Inhibition of the erbB-2 tyrosine kinase receptor in breast cancer cells by phosphoromonothioate
and phosphorodithioate antisense oligonucleotides
Inhibition of the erbB-2 tyrosine kinase receptor in breast cancer cells by phosphoromonothioate and phosphorodithioate antisense oligonucleotides
James P.
Vaughn
,
Joanne
Stekler
1
,
Samuel
Demirdji
2
,
Jeffrey K.
Mills
3
,
Marvin H.
Caruthers
2
,
J. Dirk
Iglehart
4
and
Jeffrey R.
Marks*
Department of Surgery, Box 3873, Duke University Medical Center and
1
Duke University Medical School,
Durham
, NC 27710,
USA
,
2
Department of Chemistry and Biochemistry, Box 215, University of Colorado,
Boulder
,
CO 80309-0215,
USA
and
3
Department of Radiation Oncology and
4
Departments of Surgery, Cell Biology
and Pathology, Box 3873, Duke University Medical Center,
Durham
, NC 27710,
USA
Received July 3, 1996;
Revised and Accepted September 26, 1996
ABSTRACT
Antisense activity against erbB-2 of a variety of sulfur-modified oligonucleotides was examined in a breast cancer cell line
which overexpresses this oncogene. Using a 15 base anti-erbB-2 sequence previously shown to be effective, various backbone
configurations containing phosphoromonothioate or phosphorodithioate linkages
were evaluated for antisense activity by a two-color flow cytometric assay. This sequence was effective in inhibiting the
production of erbB-2 protein when it was configured as a monothioate at each linkage and as
an alternating dithioate/phosphodiester. Both of these compounds were also able
to specifically inhibit erbB-2 mRNA expression, indicative of RNase H-mediated activity. The same sequence protected by either three
dithioate or three monothioate linkages at each end was ineffective as an
antisense reagent, suggesting that endonuclease activity is a significant
determinant of the stability of oligonucleotides. Finally, the erbB-2 sequence target was shifted in an effort to improve antisense activity.
A new lead sequence was identified that was significantly more effective in
inhibiting erbB-2 protein levels and retained activity at lower concentrations.
INTRODUCTION
Antisense oligonucleotides have long held the potential to decrease the
expression of a targeted gene by inhibiting transcription or translation and
thereby achieve a phenotypic effect based upon expression of that gene.
However, many of the effects achieved by these oligonucleotides may not be
mediated by inhibition of the target gene (
1
,
2
). For both investigational and therapeutic applications, a number of obstacles
must be overcome in order to both demonstrate and achieve specific inhibition
in living cells: (i) the compounds must be delivered and maintained at a sufficient concentration and to an intracellular compartment that makes
them available for interaction with their target; (ii) the oligonucleotides
must have sufficient stability against exonucleases and endonucleases so that they can act catalytically; (iii) they must act in a
sequence-specific manner to down-regulate only the gene that is targeted; (iv) the fraction of cells
in a given population that receives the compound and demonstrates target down-regulation must be defined.
The erbB-2 gene codes for a 185 kDa tyrosine kinase-linked transmembrane protein which is overexpressed in 30-50% of primary breast cancers (
3
-
6
). Overexpression, which is frequently due to gene amplification, is an early event in the development of many breast
cancers and is maintained during invasion and metastatic progression of the
disease (
7
). Expression of erbB-2 is low in most normal adult tissues, making it an attractive therapeutic
target (
8
). We recently described a set of methods for delivering phosphorothioate
oligonucleotides and measuring antisense activity against the human erbB-2 oncogene in breast cancer cells (
9
). Antisense oligonucleotides or control sequences are co-delivered to cells with a fluorescent-tagged oligonucleotide using cationic liposome-mediated transfer. A high concentration of the fluorescent tracer and non-tagged oligonucleotide rapidly accumulate in the
nucleus. Cells receiving the co-delivered oligonucleotides can then be identified, quantitated,
immunostained for level of the antisense target protein and physically sorted to measure RNA
levels and phenotypic changes. Because flow cytometric analysis is quantitated
on a per cell basis, simultaneous two-color analysis of the tagged oligonucleotide (a measure of dose) versus
immunodetection of the targeted gene product yields a dose-response curve for a given antisense compound.
Unmodified DNA oligomers form stable duplexes with RNA, direct RNase H activity
and have relatively few non-specific effects. However, they are extremely susceptible to degradation
by exo- and endonucleases. Therefore, a number of different modified
oligonucleotides designed to be more resistant to nucleases have been tested
for their efficacy as antisense compounds. These include methylphosphonates (
10
), phosphorothioates (
11
), propynes (
12
), phosphoramidates (
13
) and alkylphosphotriesters (
14
). We have been investigating a series of compounds that have either one
(phophorothioate, PS) or both (phosphorodithioate, PS2) of the non-bridging oxygen atoms in the internucleotide phosphodiester group replaced
by sulfur. PS oligonucleotides are the most commonly used antisense compounds,
while little biological information is currently available on PS2
oligonucleotides. Different configurations of PS2 and PS oligonucleotides were
compared using our assay system. We provide evidence that PS2 oligonucleotides
may direct RNase H-mediated cleavage of sequence-specific RNA targets
in vivo.
In addition, we have screened a series of sequences near the start site of
translation in the erbB-2 gene. Dramatically different antisense effects were observed both with
different backbone configurations and different sequences.
MATERIALS AND METHODS
Oligonucleotide synthesis
Antisense and scrambled control phosphorothioate and phosphorodithioate
oligonucleotides were synthesized on an ABI 380B DNA synthesizer (Perkin Elmer)
as previously described (
15
,
16
). After synthesis, oligomers were purified by HPLC and purity was ensured by
collection of a single peak.
31
P-NMR was performed on all syntheses to verify that thioate or dithioate
linkages were >95% of the total. A fluorescein 5'-end-labeled phosphorothioate oligomer (TCT CTC TCT CTT TTT) was
obtained from Research Genetics. Oligomers were diluted to a concentration of
20 [mu]M, filter sterilized and stored frozen in distilled water. We have observed
no evidence of degradation or loss of antisense activity with oligomers stored
in this way over the course of a 10 month period.
Cell culture and liposome-mediated delivery
SK-BR-3 cells were obtained from the American Type Culture Collection and
maintained in RPMI 1640 containing 10% heat-inactivated fetal bovine serum (FBS). For liposome delivery, cells were
preincubated in serum-free medium (Opti-Mem I; Gibco BRL). For a typical experiment, oligonucleotides were
added to Lipofectin (Gibco BRL) yielding final concentrations of 0.3 [mu]M test oligomer, 0.05 [mu]M fluorescein oligomer and 10 [mu]g/ml Lipofectin. Liposome-oligomer solutions were incubated for 30 min at room
temperature and then added at a volume of 0.3 ml/16 mm tissue culture well.
Cells were incubated with the liposome solution for 4 h, after which RPMI
containing 10% FBS was added to stop uptake of liposomes.
ErbB-2 protein detection and flow cytometry
Immunofluorescent detection of cell surface receptors was performed as
previously described (
9
). Briefly, cells were treated with 0.25% trypsin, 1 mM EDTA, counted and
aliquoted at ~50 000 cells/well into 96-well plates. Cells were pelleted at 2000
g
at 5oC, washed twice in ice-cold azide wash buffer (phosphate-buffered saline, 0.5% FBS, 0.1% azide) and then resuspended in
100 [mu]l 1% bovine serum albumin (BSA) containing the erbB-2 mouse monoclonal antibody TA-1 (Oncogene Science) at 0.25 [mu]g/ml. Cells were incubated on ice for 1 h, washed three
times in azide wash buffer, then incubated on ice in 1% BSA containing R-phycoerythrin-labeled goat anti-mouse conjugated (Molecular Probes) secondary antibody at 10 [mu]g/ml for 1 h. Cells were washed and then analyzed by flow
cytometry on a Becton Dickinson FACStar Plus. For physically sorting
populations, cells were trypsinized, pelleted and brought up to a final
concentration of 2 * 10
6
/ml in RPMI plus 20% FBS, then sorted by fluorescein content in a Becton
Dickinson FACStar Plus flow cytometer at an average flow rate of 2000 cells/s.
Northern blotting
After sorting cells based upon the level of the fluorescent tracer, total RNA
was extracted by the guanidium thiocyanate method (
17
). RNA was electrophoresed, blotted and probed for erbB-2 as previously described (
18
). Densitometry of short exposure autoradiograms was performed to quantitate the
amount of hybridization.
.
Description of US-1 compounds tested with different linkages
Name
Backbone description
S
All monothioate linkages
S-CAP
Three monothioate linkages at the 5'- and 3'-ends, phosphodiester linkages in the center
SO
Alternating monothioate and phosphodiester, monothioate at the 5'- and 3'-ends
S2O2
Alternating dithioate and phosphodiester, dithioate at the 5'- and 3'-ends
S2-CAP
Three dithioate linkages at the 5'- and 3'-ends, phosphodiester linkages in the center
S2
All dithioate linkages
Immunoprecipitation
Radioimmunoprecipitations were performed as previously described (
19
). Briefly, cells were treated with oligomers complexed with cationic liposomes
and then labeled with 50 [mu]Ci/ml [
35
S]methionine in methionine-free RPMI for 3 h. Extracts were prepared by NP-40 lysis and sonication and protein concentrations quantitated by
the Bradford Assay (BioRad). Equal amounts of protein were precleared with
protein G-Sepharose (Pharmacia) and then reacted with a combination of 1 [mu]g anti-erbB-2 antibody TA-1 (Oncogene Science) and 1 [mu]g anti-proliferating cell nuclear antigen (PCNA)
monoclonal antibody (Dako). The immune complexes were recovered by binding to
protein G-Sepharose, washed, boiled in SDS sample buffer and then electrophoresed
by 7.5% SDS-PAGE. Gels were fixed, dried and the signals quantitated by analysis on a
Molecular Dynamics PhosphorImager. The negative control for immunoprecipitation
was performed with 2 [mu]g/ml mouse IgG (Coulter Immunology). Results are reported as a ratio between
the erbB-2 and PCNA signals.
RESULTS
Antisense effects by variations of the thioated backbone
We previously identified a lead sequence for down-regulating the erbB-2 oncogene in the SK-BR-3 human breast cancer cell line (
9
). The erbB-2 gene is amplified in this cell line ~16-fold and the protein is overexpressed compared with normal
mammary epithelial cells by a factor of ~100 (
20
). This sequence targets the AUG initiation codon of the erbB-2 gene and has been designated US-1 (Table
1
). We have used a scrambled US-1 sequence as a specific control for this compound, designated US-D. The US-1 sequence was synthesized with the following configurations
for antisense testing (Table
2
): (i) all phosphoromonothioate linkages (S); (ii) all phosphorodithioate
linkages (S2); (iii) alternating monothioate and phosphodiester linkages (SO);
(iv) alternating dithioate and phosphodiester linkages (S2O2); (v) three
monothioate linkages at the 5'- and 3'-ends and phosphodiester linkages in the middle (capped
monothioates, S-CAP); (vi) three dithioate linkages at the 5'- and 3'-ends and phosphodiester linkages in the middle
(capped dithioates, S2-CAP). The US-D scrambled control sequence was also synthesized with all
monothioate, all dithioate, alternating monothioate and alternating dithioate
configurations.
Effect of sequence variations
The US-1 sequence is 15 bases long and targets the start of erbB-2 translation (Table
1
and Fig.
3
). This sequence was chosen from an initial series of sequences that included
the start of transcription and the intron 1 splice donor and acceptor sites as
other targets (J.P.Vaughn, unpublished data). Extensive testing of the US-1 sequence showed that it consistently inhibited the steady-state levels of erbB-2 and led to an accumulation of cells in the G1 phase of the
cell cycle (
9
). Since choosing the best sequence is largely an empirical exercise, we
systematically began to shift the target upstream and downstream (in 3 base
increments, all monothioates) in an effort to improve the antisense effect
(Table
1
and Fig.
3
). This series of sequences was tested in triplicate lipofections and compared
with US-1 by flow cytometry in SK-BR-3 cells (Fig.
4
A and B). Each of the antisense sequences inhibited erbB-2 protein levels, but with widely varying potency. The two compounds that
target sequences 5' of US-1 (US-3 and UT-1) were more efficient than the 3' targets (US-4 and US-5). The flow cytometric analysis
yields a dose-response curve in a single step, i.e. erbB-2 levels and the amount of fluorescent tracer are measured on a per
cell basis. Down-regulation at lower tracer levels is indicative of more effective
antisense compounds. In particular, the US-3 sequence achieved more erbB-2 inhibition than US-1 and at a lower dose. This can be seen in the flow analysis,
with more cells having a decreased erbB-2 content (
y
-axis) at lower
x
-axis values. This result was highly reproducible in a number of different
experiments. The UT-1 sequence also performed slightly better than US-1. The relative activity of the US-3 versus US-1 sequences was also compared in another erbB-2 gene-amplified cell line, SK-OV-3, derived from an epithelial
ovarian cancer. Similar results were obtained, with US-3 having greater antisense activity than US-1 (data not shown).
To further characterize and compare the relative efficiencies of the US-1 and US-3 sequences, we measured their effects on
de novo
erbB-2 protein synthesis at varying doses. SK-BR-3 cells were lipofected with the antisense compounds US-1 and US-3 (and their cognate scrambled controls US-D and SC-3 respectively). At 12 h after
lipofection, the cells were labeled for 1 h with [
35
S]methionine. Protein extracts were prepared and an immunoprecipitation
performed using both an anti-erbB-2 and an anti-PCNA antibody (Fig.
5
A). Phosphorimage quantitation of the specifically immunoprecipitating bands was
obtained and the erbB-2 signal was normalized to the level of PCNA in each lane. This ratio was
then compared with the erbB-2:PCNA ratio obtained using the scrambled control compounds (expressed as
percent of control in Fig.
5
B). While US-1 was only effective at 0.3 [mu]M (the dose used in all previous experiments), US-3 continued to inhibit erbB-2 protein synthesis at the lowest concentration used
(0.075 [mu]M). This is consistent with the flow cytometric data, which also indicated
that US-3 was effective at lower doses, i.e. at lower intracellular tracer
concentrations.
DISCUSSION
Our goal is to identify effective antisense compounds targeted to the erbB-2 oncogene, which is amplified and overproduced in a large fraction of
breast and other epithelial cancers (
3
,
21
,
22
). To this end, we have developed a rapid and sensitive assay to monitor, in
cell culture, the relationship between intracellular oligomer concentration and
down-regulation of the target gene (
9
). Using this assay, we have explored the effect of different sulfur-containing backbone modifications and different sequence targets on
inhibition of erbB-2.
We tested novel backbone configurations using combinations of phosphorodithioate
and phosphodiester linkages. Oligonucleotides with these linkages do not have
the additional chiral center resulting in the numerous potential diastereomers
characteristic of monothioate-substituted oligomers. This stereospecificity enhances the stability of
dithioate-DNA complexes compared with monothio-substituted oligonucleotides (
23
). In addition, the dithioated backbone is stable to exo- and endonucleases and can direct RNase H-mediated degradation of target RNA using HeLa cell nuclear extracts
(
23
). Examining the different backbone configurations with the same sequence, we
found that an alternating phosphodithioate/phosphodiester configuration was as
effective as monothioate linkages at each position in achieving gene-specific down-regulation. The alternating dithioate specifically reduced erbB-2 mRNA, providing further evidence that this backbone can
direct RNase H activity (
23
). Ghosh
et al.
reported that dithioate-containing (at every linkage) oligomers were as unstable as unmodified DNA
in nuclear extracts prepared from a breast cancer cell line (
24
). The current study and previous results (
23
) indicate that a phosphorodithioate at every other linkage is sufficient to
protect these oligomers from both exo- and endonuclease activity in the nucleus and cytoplasm. The measured
instability of these compounds found by Ghosh
et al.
may have been due to the use of 5'-labeled oligomers, where dephosphorylation can be interpreted as
degradation.
Other dithioate backbone configurations were also tested in the current study.
The same anti-erbB-2 sequence containing three dithioate linkages at each end (capped)
was ineffective in reducing steady-state levels of erbB-2 protein. These capped compounds were designed to be resistant to
exonucleolytic activity, which appears to be the primary type of nuclease
present in cell extracts and serum (
25
,
26
). Since both the monothioate and dithioate capped compounds had no measurable
antisense activity on the steady-state levels of the erbB-2 protein, we conclude that significant endonuclease activity is
also present. By measuring
de novo
erbB-2 protein synthesis at 12 and 24 h after delivery we were able to show a
small degree of specific inhibition only at 12 h using the capped compounds
(data not shown). This result is consistent with relatively rapid degradation
of this configuration. The alternating dithioate and full monothioate oligomers
were active at both time points in this same assay.
Comparable antisense activity was observed with the full monothioate and
alternating dithoate compounds, however, we were able to improve on this
activity by altering the sequence target. We are currently formulating the
improved sequence in the alternating dithioate backbone and testing the
efficacy of these compounds in inhibiting the growth of breast and ovarian
cancer xenografts that overproduce the erbB-2 protein, both with and without liposome-mediated delivery. Our experiments suggest that there might be a
relatively narrow window for optimal antisense effect on erbB-2. However, animal experiments have frequently demonstrated more robust
biological effects than would be predicted in tissue culture. Given the
difficulty of examining antisense effects in animals, our test system provides
a measure of confidence that observed biological effects may be due to
antisense inhibition of the targeted gene.
ACKNOWLEDGEMENTS
We thank Gudrun Huper, Lenora Blount, Michael Cook and the Duke Comprehensive
Cancer Center Flow Cytometry Shared Resource for excellent technical
assistance. We greatly appreciate helpful discussions with Rudy Juliano and
Eric Wickstrom. This work was supported by National Cancer Institute grants UO1-CA60139 to Eric Wickstrom and 1F32-CA63786 to J.P.V.
REFERENCES
1 Stein,C.A. and Cheng,Y.-C. (1993) Science, 261, 1004-1012.MEDLINE Abstract
15 Beaton,G., Dellinger,D., Marshall,W.S. and Caruthers,M.H. (1991) In Eckstein,F. (ed.), Oligonucleotides and their Analogues, A Practical Approach. IRL Press, Oxford, UK, pp. 109-135,
