ABSTRACT
RNA sequences containing 2
'
-amino pyrimidines that bind with high-affinity to human thyroid stimulating hormone (hTSH) were isolated
from a random sequence library by an
in vitro
selection-amplification procedure. A representative RNA ligand (T-15) has an equilibrium dissociation constant (
K
d) of 2.5 nM for its interaction with hTSH and can discriminate between other
members of the glycohormone family; no detectable binding was observed at low
micromolar concentrations of hCG (human chorionic gonadotropin), while measured
K
d values for the interactions with hLH (human leutinizing hormone) and hFSH
(human follicle stimulating hormone) were >1
[mu]
M and
~
0.2
[mu]
M, respectively. The detection of hTSH in a dot blot assay with radiolabeled T-15 RNA was demonstrated.
Specific molecular recognition is central to the detection of disease causing
agents. Although most diagnostic tests are based on antibody ligands, other
molecules capable of fulfilling the requirements for specific molecular
recognition are being discovered (
1
-
3
). The
S
ystematic
E
volution of
L
igands by
EX
ponential enrichment (SELEX) process, which utilizes
in vitro
selection and amplification of nucleic acids, is a combinatorial approach for
the isolation of specific sequences with unique properties of interest from
random sequence libraries of oligonucleotides (
4
-
6
). The complexity (or sample size) of a typical starting library is ~10
14
molecules. The outcome of an
in vitro
selection experiment is a population of sequences whose functional properties
were dictated by the selection criterion employed in the process. Examples
include high-affinity nucleic acid sequences selected for a variety of target molecules
(
4
-
19
) and sequences that can either undergo autocatalysis of novel reactions (
20
,
21
) or function as an enzyme in the chemical transformation of a substrate
molecule (
22
).
Recently, random sequence oligonucleotide libraries consisting of 2'-amino pyrimidines have been used to isolate modified
oligonucleotide ligands with enhanced stability in biological fluids (
12
,
17
). Such modified oligonucleotide ligands are attractive candidates for both
therapeutic and diagnostic applications. SELEX-derived small oligonucleotides (~20 kDa) typically exhibit affinities similar to those of relatively
large antibodies (~160 kDa). This feature, along with the availability of modified
oligonucleotide random sequence pools with enhanced nuclease stability,
prompted us to exploit the SELEX process for the isolation of nucleic acid
ligands with high affinity for human thyroid stimulating hormone (hTSH), a
hormone commonly used as a diagnostic marker for thyroid abnormalities.
hTSH (
M
r
27.7 kDa) is a glycohormone secreted by the pituitary that controls the
synthesis of hormones by the thyroid gland. Measurement of serum hTSH levels is
important in the diagnosis of both pituitary and thyroid disorders such as
hyperthyroidism and hypothyroidism (
23
,
24
). Pituitary-derived hTSH, hLH (leutinizing hormone) and hFSH (follicle stimulating
hormone), along with hCG (chorionic gonadotropin) secreted by placenta,
constitute a family of glycohormones (
23
-
27
). Members of this glycohormone family have very similar structural properties.
Each hormone is a heterodimer composed of non-covalently associated subunits ([alpha] and [beta]). All four members have an identical [alpha] subunit. Each member of the family exerts its
biological response via interacting with a cell surface receptor. While the [beta] subunits confer specificity, there is a high degree of sequence
similarity among the members. The sequence similarity within the first 114
amino acids of the [beta]-subunit of hCG is 85% with hLH, 46% with hTSH and 36% with hFSH (
27
). Further, the [beta]-subunit specific to one member and the [alpha]-subunit derived from a different member have been
reconstituted into a heterodimer that elicited the biological response of the
hormone derived from the [beta]-subunit (
23
,
25
). Due to these structural similarities, the development of immunological assays
for the specific detection of each hormone has been challenging (
25
).
In this study, the SELEX process was used to isolate high affinity RNA ligands
for hTSH from a random sequence RNA pool containing 2'-amino pyrimidines with a complexity of ~10
14
individual sequences. After nine rounds of selection and amplification of hTSH-bound RNA, two families of sequences emerged. A representative RNA ligand
(T-15) of one family characterized by a predicted stem-loop structure bound hTSH with an equilibrium dissociation constant
(
K
d
) of 2.5 nM. Importantly, this ligand bound much more weakly to other members of
the glycohormone family. The use of radiolabeled ligand T-15 to detect hTSH in a dot blot assay suggests that SELEX-derived oligonucleotide ligands have potential use as diagnostic
reagents.
DNA sequences were synthesized by standard cyanoethyl phosphoramidite chemistry.
After deprotection, DNA sequences were purified by gel electrophoresis under
denaturing conditions. 2'-NH
2
-modified UTP and CTP were synthesized as described previously (
28
). All RNA sequences used in the study were prepared by
in vitro
transcription of DNA templates (
29
) and purified by denaturing gel electrophoresis. hLH (
M
r
35 500) and hFSH (
M
r
38 250) were from Becton Dickinson (Research Triangle Park, NC). hTSH (
M
r
27 700; 8 IU/mg) was from either Becton Dickinson or Vitro Diagnaostics
(Littleton, CO). hCG (
M
r
42 000; 14 000 IU/mg) was from Vitro Diagnaostics. The [alpha]-subunit and the [beta]-subunit of hTSH were obtained from Calbiochem (La
Jolla, CA). Enzymes and other chemicals of analytical grade were purchased from
commercial sources.
Five nanomoles of gel-purified, synthetic template DNA containing 40 nucleotide (nt) contiguous
random sequence flanked by defined primer annealing sequences [5'-GGGAGGACGATGCGG-(N)
40
-CAGACGACTCGCCCGA-3'] were amplified by four cycles of the polymerase chain
reaction (PCR) with primers 5'-TAATACGACTCACTATAGGGAGGACGATGCGG-3' and 5'-TCGGGCGAGTCGTCTG-3'. Approximately 800 pmol of
the PCR-derived template DNA (~5 * 10
14
molecules) were transcribed
in vitro
by T7 RNA polymerase (1000 U) in a 3-ml transcription reaction consisting of 40 mM Tris-HCl (pH 8.0), 12 mM MgCl
2
, 1 mM Spermidine, 5 mM DTT, 0.002% Triton X-100 (v/v), 4% polyethylene glycol (w/v) and 2 mM each of ATP, GTP, 2'-NH
2
CTP and 2'-NH
2
UTP (
17
). Full-length transcription products were purified on 8% polyacrylamide gels
under denaturing conditions, suspended in TEM buffer (binding buffer; 10 mM
Tris-HCl, 0.1 mM EDTA, 2.5 mM MgCl
2
, pH 6.8), heated to 70oC and chilled on ice. The RNA was then incubated with hTSH at 37oC for 15 min. The RNA-protein mixture was filtered through a pre-wet nitrocellulose filter and the filter was immediately
washed with 5 ml binding buffer. Bound RNAs were eluted from the filter (
6
), recovered by ethanol precipitation and reverse transcribed by avian
myeloblastosis virus reverse transcriptase (Life Sciences) at 48oC for 45 min with the DNA sequence 5'-TCGGGCGAGTCGTCTG-3' as the primer. Following PCR amplification of
the cDNA, the resulting duplex DNA template was transcribed
in vitro
to obtain RNA for the next round of selection.
The concentration of hTSH in the binding reaction was decreased gradually in
successive rounds from 3 [mu]M to 200 nM to progressively increase selective pressure. The selection
process was repeated until the affinity of the enriched RNA pool for hTSH was
substantially increased. At that point, cDNA was amplified by PCR with primers 5'-CCG
Internally-labeled RNA was synthesized by using [[alpha]-
32
P]ATP in
in vitro
transcription reactions. Full-length transcripts were purified by polyacrylamide gel electrophoresis
under denaturing conditions. RNA suspended in TEM buffer (~5 nM) was heated to 80oC, chilled on ice and then transfered to room temperature. Protein
binding was carried out with low concentrations of radiolabeled RNA (typically ~10 pM) to satisfy saturation binding at protein excess conditions. RNA was
incubated with varying amounts of hormone in 50 [mu]l TEM buffer containing 0.01% hSA for 10 min at 37oC. RNA-protein mixtures were passed through pre-wet nitrocellulose filters (0.2 [mu]m) and the filters were immediately washed with 5 ml
binding buffer. Radioactivity retained on filters was quantitated by liquid
scintillation counting. The quantity of RNA bound to filters in the absence of
hormone was used for background correction. The percentage of input RNA
retained on each filter was plotted against the corresponding log protein
concentration. The non-linear least-squares method was used to obtain the dissociation constant (
K
d
) (
11
).
For thermal melting, the full-length sequence T-15 was synthesized by
in vitro
transcription and purified by denaturing polyacrylamide gel electrophoresis.
RNA was heated in TEM or sodium phosphate buffer to 95oC and cooled to room temperature prior to the determination of its melting
profile. Thermal melting profiles of RNA were obtained on a Cary Model 1E
spectrophotometer by recording the absorbance at 260 nm while the sample was
heated at 1oC/min. The melting point (
T
m
) was calculated using the first derivative values.
hTSH was suspended in TEM buffer (300 [mu]l) containing 0.1% hSA (w/v) and applied to pre-wet nitrocellulose filters (0.45 [mu]M; BioRad) under suction. Gel-purified, internally labeled RNA T-15 was then applied to the blots in 50 [mu]l TEM buffer (~0.5 pmol/[mu]l) and filtered gently. Filters were
immediately washed three times with 300 [mu]l of the same buffer followed by two times with 300 [mu]l 0.5 M urea in the same buffer to eliminate most of background binding
of RNA to nitrocellulose filters. The blots were dried and analyzed with a
PhosphorImager and by autoradiography.
A selection carried out in the low ionic strength TEM buffer resulted in a
progressive increase in affinity of the enriched RNA pools for hTSH. Sequences
obtained from the 9th round PCR products are shown in Figure
1
. There were 21 unique sequences among the 37 sequences analyzed. The selected
purine-rich RNA sequences share a somewhat conserved 9 nucleotide sequence; 5'-GG/
C
A/
U
GGG/
AC
U/
A
A/
GC
G/
C
-3' (boxed in Fig.
1
). These sequences were classified into two families based on predicted
secondary structures. The 5' and 3' halves of the variable (or randomized) regions of sequences
grouped into family I share a high degree of complementarity in base pairing
(indicated by half arrows in Fig.
1
), suggesting their possible existence as a stem-loop structure in solution. A portion of the 5' half of the variable region of sequences classified into family
II are partially complementary to the 3' fixed sequence (primer annealing site). The predicted family II stem-loop structures have less structured stems and larger loops than
the corresponding structures predicted for family I RNA sequences.
A nitrocellulose filter binding technique was used to screen several
representative ligands from the two sequence families for their ability to
interact with hTSH. The calculated
K
d
values of these ligands range from 2.5 to 100 nM, whereas the
K
d
of the unselected random sequence pool is >= 2.5 [mu]M. Of the sequences analyzed, RNA sequence 15 (referred to as ligand T-15) exhibited the highest affinity for hTSH (
K
d
= 2.5 +- 0.6 nM). Ligand T-15, the most abundant sequence that represented 24% of the
affinity-selected pool was chosen for further study, based on its high-affinity. Ligand T-15 bound hTSH with an at least three orders of magnitude
higher affinity than the unselected random sequence pool (Fig.
2
a). This ligand did not exhibit high-affinity binding to hTSH in the absence of Mg
2+
ions (Fig.
2
a), indicating the crucial role played by Mg
2+
ions in the RNA-protein interaction.
As discussed above, due to the presence of complementary bases in the 5' and the 3' regions of the variable 40 nt stretch, a stem-loop structure can be predicted for ligand T-15 (Fig.
4
a). Alternatively, the existence of several guanine dinucleotide repeats in the
sequence may allow the RNA to fold into intermolecular or intramolecular G-quartet structures. Among other secondary structures such as stem-loops and pseudoknots, G-quartet structures were also common in ligands that have been
identified by
in vitro
selection (
9
,
16
-
18
,
35
). Guanines involved in G-quartet structures are resistant to RNAse T1 cleavage (
17
), especially under conditions that favor G-quartet formation (
36
,
37
). To investigate whether ligand T-15 assumes G-quartet structures, RNAse T1 cleavage patterns were obtained in
several buffers [with KCl (facilitates G-quartet formation) and LiCl (does not facilitate G-quartet formation)] as described in (
17
). The cleavage pattern of ligand T-15 was identical in all three buffers examined (standard RNase T1 buffer
containing urea, and TEM buffer containing either KCl or LiCl) (data not shown), suggesting that this ligand does not assume the G-quartet structure. As an alternative to RNAse T1 cleavage susceptibility,
chemical probing with dimethyl sulfate (DMS), which reacts primarily at the N7
of guanines (
38
), can be used to confirm the possible existence of G-quartet structures (
36
,
37
). However, DMS treatment of the T-15 RNA resulted in smeared bands on sequencing gels, hampering the
interpretation of results (data not shown). This unexpected observation could
be due to the 2'-amino groups in RNA that may potentially react with DMS.
Ligand T-15 containing 2'-amino pyrimidines has a
T
m
(the temperature at which 50% of RNA is denatured) of 42oC in 0.1 M sodium phosphate buffer (Fig.
4
b; dotted line), indicating that the RNA is folded into a secondary structure. A
similar
T
m
value was observed in TEM buffer, but the melting transition occurs over a
broader range than that observed in the phosphate buffer (Fig.
4
b; solid line).
We investigated whether the RNA T-15 ligand could be used in a dot blot assay to detect hTSH, analogous to
immunoblots in which antibodies are used to detect and quantitate their cognate
targets. Different concentrations of hTSH [from 800 nM (177 [mu]IU/ml) to 50 nM (11 [mu]IU/ml)] were applied to a nitrocellulose membrane. The membrane was then
exposed to radiolabeled RNA (either affinity-selected T-15 or unselected random sequence library). A representative
autoradiogram of a dot blot is shown in Figure
5
a. PhosphorImager quantitation of the signal as a function of the input
concentration of hTSH is shown in Figure
5
b. The signal obtained with the radiolabeled unselected random sequence pool
(used as a control) did not correlate with the amount of hTSH on the blot (Fig.
5
b; open circles). However, with the affinity-selected radiolabeled ligand T-15, the signal correlated with the concentration of hTSH used. A
linear relationship between the signal and the amount of hTSH up to 50 [mu]IU/ml was observed.
In spite of the high degree of structural similarity present among the four
members of the glycohormone family, they elicit different biological effects by
interacting with three distinct receptors. HCG and hLH bind to the same
receptor (
27
). The existence of hormone-specific receptors (natural ligands) suggests that each member contains a
unique set of epitopes for specific molecular recognition; feasibly even with
artificial ligands. In fact, with a great deal of effort, antibodies specific
for each member have been raised (
25
). The identification of such antibodies has led to the development of
diagnostic assays for the detection of the individual hormones in biological
fluids.
The work described here was initiated to use the SELEX process to select
oligonucleotide ligands that bind hTSH with high affinity and specificity.
Clone 15 (T-15), the highest affinity ligand in the selected pool, has a
K
d
of 2.5 nM for its interaction with hTSH. The specificity of ligand T-15 for hTSH was indicated by its inability to bind with high-affinity to hCG, hLH and hFSH, especially in the presence of
competing tRNA.
These results were obtained under direct selection conditions, where no specific
counterselection against ligands with affinity for closely related members of
the glycohormone family was incorporated. This outcome is intriguing,
especially given the existence of an identical [alpha]-subunit in all four hormones. Furthermore, the crystal structure of
one of the member (hCG) indicates that the two subunits are highly structured
and have similar topology (
27
). Affinity selection described in this study demonstrates that the RNA selected
for hTSH recognizes a site that is unique to this hormone and does not readily
identifiable in other members. It should be noted that under similar
conditions, without counterselection, RNA ligands isolated by affinity-selection for hCG bound hLH with equal affinity (in addition to the
presence of identical [alpha]-subunits, the [beta]-subunits of the two hormones are ~85% similar), but with low affinity to the other
two members (Y. Lin, D. Nieuwlandt and S. D. Jayasena; unpublished results).
The selection of RNA ligands specific for hCG required a counterselection
procedure to eliminate RNAs with affinity for hLH (West
et al
., in preparation). The counterselection strategy has been successfully used to
isolate RNA ligands highly specific for theophylline and can discriminate
between binding to caffeine, an analogue that differs from theophylline by a
single methyl group (
13
), as well as ligands that discriminate in binding to arginine and citrulline (
19
,
41
). These studies suggest that a direct SELEX process with no counterselection
may be sufficient for the isolation of oligonucleotides that can discriminate
between related targets, but highly discriminatory ligands to very similar
targets may require counterselection strategies.
Ligand T-15 did not show high-affinity binding to individual subunits of hTSH, indicating the
requirement for the intact heterodimer for effective binding. Antisera specific
for a given member have been raised by using the respective [beta]-subunits (
25
). The difference in the molecular recognition properties of antibodies and
SELEX-derived nucleic acid ligands may be due to the way in which the target is
presented during the two selection processes. Most antibodies are known to
recognize linear epitopes within a protein, presumably due to the presentation
of peptide fragments by antigen-presenting cells. However, in the SELEX process an intact protein is
repeatedly presented to increasingly enriched pools of oligonucleotides. Hence,
SELEX-derived oligonucleotide ligands tend to recognize conformational epitopes.
Ligand T-15 can potentially form a stem with 14 base pairs of which 9 base pairs
form an uninterrupted helix. The melting temperature observed for this ligand
is substantially lower than what would be observed for an unmodified 2'-OH RNA sequence. In agreement with the
T
m
observed for sequence T-15 containing 2'-amino pyrimidines, the presence of 2'-amino pyrimidines in oligonucleotides has been
shown to lower their melting transitions significantly (
42
,
43
). Decreased stability of helices containing 2'-amino pyrimidines may direct SELEX-derived ligands to adopt alternate structures such as G-quartets (
17
) or may not assume a known secondary structure (
12
). However, as evidenced by ligand T-15, the selection of ligands which adopt helical structures is also
possible from random sequence pools containing 2'-amino pyrimidines, suggesting that this particular modification may
not significantly decrease the shape repertoire available for molecular
recognition.
Affinity selections on a single target have been carried out with chemically
different oligonucleotide libraries such as DNA, RNA and those with other
chemical modifications (
8
,
9
,
15
,
17
,
18
,
35
,
44
). `Winning' ligands resulting from such selections were different with respect
to their nucleotide sequence and predicted secondary structures. Furthermore,
RNA versions of DNA ligands (or vice versa) were not functional in binding to
the cognate target (
17
,
35
,
44
). Similar to these previous observations, the high-affinity binding of ligand T-15 was abolished when the 2'-amino pyrimidines of the sequence were replaced by 2'-hydroxy pyrimidines; the
K
d
of T-15 in its all 2'-hydroxy form is similar to that of the unselected random
sequence pool (data not shown). These results indicate that the chemical
composition of the initial library that governs the folding of individual
sequences remains critical for the function of selected ligands. The effect of
particular chemical modifications, 2'-amino modification in the present study, on target recognition may
be a direct participation in the interaction with the target and/or an indirect
effect contributing to the overall folding of the ligand.
The linear sequence information of oligonucleotide probes is used in the
detection of complementary sequences in northern and Southern blotting. This
technique has been somewhat modified to detect proteins that naturally bind to
nucleic acid sequences on transfer blots by using nucleic acid probes (
45
). The SELEX process permits the isolation of high-affinity nucleic acid sequences for a wide range of targets, including non-nucleic acid binding proteins. Such ligands can potentially be
useful for the detection of non-nucleic acid binding proteins on blots. In this report, we tested this
idea by demonstrating the detection of hTSH by RNA ligand T-15. Using a dot blot format, native hTSH was detected with the
radiolabeled ligand as the signal.
Chemilumeniscent substrates have been useful for the detection of alkaline
phosphatase at 10
-19
-10
-15
M (
46
). Synthetic oligonucleotides conjugated to alkaline phosphatase have been used
in non-radioactive hybridization assays (
47
). Thus, the sensitivity of the dot blot assay described here can potentially be
increased by using RNA-enzyme conjugates that utilize fluorescence or luminescence substrates.
The results presented in this report demonstrate that the SELEX process can be
used to identify
nucleic acid ligands with high affinity and specificity for a given member of a
protein family. Data presented here as well as accumulating evidence (
48
,
49
) suggest that SELEX-derived nucleic acid ligands can be used in diagnostic assays either in
place of or in combination with antibodies. Analytes that are not readily
immunogenic as well as those to which high-affinity antibodies are not currently available may be attractive targets
for the SELEX process. The ability to obtain amplifiable oligonucleotide
libraries decorated with both hydrophilic and hydrophobic functional groups (
44
,
50
) offers an exciting opportunity to identify candidate ligands for both
diagnostic and therapeutic applications (
51
).
We thank Barry Polisky, Larry Gold, Garry Kirschenheuter and Ginny Orndorff for
critical reading of the manuscript. Glenn Vonk at Becton Dickinson Research
Center is gratefully acknowledged for providing the hTSH, hLH and hFSH used in
this study. We also thank colleagues in the chemistry division at NeXstar
Pharmaceuticals for synthesizing 2'-NH
2
-UTP and 2'-NH
2
-CTP.
REFERENCES
Return