ABSTRACT
Through a mutational analysis of a differentially regulated enhancer, we present evidence that supports a role for the transcription factor YY1 in tumor suppression in HeLa/fibroblast
somatic cell hybrids. The human ST5 gene was previously shown to be expressed
as three RNA species, 4.6, 3.1 and 2.8 kb in length. Whereas the two larger
species are expressed at similar levels in all cell lines examined, the 2.8 kb
mRNA is expressed specifically in non-tumorigenic hybrids. In this study, the basis for the differential
expression of this mRNA species was investigated. The message was shown to
originate from a promoter located within an intron of the ST5 gene. An enhancer
located
~
1500 nt upstream of the start site was required for cell type specific
expression. Mutational analysis of this enhancer revealed an AP1 site and five
YY1 sites which were necessary for full enhancer activity. Levels of YY1 DNA
binding activity were found to be as much as 6-fold higher in the non-tumorigenic cells relative to the tumorigenic cells, while AP1
activity was similar in both cell types. These results suggest that a signaling pathway targeting YY1 may play an important role in tumor suppression in HeLa-fibroblast hybrids.
Somatic cell hybrids formed by fusion of HeLa cells to non-tumorigenic fibroblasts provide a model system for studying alterations in
gene expression in carcinogenesis (
1
-
3
). The phenotypes of the hybrid cell lines derived from such fusions parallel
that of cells at various stages of carcinogenic progression in the cervix, the
anatomical site of origin of the HeLa cell line. In cervical carcinogenesis,
expression of the transforming genes of a human papillomavirus (HPV) leads to
abnormalities in cell growth control, but is not in itself sufficient to
convert the benign keratinocyte into a carcinoma (
4
). Similarly, transfection of HPV DNA into primary cultures of human
keratinocytes confers enhanced growth potential on the cells, but fails to
produce a fully malignant phenotype (
5
). Such HPV `transformed' keratinocytes grow with a flat morphology in culture
and are non-tumorigenic upon injection into nude mice. In contrast, most cervical
carcinoma cells prove to be tumorigenic in mice and grow with a characteristic
transformed morphology in culture.
In the HeLa-fibroblast system, the hybrid cell lines initially demonstrate a non-tumorigenic phenotype. On prolonged passage these non-tumorigenic hybrids give rise to segregants which have regained the
tumorigenic phenotype. In their morphology and lack of tumorigenicity, the non-tumorigenic hybrids resemble HPV transformed keratinocytes. In addition,
these hybrids express the transforming genes of HPV18 at the same or slightly
lower levels than the parental HeLa cell line (
6
). The tumorigenic segregants, in an analogous manner, show phenotypic
similarities with cervical carcinoma cell lines. However, whereas cervical cancer cells might
be expected to have undergone multiple genetic alterations relative to a benign keratinocyte, the tumorigenic segregant lines are more likely to differ from their non-tumorigenic parents predominantly in the expression of those genes required for the
regulation of tumorigenicity (
1
,
2
,
7
).
The specific molecular events leading from the HPV transformed keratinocyte to carcinoma remain poorly understood. Some experimental evidence
suggests that mutational activation of a
ras
oncogene can in some cases suffice to cause this phenotypic transition (
8
,
9
). Another series of studies, using the somatic cell hybrids as a model system,
has led to the suggestion that the transformation results from alterations in a
regulator of HPV transcription such that deregulated expression of the
transforming proteins of HPV leads to the fully malignant phenotype (
4
,
6
,
10
).
In this study, we present evidence that at least some of the alterations in gene
expression that distinguish the tumorigenic from the non-tumorigenic hybrids can be attributed to transcriptional regulation through YY1, a factor which has in fact been implicated as a
regulator of HPV transcription. YY1 (also referred to as NF-E1, [delta] and UCRBP) is a bifunctional transcription factor which can act as either an activator or repressor of
transcription depending on the specific context of its binding site. In
addition to its effects on the long control regions (LCRs) of HPV16 and 18, YY1
has been implicated in the regulation of several cellular and viral regulatory
regions [reviewed in (
11
)]. The YY1 protein contains distinct activator and repressor domains (
12
). In the regulatory region of HPV 18, YY1 functions either as an activator or repressor depending on the presence or absence of a `switch' sequence (
13
). A role for YY1 in cervical carcinogenesis has been suggested by the observation of several cervical carcinomas in which the HPV 16 promoter has escaped
cellular repression via deletion of YY1 binding sites (
14
).
We previously reported the identification of a cellular gene which showed a striking pattern of differential regulation that correlated with
tumorigenicity (
15
). To avoid ambiguity in nomenclature, this gene, which we previously named
HTS1, will be designated by its locus name, ST5 (
16
,
17
). Non-tumorigenic hybrids express three ST5 RNA species, 4.6, 3.1 and 2.8 kb in
length. The tumorigenic hybrids express similar or slightly reduced levels of
the two larger species, but express very markedly reduced levels of the 2.8 kb
transcript. This pattern of expression was of interest because the tumorigenic
and non-tumorigenic hybrids are closely related genetically, and show extremely
few differences in gene expression (
3
,
18
). The correlation between expression of the 2.8 kb transcript and the non-tumorigenic phenotype applied not only to the HeLa-fibroblast system but
also to the non-tumorigenic HPV immortalized keratinocytes, which expressed the
transcript, and to several cervical carcinoma cell lines, which showed markedly
reduced levels of expression (
15
).
Regardless of whether the protein encoded by the 2.8 kb RNA plays a direct role
in determining the non-tumorigenic phenotype, the pattern of expression of this transcript
suggested that the mechanism responsible for its regulation might play an
important role in the biology of tumor suppression in this system. In this
report, we present the structure of the three full length ST5 mRNA species and
characterize the mechanism of regulation of the 2.8 kb transcript. This RNA was
found to be driven by a promoter located within an intron of the ST5 gene and
to initiate near the 3' splice site of this intron. Sequential deletions of this promoter
demonstrated the presence of an enhancer ~1500 nt upstream of the initiation site. A mutational analysis of this
enhancer revealed a cluster of five YY1 sites which play a critical role in
mediating the specific expression of the 2.8 kb transcript in non-tumorigenic cell lines. Our results suggest that regulation of
transcription by YY1 may be a mechanism of action of the tumor suppressor gene
active in the HeLa-fibroblast hybrids.
Four successive applications of the 5'-RACE procedure (
19
) yielded 4277 nt of ST5 cDNA sequence (GenBank #U15131). This sequence was
judged to be the full length or nearly full length 4.6 kb transcript by the
following criteria: (i) 4277 is close to the length of 4.6 kb estimated from Northern blots; (ii) primer extension experiments with several oligonucleotides yielded similar banding patterns consistent with 5' termini at or just beyond the first nucleotide in the sequence; and (iii) additional rounds of 5'-RACE failed to yield additional upstream sequence.
Cell lines were maintained in DMEM plus 10% fetal calf serum. Cells were plated out at 5 * 10
5
cells per 10 cm plate the day before transfection. Calcium phosphate precipitates (1.0 ml) were prepared containing
20 [mu]g of the enhancer construct and 1 [mu]g of pCMV-[beta]gal to control for transfection efficiency. The transfected
cells were subjected to a 1 min glycerol shock 4 h after adding the
precipitate. Extracts were prepared 72 h post transfection by three freeze-thaw cycles in 150 [mu]l 0.25 M Tris, pH 8.0. Acetylation assays were carried out in a final
volume of 75 [mu]l that included 20 [mu]l lysate (50-70 [mu]g protein), 0.1 [mu]Ci [
14
C]chloramphenicol (Amersham), 0.25 M Tris, pH 8.0, and 1 mM acetyl-CoA. Except where indicated otherwise, reactions were incubated for 2 h at
37oC, extracted with ethyl acetate, and analyzed by thin layer chromatography in 96% chloroform/4% methanol. Reaction products were quantitated on a Molecular Dynamics PhosphorImager and
percent acetylation was standardized to [beta]-galactosidase activity in the same extract. For each set of CAT assays presented,
the activity indicated for the wild type construct represents the actual %-acetylation observed. The tumorigenic and non-tumorigenic hybrids used in this study demonstrated very similar
transfection efficiencies, as assessed by quantitation of [beta]-galactosidase activity in parallel transfections of the two cell
lines.
The partial genomic clone of the ST5 gene was isolated from a Charon 4A human
genomic library (
21
) purchased from ATCC (#37333).
32
P-end labeled oligonucleotide primer (1 ng) was mixed with 5 [mu]g polyadenylated RNA in 18 [mu]l reactions containing 0.5 mM dNTPs and RT buffer (GIBCO/BRL).
After incubation at 95oC for 3 min and 50oC for 60 min, 200 U of SuperScript reverse transcriptase (GIBCO/BRL)
was added and the incubation was continued for an additional 60 min at 50oC. After phenol extraction and ethanol precipitation, the reaction products were separated on 6% acrylamide/7 M urea sequencing gels and detected by exposure to a PhosphorImager screen (Molecular Dynamics).
The probe used for S1 protection assays was an oligonucleotide 60 bp in length
complementary to nt 1579-1638 of the cDNA sequence. Hybridizations were performed at 50oC for 2 h in 20 [mu]l reactions containing 1 ng of the
32
P-end labeled oligonucleotide, 20 [mu]g cytoplasmic RNA from the indicated cell line, 1 M NaCl, 1 mM EDTA
and 133 mM HEPES, pH 7.2. S1 digestion was carried out at 13oC for 2 h by bringing the reaction volume to 150 [mu]l with 0.25 M NaCl; 30 mM sodium acetate, pH 4.5; 1 mM ZnSO
4
and adding 400 U S1 nuclease (Boehringer-Mannheim). Reactions were stopped by the addition of EDTA to 10 mM followed by phenol/chloroform extraction and ethanol precipitation. Products were analyzed by electrophoresis on an 8% sequencing gel. Bands were visualized with a Molecular Dynamics PhosphorImager.
Unidirectional Exonuclease III deletions were constructed by using the Exo/Mung kit (Stratagene) according to the manufacturer's instructions. The
panel of clustered point mutations in the 85 nt enhancer fragment was constructed by reconstruction of this region with three overlapping pairs of oligonucleotides. The structure of each mutant was confirmed by DNA sequencing with the Sequenase kit (Amersham).
Nuclear extracts used for electrophoretic mobility shift assays (EMSA) were
prepared by the method of Dignam
et al
. (
22
). Reaction mixtures (15 [mu]l) contained 1 ng of
32
P-end labeled probe, 1.5 [mu]g poly d(I-C) and 4 [mu]g of protein in 10 mM Tris, pH 7.5, 4% glycerol, 1 mM EDTA and 150
mM NaCl (final concentration). Reactions were incubated at 4oC for 15 min, loaded onto a pre-run 4% polyacrylamide gel in 0.25* TBE, and run at 250 V and 4oC until bromophenol blue run in a separate lane was 1-2 cm from the bottom of the gel. The competitors were
all double stranded oligonucleotides, and were added to the reactions where
indicated prior to the addition of nuclear extract.
The sequences of the 4.6, 3.1 and 2.8 kb ST5 cDNAs have been assigned GenBank accession numbers U15131, U15780 and U15779. 2.3 kb of sequence from the promoter containing intron extending upstream from
the splice junction and containing the enhancer sequence has been submitted to
GenBank with the accession number U15132.
The sequence of the full length ST5 cDNA, corresponding to the 4.6 kb transcript
detected on Northern blots, was obtained as described in Materials and Methods.
A series of RT-PCR and Northern blot experiments were then performed to determine the
relationship of the smaller two RNA species to the 4.6 kb message. The results
support the structures shown in Figure
1
A. To identify alternatively spliced species, RT-PCR reactions were carried out with an upstream primer containing
sequence near the 5' end of the message and a downstream primer from one of several regions
within the cDNA. With a downstream primer taken from within the 3'-terminal 2823 nt, these PCR reactions produced two products, the larger one of which had the size and
sequence expected from the full length cDNA sequence. Sequencing of the smaller product revealed a deletion of nt 195-1454, suggesting the existence of an alternatively spliced product
lacking this region. Consistent with this interpretation, a probe containing
only this region recognized exclusively the 4.6 kb ST5 message on a Northern
blot (Probe 2 in Fig.
2
). A probe upstream of this region (Probe 1) recognized both the 4.6 and 3.1 kb
RNA species but not the 2.8. Several probes tested from the downstream 2687 nt recognized all three species. These results indicated that the 4.6 and
3.1 kb ST5 transcripts represent alternatively spliced mRNAs with common 5' and 3' ends.
The absence of the Probe 1 sequence in the 2.8 kb transcript, together with the
lack of evidence for an alternative 5' exon after sequencing multiple RACE products, suggested that this transcript might be driven by a promoter located internally in the gene. To
investigate this possibility, a probe consisting of nt 1593-1958 [the most upstream portion of the previously reported sequence (
15
)], which recognized the 2.8 kb species on a Northern blot, was used to isolate a partial genomic clone of the ST5 gene. Two positive phage [lambda] clones were isolated, one of which contained 9.5 kb of genomic sequence upstream of the
Bgl
II site at nt 1593. By sequencing, restriction mapping and hybridization to cDNA
probes, this clone was found to contain two exons of the ST5 gene, designated
exon A and exon B (Fig.
1
B). On Northern blots (Fig.
2
), a probe consisting of the exon B sequence recognized all three ST5 RNAs. In
contrast, the 136 nt exon A sequence detected the 4.6 and 3.1 kb species but
did not recognize the 2.8 kb RNA. These results suggested that the 5' end of the 2.8 kb message was within or immediately upstream of exon B.
The localization of the 5' end was defined more precisely with the primer extension and nuclease protection experiments presented in Figure
3
B and C. The relative levels of the three mRNA species in tumorigenic and non-tumorigenic hybrids are illustrated in the Northern blot shown in Figure
3
A. For primer extension (Fig.
3
B), a 30 nt antisense oligonucleotide primer yielded major primer extension
products mapping putative start sites to positions -2 and +15 relative to the 3' splice junction. Neither band was detected in the absence of RNA
(lane 1), or with RNA from the tumorigenic hybrid (lane 3). These results
parallel the differential expression of the 2.8 kb RNA in the tumorigenic and
non- tumorigenic hybrids (Fig.
3
A), and support the conclusion that the bands detected correspond to the 5' termini of this RNA species. In the nuclease protection assay (Fig.
3
C), an antisense oligonucleotide probe complementary to nt 1579-1638 of the cDNA, and spanning the splice junction at nt 1590-1591, was incubated either with no RNA or with RNA from the
tumorigenic or non-tumorigenic hybrids followed by digestion with nuclease S1. A band
corresponding to protection of the full length oligonucleotide probe, resulting from the 4.6 and 3.1 kb transcripts, was detected with somewhat greater intensity with RNA from the non-tumorigenic hybrid, consistent with the expression pattern of these transcripts (Fig.
3
A). In contrast, a cluster of digestion products centered 2-3 nt upstream of the splice site showed marked specificity for the non-tumorigenic cell line, as expected for the 2.8 kb transcript. The
primer extension and S1 protection data argue for a major initiation site of
the 2.8 kb RNA 2-3 nt upstream of the splice junction.
The localization of the initiation site for the 2.8 kb mRNA suggested that the
promoter driving transcription of this message might lie within the intron
between exons A and B. To test this possibility, a 2.6 kb
Pvu
II fragment containing the downstream portion of this intron (Fig.
1
) was cloned into pSV0CAT. The 2.6 kb
Pvu
II fragment drove expression of the CAT gene in transient expression assays in
an orientation dependent manner (Fig.
4
). Furthermore, this activity demonstrated a marked specificity for the non-tumorigenic relative to the tumorigenic HeLa-fibroblast hybrid. Control
expression plasmids, including pSV2CAT, an HPV18 LCR-CAT construct and pCMV-[beta]gal, the plasmid used for standardization of the CAT data,
resulted in similar levels of CAT activity in the two cell lines. These
results, together with the primer extension, nuclease protection, and Northern
blot data, strongly support the conclusion that the promoter for the 2.8 kb
transcript lies within the intron between exons A and B. Furthermore, since
cell type specific regulation can be mediated by the untranscribed upstream
sequence, the results suggest that regulation of this transcript occurs at the
level of transcription.
A series of deletions were introduced into the 2.6 kb
Pvu
II promoter fragment to identify elements required for activity. The sequences
of the regions found to be contributory to promoter activity are given in
Figure
5
. Deletions were generated upstream and downstream of a unique
Spe
I site located 836 nt upstream of the splice junction by exonuclease III
digestion. Deletions upstream of the
Spe
I site were either without effect or resulted in a modest stimulation of
activity (with no loss of specificity) until nt -1550 relative to the splice junction (Fig.
4
). Deletions extending upstream of this point yielded markedly reduced activity. These deletion
plasmids thus suggested that a region located upstream of nt -1550 is required for maximal expression in the non-tumorigenic hybrid cell line.
To determine whether the property of specific expression in the non-tumorigenic cells mapped to any of the regions critical for activity,
portions of the promoter were cloned in front of the enhancerless SV40 promoter
in the plasmid pA10CAT and transiently transfected into tumorigenic and non-tumorigenic HeLa-fibroblast hybrids (data not shown). The promoter proximal region containing the G-rich region had no detectable effect on the basal activity of the
vector in either cell line. A
Pst
I fragment containing nt -1497 to -1826 resulted in a 20-30-fold stimulation of CAT activity which was specific
for the non-tumorigenic cell line and independent of the orientation of the fragment.
Other fragments derived from the promoter sequence had no effect on promoter
activity. Specific expression of the internal promoter therefore maps to an
enhancer element located between positions -1497 and -1826 relative to the splice junction (Fig.
5
A).
To localize specific elements required for activity, exonuclease III deletions
were made from either end of this 330 nt fragment and tested in the transient
expression assay. Deletions from -1826 to -1681 had no effect on enhancer activity. The results of additional
deletions with schematics showing the AP1 and YY1 sites subsequently
demonstrated to be important (as shown below) are given in Figure
6
. Deletions downstream from nt -1681 resulted in an ~20-fold reduction in activity with deletions as small as 17 nt,
which eliminated the AP1 site. Deletions from the opposite end of the fragment
resulted in a more gradual loss of activity. Fifty percent of the activity
remained with a deletion extending to a point just downstream of the YY1
cluster. Further deletion of the region containing four YY1 binding sites led
to an additional 10-fold loss of activity, and deletion of all five YY1 sites reduced the
activity 2-3-fold further.
Several probes derived from the critical regions of the enhancer were employed
in mobility shift assays with nuclear extracts from the tumorigenic and non-tumorigenic hybrids. A probe containing the AP1 consensus sequence demonstrated a prominent shifted band, present
with approximately equal intensity with extracts from either cell line (Fig.
8
A). This band was efficiently competed by an excess of the probe or an AP1 consensus competitor, but was not competed
appreciably by the probe sequence with a mutation in the AP1 consensus.
We have studied the mechanism of activation of the ST5 2.8 kb transcript in non-tumorigenic cells and have obtained evidence suggesting a critical role
for the transcription factor YY1. Although this factor has been implicated in the regulation of several genes, we believe that this is the first study to suggest a specific role for
YY1 as a mediator of tumor suppression. The initial studies to characterize the
2.8 kb transcript involved defining the initiation site and obtaining a genomic
clone of the promoter. Our results demonstrate that this transcript is driven
by a promoter located within an intron of the ST5 gene, with transcriptional
initiation occurring very close to the 3' splice junction. The sequence upstream of the splice site contained a G-rich region, which was shown to be required for promoter activity.
Although no perfect TATAA motif was present in the -30 region, the sequence TCTAA was present at that location, and potential
TATAA motifs were noted further upstream. Three regions required for promoter
activity were identified: the G-rich region, the sequence between this region and the start site, and an
enhancer element located ~1500 nt upstream of the start site. Each functional region of the promoter was tested separately for the ability to confer cell type specific expression on a heterologous promoter. These experiments localized this activity to the upstream enhancer element.
This work was supported by Public Health Service grant R01 CA-64114 from the National Cancer Institute. The early phases of this work
were supported by NIH intramural funds to the Laboratory of Tumor Virus
Biology, National Cancer Institute. J.L. gratefully acknowledges the support of
Dr Peter M. Howley during the initial phases of this project.
*To whom correspondence should be addressed at: Department of Cellular
Pathology, Armed Forces Institute of Pathology, 14 Street and Alaska Avenue,
NW, Washington, DC 20306-6000, USA. Tel: +1 202 782 2562; Fax: +1 202 782 7623;
Email:lichy@email.afip.osd.mil
REFERENCES
Return