Assessment of major and minor groove DNA interactions by the zinc fingers of
Xenopus
transcription factor IIIA
Assessment of major and minor groove DNA interactions by the zinc fingers of Xenopus transcription factor IIIA
Steven J.
McBryant
,
Benjamin
Gedulin
+
,
Karen R.
Clemens
,
Peter E.
Wright
and
Joel M.
Gottesfeld*
Department of Molecular Biology, The Scripps Research Institute, 10666 North
Torrey Pines Road,
La Jolla
, CA 92037,
USA
Received March 12, 1996;
Revised and Accepted May 22, 1996
ABSTRACT
Zinc finger proteins of the Cys2His2 class are DNA sequence-specific transcription factors. Previous structural studies of zinc finger protein-DNA complexes have shown that amino acids in the finger tip and
[alpha]
-helix regions within individual finger domains make base-specific contacts with the major groove of DNA. The nine finger
protein transcription factor IIIA (TFIIIA) from
Xenopus
oocytes binds a 43 base pair region of the 5S RNA gene through major groove
interactions with two sets of three fingers (fingers 1-3 and 7-9) and with finger 5. Previous studies have suggested that zinc
fingers 4 and 6 each bind in or across the minor groove to bridge these major
groove-binding zinc fingers. Here it is shown that a polypeptide containing zinc
fingers 1-5 (zf1-5) binds oligonucleotides with modifications in the major groove
of the finger 4 binding site with wild-type affinity. Mutagenesis and binding site selection studies were
performed to determine whether high affinity DNA binding by zf1-5 requires a particular sequence in the binding site for finger 4.
Several mutations in this region of the 5S gene reduced the DNA-binding affinity of zf1-5; however, selection and amplification binding assays did not
recover the wild-type finger 4 binding site sequence from a pool of mixed sequence
oligonucleotides. Rather, a purine-rich sequence on the top strand was highly selected within the finger 4
binding site. We suggest that high affinity DNA binding by zinc finger 4 may be
dictated by a sequence-specific DNA structure rather than by a unique DNA sequence. Deletion of
finger 4 from zf1-5 results in a protein with poor binding affinity, demonstrating the
importance of finger 4 in proper alignment of neighboring fingers with the DNA,
and/or the importance of correct protein-protein interactions between fingers.
INTRODUCTION
The structures of individual zinc finger domains of the Cys2His2 class have been
solved by nuclear magnetic resonance methods (
1
-
4
) and structures of two-, three- and five-finger protein-DNA complexes have been solved by X-ray crystallography (
5
-
7
). In these structures, the ~30 amino acid zinc finger consists of a short antiparallel [beta]-turn, a finger tip region and an [alpha]-helix. Zinc coordination is through a pair of
cysteine residues on the [beta]-turn and a pair of histidines on the [alpha]-helix. In the co-crystal structures, specific protein-DNA contacts are made by amino acids
immediately preceding the [alpha]-helix and at amino acid positions +2, +3 and +6 of the helix (
5
-
7
). In each of these structures, the zinc finger domains lie in the major groove
of the DNA. Binding site selection studies (
8
-
12
) and structural studies (
7
) of the zif268/Sp1 class of Cys2His2
zinc finger proteins have suggested that each finger recognizes a 3 bp subsite
within the protein binding site; however, other zinc finger proteins bind
different lengths of DNA and likely utilize additional types of DNA contacts (
6
,
7
).
The nine-finger protein transcription factor IIIA (TFIIIA) from
Xenopus
binds to a 43 bp internal control region (ICR) within the coding sequence of
the 5S ribosomal RNA genes (
13
-
15
). TFIIIA is unique among the zinc finger proteins in that it binds with high
affinity to both the 5S RNA gene and to the 5S rRNA transcript (
16
-
18
). Considerable effort has gone into defining both the nucleic acid sequences
and specific fingers that are involved in high affinity DNA and RNA binding.
For DNA binding, a series of deletion, linker scanning, substitution and point
mutagenesis studies have revealed that the ICR is not contiguous but is composed of three important sequence elements: a 5' A-box region, an intermediate element, and a 3' C-box element (
19
-
23
). Each of these elements is involved in base-specific contacts with particular fingers or sets of fingers (
24
,
25
). Studies with truncated TFIIIA polypeptides demonstrate that the N-terminal three fingers interact with the 3' C-box through base-specific major groove contacts and bind with only a 2- to 3-fold lower affinity than full length TFIIIA
(
26
-
28
). Finger 5 interacts with the major groove of the intermediate element and the
C-terminal three fingers, seven through nine, interact with the major groove
of the 5' A-box (
29
).
Mutagenesis studies have shown that fingers 4 and 6 each contribute to the DNA
binding affinity of the protein. Amino acid substitutions in the [alpha]-helix of finger 4 and in the finger tip region of finger 6 (
30
) each reduce DNA binding affinity. Similarly, broken finger mutations in these
fingers (zinc-coordinating histidine to asparagine mutations;
31
) also reduce DNA-binding affinity. Experiments with truncated TFIIIA polypeptides have
shown that the binding sites for fingers 4 and 6 each span nearly a full turn
of DNA helix (
29
); however, mutagenesis, chemical modification and missing contact experiments
have failed to detect base-specific contacts over these regions (
21
-
25
,
32
). In particular, methylation protection (
25
) and methylation interference experiments (
24
,
29
) failed to detect any major groove guanine contacts within the binding sites
for fingers 4 or 6. However, strong hydroxyl radical protection of the finger 4
and 6 binding sites by TFIIIA has been observed (
26
,
33
,
34
). Examination of the hydroxyl radical protection patterns on the two DNA
strands indicates an offset in the center of the protected regions by 3
nucleotides (nt) in the 3' direction. This offset is indicative of a minor groove interaction of
these zinc fingers (
32
-
34
). Based on these results, we and others proposed that fingers 4 and 6 each bind
either in or across the minor groove to link the major groove binding fingers (
29
,
32
), but no direct structural data are yet available to demonstrate minor groove
binding by individual zinc fingers within a multi-finger protein.
In this study, we have monitored the binding of a five-finger polypeptide derived from TFIIIA (zf1-5, containing the five N-terminal zinc fingers) to a series of oligonucleotides and
DNA fragments harboring modified bases or mutations in the binding site for
finger 4. Our results are consistent with the model for the TFIIIA-DNA interaction (
29
,
32
) in which finger 4 either binds in or crosses the minor groove. Moreover, our
results point to the importance of zinc finger 4 in the DNA-binding activity of TFIIIA.
MATERIALS AND METHODS
Zinc finger proteins
TFIIIA was purified from the ovaries of immature
Xenopus laevis
females as previously described (
35
), with the exception that heparin agarose was used in place of Bio-Rex 70. Specific zinc-finger regions of TFIIIA-coding sequence were amplified from the TFIIIA cDNA by the
polymerase chain reaction and the PCR products were cloned in the expression
vector pRK172 (
29
). The protein-coding sequences were verified by dideoxy DNA sequencing. TFIIIA deletion
polypeptides were expressed in BL21(DE3) cells (Novagen) and purified by
heparin-agarose ion exchange chromatography using a denaturation/renaturation
protocol (
29
). Proteins were solubilized in a solution containing 50 mM Tris-HCl, pH 7.8, 100 mM NaCl, 50 [mu]M ZnCl
2
, 10% (v/v) glycerol, 10 mM dithiothreitol and 6 M urea. Insoluble material was
removed by centrifugation at 10 000
g
for 10 min at 4oC. The urea concentration was reduced to 1.5 M by dilution with the buffer
above without urea and soluble protein was applied to the agarose resin.
Unbound proteins were eluted with this buffer (lacking urea), and the bound
zinc-finger proteins were renatured by washing the resin with buffer lacking
urea. The zinc-finger proteins were eluted from the resin with a linear gradient of 100
mM to 1 M NaCl in buffer lacking urea. This purification method yields protein
preparations of >90% purity (as judged by SDS-PAGE and Coomassie staining). Protein concentrations were estimated by
densitometry of the stained gels using bovine serum albumin as a standard.
Expression of zf1-3/5
An expression clone for the zf1-3/5 protein sequence shown in Table
4
was generated by PCR amplification of the zf1-3 and finger 5 coding regions from the TFIIIA cDNA with the following
sets of primers: ([Delta]N finger one start, initiation codon in bold,
Nde
I site underlined) 5'-GCCGGTGG
CAT
ATG
AAGCGGTACATC-3'; (finger 3 stop, TGE codons underlined,
Bsa
I recognition site in bold) 5'-CGAC
CGTCTC
T
CTCGCCATA
GAATCTGTTAAAGTGCTTCTTCATG-3'; (finger 5 start, EKP codons underlined,
Bsa
I recognition site in bold) 5'-GAC
CGTCTC
AC
GAGAAGCCA
TACGAATGTCCTCATGAA GGC-3'; (finger 5 stop, stop codon in bold,
Eco
RI site underlined) 5'-GCAAAAA
GAATTC
TTA
GCCTGCATGGACTTTTTCATGACG-3'.
Bsa
I cleaves 1 nt 3' to the recognition sequence leaving a four-base 5' overhang. The two PCR products were digested with the
restriction enzymes listed above and ligated with T4 DNA ligase, creating the
codons for the linker amino acid sequence TGEKP. The ligation product was
cloned in the expression vector pET24a, which had been digested with
Nde
I and
Eco
RI. The sequence of the protein-coding region was verified and protein expression and purification were as
described above. Binding site oligonucleotides with GGT sites at various
distances from the zf1-3 binding site were selected from clones of the finger 4 binding site
random pool oligonucleotide (Table
4
and below).
DNAs, oligonucleotides and binding assays
Plasmids containing a synthetic copy of the oocyte-type 5S RNA gene cloned in pUC18 and the pXlo73-76 mutant were obtained from Dr P. Romaniuk (University of Victoria,
British Columbia, Canada;
22
). Oligonucleotides were synthesized with the appropriate phosphoramidites and were purchased from Genosys (The Woodlands, TX). Nucleotide positions within the oligonucleotides and PCR products indicated in the text and tables correspond to positions
within the
Xenopus
5S RNA gene coding sequence. The 7-deaza-deoxyadenosine substituted oligonucleotides were the generous gift
of D. Hornby (University of Sheffield, UK). Oligonucleotides were end-labeled with [[gamma]-
32
P]ATP and polynucleotide kinase, annealed to give the double-stranded products, and used in gel mobility shift assays with varying
amounts of the recombinant proteins, as described (
31
). The binding buffer for all experiments consisted of 20 mM HEPES-KOH (pH 7.5), 10% (v/v) glycerol, 50 mM KCl, 1 mM MgCl
2
, 25 [mu]M ZnCl
2
and 5 mM dithiothreitol. Incubations were at ambient temperature (22-24oC) in reaction volumes of 20 [mu]l. Reactions also contained 200 ng poly (dI-dC) unless stated otherwise. Electrophoresis conditions
were as described (
31
). The gel autoradiograms were scanned with a laser densitometer and plots of
fraction bound DNA versus protein concentration were fit to a single component
pseudo first-order reaction with the equation: fraction bound = [protein]/[protein] +
K
d
, using Kaleidagraph software (Synergy Software, Reading PA).
Selection and amplification binding (SAAB) assays
The top strand sequence of the 46 bp double-stranded oligonucleotide used for SAAB selection (
36
) is shown in Table
3
. PCR primers for amplification of protein-bound DNA were 5'-CCG
GAATTC
CGGCCAGG-3' and 5'-CG
GGATCC
CGAGGCGGTCTCCC-3'.
Eco
RI and
Bam
HI sites are underlined. These primers give rise to a 57 bp PCR product, which
was end-labeled and used in subsequent rounds of selection and amplification. SAAB
assays were performed with both zf1-5 and zf1-3/5 proteins and binding assays were performed by non-denaturing gel electrophoresis. At each round of selection,
a protein titration was performed with the labeled PCR product, and the amount
of protein that gave 3-5% binding was used for subsequent selection. DNA was recovered from the
protein-DNA complex by elution from a gel slice into a buffer containing 0.5% SDS (w/v), 10 mM Tris-HCl, pH 7.6, 15 mM EDTA, 0.5 M NaCl. Fragments of polyacrylamide were removed by filtration and
the DNA was extracted with phenol, precipitated with ethanol and used in PCR
reactions. After five rounds of selection and amplification, the final PCR
product was digested with
Eco
RI and
Bam
HI and cloned in similarly-digested pUC18 DNA. The insert regions of randomly selected clones were
sequenced by the dideoxy method using M13 -20 and -40 universal primers. The random pool oligonucleotide was also
amplified by PCR and the PCR product was cloned in pUC18 as described above.
Fifty clones were sequenced and the insert regions in selected clones were
amplified by PCR for use in binding reactions.
RESULTS
Binding studies with modified oligonucleotides
To assess major groove versus minor groove contacts by finger 4, we monitored
the affinity of a recombinant polypeptide containing the five N-terminal zinc fingers, zf1-5 (
29
), to a series of double-stranded oligonucleotides in which specific modified bases have been
introduced in the finger 4 binding site (Table
1
). For these experiments, we compared the binding affinities of zf1-5 and zf1-3, since the binding site for fingers 1-3 was not altered in the modified oligonucleotides, and
hence, no change in zf1-3 affinity was expected. Extent of binding was determined using gel
mobility shift assays with each double-stranded oligonucleotide and various amounts of the recombinant proteins. The fractions of bound and free DNA at each protein concentration were determined by laser densitometry of the gel autoradiograms. Examples
of binding titrations are shown in Figures
1
and
2
. Oligonucleotides were synthesized in which each of the adenine residues
between ICR positions +72 to +77 have been replaced by 7-deaza-adenosine. If major groove interactions with the N-7 position of adenine are important for finger 4 binding, then
this oligonucleotide should have a reduced affinity for zf1-5. Gel mobility shift assays with this double-stranded oligonucleotide, and with heteroduplex oligonucleotides in
which the A residues on only one strand have been replaced with 7-deaza-adenosine, do not show significant reduction in zf1-5 affinity for this DNA compared to the unmodified fragment
of similar length (Table
1
). The heteroduplexes exhibit a small increase in affinity for zf1-5 compared to the unmodified oligonucleotide. This increase in affinity
could be due to a change in DNA conformation adopted by these modified
oligonucleotides that results in a higher affinity binding site for finger 4
(see below).
Binding affinities of zinc finger proteins with modified oligonucleotides
a
Affinities were measured by protein titration with a gel mobility shift assay
(28) and are expressed as the ratio of the apparent dissociation constants +- relative errors, as determined with the Kaleidagraph program. Values
above 1.00 indicate affinities higher than that for the wild-type DNA sequence (NC-reg/C-reg) while values below 1.00 indicate affinities below that
for the wild-type sequence.
b
NC and C denote non-coding (top) and coding (bottom) strands; reg, unmodified oligonucleotide;
da, 7-deaza-adenosine substituted strand. Nucleotide positions indicated above refer to positions within the 120 nt 5S
ribosomal RNA gene.
Another DNA modification to test for major groove versus minor groove
interaction is replacement of the adenines at positions A74 and A77 on the top
strand and A73 and A76 on the bottom strand with deoxyinosine (Table
1
; Fig.
1
). The corresponding base-pairing thymines have been replaced with cytosine. The dI-dC base pair has the same functional groups as A[middot]T base pairs in the minor groove but resembles G[middot]C base pairs in the major groove. If major groove
interactions with either the As or Ts are important for finger 4 recognition
then this modified oligonucleotide should have a reduced affinity for zf1-5. If, however, only minor groove interactions are important, then this
oligonucleotide should bind with comparable affinity as the wild-type sequence. This approach has been used previously to demonstrate the
minor-groove binding activity of the TATA-box binding protein (
37
) and Hin recombinase (
38
). DNA binding assays do not detect a significant decrease in binding affinity
for either zf1-3 or zf1-5 with this dI-dC oligonucleotide compared to the unmodified
oligonucleotide (Fig.
1
; Table
1
), confirming that finger 4 does not participate in major groove contacts.
Effects of base substitutions within the finger 4 binding site
Previous studies with full-length TFIIIA have failed to demonstrate any significant effects of sequence changes in either the finger 4 or 6
binding sites (
19
-
22
). However, single amino acid changes in these fingers (
30
) and broken finger mutations (
31
) have documented the importance of fingers 4 and 6 in DNA binding. Since zf1-5 binds 5S DNA with a higher affinity than the full-length nine finger protein (
30
), we monitored the affinity of zf1-5 for DNA fragments harboring mutations in the finger 4 binding site
(Table
2
). One such mutation is contained in the clone pXlo73-76 (
22
), where the finger 4 binding site has been altered at nucleotide positions
73-76 from TAGT to GGAC (on the top strand). For these experiments, we
amplified a 57 bp DNA fragment from the wild-type pXlo and pXlo73-76 clones by PCR and used these fragments in binding assays with zf1-5 (Fig.
2
A). The 73-76 mutation causes a ~2-fold reduction in zf1-5 binding affinity, corresponding to a loss of free
energy of ~0.5 kcal/mol (Table
2
). We also measured the affinity of zf1-5 to other DNAs harboring mutant finger 4 binding sites. A double-stranded oligonucleotide was synthesized with each of the four
deoxynucleotides included at nucleotide positions +72 to +78 of the 5S RNA gene
(shown at the top of Table
3
). The important contacts for zinc fingers 1-3 (the C-box) and the pair of G residues required for finger 5 binding at
the intermediate element (
23
,
29
) are preserved in this oligonucleotide. The binding site for finger 4, however,
has been randomized. This mixture of oligonucleotides bound zf1-5 with an apparent dissociation constant of 6 nM, corresponding to ~12% of the affinity of the wild-type sequence and to a 1.2 kcal/mol loss in binding free energy (Fig.
2
B; Table
2
). The affinity of zf1-5 for this mixed sequence oligonucleotide is similar to the affinity of
zf1-3 for oligonucleotides of similar length (
28
,
31
). This loss in affinity suggests that the sequence of the finger 4 binding site
is indeed important for zf1-5 binding and that randomization of this sequence results in an affinity
similar to that observed for zf1-3 (
31
). We also tested the affinity of several DNA fragments with different sequences
in the finger 4 binding site. PCR products were derived from individual clones
of the mixed sequence oligonucleotide (sequences 1-3, Table
2
) and used in binding titrations with zf1-5. The A+T-rich sequence, ATTTTA at positions +73 to +78 (sequence 1), binds
zf1-5 with 7.5% of the affinity of the wild-type sequence. Two additional DNA fragments that had mutant finger
4 binding sites were also tested for zf1-5 binding affinity (sequences 2 and 3). These sequences had all four
nucleotides on both strands; however, these sequences were more G+C-rich (3 and 4 G+C bp out of 7 bp, respectively) than the wild-type sequence (2 G+C bp out of 7 bp). The PCR products from these
two clones also bound zf1-5 with low affinity (Table
2
). In contrast, another G+C-rich DNA sequence bound zf1-5 with high affinity (sequence 4, Table
2
). These data suggest that optimal binding of finger 4, at least in the context
of zf1-5, does not absolutely require the wild-type sequence and that other DNA sequences can substitute for the
wild-type sequence for proper zf1-5/DNA interactions.
Binding affinities of zf1-5 with DNAs harboring mutant finger 4 binding sites
a
Binding site DNA fragments were 57 bp PCR products obtained from the indicated
DNAs (wild-type pXlo and pXlo73-76), the mixed sequence oligonucleotide shown in Table 3 (random
pool), from individual pUC18 clones isolated from this oligonucleotide mixture
(sequences 1-3), or from a selected clone from the SAAB assay (sequence 4). The
sequence of the finger 4 binding region is shown.
Nucleotides that differ from wild type are underlined.
b
Measured by protein titration.
c
[Delta][Delta]Go values were calculated from the apparent dissociation
constants measured for each DNA fragment according to the equation [Delta]Go = -
RT
ln(1/
K
d
), where T = 295 K, and are compared with the dissociation constant determined
for the wild-type sequence.
d
Average of three determinations +- SD.
If base-specific recognition is involved in the binding of finger 4 to
DNA, then a selection and amplification binding (SAAB) assay (
36
) should recover the finger 4 binding site from a mixed sequence pool of
oligonucleotides selected with zf1-5 for high-affinity binding sites. If high affinity binding does not require a
unique sequence in the finger 4 binding site, then no sequence selection will
be observed in this assay. The randomized finger 4 binding site oligonucleotide
mixture (Table
3
) was used for binding site selection. Since zinc fingers 1-3 impart >90% of the binding energy for the full-length protein-DNA interaction (
28
,
31
), binding site selection assays were performed at a very low concentration of
protein, such that only 3-5% of the oligonucleotide is bound to the protein. Bound DNA was
recovered from a mobility shift gel and amplified by PCR using appropriate
primers. After each round of selection and amplification, a protein titration
experiment was performed to monitor changes in affinity of the selected DNA for
zf1-5. As discussed above, the starting mixed sequence oligonucleotide bound
zf1-5 with an apparent dissociation constant of 6 nM (Table
2
). After five rounds of selection and amplification, the affinity of zf1-5 for the selected DNA reached that of the wild-type sequence (~ 1 nM, data not shown). The fifth round PCR product was cloned
and 38 representative clones were sequenced. Table
3
lists a compilation of the frequency of occurrence of each nucleotide at each
position within the randomized finger 4 binding site. Surprisingly, the wild-type sequence was not recovered; however, the selection protocol recovered
a very purine-rich, and generally G-rich, sequence on the top strand. The G-richness of the selected sequence decreases in a 5' to 3' polarity.
The sequences of 38 clones obtained after five rounds of SAAB selection are
shown along with the frequencies of nucleotides selected at each position.
As a control, the mixed sequence oligonucleotide was amplified by PCR and the unselected PCR product was cloned. Sequence analysis of randomly
selected clones showed that the observed guanine preference in the SAAB assay
was not due to a drastic bias for this nucleotide in the starting
oligonucleotide mixture (data not shown). To ensure that the selection procedure recovered sequences with high binding affinity, we monitored the affinity of zf1-5 for a highly selected consensus sequence from the SAAB experiment
(sequence 4 shown in Table
2
). The PCR product obtained from this clone bound zf1-5 with high affinity (Fig.
3
B; Table
2
), verifying that the selection protocol enriched for DNA sequences with a
higher binding affinity for zf1-5 than the starting oligonucleotide mixture. The selected G-rich sequence in the finger 4 binding site creates a new sequence that is identical in eight out of 14 positions (ICR positions 68-82) to the sequence of the binding site for fingers 1-3 (the C-block). This suggests that the SAAB assay may
have selected for an additional zf1-3 binding site. To assess this possibility, we monitored the affinity and
nature of the complexes formed between the highly selected sequence 4 (shown in Table
2
) and zf1-3. Similar to the wild-type 5S gene sequence, only one high affinity binding site for zf1-3 is present in this sequence (as determined by non-denaturing gel electrophoresis, data not shown).
Moreover, the affinity of this sequence for zf1-3 is similar to that of the wild type (5-7 nM). These controls suggest that the SAAB assay selected for
high affinity binding sites for zinc finger 4. The finding that the wild-type sequence was not recovered by the SAAB assay suggests that finger 4
does not require a specific sequence for high affinity DNA-binding. The result of this selection protocol suggests that a purine-rich/G-rich sequence can also suffice for optimal binding of zf4 in
the zf1-5 protein.
SAAB selection of finger 5 binding sites with zf1-3/5
The predicted amino acid sequence of zf1-3/5 and the sequence of the aptamer oligonucleotide used for SAAB
selection are shown at the top of the table. The sequences of 26 clones
obtained after three rounds of SAAB selection with zf1-3/5 are shown along with the frequencies of nucleotides selected at each
position.
To further examine the role of finger 4 in DNA binding, we designed an
expression vector for a zf1-3/5 protein (see Materials and Methods), in which the finger 4 coding
sequence was deleted and the coding sequences for fingers 3 and 5 were joined
by the sequence TGEKP. This linker amino acid sequence is found in the zf1-3 linkers and in the linkers of numerous other zinc finger proteins (
39
). The predicted protein sequence is shown at the top of Table
4
. We expected that this protein would optimally bind a DNA sequence with the
native finger 5 binding site located just upstream from the binding site for
fingers 1-3, such that contiguous major groove interactions with these four fingers
would be favored. Previous binding studies with truncated TFIIIA polypeptides (
29
,
34
) have shown that finger 5 interacts with the intermediate element of the ICR.
In vitro
transcription assays originally identified this element as nucleotide positions
67-72 (
40
). However, more recent mutagenesis and TFIIIA binding experiments (
23
) have indicated that finger 5 requires only the pair of guanines at nucleotide
positions 70 and 71 (on the non-coding strand) for wild-type TFIIIA binding affinity. Zf1-3/5 protein was expressed in bacteria, purified and used in a
SAAB selection assay with the same mixed sequence oligonucleotide that was used for zf1-5 selection (Table
3
). Zf1-3/5 bound the starting input oligonucleotide with an apparent
dissociation constant of 9.1 +- 2.6 nM in a protein titration experiment (data not shown). Limiting
amounts of protein were used at each round of selection such that <5% of the input DNA was converted into complex with zf1-3/5. After three rounds of selection and amplification, the final PCR
product was cloned in pUC18 and 26 individual clones were sequenced. Table
4
shows a compilation of the frequency of occurrence of each nucleotide at each
of the seven randomized positions in the starting mixed sequence
oligonucleotide. Surprisingly, the wild-type finger 5 binding site sequence was not selected at a position near to
the zf1-3 binding site. As for the SAAB selection with zf1-5, guanines on the top strand near the 5'-end of the randomized sequence were strongly
selected.
Binding assays were also performed with PCR products from selected clones that
contain the wild-type finger 5 binding site at different positions relative to the binding
site for fingers 1-3 (nucleotide positions +80 to +92; Table
5
). When the sequence GGT was adjacent to, or spaced 2 or 3 nt upstream from
position +80, very poor binding was observed, corresponding to a ~2 kcal/mol loss in binding free energy, compared with zf1-3. When the GGT sequence was spaced 1 nt upstream from position +80,
high affinity binding by zf1-3/5 was observed. However, this binding affinity was similar to that of
the starting mixed sequence DNA used in the SAAB assay. This DNA bound zf1-3/5 with a 0.5 kcal/mol loss in free energy, relative to zf1-3, suggesting that finger 5 slightly decreases the binding energy
of zf1-3 when placed out of context in this artificial protein. These results
suggest that finger 4 is necessary for optimal interactions of fingers 1-3 and 5 with the 5S RNA gene.
Binding affinities of zf1-3/5 with DNAs harboring selected binding sites
+72 to +79 Sequence
a
Apparent
K
d
[Delta][Delta]G versus zf1-3
for zf1-3/5(nM)
b
(kcal/mol)
c
......GG
GGT
GTC....
86 +- 9
+1.7
......GGA
GGT
AC....
103 +- 5
+1.8
......CAAC
GGT
T.....
12.3 +- 1.3
+0.5
......CGCAG
GGT
.....
110 +- 9
+1.8
a
Binding site DNAs were 57 bp PCR products obtained from selected clones of the
random sequence pool oligonucleotide (shown in Table 3) cloned in pUC18.
b
Measured by protein titration.
c
[Delta][Delta]Go values were calculated from the measured apparent
dissociation constant for the reaction of each DNA fragment with zf1-3/5 protein relative to an apparent dissociation constant of 5 nM for zf1-3 measured with a DNA fragment of similar length as described in
Table 2.
DISCUSSION
The results presented in this paper are consistent with the model for the TFIIIA-DNA complex (
29
,
32
,
34
) in which fingers 4 and 6 each bind in or across the minor groove to bridge
adjacent fingers that bind in the major groove. These results are also consistent with previous methylation protection (
25
) and interference (
24
,
29
) experiments which failed to show any major groove guanine contacts over the finger 4 binding site (
29
). Our current results are also in agreement with hydroxyl radical footprinting
experiments with zf1-3 and zf1-4 proteins that directly demonstrated protection of the minor
groove by finger 4 (
34
).
The binding experiments with modified bases in the finger 4 binding site are
also consistent with placement of finger 4 in the minor groove: replacement of
each of the A[middot]T base pairs in the finger 4 site with I[middot]C base pairs only reduces binding affinity of zf1-5 by ~30%, which corresponds to a loss in binding free
energy of 0.2 kcal/mol. Previous studies with TBP and with Hin recombinase have
also shown that the A[middot]T to I[middot]C modification does not result in a significant loss in the
binding affinity of either minor groove-binding protein (
37
,
38
). The I[middot]C base pair has a similar minor groove structure to the A[middot]T base pair but the major groove of the I[middot]C base pair resembles the surface of the G[middot]C base pair. At a minimum, this result with zf1-5 indicates that finger 4 does not
participate in A or T contacts in the major groove at nucleotide positions 73,
74, 76 or 77 of the ICR. Similarly, replacement of each of the adenine residues
at these positions (and A72) with 7-deaza-adenosine resulted in a ~30% loss in binding affinity of zf1-5. This modification removes a nitrogen atom in the
major groove that could conceivably participate in hydrogen bond formation with
amino acids in the finger tip or [alpha]-helix of finger 4. For example, in the GLI-DNA structure, Ser 147 of finger 5 forms a hydrogen bond to N7 of adenine in the
major groove (
7
). The small loss in binding affinity (0.2 kcal/mol) due to the deaza
modification is not consistent with H-bond contacts with adenine residues in the major groove. Taken together
then, these results are consistent with finger 4 lying in or crossing the minor
groove to bridge the binding of fingers 1-3 and finger 5 in the major groove.
Mutation of the finger 4 binding site (in pXlo73-76) resulted in a modest 2-fold loss in zf1-5 binding affinity. This drop in affinity corresponds to a
loss in binding free energy of 0.4 kcal/mol, which is not consistent with a
loss of a hydrogen bond contact. This finding is also consistent with the
results of missing contact experiments which found energetically important
contacts only in the A-block, intermediate element and C-block (
32
). The 73-76 mutation did not result in a loss of binding affinity of full-length TFIIIA as measured with a nitrocellulose filter binding
assay (
22
). This difference between the present and previous results is likely due to the
fact that zf1-5 has a higher affinity for DNA than does full-length TFIIIA (
30
) and this increased affinity allows for a more sensitive assay for small
contributions of DNA sequence (or DNA structure) to the interaction of finger 4
with DNA. Additionally, the contribution of finger 4 to the binding affinity of
a five finger protein may be more significant than that in the full-length nine finger protein.
In contrast to this relatively small loss in affinity with the 73-76 mutation, randomization of the sequence of the finger 4 binding site
results in an 8-fold loss of affinity for zf1-5, reducing the affinity to that of zf1-3. Other mutations in the finger 4 binding site also exhibit
significant losses in binding affinity (Table
2
). These results initially suggested that high affinity zf1-5 binding requires the wild-type sequence within the finger 4 binding site; however, the result
of the SAAB assay indicated that this was not the case. The finding that the
wild-type sequence in the finger 4 binding site was not recovered in this
experiment suggests that the precise nucleotide sequence is not absolutely
required for the high affinity binding by zf1-5. Clearly, the purine/G-rich sequence selected in this assay can bind zf1-5 with comparable affinity as the wild-type sequence (Table
3
; Fig.
2
B). This observation prompts us to suggest that the DNA structure of the finger
4 binding site, rather than the nucleotide sequence, dictates high affinity
zinc finger interactions.
Our results with the zf1-3/5 protein demonstrate the importance of finger 4 for providing optimal DNA contacts for adjacent zinc
fingers. Based on previous phage display selection experiments (
8
-
12
), we expected that finger 5 would bind to its native binding site when this DNA
sequence was placed either adjacent to or a few (one to three) nucleotides 5' to the binding site for fingers 1-3. Moderate binding affinity was observed when the native finger 5
binding site was located with a single nucleotide spacing; however, this
affinity was lower than the affinity of fingers 1-3 in isolation (zf1-3). Thus optimal interactions between zinc fingers and DNA may
also be dependent upon finger-finger interactions in addition to the zinc finger-DNA contacts revealed in the three co-crystal structures published to date (
5
-
7
). Future structural studies will be needed to reveal the path of the zinc
fingers in the major and minor grooves and the precise DNA conformation within
the TFIIIA complex.
ACKNOWLEDGEMENTS
Supported by grants from the National Institute of General Medical Sciences to
J.M.G. (GM26453 and GM47530) and to P.E.W. (GM36643). K.R.C. received support
from a National Institutes of Health postdoctoral fellowship award (F32 CA09023). We thank Dr P. Romaniuk for his suggestions regarding controls for the SAAB
assays and discussions of this work.
