border membrane (12) . 
-Interferon (Collaborative Research Inc., Bedford,
MA) was applied to Caco-2 cells at 2000 units/ml as described
previously (12) immediately following transfection, with a
fresh medium change. LUC activity was measured as described by DeWet et al.(17) . -Galactosidase activity (23) in these same extracts was measured to correct for
variations in transfection efficiency. Relative luciferase activity was
calculated as the ratio between LUC and -galactosidase activities
for each transfection and then reported as -fold above background
(considered as activity of the parental promoterless vector, pXP2).
Each plasmid was tested in duplicate plates and in three to five
different transfection experiments.
Nuclear Extracts and DNase I Footprinting
Assays
Nuclear extracts were prepared as described by Therrien et al.(27) . All buffers contained leupeptin,
pepstatin A, and aprotinin at concentrations of 1 µg/ml, and
phenylmethylsulfonyl fluoride at 1 mM. Extracts were assayed
for protein content by the Bio-Rad protein assay and stored at
-75 °C.Gel Mobility Shift Assays
Probes used for gel
mobility shift assays corresponded to the protected sequences in the
DNase I footprinting experiments, including a few nucleotides on either
side. Their sequences and that of other competitor oligonucleotides
used are shown in Table 1. The double-stranded oligonucleotides
were synthesized (Sheldon Biotechnology Centre, McGill University),
purified by gel electrophoresis, phosphorylated with T4 polynucleotide
kinase (Pharmacia Biotech Inc.) in the presence of
[-
P]ATP (Amersham Corp.), and purified from
unincorporated radioactivity by passage through NICK columns
(Pharmacia). About 2 fmol (15,000 cpm) of 5`-end-labeled,
double-stranded oligonucleotides were then incubated with 20-60
µg of various nuclear extracts and 4 µg of poly(dI-dC) in
binding buffer (29) at 23 °C for 15 min. For competition
experiments, the competitor oligonucleotides and extracts were mixed
together for 15 min at 23 °C, then probe was added for an
additional 15-min incubation. All samples were subjected to
electrophoresis on 5% nondenaturing polyacrylamide gels buffered with 1
TGE (Tris/glycine/EDTA)(23) . The gels were
autoradiographed at -75 °C with an intensifying screen.
Purified human AP-2 (30) and Sp1 (31) transcription
factors were obtained from Promega Corp. (Madison, WI); rabbit
polyclonal IgG specific to the human p95 and p106 Sp1 proteins was
obtained from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Human
USF cDNA (24, 25) in a T7-driven plasmid and a
USF-specific antibody were gifts of Dr. R. G. Roeder (The Rockefeller
University, New York, NY). USF protein was obtained directly from USF
cDNA by T7 RNA polymerase (New England Biolabs) transcription in
vitro, followed by translation in vitro using rabbit
reticulocyte lysate (Promega).
Functional Analysis Reveals Negative and Positive
Cell-specific Regulatory Elements
A genomic CEA clone
containing 3.5 kilobase pairs of sequence upstream of the initiation
codon was isolated from a normal human epithelial cell library and was
found to be virtually identical in the 5` noncoding sequence reported
by Schrewe et al.(15) for the CEA gene. In
the present report, we specifically localize the regulatory elements in
the CEA upstream region by various 5` deletion constructs in several
CEA-producing and non-producing cell lines. A very high CEA-producing
human colon carcinoma cell line, SW403, showed high levels of
luciferase expression when driven by a 403-bp upstream region of the
CEA gene (the p403LUC construct in Fig. 1): 91-fold above
background relative to only 2-fold and 5.5-fold above background in
CEA-non-expressing rodent LR-73 and human HeLa R19 cells, respectively.
In a low CEA-expressing human hepatocarcinoma cell line, HepG2, the
403-bp minimal promoter gave 7.7-fold above background activity,
whereas in HT29 cells, a medium CEA-expressing human colon carcinoma
cell line, it was surprisingly not active (1.8-fold above background)
This was perhaps due to the inherently low transfection efficiency for
HT29 cells, despite corrections for efficiency.279LUC and
p1098a279LUC). Hence, a silencer region, which can down-regulate CEA gene transcription, must lie within the -1098 to
-403 bp sequence. This silencer can nevertheless be completely
overcome by the upstream placement of a second 279-bp (from -403
to -124) region at -1098 bp (construct p279+1098LUC) (Fig. 1).Differentiation Dependence of Regulatory
Elements
To determine whether the transcriptional control of CEA
expression seen previously with the differentiation of Caco-2 cells (12) could be accounted for by the CEA gene upstream
regulatory region, all promoter constructs were transiently transfected
into undifferentiated, subconfluent Caco-2 cells and into fully
differentiated Caco-2 cells. The last two columns in Fig. 1summarize the results. The 403-bp minimal promoter was
found to be 3 times more active in differentiated than in
undifferentiated Caco-2 cells, thus correlating exactly with the 3-fold
increase seen in CEA mRNA in differentiated monolayers(12) .
CEA regulatory elements found within this 403-bp region could therefore
be responsible for differentiation-dependent CEA expression in Caco-2
cells.-interferon increases levels of CEA mRNA quite dramatically in the
Caco-2 cell line(12) , this cytokine was applied for 3 days to
examine changes in promoter activity. However, no effects on promoter
activity were seen (data not shown), despite the presence of several
possible -interferon activation sites and
-interferon-stimulated response elements in the upstream promoter
region (-835 to -1650).DNase I Footprinting
To localize the regulatory
elements responsible for the activities seen with the luciferase
constructs in Fig. 1, we performed DNase I protection
experiments using end-labeled fragments as probes incubated with
nuclear extracts from various CEA-expressing and non-expressing cell
lines. In all cases, any observed footprints were considered valid only
if demonstrated on both strands in repeated experiments. Within the
-403 to -124 upstream region, the deletion of which
abolished all promoter activity (Fig. 1; p124LUC construct), we
identified four sites protected from DNase I digestion by various
CEA-expressing cell line nuclear extracts but not by CEA non-expressing
LR-73 extracts (Fig. 2, B and C). Footprints
(FP) FP1 and FP4 were clearly visible with nuclear extracts from most
of the CEA-expressing lines; their positions in the upstream sequence
are shown in Fig. 2A and in Fig. 1(ovals). Comparison of the luciferase activities
of p300LUC with p280LUC, containing and lacking the FP4 element,
respectively, indicates an approximately 2-fold contribution of this
element to the minimal promoter activity.
) produced
by binding of nuclear proteins to the FP2 and FP3 elements are visible. D, localization of binding sites in the silencer region with a
3`-end-labeled -835 to -403 probe (Probe III). In addition
to SW403 nuclear extract, a nuclear extract prepared from HT29 cells
was assayed for comparison. Both extracts reveal DNase I hypersensitive
sites () flanking the FP5 element.
USF Binds to the CEA FP1 Element
Comparison of the
protected sequence in FP1 to a transcription factor data base
identified several candidate binding factors. These included the
upstream stimulatory factor (USF) (34) and hepatic nuclear
factor 4 (HNF-4)(26) , also known as liver-specific factor A1
(LF-A1)(35) . Two complexes with labeled FP1 oligonucleotide (Table 1), C1 and C2, were visible by band-shift assays using
SW403 (Fig. 3A) or HepG2 (Fig. 3C)
nuclear extracts. Formation of both complexes were competed effectively
by 5- and 10-fold molar excesses of either unlabeled FP1
oligonucleotide (Fig. 3A, lanes 2-4) or
of GAL2 oligonucleotide, whose sequence represents a USF-binding
element in the GAL2 gene (36) (Fig. 3A, lanes5 and 6). The two complexes were not
affected by an excess of an oligonucleotide representing the
HNF-4/LF-A1 motif taken from the 1-antitrypsin gene (35) (1-AT, Fig. 3A, lane 7),
however, suggesting that the FP1 element is not an HNF-4/LF-A1 site.
This corroborates our studies in which 10 µg of p403LUC and up to
20 µg of functional HNF-4 cDNA were co-transfected into SW403,
Caco-2, HepG2, and LR-73 cells. Activation of the CEA promoter in
p403LUC by HNF-4 could not be detected (data not shown), even though
USF was capable of activating p403LUC in the same experiment (see
below).
USF Activation of the CEA Gene Promoter in Vivo
To
test directly whether USF could modulate CEA gene promoter activity in vivo, an expression plasmid carrying USF cDNA, pRSV.USF,
together with the CEA minimal promoter construct, p403LUC, were
co-transfected into various cells. Co-transfection with 15 µg of
pRSV.USF produced 2.6-fold greater LUC activity in SW403 cells and
3.1-fold greater activity in undifferentiated Caco-2 cells than the
level produced by co-transfection with 15 µg of pUC, the parental
vector of pRSV.USF, lacking USF cDNA; in addition, the effect of
pRSV.USF was concentration-dependent (Table 2). In contrast, the
activity of the basic luciferase vector, pXP2, was not influenced by
co-transfected pRSV.USF (data not shown), indicating that USF-mediated
transactivation was not due to sequences within the luciferase vector.
In addition, as mentioned above, co-transfection with functional HNF-4
cDNA also had no effect on the activity of p403LUC. Thus it seems
likely that CEA promoter activity is partially controlled by the level
of USF in differentiated Caco-2 cells and colon tumors. However, USF is
a ubiquitously expressed factor and CEA shows a much more restricted
tissue-specific expression pattern. Therefore, other elements and trans-acting factors of the CEA gene promoter must interact to
achieve tissue-specific expression.Sp1, Sp1-like, and an Unknown Nuclear Factor Bind the CEA
FP2 and FP3 Elements
Gel mobility shift assays with
oligonucleotides representing the FP2 and FP3 regions are shown in Fig. 4and Fig. 5, respectively. Three retarded complexes,
C1, C2, and C3, were seen with the FP2 probe (Fig. 4A).
Two of these, C1 and C2, comigrated with the two retarded complexes
seen with the FP3 probe (Fig. 5A). FP2 and FP3 compete
with each other for the formation of C1 and C2 (Fig. 4A and Fig. 5A), as would be expected from the high
sequence homology between these two adjacent sites. The C3 complex
appears to be due to an unrelated factor specifically recognizing the
FP2 element, since excess FP2 oligonucleotide, but not FP3
oligonucleotide, competes for its formation (Fig. 4A, lanes 4 and 6). The FP2 sequence exhibited some
homology to the AP1 consensus recognition site, but molar excesses of
an oligonucleotide representing an AP1 site did not compete with the
FP2 probe (Fig. 4A, lanes 9 and 10)
or with the FP3 probe (Fig. 5A, lanes 9 and 10). The FP3 sequence showed some homology to a PEA3
site(32, 37) , but an excess of an oligonucleotide
with the PEA3 sequence from the polyoma enhancer (PY) did not inhibit
complex formation with either the FP3 probe (Fig. 5, lanes
11-13) or the FP2 probe (data not shown). Positive
competition by an oligonucleotide representing an Sp1 binding site (38) with both FP2 and FP3 as probes (Fig. 4A and Fig. 5A, lanes7 and 8), however, indicated that this factor may be responsible for
complexes C1 and C2, but not C3. Although the FP2 and FP3 sites are
remarkably GA-rich, rather than the typical GC-rich Sp1 site
sequence(38) , purified Sp1 protein did form complexes with
both the FP2 and FP3 probes, which comigrated with the C1 complex in
SW403 extract (Fig. 4B, lanes 11 and 12; Fig. 5B, lanes 14 and 15). Antibody specific to Sp1 supershifted the Sp1 C1
complexes formed with the FP2 and FP3 elements, as expected, but also
the FP2 and FP3 C1 complexes formed with SW403 nuclear extracts (Fig. 4B, lanes 14 and 15; Fig. 5B, lanes 17 and 18). The C2 and
C3 complexes remained unaffected. Since Sp1 protein did not produce a
band comigrating with the C2 complex and the latter was not affected by
an Sp1-specific antibody, but Sp1 oligonucleotide did compete for the
formation of the C2 complex, we can only surmise that the second
complex is due to a Sp1-like factor, and not Sp1 itself. Thus, the FP2
and FP3 regulatory elements bind Sp1 as well as a Sp1-related factor.
In addition, another unknown factor binds exclusively to the FP2 site
to produce the C3 complex. Similar complexes were obtained with Caco-2
extracts (Fig. 6).
Levels of Nuclear Factors in Differentiating Caco-2
Cells
Caco-2 cells show increased levels of CEA mRNA (12) and increased transcriptional activity of transfected CEA
promoter-luciferase constructs (Fig. 1) with differentiation.
Equivalent amounts of undifferentiated and differentiated Caco-2
nuclear extracts (based on equivalent numbers of nuclei that differed
by only 10% in the concentration of total nuclear proteins; see
``Materials and Methods'') were examined for the levels of
transcription factors by gel mobility shift assays using labeled
oligonucleotides representing three of the five CEA cis-acting
regulatory elements identified above (Fig. 6).
)
We are indebted to Drs. P. Pognonec, T. Gutjahr, and
R. Roeder (Rockefeller University, New York) for USF cDNA and anti-USF
antibody, and to Dr. F. M. Sladek (University of California, Riverside,
CA) for HNF-4 cDNA. We also thank Dr. N. Beauchemin for stimulating
discussions and critical advice and Drs. M. Chamberlin and J. Chou
(National Institutes of Health, Bethesda, MD) for helpful discussions
of unpublished work.
©1995 by The American Society for Biochemistry and Molecular Biology, Inc.
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  P. Jay, P. Berta, and P. Blache Expression of the Carcinoembryonic Antigen Gene Is Inhibited by SOX9 in Human Colon Carcinoma Cells Cancer Res., March 15, 2005; 65(6): 2193 - 2198. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Li, Y. Chen, J. Dilley, T. Arroyo, D. Ko, P. Working, and D.-C. Yu Carcinoembryonic antigen-producing cell-specific oncolytic adenovirus, OV798, for colorectal cancer therapy Mol. Cancer Ther., October 1, 2003; 2(10): 1003 - 1009. [Abstract] [Full Text]
|
|
|
|
 |
|
 A. C. Keates, S. Keates, J. H. Kwon, K. O. Arseneau, D. J. Law, L. Bai, J. L. Merchant, T. C. Wang, and C. P. Kelly ZBP-89, Sp1, and Nuclear Factor-kappa B Regulate Epithelial Neutrophil-activating Peptide-78 Gene Expression in Caco-2 Human Colonic Epithelial Cells J. Biol. Chem., November 16, 2001; 276(47): 43713 - 43722. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Q. Huang, J. K. Hu, F. Lohr, L. Zhang, R. Braun, J. Lanzen, J. B. Little, M. W. Dewhirst, and C.-Y. Li Heat-induced Gene Expression as a Novel Targeted Cancer Gene Therapy Strategy Cancer Res., July 1, 2000; 60(13): 3435 - 3439. [Abstract] [Full Text]
|
|
|
|
|
|
H. Ding, A. M. Benotmane, G. Suske, D. Collen, and A. Belayew Functional Interactions between Sp1 or Sp3 and the Helicase-like Transcription Factor Mediate Basal Expression from the Human Plasminogen Activator Inhibitor-1 Gene J. Biol. Chem., July 9, 1999; 274(28): 19573 - 19580. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. W. Weston, K. M. Hiller, J. P. Mayben, G. A. Manousos, K. M. Bendt, R. Liu, and J. C. Cusack Jr. Expression of Human {{alpha}}(1,3)Fucosyltransferase Antisense Sequences Inhibits Selectin-mediated Adhesion and Liver Metastasis of Colon Carcinoma Cells Cancer Res., May 1, 1999; 59(9): 2127 - 2135. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A.-M. Eades-Perner, J. Thompson, H. van der Putten, and W. Zimmermann Mice Transgenic for the Human CGM6 Gene Express Its Product, the Granulocyte Marker CD66b, Exclusively in Granulocytes Blood, January 15, 1998; 91(2): 663 - 672. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. Finkenzeller, B. Kromer, J. Thompson, and W. Zimmermann cea5, a Structurally Divergent Member of the Murine Carcinoembryonic Antigen Gene Family, Is Exclusively Expressed during Early Placental Development in Trophoblast Giant Cells J. Biol. Chem., December 12, 1997; 272(50): 31369 - 31376. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. P. Koritschoner, J. L. Bocco, G. M. Panzetta-Dutari, C. I. Dumur, A. Flury, and L. C. Patrito A Novel Human Zinc Finger Protein That Interacts with the Core Promoter Element of a TATA Box-less Gene J. Biol. Chem., April 4, 1997; 272(14): 9573 - 9580. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Velcich, L. Palumbo, L. Selleri, G. Evans, and L. Augenlicht Organization and Regulatory Aspects of the Human Intestinal Mucin Gene (MUC2) Locus J. Biol. Chem., March 21, 1997; 272(12): 7968 - 7976. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I. Castagliuolo, L. Valenick, J. Liu, and C. Pothoulakis Epidermal Growth Factor Receptor Transactivation Mediates Substance P-induced Mitogenic Responses in U-373 MG Cells J. Biol. Chem., August 18, 2000; 275(34): 26545 - 26550. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Suzuki, T. Tsunoda, J. Sese, H. Taira, J. Mizushima-Sugano, H. Hata, T. Ota, T. Isogai, T. Tanaka, Y. Nakamura, et al. Identification and Characterization of the Potential Promoter Regions of 1031 Kinds of Human Genes Genome Res., May 1, 2001; 11(5): 677 - 684. [Abstract] [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| All ASBMB Journals | Molecular and Cellular Proteomics |
| Journal of Lipid Research | ASBMB Today |