Download Identification of Genes Overexpressed in Tumors

[CANCER RESEARCH 54, 5217-5223,
October 1, 1994]
Identification of Genes Overexpressed in Tumors through Preferential Expression
Screening in Trophoblas&
Dorine Chassin, Jean-Louis Bénifla,Claire Delattre, HervéFernandez, Danièle Ginisty, Jean-Louis Janneau,
Michel Prade, Genevieve Contesso, Bernard Caillou, Michel Tournaire, RenéFrydman, Dominique Elias,
Pierre
Bedossa,
Jean-Michel
Laboratoire
d'!minunologie
des Twneurs,
Paris ID. C., C. D., D. G., J-L J., J-M.
Bidart,
Dominique
Bellet,
and Ahmet
CNRS URA 1484, Université René Descartes,
B., D. B., A. K.J; Maternité, Hôpital Bichat,
Koman2
Faculté des Sciences Pharmaceutiques
et Biologiques,
4, avenue de l'Observatoire,
75018 Paris (J-L B.]; Service d'Inimunologie
Moléculaire, Institut Gustave-Roussy,
75006
94805
Villeju(f [f.M. B., D. B.J; Maternité,Hbpital Antoine Béclère,92140 Clamart [H. F., R. F.]; Services d'Anatomie Pathologique B [M. P.], C 1G. C.], and A [B. CI, Institut
Gustave-Roussy, 94805 Villejuif, Maternité,Hôpital St. Vincent de Paul, 75014 Paris [M. T.]; Service de Chirurgie, Institut Gustave-Roussy, 94805 Villeju@f[D. E.J; and Service
d'Anatomie
Pathologique,
Hôpital Kremlin-Bicêtre,
94270
Kremlin-Bicêtre
[P. B.], France
ABSTRACT
Early trophoblastic
cells share several features with neoplastic cells.
Based on that observation, we attempted to identify genes overexpressed
in tumors by analyzing genes preferentially expressed in trophoblasts. A
subtracted library enriched in complementary DNA from early cytotro
phoblasts was constructed, and the expression level of selected recombi
nants was analyzed on a large panel of normal and tumor tissues. The
library was prepared using a polymerase chain reaction-based
comple
mentary DNA subtraction method with 6-week amenorrhea
cytotropho
blast endoplasmic
reticulum-bound
RNA as target, and a mixture of
complementary DNA prepared from terminal placenta and activated
T-lymphocytes
as driver. Two rounds of screening were performed to
isolate dones preferentially
expressed in early placenta. From a total
number of recombinant
clones estimated at 32,000 in the subtracted
library, 594insertswereanalyzedby Southernblotand21sequences
were
isolated as corresponding to genes highly expressed in early placenta.
Eleven encoded known molecules, such as carcinoembryonic antigen,
human chorionic gonadotropin, and mitochondrial rRNAs. Ten sequences
represented novel genes. Northern blot analysis confirmed that most of
these genes were preferentially expressed in early trophoblast in compar
ison to terminal placenta. Three clones that gave detectable hybridization
si_
on total RNA were extensively studied and were found to be
overexpressed in various tumors. Two of these clones, designated B9 and
E4, were later identified
as corresponding
to genes coding
for the putative
ribosomal protein S18 and the bifunctional enzyme ADE2H1 involved in
purine biosynthesis, respectively. Expression of the third clone, E9, was
Increased up to 10-fold in breast cancer tissues in comparison with normal
counterparts.
Present
results confirm
that many genes expressed
in the
trophoblast are overexpressed in malignant cells. This approach could
provide a general targeted method for the identification of genes overex
pressed in various neoplastic
cell types.
Received 5/20/94; accepted 8/4/94.
The costs of publication of this article were defrayed in part by the payment of page
charges. This article must therefore be hereby marked advertisement in accordance with
18 U.S.C. Section 1734 solely to indicate this fact.
was
supported
Tissues. Early placentae (5 to 12 weeks of pregnancy) from women Un
dergoing legal abortions, normal term placentae from uncomplicated caesarean
section, and surgical samples of normal and tumoral tissues (maximal size, 1
cm3) were immediately frozen (within 10 mm after surgical intervention) and
stored in liquid nitrogen until RNA preparation. This tissue collection was
obtained and used in accordance with protocols previously approved by the
Bichat).
Cells. Human cell lines used during this study and their histological origin
In the search for molecules preferentially expressed in tumors and
implicated in the neoplastic process, the placenta represents a model
of particular interest. Although the placenta is of normal tissue, its
constituent cells share several common features with neoplastic cells,
associated in particular with highly mitotic, invasive cytotrophoblasts
able to penetrate the maternal tissue. Trophoblasts are plentiful
sources of growth factors, hormones, and growth factor receptors, and
several pertinent observations suggest autocrine growth control (1—5).
Furthermore, most known protooncogenes have also been found to be
expressed in placenta (6, 7). Their invasiveness, high cell prolifera
study
MATERIALS AND METHODS
human studies committees of each contributing hospital (Institut Gustave
Roussy, HôpitalAntoine Béclère,
HOpital St. Vincent de Paul, and Hôpital
INTRODUCTION
1 This
tion, immune privilege, and lack of cell contact inhibition, particularly
during the first trimester of pregnancy, have led to the definition of the
trophoblast as a pseudomalignant type oftissue (8, 9). Taken together,
these features suggest that genes preferentially expressed in tropho
blastic cells might also be preferentially expressed in neoplastic cells.
Furthermore, while the placenta resembles a locally invasive tumor,
trophoblast invasion remains under strict control during normal preg
nancy (9). Trophoblastic molecules involved in this control are po
tential candidates as tumor suppressors and might be absent or mod
ified in tumor cells. These observations led us to investigate gene
expression in trophoblasts as a source of genes generally implicated in
cell growth control and tumor development.
We report here the isolation of genes expressed in early tropho
blasts using a PCR3-based cDNA subtraction method. Analysis of
their distribution on a large panel of normal and tumoral cell lines and
tissues has demonstrated that this approach provides a suitable method
for identifying genes overexpressed in various neoplastic cell types.
Among these genes, we identified gene E9, which is preferentially
expressed in tumoral breast.
by grants
from
the Association
pour
la Recherche
sur Ic
Cancer (ARC), Vilejuif, France.
2To whom requestsfor reprintsshould be addressed,at Laboratoired'Immunologie
des tumeurs, CNRS URA 1484, Université
René
Descartes, Faculté
des Sciences Phar
maceutiques et Biologiques, 4, avenue de l'Observatoire, 75006 Paris, France.
are: gestational choriocarcinoma (JAr and JEG-3); hepatocellular carcinoma
(PLC/PRF/5 and Hep G2); colon adenocarcinoma
(LS18O); ovarian carcinoma
(OV1/p and OV1/VCR; the OV1IVCR cell line was derived from OV1/p and
was resistant to vincristin); epidermoid carcinoma (A431); lung carcinoma
(A427); cervix epitheloid carcinoma (HeLa); mammary carcinoma (McF7,
MDA-MB-361, SK-BR-3, BT-20, and BT-474); normal breast transformed
by SV-40 (HBL-100); a neuroblastoma cell line (SHSY-5Y); and a normal
fibroblast cell line (CCL-137). All cell lines were obtained from the
American Type Culture Collection (Rockville, MD), with the exception of
IGR/OV1 (OV1/p; Ref. 10) and 0V1/VCR cell lines, kindly provided by
Dr. Jean Bénard
(Institut Gustave Roussy, Villejuif, France), and SH-SY5Y
and CCL-137, kindly provided by Dr. Pierre-Olivier Couraud (Institut
Cochin de Génétique
Moléculaire,
Paris, France). Cell lines were grown in
Dulbecco's modified Eagle's medium or RPM! 1640 (GIBCO-BRL Labo
ratories, Gaithersburg, MA) supplemented with heat-inactivated
10% fetal
calf serum (GIBCO-BRL Laboratories, Gaithersburg, MA), 10 g@Mnones
sential amino acids, 4 mM glutamine, 100 units/mI penicillin, and 100
3 The
abbreviations
used are: PCR,
polymerase
chain
reaction;
cDNA,
complementary
DNA; MB-RNA, membrane-bound polysomal RNA; SSC, standard saline citrate.
5217
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1994 American Association for Cancer Research.
GENEEXPRESSIONIN TUMORSANDTROPHOBLASTh
@n/mlstreptomycin
at 37°C with 5% CO2. OV1/VCR
Northern Blot Analysis. A total of 5 @g
of total RNA from various tissues
and cell lines were analyzed by electrophoresis on 1% agarose gels containing
derived from the
OV1/p cell line was grown as described (11).
Isolation of Cytotrophoblasts. Cytotrophoblasts were purified as de
scribed by Kliman et al. (12). Briefly, villous tissue from first trimester
2.2 M formaldehyde.
placentae was dispersed with trypsin and deoxyribonuclease I. The dispersed
cells were then purified on a 5—70%Percoll step gradient (Pharmacia, Uppsala,
Sweden). The band at 1.040—1.060 g/ml density was collected. Microscopic
examination revealed it to be comprised of cytotrophoblasts with less than 5%
contamination by nontrophoblastic cells such as macrophages, fibroblasts, and
endothelial cells.
RNA Preparation.
Total RNA was prepared from preconfluent cell cul
tures or frozen tissues using guanidinium isothiocyanate and cesium chloride
gradient centrifugation as adapted from the protocol of Chirgwin et a!. (13).
Endoplasmic reticulum MB-RNA was purified as described previously (14).
@
RNA was then transferred to a reinforced nitrocellulose
membrane (Schleicher and Schuell, Dassel, Germany) with 20X SSC transfer
buffer for 18 h. Transferred RNAs were cross-linked to the filter by UV light
(Stratalinker; Stratagene, La Jolla, CA) before hybridization. The probes used
to hybridize the membrane were excised inserts labeled with 5'-[a@2PJdCFPby
random priming or single-stranded probes generated by PCR using one uni
versal antisense primer. Hybridization was carried out overnight at 42°C,
followed by stringent washing in 0.1X SSC at 50°Cand then autoradiography.
DNA Sequencing and Computer Analysis. PlasmidDNA was prepared
After trypsinization and purification on Percoll, cytotrophoblasts were homog
by the boiling miniprep method (18). Double-stranded sequencing was per
formed with Sequenase, version 2.0 (United States Biochemical, Cleveland,
OH). Primers were either universal primers (M13, T3, and ‘P7)
or internal
specific primers. Reaction products were run on 6% acrylamide/50% urea gels.
enized in a glass dounce, and MB-RNA was isolated on a sucrose gradient
Sequences obtained were compared to sequences from the GENBANK,
including vanadyl ribonucleoside complexes as ribonuclease inhibitors.
EMBL, and SW!SSPROT databases, with the French C!T12/BISANCE server
or with GENEWORKS software (Intelligenetics, Mountain View, CA).
Synthesis of cDNA and Amplification by PCR. A total of 0.1
of
MB-RNA dissolved in 5 p3 diethyl pyrocarbonate-treated water was denatured
with 0.1 M MeHgOH and f3-mercaptoethanol. First-strand cDNA was synthesized
RESULTS
with the avian myeloblastosisvirus reverse transcriptase(Copy kit; In vitrogen,
San Diego, CA), using the modified dT-primershown below (15). RNA-CDNA
Isolation of cDNA clones. We used 6-week amenorrheae placenta
as the richest source of highly proliferative, invasive cytotrophoblasts
in 2% agarose.
A tail of oligodeoxyguanosine
was added to the 3'-end
of the
for synthesis of target cDNA used in the subtraction experiment
first-strand cDNA with tenninal deoxynucleotidyl transferase (ff1, New Haven,
described here. Fig. 1 shows the schematic outline of the cloning
CD, andtheRNAwaseliminatedby alkalitreatment.ThecDNAswereamplified procedure. cDNA was synthesized from both terminal placenta and
heteroduplexes ranging from 0.5 to 2 kilobases were electroeluted after migration
with nonspecific primers, including restriction sites for Nod and Sal]. The se
quence of the “T-primer―
used for reverse transcriptionand PCR was 5'-GACT
CGAGTCGACATCGA11ITT
1F1T1T111FF
1-3'.
The
GCATCGGCGCGcICCGCGGAGGCCCCCCCCCCCCCC-3'
“C-primer―was 5'-
as described by
Preparation of th. subtracted cDNA library
Loh et aL (16).
The reaction mixtures consisted of 1.25 m@ieach of the four deoxyribonu
cleotide triphosphates, 0.5 p@M
of each primer, and 2.5 units of Taq polymerase
(Perkin-Elmer Cetus, Norwalk, CT) in 50 m@iKCI, 10 m@ Tris-HCI (pH 8.3),
Preparation of target cDNA
Preparation of driver eDNA
Isolation of cytotrophoblasts on
Percoll gradient
3.5 mM MgCl2 and 0.01% gelatin. Amplification was performed in a Hybaid
thermocycler for 25 cycles consisting of 20 s at 94°C,30 s at 55°C,and 1 mm
at 72°C.The products were loaded onto a 1% low-melting agarose gel (FMC
BioProducts, Rockland, ME). Then, the 0.5—2-kilobase region was excised and
reamplified under the same conditions. The products were precipitated, di
gested with Notl and Sail, size selected, and electroeluted as above.
Purification of MB-RNA
Total cDNA from terminal placenta and from T-lymphocytes activated with
Purification of poly(A)RNA
from terminal placenta and
activated T lymphocytes
1“
strandcDNA synthesis
phytohemagglutinin and cultivated for several days in the presence of 1L2 was
prepared using a double-stranded cDNA synthesis system (In vitrogen).
Construction of the Subtracted Library. Subtractive hybridization was
+
dG tailing
performed basically as described by Klickstein (17); 0.2 pg of PCR-amplified
early cytotrophoblast cDNA digested with NotI and Sal! was mixed with 8 @g
of activated T-lymphocyte cDNA and 8 @gof terminal placenta cDNA
PCR amplification with Noti and
digested with Alit] and RsaI (New England Biolabs, Beverly, CA), and then
SaIl primers
Synthesis of double-stranded
cDNA
dissolved in 40 pi of hybridization buffer [50% deionized formamide, 10 mM
sodium phosphate buffer (pH 7), 5X SSC, 0.1% sodium dodecyl sulfate, and
10 mM EDTA), denatured 5 mm at 98°C,and incubated 24 h at 37°C.Common
sequences
between
target
and driver
reannealed
to form
duplexes
Noti and £@digestion
between
short (AlulRsa, driver) and long (Not/Sal, target) fragments, inhibiting forma
tion of cohesive ends. Uninhibited reannealing of complementary strands of
cDNAs specifically expressed in early cytotrophoblasts allowed the regener
ation of NotI and Sail cohesive ends for unidirectional cloning into respective
sites in the pBSK
11+ phagemid
Screening of the Library.
vector
(Stratagene,
@_
+
+
+
La Jolla, CA).
The subtracted cDNA library was plated at low
Denaturation and reannealing
density, and colonies were picked for growth in duria-Bertani medium and
PCR amplification. Plasmid DNA prepared by the boiling miniprep method
(18) and digested with Not! and Sal! or PCR products of inserts amplified
using the original primers were analyzed by Southern blot from 1.2% agarose
gels. The gels were soaked in 0.4 M NaOH and sandwiched between two nylon
membranes (Hybond N; Amersham, Las Ulis, France) in order to obtain two
copies. The probes used for Southern hybridization were total cDNAs synthe
sized from both terminal placenta and activated T-lymphocytes, labeled with
5'-[a32P]dCrP (Amersham) by random priming. Hybridization was carried
out overnight at 42°C,followed by stringent washing in 0.1X SSC at 50°Cand
then autoradiography.
Alul and Asal digestion
Ligation into pBSKII+digested by Notl and Sail
Transformation into Xl1-blue host bacteria
Fig. 1. Schematic outline of the procedure for the construction of the subtracted cDNA
library enriched for sequences expressed in early cytotrophoblasts (6-week amenorrhea).
A detailedexplanationof the subtractedcDNA libraryscreeningis providedin the text
MB-RNA, endoplasmic reticulum membrane-bound polysomal RNA.
5218
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1994 American Association for Cancer Research.
GENE EXPRESSION IN TUMORS AND TROPHOBLASTS
A.
1_@2
Fig. 2. Northern blot analysis of the isolated cDNA
clones on RNA from terminal placenta, early placenta,
and activated T-lymphocytes. A, five @g
of total RNA
prepared from terminal placenta (Lane I), 7.5-week
amenorrhea placenta (Lane 2), and activated T-lym
phocytes (Lane 3) were hybridized with 32P-labeled
probes for selected cDNA clones from the subtracted
library. The autoradiographs were obtained after expo
sure with an intensifying screen at —80°C
for 3 days
with probes B9 and E9 and 2 days with probe E4. The
probe used is indicated above each blot. B, control
hybridization with an a-tubulin probe. C, autoradio
grams were quantified by scanning densitometry. Val
ues were normalized to the intensity of each respective
28S rRNA band after scanning densitometry on photo
graph of ethidium bromide-stained nitrocellulose mem
branes, and ratios were given in relation to terminal
placenta. Normalization to a-tubulin hybridization sig
nals gave equivalent results; TP, terminal placenta; EP.
early placenta; ATL activated T-lymphocytes.
E4
B9
I
3
23
123
28S.
28S
18S
I 8S.
B.
IT
F
—@
C.
0
0
C
-e
0
Cl)
-a
a)
>
a)
cc
TP
EP
TP
EP
ATL
TP
EP
inserts were analyzed by Southern blot. Among these inserts, 72
recombinant clones which showed weak or no hybridization signals
with radiolabeled total cDNA probes from terminal placenta (T probe)
and activated T-lymphocytes (A probe) were selected. Clones were
sibling grouped to 13 discrete sequence populations by matrix cross
hybridization of dot blots. The frequency of clone duplication ranged
from 1 to 48. Following this first step of screening, we plated 2,500
colonies of the subtracted cDNA library and performed a second
screening. In light of the redundancy of clones isolated during the first
screening, we modified our screening procedure. A cDNA probe
which encompassed each previously selected clone was synthesized
(M probe); 253 inserts from white colonies were amplified by PCR
and analyzed on duplicate Southern blots. The two blots were hybrid
ized with the M probe and a mixture of the A+T probes, respectively.
Eighteen clones which did not hybridize with these probes were
isolated. Sibling grouping defined eight new discrete sequence pop
ulations. The frequency of duplication for these clones ranged from
one to four.
Sequence Analysis of Selected cDNA Clones. Partial DNA se
quencing was performed on the 21 selected cDNA clones. In general,
250—350bases were sequenced for each, sufficient for determining
identity to known genes. Database interrogation originally revealed 10
novel sequences. The presence of long open reading frames and
polyadenylation signals suggested functional transcripts. The other 11
cDNA clones corresponded to known molecules, as summarized in
Table 1. Clones B9 and E4, originally among the 10 novel sequences,
are listed with their determined identities. Clones of interest corre
sponding to novel sequences were completely sequenced from both
strands.
Northern Blot Analysis of Tissue Distribution. The enrichment
attained by subtractive hybridization was confirmed by comparing the
5219
activated T-lymphocyte polyadenylated RNA and used as a driver in
an attempt to diminish genes expressed at all stages of placental
development and those encoding molecules such as cytokines previ
ously identified in T-lymphocytes. The total number of recombinant
clones was estimated at 32,000. The size of the inserts ranged between
500 and 1,000 base pairs. Two rounds of screening were performed to
isolate clones preferentially expressed in early placenta. In the first
screening, 5,000 colonies were plated. A total of 341 randomly chosen
clonesClone
@
Table 1 Identification of cDNA
transcript(s)Csymbol cDNA inserts'
(kilobases)A4
Frequency―
pairs)
mRNA
Homologous gene
NDd£5
2
430
NDE9
1
400
Novel
Novel
0.7Fl
1
580
Novel
0.8;1.25D8
2
308
Novel
0.6;1.57B6 1
ND8A6
1
1.48D1
1
ND5A8
1
1.6AlO
1
NDAl
16
NDFl
53
NDE4
1
2
1.5;3.1FlO 1
ND7D5
1
900
300
500
600
1000
430
610
203
780
470
Novel
Novel
Novel
Novel
Cytochrome C oxidase I
128 mitochondrial rRNA
16S mitochondrialrRNA
95 mitochondrial rRNA
ADE2-1
a-hCG
1.68B6
1
580
CEA/psgll
NDE8
NDB9
0.7BlO
NDE6
NDa
1
1
1
1
1
800
545
550
330
435
Fibronectin
Nucleophosmin
Ribosomalprotein518
Ubiquitin
Vinculin
Number
ofsiblings.
ofinserts.C
b Insert
size
Transcript
d Not
as determined
size
determined;
as
by
determined
CEA,
sequencing
from
Northern
carcinoembryonic
or
electrophoresis
blots.
antigen.
of
PCR
products
ATL
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1994 American Association for Cancer Research.
I@.
j ES
GENE EXPRESSION IN TUMORS AND TROPHOBLASTS
E4
A.
Fig. 3. Northern blot analysis of cDNA clones on
RNA from normal and tumoral tissues. A, five @g
of
total RNA prepared from human non-cancerous sur
gical specimen (N), and cancer tissues (1), when
possible from the same patient, were hybridized with
32P-labeledcDNA probes for clones B9, E4, and E9.
Autoradiographs were obtained after exposure for 1
day with probe B9 and 2 days with probes E9 and E4.
Only significant lanes are shown. The probe used is
indicated above each blot. B, control hybridization
p iesJ
a.
with a-tubulin (colon, thyroid, and bladder) or 13-ac
grams presented below each Northern blot indicate
the densitometric ratios between B9, E4, or E9 and
285 rRNA levels. Autoradiograms were quantified by
scanning densitometry.
@
18
25
I 16
14
12
Values were normalized to
the intensity of each respective 285 rRNA band after
scanning densitometry on photograph of ethidium
bromide-stained nitrocellulose membranes, and ratios
were given in relation to the lowest value. Normal
ization to control probes gave similar results.
30
2C
tin probes (rectum, stomach, and breast). C, histo
10
18
(
flH[flflflr.,flflhlflfl
NTNTNTNTNTNT
@e@umsih5F@
N
T
N
T
N
T
N
T
N
T
NT
Th@3@ 5@
level of expression of the transcripts in early placenta versus both
terminal placenta and activated T-lymphocytes by Northern blotting,
as shown for clones B9, E4, and E9 taken as representative (Fig. 2).
While B9 and E4 expression was also high in lymphocytes, a general
survey of several clones, including mitochondrial sequences, revealed
mean densitometric scan ratios of 2.2:1:1.4 between hybridization
signals on equal amounts of early placenta, terminal placenta, and
lymphocyte total RNA, respectively. We then analyzed the expression
level of B9, E4, and E9 on a large panel of tumoral and corresponding
normal tissues, in order to determine the potential implication of these
genes in the neoplastic process.
Identification of B9, E4, and E9 Clones. Clone B9 hybridized
with a 0.7-kilobase transcript abundantly expressed in early tropho
blast. Moreover, expression of B9 was increased in tissue specimens
from colon, rectum, stomach, breast, thyroid, and bladder cancers in
comparison with their normal counterparts (Fig. 3). Finally, B9 was
expressed sparsely in terminal placenta and at varying levels in most
other normal tissues. Sequence analysis revealed 549 nucleotides
containing a polyadenylation signal and an open reading frame of 152
amino acids. In a first attempt, database comparison using the FASTA
program (19) did not reveal any significant homologies to known
sequences. Subsequent search showed 90.7% nucleic acid and 100%
amino acid homology with the mouse KE3 sequence. The correspond
ing rat amino acid sequence, as predicted from cDNA, was also found
to be identical (20). The predicted KE3 gene product has been
described as the ribosomal protein S18, based on homology with
bacterial ribosomal protein 513 and the electrophoretic properties of
the in vitro translation product (21). On the basis of this identity, we
concluded that the B9 clone encoded the human ribosomal protein
S18 (22). The sequence of this clone also showed a structural homol
ogy with fos-related molecules based on the heptad repeat of leucine
residues (leucine zipper) preceded by a basic region (23). PROSITE
analysis revealed six putative phosphorylation sites.
Clone E4 hybridized with two transcripts at 1.5 and 3.1 kilobases,
for which the expression level was higher in early trophoblast than in
terminal placenta. It was also preferentially expressed in numerous
tumoral tissues such as colon, rectum, stomach, breast, thyroid, and
bladder, in comparison with their normal counterparts (Fig. 3). DNA
sequence analysis revealed a 780-base pair insert with a polyadeny
lation signal and an open reading frame of 185 codons. Initially
studied as a novel sequence, repeated database searches using the
FASTA program revealed a recent entry by a match of 125 of 130
amino acids, described as the bifunctional enzyme ADE2H1 involved
in de novo purine biosynthesis (24). We found five amino acid
differences between our sequence deduced from a clone of the sub
tracted cDNA library and the published sequence. As the cDNA clone
corresponding to the published sequence was isolated from a HeLa
cell cDNA library, we investigated the potential existence of muta
tions. We performed reverse transcription-PCR on RNA samples from
both early trophoblast and HeLa cells. The products were analyzed by
denaturing-gradient gel electrophoresis. The pattern of migration did
not suggest any sequence differences between normal and tumoral
cells (data not shown).
Clone E9 was preferentially expressed in early placenta in compar
ison with terminal placenta and activated T-lymphocytes. Northern
blot analysis revealed a transcript of 0.7 kilobase. The nucleotide
sequence of the 580-base pair insert included a polyadenylation signal
and several small open reading frames ranging from 35 to 67 amino
acids. PROSITE analysis revealed an N-glycosylation site and poten
tial phosphorylation sites. Comparison of both nucleic acid and the
predicted polypeptide sequences with GENBANK and SWISSPROT
databases, respectively, revealed no significant homology with previ
ously known sequences. The E9 sequence has been submitted to
GENBANK (GSDB) with the accession number 134839. In order to
investigate its potential association with the tumoral phenotype, we
analyzed its distribution in various normal and tumoral tissues by
Northern blotting (Fig. 3). E9 was overexpressed in tumoral rectum
and bladder, and more strikingly, in tumoral breast tissues in corn
parison with their normal counterparts. When comparing the expres
sion level of E9 in tumor and in the adjacent normal tissue from five
patients, we found it elevated 7-, 5.2-, 3.2-, and 330-fold in patients A,
B, D, and E, respectively (Fig. 4). Northern blot analysis of its
expression in normal and tumor cells also revealed a high level of
transcripts in tumoral breast cell lines (Fig. 5).
DISCUSSION
In an attempt to identify genes preferentially expressed in both
early trophoblastic cells and tumors, we constructed a subtracted
cDNA library and analyzed the expression level of the cDNA clones
selected after differential screening on a large panel of tumoral and
normal tissue samples. Throughout recent decades, numerous tech
niques based on tumor cells as sources of investigation have been used
in the search for either tumor-associated antigens or genes involved in
neoplastic growth. Immunological approaches have led to the finding
of a small number of highly immunogenic antigens usually abun
dantly present in normal tissues (25). Molecular biology methods led
to the characterization of a limited set of genes differentially cx
pressed in tumors of a given histological origin (26, 27). In the search
for an approach enabling characterization of classes of molecules
5220
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1994 American Association for Cancer Research.
GENE EXPRESSION IN TUMORS AND TROPHOBLASTS
corresponding to secreted molecules such as carcinoembryonic anti
gen and human chorionic gonadotropin were indeed found. However,
the presence of cytoplasmic mRNA, such as that for nucleophosmin,
was not completely eliminated in this enrichment method. We used a
molecular subtraction approach in order to construct a library enriched
.@
in sequences preferentially expressed in early cytotrophoblasts versus
terminal placental cells and activated T-lymphocytes. Molecular sub
.@
traction has proven to be a powerful tool for isolation of cDNA clones
encoding preferentially expressed transcripts (28). Classical ap
.@
proaches require large amounts of target RNA for cDNA synthesis,
limiting their use on small tissue samples. We developed a PCR-based
technique in which small amounts of polysome-bound RNA could be
T;..I.
:.@.:..i
used as target. We chose to amplify mRNA size-selected at 0.5 to 2
@N T N T N T N TNT
kilobases, corresponding to the average size of mRNA and also
remaining within reasonable limits for PCR amplification. Ultimately,
A
B
C
D
E
the mean insert size was 0.5—1kilobase, reflecting PCR and cloning
Fig. 4. Relative abundance of E9 mRNA in normal and tumoral breast samples. Five
bias. Subtraction was performed with cDNA from both terminal
ILg oftotal RNA prepared from five pairs (A-E) of adjacent normal human breast tissues
(N) and human breast cancer tissues (7) were hybridized with 32P-labeled cDNA probe
placenta and activated T-lymphocytes as driver. Northern blot analy
for clone E9 (each paired sample was obtained from the same patient). Autoradiograms
sis revealed that subtraction was less efficient for eliminating tran
were quantified by scanning densitometry. Values were normalized to the intensity of each
respective
28S rRNA band after scanning
densitometry
on photograph
of ethidium
scripts expressed in activated T-lymphocytes. As the total amount of
bromide-stained nitrocellulose membranes, and ratios were given in relation to the lowest
nucleic acid is limited in a concentrated hybridization mix, adding two
value. The results were confirmed by normalization to 13-actinhybridization signals.
types of driver cDNA lowers their respective target:driver ratio, and it
Breast cancer types and tumor stage are: A, intraductal comedocarcinoma and invasive
intralobular carcinoma (stage T,/N@JM.@,);
B, infiltrating ductal carcinoma, (stage T@/N@/ is likely that subtraction was more efficient for molecules present in
M0); C, infiltrating ductal carcinoma (stage T@/N@,fM.,);
D, infiltrating ductal carcinoma
both terminal placenta and lymphocytes.
(stage T,/N,fM@);E, infiltrating ductal carcinoma (stage T@/N@JM.,).
From a methodological point of view, our PCR-based DNA-driven
cDNA subtraction method was neither exhaustive nor meant to be
involved in either cell growth control or tumor development, we chose
representative of the entire RNA population. Subtraction with double
cytotrophoblastic cells, which display a pseudomalignant phenotype,
stranded cDNA has been described as giving enrichment of up to 50%
as a source of genes encoding such molecules. Indeed, the placenta, at (17), and PCR amplification of heterogeneousmolecule populations
an early stage of development, shares several common features with
tends to differ for each molecular species and is irreproducible be
neoplastic cells.
tween identical samples due to template size, sequence, and abun
Endoplasmic reticulum-bound RNA was used in order to enrich for dance. Amplification may favor the presence of some cDNA species
transcripts encoding membrane or secreted proteins, and clones
to the detriment of others. While some sequences may be lost, rare
P
@
A.
1.
28S@
18s-
Fig. 5. Northern blot analysis of E9 RNA in var
ious normal and tumoral cell lines and lymphocytes.
@
A. 1-5
of total RNA prepared from human nor
mal and tumoral cells were hybridized with 32Plabeled cDNA probe E9. The autoradiographywas
for 1 day at —80°C.
ATL. activated T-lymphocytes
(phytohemagglutinin + interleukin 2); T LYMPH,
T-lymphocytes. Two to five @gof total RNA pre
pared from six breast tumoral cell lines were hybrid
ized with 32P-labeledcDNA probe E9. The exposure
a
C.
40
8
30
6
wasfor oneday at —80°C.
B, controlhybridization
@
with an a-tubulin probe. C, histograms presented
below each Northern blot indicate the ratios between
E9 and 28S rRNA levels. Autoradiogramswere
quantified by scanning densitometry. Values were
normalized to the intensity of each respective 28S
rRNA band after scanning densitometry on photo
graph of ethidium bromide-stained nitrocellulose
membranes and ratios were given in relation to the
lowest value. Normalization to a-tubulin hybridiza
tion signals gave equivalent results.
120
4
@1o
C'
@
nfl
!
nn@n
2
H
a@
0
.@@j>-.J
@ii@!a@@E
00<@;
I
in
<@-OC.)
@i@mm
5221
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1994 American Association for Cancer Research.
m
@-—
I
GENE EXPRESSION IN TUMORS AND TROPHOBLASTS
molecular species of potential interest might be amplified; among the
eight clones, the sequences of which did not reveal any significant
homology to sequences in the databases, it is noteworthy that
transcripts of four clones were not detected by Northern blot of total
RNA. This observation, as well as the presence of polyadenylation
signals, suggests that these clones correspond to rare transcripts.
Partial sequence analysis of the selected cDNA clones, which
revealed their identity, enabled us to further evaluate the efficacy of
the subtraction. Among the 21 cDNA clones, 11 were first identified
as corresponding to known molecules. One-half of these known genes
were mitochondrial. The abundance of mitochondrial sequences in
subtracted libraries is a well-kown nuisance. In this case, the partic
ularly high numbers of mitochondrial sequences may have been due to
entrapping of mitochondria along with the membrane in the sucrose
gradient performed to isolate RNA. In addition, it is striking that
mitochondrial sequences were most often preferentially expressed in
the early trophoblast used as the target cDNA in comparison to
terminal placenta used as the driver cDNA, as assessed by Northern
blot analysis. The most likely explanation is that variations in expres
sion of mitochondrial sequences are related to energy metabolism in
the trophoblastic tissue. It is also noteworthy that mitochondrial genes
have been implicated in cancer development, not only by insertion
into the nuclear genome, disrupting normal genes, but also through
increased expression (29, 30). Among several alterations in chemi
cally induced rat hepatomas, very high levels of mitochondrial
transcripts have been noted. Interestingly, overexpression of NDS
transcripts in rat hepatomas occurs during the early steps of carcino
genesis and therefore does not result simply from ultrastructural
changes following neoplasia. Thus, an increased level of expression of
mitochondrial transcripts in early trophoblasts could correlate with its
pseudomalignant pattern.
Among the known genes found in the present study, several have
been described as being related to cancer. In addition to the obvious
carcinoembryonic antigen/pregnancy-specific
protein family of cell
adhesion molecules (31), ubiquitin mRNA is known to be one of the
major stress-induced molecules (32), and ubiquitin immunoreactivity
has been found to be increased in malignancy (33).
PA, first considered as an unkown gene expressed in early placenta,
was finally identified as encoding an enzyme involved in de novo
purine biosynthesis, ADE2H1. The observed sequence differences
between ADE2H1, cloned from HeLa, and PA might stem from
cloning artifacts or sequencing errors, although PCR-induced se
quence artifacts were not noticed when comparing other known se
quences. Denaturing-gradient gel electrophoresis analysis performed
after reverse transcription-PCR on RNA samples from both early
trophoblast and HeLa cells may not have been adequate to differen
tiate between the different transcripts expressed in each cell type, and
very low levels of a highly homologous transcript cannot be excluded.
The high level of expression of an enzyme implicated in nucleotide
synthesis correlates to the highly mitotic and proliferative status of
cytotrophoblasts of first trimester placentae. Its overexpression in
various cancer tissues might also reflect the highly proliferative
phenotype of neoplastic cells.
The B9 gene product has been defmed as human ribosomal protein
518, and its expression level is significantly increased in some tumors.
Interestingly, numerous ribosomal proteins have been demonstrated to
be overexpressed in tumors. These include mRNA for ribosomal
proteins S3 and 519, which were shown to be expressed at increased
levels in colon carcinomas and polyps (34, 35). Since the index of
proliferation of colorectal carcinomas is not significantly different
from that of normal mucosa, a higher percentage of dividing cells in
tumor tissue might not be the only factor. A high level of expression
of ribosomal protein L19 has also been described in breast tumors that
overexpress erbB-2 (36). In the case of ribosomal protein 518, the
homologous mouse KE3 gene belongs to a group of six genes (KEJ—5
and SET) located in the H-2K region of murine major histocompati
biity complex (37). A homologue of this H-2K region has also been
localized within the human major histocompatibiity complex region
known to contain several genes encoding proteins with important
biological activities (38). This possible localization, in addition to
pattern homologies of B9 to the fos and jun gene families, should
stimulate further studies so as to explore a possible complementary
role of the B9 gene in transcriptional regulation.
Finally, the study of clone E9 confirmed that this experimental
approach enabled the identification of new genes overexpressed in
tumor tissues. Indeed, the higher expression of E9 in tumoral tissues
in comparison with their normal counterparts was particularly striking
in breast carcinomas. The high level of expression of the E9 gene in
six tumoral breast cell lines suggested that it might be expressed by
neoplastic cells themselves rather than by stromal cells surrounding
the tumor. Recently, a new member of the family of metalloproteinase
enzymes which degrade the extracellular matrix, stromelysin-3, was
found to be overexpressed in breast adenocarcinomas by a subtractive
hybridization method (39). In contrast to E9, its expression was
restricted to fibroblasts immediately surrounding neoplastic cells of
the invasive component of the tumor. The identification of the E9
gene, overexpressed by neoplastic breast cells, will enable us to
investigate its function with regard to malignant transformation. Ex
periments are in progress to characterize the complete genomic
sequence of the E9 gene and its translated product.
In conclusion, the PCR-based DNA-driven cDNA subtraction
method described here appears to be suitable for construction of a
cDNA subtraction library from small amounts of RNA. This study
demonstrates that the trophoblast is a source of genes preferentially
expressed in various tumor cells and confirms the pseudomalignant
nature of early placenta.
ACKNOWLEDGMENTS
We thank P-O. Couraud,T. Hercend, T. Poynard,D. Vidaud, and M.
Vidaud for helpful discussions. We are indebted to the surgeons and pathol
ogists of the Institut Gustave-Roussy for their contribution, and we acknowl
edge S. Richon, J. Bombled, and A. Girot for their excellent technical
assistance.
REFERENCES
1. Goustin,A. C., Betsholtz,C., Pfeifer-Ohlamn,S., Persson,H., Rydnert,J., Bywater,
M., Holmgren,G., Heldin, C-H., Westermark,B., and Ohlsson, R. Coexpressionof
the sis and myc proto-oncogenesin developing humanplacentasuggests autocrine
control of trophoblast growth. Cell, 41: 301—312,
1985.
2. Holmgren, L, Claesson-Welsh,
L, Heldin, C-H., and Ohlsson, R. The expression of
PDGF a- and 13receptors in subpopulations of PDGF producing cells implicates
autocrine stimulatory loops in control of proliferation of cytotrophoblasts which have
invaded the maternal endometrium. Growth Factors, 6: 219-231,
1992.
3. Reshef,E., Lai, M., Rao,C. V., Pridham,D. D., Chegini,N., andLuborsky,J.L The
presence of gonadotropin receptors in nonpregnant human uterus, human placenta,
fetal membranes, and decidua. J. Clin. Endocrinol. Metab., 70: 421—429,1990.
4. Horowitz, G. M., Scott, R. T., Jr., Drews, M. R., Navot, D., Hofman, G. E.
Immunohistochemicallocalizationof transforminggrowth factor-a in human endome
trium, decidua, and trophoblast. J. tlin. Endocrinoi Metab., 76: 786—791,1993.
5. Hofmann, 0. E., Draws, M. R., Scott, R. T., Jr., Navot, D., Heller, D., and Deigdish,
L Epidermalgrowth factor and its receptor in human implantationtrophoblast:
immunohistochemical evidence for autocrine/paracrine function. J. Clin Endocrinol.
Metab., 74: 981—988,
1992.
6. Adamson, E. D. Expression of proto-oncogenes in the placenta. Placenta, 8:
449—466,1987.
7. Ohlsson, R., and Pfeifer-Ohlsson,
S. B. Cancer genes, proto-oncogenes,
and devel
opment. Exp. Cell Rex., 173: 1—16,
1987.
8. Ohlsson, R. Growth factors, protooncogenes and human placental development. Cell.
Differ. Dcv., 28: 1—15,
1989.
9. Strickland,S., and Richards,W. G. Invasionof the trophoblasts.Cell, 71: 355—357,
1992
5222
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1994 American Association for Cancer Research.
GENE EXPRESSION IN TUMORS AND TROPHOBLkSTS
down expressed in Wilms' tumor by a subtractive hybridization approach. Cancer
Res., 53: 2888-2894, 1993.
10. Bénard,
J., Da Silva, J., Dc Blois, M. C., Boyer, P., Duvillard, P., Chine, E., and Riou,
G. Characterization of a human ovarian adenocarcinoma line, IGR OV1, in tissue
culture and in nude mice. Cancer Res., 45: 4970—4979,1985.
11. Bénard,
J., Da Silva, J., Teyssier, J. R., and Riou, 0. Overexpressionof MDRJgene
in a multiple-drug-resistant human ovarian carcinoma cell line. tnt. J. Cancer, 43:
471—477,1989.
12. Kliman, H. J., Nestler, J. E., Sermasi, E., Sanger, J. M., and Strauss, J. F., III.
Purification,
characterization,
and in vitro differentiation
of cytotrophoblasts
27. Hutchins, J. T., Deans, R. I., Mitchell, M. S., Uchiyama, C., and Kan-Mitchell,
28.
from
human term placentae. Endocrinology, 118: 1567—1582,
1986.
13. Chirgwin, J. M., Przybyla, A. E., MacDonald, R. J., and Rutter, W. J. Isolation of
biologically active ribonucleic acid from sources enriched in ribonuclease. Biochem
istry, 18: 5294—5299,1979.
14. Mechler, B. M. Isolation of messenger RNA from membrane-bound polysomes. In:
29.
S. L. BergerandA. R. Kimmel(eds.),Guideto MolecularCloningTechniques,pp.
31.
241—248.
New York: Academic Press, Inc., 1987.
15. Frohman, M. A., Dush, M. K., and Martin, U. R. Rapid production of full-length
cDNAs from rare transcripts: amplification using a single gene-specific oligo
nucleotide primer. Proc. Nail. Acad. Sci. USA, 85: 8998—9002,1988.
16. Loh, E. Y., Elliott,I. F., Cwirla,S., Lather,L L, andDavis, M. M. Polymerasechain
reaction with single-sided specificity: analysis of TecH receptor 8 chain. Science
(WashingtonDC),243:217—220,
1989.
17. Klickstein, L B. Production of a subtracted cDNA library. In: F. M. Ausubel, R.
Brent, R. H. Kingston, D. D. Moore, J. G. Scidman, J. A. Smith, and K. Strulil (eds.),
Current Protocois in Molecular Biology, pp. 5.8.9—5.8.15.New York: Wiley
Interscience, 1989.
18. Del Sal, 0., Manfioletti, G., and Schneider, C. A one-tube plasmid DNA mini
preparationsuitablefor sequencing.NucleicAcidsRca.,16: 9878,1988.
19. Lipman,D. J., and Pearson,W. R. Improvedtools for biological sequence compar
hon. Proc. Natl. Acad. Sci. USA, 85: 2444-2448,
1988
30.
32.
J.
Novel gene sequences expressed by human melanoma cells identified by molecular
subtraction. Cancer Res., 51: 1418—1425,1991.
Hedrick, S. M., Cohen, D. I., Nielsen, E. A, and Davis, M. M. Isolation of cDNA
clones encoding T-cell-specific membrane-associated protein. Nature (Lond.), 308:
149—153,1984.
Corral, M., Defer, N., and Paris, B. Isolation and characterization of complementary
DNA clones for genes overexpressedin chemically induced rat hepatomas.Cancer
Res., 46: 5119—5124,1986.
Corral, M., Paris, B., and Baffet 0. increased level of the mitochondrial ND5
transcript in chemically induced rat hepatomas. Exp. Cell Res., 184: 158—166,1989.
Santa, B., Abraham, F., Serge, J., Nicole, B., Kinji, S., and Oifford, P. S. Carcino
embryonic antigen, a human tumor marker, functions as an intercellular adhesion
molecule. Cell, 57: 327—334,1989.
Fornace, A. J., Jr., Alamo, I., Jr., Hollander, M. C., and Lamoreaux, E. Ubiquitin
mRNA is a major stress-induced transcript in mammalian cells. Nucleic Acids Res.,
17: 1215-1230, 1989.
33. Ishibashi, Y., Takada, K., Joh, K, Ohkawa, K., Aoki, T., and Matsuda, M. Ubiquitin
immunoreactivity in human malignant tumors. Br. J. Cancer, 36: 320—322,1991.
34. Kondoh, N., Schweinfest, C. W., Henderson, K. W., and Papas, T. S. Differential
expression ofSl9 ribosomal protein, laminin-binding protein, and human lymphocyte
antigen class I messenger RNAs associated with colon carcinoma progression and
differentiation. Cancer Res., 52: 791—796,1992.
35. Pogue-Geile, K., Geiser, J. R., Shu, M., Miller, C., Wool, I. 0., Meisler, A. I., and
Pipas, J. M. Ribosomal protein genes are overexpressed in colorectal cancer: isolation
of a cDNA clone encoding the human S3 ribosomal protein. Mol. Cell. Biol., 11:
3842—3849,
1991.
36. Henry, J. L., Coggin, D. L., and King, C. R. High-level expression of the ribosomal
protein L19 in human breast tumors that overexpress erbB-2. Cancer Res., 53:
20. Chan, Y-L, Paz, V., and Wool, I. 0. The primary structure of rat ribosomal protein
S18. Biochem. Biophys. Res. Commun., 178: 1212—1218,
1991.
1403—1408,1993.
21. MacMurray, A. J., and Shin, H-S. The murine MHC encodes a mammalian homolog
of bacterial ribosomal protein S13. Mamm. Genome, 2: 87—95,1992
37. Janatlpour, M., Naumov, Y., Ando, A, Sugimura, K., Okamoto, N., Tsuji, K., Abe,
22. Chassin, D., Bellet, D., and Koman, A. The human homolog of ribosomal protein
K., and Inoko, H. Search for MHC-associatedgenes in human: five new genes
S18. Nucleic Acids Rca., 21: 745, 1993.
centromeric to HIA-DP with yet unknown functions. Immunogenetics, 35: 272—278,
23. Kerppola, T. K, and Curran, T. Transcription factor interactions: basics on zippers.
1992.
Cun. Opin. Struct. Biol., 1: 71—79,
1991.
38. Abe, K., Wei, J-F., Wei, F-S., Hsu, Y-C., Uehara, H., Artzt, K., and Bennett, D.
24. Minet, M., and Lacroute, F. Cloning and sequencing of a human eDNA coding for a
Searching for coding sequences in the mammalian genome: the H-2K region of the
multifunctional polypeptide of the purine pathway by complementation of the ade2—
mouse MHC is replete with genes expressed in the embryos. EMBO 3., 7: 3341—3449,
101 mutant in Saccharomyces cerevisiae. Curr. Genet., 18: 287—291,1990.
1988.
25. Reisfeld, R. A., and Cheresh, D. A. Human tumor antigens. Adv. Immunol., 40:
39. Basset, P., Bdllocq, J. P., Wolf, C., Stoll, I., Hutin, P., Limacher, J. M., Podhajcer,
323—377,1987.
0. L, Chenard,M. P., Rio, M. C., andChambon,P. A novel metalloproteinasegene
26. Austruy, E., Cohen-Salmon, M., Antignac, C., Béroud,
C., Henry, I., Van Cong. N.,
specifically expressed in stromal cells of breast carcinomas. Nature (Lond.), 348:
Brugiêres,L, Junien, C., and Jeanpierre, C. Isolation ofkidney complementary DNAS
699—704,1990.
5223
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1994 American Association for Cancer Research.
Identification of Genes Overexpressed in Tumors through
Preferential Expression Screening in Trophoblasts
Dorine Chassin, Jean-Louis Bénifla, Claire Delattre, et al.
Cancer Res 1994;54:5217-5223.
Updated version
E-mail alerts
Reprints and
Subscriptions
Permissions
Access the most recent version of this article at:
http://cancerres.aacrjournals.org/content/54/19/5217
Sign up to receive free email-alerts related to this article or journal.
To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Department at [email protected].
To request permission to re-use all or part of this article, contact the AACR Publications
Department at [email protected].
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1994 American Association for Cancer Research.