| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J. Biol. Chem., Vol. 278, Issue 37, 35597-35608, September 12, 2003
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
¶ ||
From the
Department of Chemistry and
¶Curriculum in Materials Science, the University
of North Carolina, Chapel Hill, North Carolina 27599
Received for publication, April 29, 2003 , and in revised form, June 13, 2003.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
residues 543–546), which lies across from the
downstream DNA, as the putative allosteric NTP binding site. We present a
structural model in which the NTP binds to the streptolydigin loop and upon
pairing with the +1 DNA base in the unactivated state or the +2 DNA base in
the activated state facilitates translocation via a ratchet motion. This model
provides an alternative mechanism for pausing as well as a structural
explanation not only for our kinetic data but also for data from elongation
studies on yeast RNAP II. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
A common regulatory signal in transcription is pausing. Pause and termination signals are multipartite and depend on a nascent RNA hairpin, the 3'-proximal region of the DNA, the 3'-RNA nucleotide, the incoming nucleoside triphosphate (NTP), and the DNA sequence downstream from the 3'-end of the RNA (11–16). In fact, downstream DNA and the 3'-proximal regions appear to play important roles in the formation of a paused complex (12, 17–20). Interestingly, changing the DNA downstream adjacent to pause and termination sites can affect their pause and termination efficiencies (14, 18, 21–23). Even subtle single base pair changes in this downstream DNA can strongly affect these efficiencies (21, 22). As such, it is important to identify and examine downstream DNA sequences that may be important in the regulation of transcription elongation. In this report, we investigate the effects of downstream sequences on the detailed kinetics of nucleotide incorporation.
We have previously presented compelling evidence for the existence of an allosteric binding site on Escherichia coli RNAP that is specific for the incoming NTP (1). Binding of the templated NTP to the allosteric site shifts RNAP from an unactivated (slow) to an activated (fast) state. We demonstrated that the kinetics of nucleotide addition to the RNA chain follows a non-essential allosteric activation mechanism where nucleotide addition can occur in either the fast or slow states, with the transition to the fast state being induced by nucleotide binding to the allosteric site (1). The allosteric site provides a means of regulation for the enzyme by "trapping" complexes into the slow state in the absence of or at low concentrations of the correct NTP.
To assess the effects of downstream DNA sequences on the non-essential
allosteric activation model and, therefore, on the distribution between the
slow and fast states, we have constructed mutant DNA templates and examined
the effects of different downstream DNA sequences on the rate of nucleotide
addition to the RNA chain. We demonstrate that the DNA sequence 1 bp
downstream of the site of incorporation (+2) can affect the rate of nucleotide
incorporation. Functionally, these data implicate a conserved loop on the
subunit as being important in elongation. Detailed sequence and
structural analyses reveal that this loop could comprise the allosteric
nucleotide binding site. Combining our kinetic data with molecular modeling,
we present a model for nucleotide synthesis and RNAP translocation involving
this loop.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
PR
promoter, and coded for a transcript in which the first cytosine is at
+25.
In Vitro Transcription Reactions
RNAP (60 nM) and 5'-biotinylated DNA template (60
nM) bound to streptavidin-coated magnetic beads were incubated 10
min at 37 °C in 1x TB (30 mM HEPES (pH 8.0), 10
mM Mg+2 glutamate, 200 mM
K+ glutamate, 25 µg/ml bovine serum albumin, and 1 mM
dithiothreitol) to form open complexes. Complexes stalled at +24 were formed
by adding 15 µM UTP, 20 µM ATP, and 20
µM [
-32P]GTP (150 Ci/mmol) and incubating at
room temperature for 1 min. The complexes were washed seven to ten times using
ice-cold 1x TB by holding the reaction tube next to a strong magnet to
retain the complexes. The complexes were then resuspended in ice-cold 1x
TB. In some experiments, the complexes were then kept on ice until used in the
rapid kinetic experiments. In most of the experiments, 10 µM CTP
was added to "walk" the complexes to +25 at room temperature, and
then the complexes were purified again after 1 min as described above and kept
on ice until used in the rapid kinetic experiments. The rapid kinetic
experiments were performed at room temperature (
23 °C), using a
KinTek Rapid Quench Flow apparatus. For each time point, 20 µl of the
purified elongation complexes was injected into one reactant loop, and 20
µl of the designated NTP(s) was injected into the other reactant loop. The
reactants were mixed for the indicated times and subsequently quenched with
0.5 M EDTA. Each time point represents a separate experiment. To
assure that the results were not dependent on the length of time the complexes
remained on ice, time points were taken in different orders. At designated
times during the reactions, a portion of the purified elongation complexes was
removed and extended to full-length by the addition of 1 mM of all
four NTPs (chased) to establish that the complexes were still active. The NTP
concentrations reported are final concentrations after mixing in the quench
flow. Prior to running the samples on 8 M urea, 20% polyacrylamide
gels, the EDTA was removed from the complexes with the aid of a magnet as
described above, and the products were resuspended in 100% formamide.
Data Analysis
Quantification and Normalization of Rate Data—The amount of
radioactivity in each lane of the gel was measured on an Amersham Biosciences
PhosphorImager and analyzed with ImageQuant software. The percentage of
complexes at each position on the template was calculated by dividing the
amount of radioactivity in the indicated band by the total amount of
radioactivity in all the bands 
24 nucleotides in length. To compare data
from different experiments, it was necessary to normalize the data such that
at 0 time, there was 0% incorporation, and, upon completion, there was 100%
incorporation. The data were normalized to 100% by dividing each time point by
the highest percentage of complexes that incorporated beyond either +24 or
+25, as described previously
(1). The experiments were
conducted two to six times for each concentration.
Fits of the Kinetic Data to the Mechanism—For both the wild type and A27c templates, each data set was fit to both single- and double-exponential equations. The quality of the double-exponential fits was significantly better than that of the single-exponential fits, because the data were clearly biphasic. The data from the double-exponential fits of the individual rate curves were used as a starting point to obtain initial values for binding constants to the catalytic and allosteric sites and the rate constants for the unactivated and activated states as previously described (1). Dynafit (26) was used to attempt to fit the data to the previously described non-essential allosteric activation mechanism in which all NTP binding steps were assumed to be in rapid equilibrium (1). All [ATP] were fit simultaneously, and all binding constants and the polymerization rate constants were allowed to vary over several orders of magnitude. Despite these procedures, we were not able to find a single set of constants that produced reasonable fits to the data for all [ATP] using the published mechanism. This result is not surprising, because this mechanism does not account for the biphasic nature of the kinetics.
To fit the biphasic curves, we needed to refine the mechanism (see text). Fig. 2A shows the fewest number of changes that could be made to the mechanism that would allow it to fit the biphasic data. We fit the data to this new mechanism both by computer using Dynafit (26) and by hand using KinSim (27) as previously described (1). The rapid equilibrium-binding step to the allosteric site after the catalytic site is filled (gray reaction arrows) was removed. Simulations performed using KinSim (27) demonstrated that including this step with all possible permutations of the constants destroyed the biphasic nature of the curves produced; therefore, it was essential to remove this step. The only rapid equilibrium step that remained was Kallos (Fig. 2, A and B). Nucleotide binding to the catalytic site was no longer a rapid equilibrium step. There may have been a conformational change that occurred just before these binding steps; however, we could not observe it.
|
|
residue 326 and at
the catalytic site. This complex was then minimized using CNSsolve to remove
any bad contacts. A UTP that would base pair with the +1 downstream DNA base
(ATP) was then built using Sybyl and docked in to the model of the elongation
complex. This complex was then minimized using Sybyl to remove bad contacts
between the UTP and its binding loop. The same procedure was used to make a
model with Thermus aquaticus RNAP.
|
Sequence and Structural Alignments
The alignments shown in Fig.
5 were first done using Vector NTI. These alignments were then
checked by superimposing the structures of T. thermophilus, T.
aquaticus, and Saccharomyces cerevisiae RNA polymerase II in
InsightII. The specific areas that were analyzed are those that are described
in Fig. 5. It was then
discovered that these alignments were not correct and that the previously
published alignments (28) were
not correct. The Vector NTI alignments were then corrected accordingly based
on the observed structural alignments.
|
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
|
Refining the Kinetic Mechanism of Nucleotide
Incorporation—Previously, we demonstrated the existence of an
allosteric binding site by measuring the kinetics of CMP addition at position
+25 and of AMP addition at position +26 on the DE13 template
(1). We characterized the
kinetics of CMP addition at position +25 as a function of [CTP]
(1). The data were fit to a
non-essential activation mechanism, in which the substrate acts both as a
substrate and an allosteric effector. In this model, a rapid equilibrium
between the activated (fast) and unactivated (slow) states was assumed.
Although this mechanism (Fig.
2A and Ref.
1) is sufficient to describe
the concentration dependence of the kinetics, it can not reproduce the
biphasic nature of the kinetic curves at intermediate concentrations of CTP,
because the rapid equilibrium assumption produces single-exponential kinetics.
Because the rate of synthesis in the slow state and the fast state differ by
only 30-fold, single exponentials were sufficient to get reasonable fits.
To further characterize the kinetic mechanism, we investigated the kinetics of
AMP addition at position +26 as a function of [ATP] on the DE13 template and
several variants of it.
As can be seen from inspection of the data in Fig. 3, double-exponential fits were required to obtain reasonable fits to the kinetic data for AMP incorporation at position +26. The enhanced biphasic nature of the kinetic data at position +26 relative to position +25 results from greater differences between the rates of synthesis in the slow and fast states. The biphasic nature of the curves indicates that there is not a rapid equilibrium between the activated (fast) and unactivated (slow) states as was previously assumed (1). Not surprisingly, we were not able to find a set of rate and binding constants that could fit the data at all ATP concentrations using the previous mechanism. Accordingly, we set out to modify this rapid equilibrium non-essential activation mechanism to accommodate these new data. Fig. 2A shows one of the two simplest kinetic mechanisms that could fit the data. In this mechanism, like the previous one, there is still a rapid binding equilibrium of the templated NTP to the allosteric site; however, unlike the previous one, NTP binding to the catalytic site is not in rapid equilibrium, and once an NTP became bound in the catalytic site there is not a rapid equilibrium between the unactivated (En:SC) and activated (E*n:SA:SC) states. The rapid equilibrium binding of the NTP to the allosteric site is required, because if binding were slow, a fast synthesis rate would not occur, and if dissociation of the NTP from the allosteric site were slow, all of the complexes would end up in the fast state. In contrast, if NTP binding to the catalytic site were in rapid equilibrium, biphasic kinetics would not be observed.
|
Inspection of Fig. 4 (A and B) reveals that the new mechanism (Fig. 2A) could fit the data for all [ATP] (for position +26) and all [CTP] (for position +25) (taken from Ref.1), with a single set of rate and binding constants for each template position (shown in Table I). This mechanism fits the data for position +25 better than the previously published mechanism (Fig. 4B) (1). As a result of the better fits, we were able to fit more data sets for position +25 to this non-equilibrium non-essential activation mechanism than to the previous mechanism.
|
|
We were also able to fit the data to the mechanism shown in Fig. 2B using the same constants shown in Table I except that the new values for k'unact are shown in braces. The mechanisms depicted in Fig. 2 (A and B) are kinetically indistinguishable. In the mechanism shown in Fig. 2A, it is assumed that there is an equilibrium between the pre- and post-translocated states with the NTP binding to RNAP in the post-translocated state (EnT); whereas, in the mechanism shown in Fig. 2B, the NTP binds to the allosteric site in the pre-translocated state of synthesis (EnPT) and induces translocation. In this latter model, the unactivated state is represented by the allosteric NTP switching over to the catalytic site, thus becoming the catalytic NTP. The activated state is represented by a second NTP binding to the catalytic site after NTP binding to the allosteric site induces translocation. The main difference between these two mechanisms is that in the former case (Fig. 2A), an equilibrium between the pre- and post-translocated states is assumed and the first incoming NTP can bind to either the catalytic site or the allosteric site; whereas, in the latter mechanism (Fig. 2B), the incoming NTP must first bind to the allosteric site and induce translocation from the pre- to the post-translocated state before the catalytic site can be bound. These mechanisms will be discussed further in the context of the RNAP crystal structures.
Using both Dynafit (26) and KinSim (27), we tried to fit the data to several other mechanisms, including an ordered binding mechanism in which the incoming NTP must first bind to the catalytic site before the allosteric site can be filled, and an essential activation mechanism in which catalysis occurs only in the activated state (1). Despite all permutations, the alternative mechanisms could not fit the data. The mechanisms presented are the two simplest mechanisms that can fit the data.
To determine the rate and binding constants
(Table I) for the mechanisms
shown in Fig. 2, we fit the
data both by hand and by computer. The manual fits were done by simulating the
kinetic data for each [NTP] using the program KinSim and manually changing the
kinetic constants to obtain the best fits visually. The computer fits were
done using the program Dynafit, which uses a Markov algorithm to fit the data,
and the data for all [NTP] were fit simultaneously; that is, the fits were
globally optimized. In both cases, we varied the starting values of the rate
and binding constants over 5 orders of magnitude, and in both cases, the
constants converged to the values shown in
Table I. These results strongly
suggest that there is only a single set of rate and binding constants that can
fit the data; i.e. these constants are unique.
The Unactivated State Is the Primary Regulatory State— Overall, the rate of synthesis is slower at position +26 (wild type template) than at position +25 (Fig. 4, A and B, and Table I) signifying that the rates of addition can differ between template positions in a given elongation complex as has been observed in other studies of elongation complexes (4–6, 29–31). Specifically, the catalytic rate in the activated state, kfast, is seven times faster for position +25 than for position +26. The catalytic rate in the unactivated state, kslow (shown in bold face italics in Table I) for position +25 is at least 30 times faster than position +26 in the wild type template. The zero in parenthesis indicates that we were able to fit these data with a kslow of less than the indicated number, including a value of zero. These data correlate with our misincorporation studies in which synthesis is not observed in the unactivated state (2). The fact that kslow could be assigned a value of zero and still produce reasonable fits indicates that the unactivated state is not a productive synthesis state for this template position. Significantly, however, we could not obtain reasonable fits without requiring that NTPs bind to the catalytic site in this state. Consequently, our kinetic data indicate that this state is capable of binding NTPs to the catalytic site; however, it is not contributing to synthesis. Instead, the unactivated state is acting in an inhibitory manner at this template position, trapping complexes on a non-synthesis pathway and preventing them from getting on the activated path, which is the only synthesis pathway at this position. This condition is not true for the +25 template position, because we were not able to fit the data without a non-zero value for the rate of synthesis in the unactivated state. At low [CTP], there is a significant population of complexes in the unactivated state undergoing synthesis.
The significant differences in the catalytic activity of the unactivated state between template positions +25 and +26 suggest that this state may play important roles in the regulation of transcription elongation. The unactivated state can serve as a precursor state to a multitude of other conformationally different states (1, 2, 10). For example, pause, arrest, and termination states all appear to arise from the unactivated (slow) state of synthesis (6, 9, 10, 32, 33). Backtracked states have also been observed to come off of the unactivated pathway of synthesis (10, 32, 34). This state appears to be susceptible to GreA- and GreB-induced transcript cleavage and may be sensitive to NusA and NusG (2, 10). Synthesis in this state proceeds slowly with a significantly higher fidelity than the activated (fast) state (10). In addition, complexes in the activated state of synthesis are more resistant to pause and termination signals (4, 35).
The unactivated state can generally be thought of as a synthesis state as well as a regulatory state, capable of determining whether complexes will continue on a synthesis pathway or undergo a regulatory or rescue event (1, 2, 7, 10). We have demonstrated that, for a different template position, the unactivated state becomes a non-productive synthesis state, serving as a trap for complexes that can bind NTPs and acting in an inhibitory fashion. This state probably acts to slow down synthesis at pause or regulatory sites. These results support the idea that the unactivated state is the major regulatory state during elongation.
The Rate of RNA Synthesis Is Regulated by the Identity of the
+2 DNA Base—The difference in catalytic rates among different
template positions (Table I)
prompted us to examine the effects of downstream DNA on the kinetics of
nucleotide addition. Using site-directed mutagenesis, we constructed several
single-base pair substitutions, shown in
Fig. 1A, within the
DE13 (wild type) template sequence. Fig.
3A shows the kinetic curves for AMP incorporation at
position +26 at 5 µM ATP for four templates, each of which have
a different base at position +27. Notably, the rates of incorporation for the
two templates with purine bases in the template strand at this position (A27c
and A27u) exhibit markedly reduced rates relative to those with pyrimidines at
this position (A27 and A27g). Specifically, by 1 s, 50% of the complexes
formed with the wild type and A27g templates have incorporated an AMP at
position +26, whereas only 25% of the complexes formed with the A27c and
A27u templates have reached this position
(Fig. 3A). The
similarity of the rate of AMP incorporation for the A27c and A27u templates
indicates that the presence of a purine in the template strand one base pair
downstream of the site of AMP incorporation causes the RNAP to slow down
synthesis in this sequence context. This substitution is the only variable in
the experiments between the wild type template and the A27c and A27u
templates. Interestingly, the rate of incorporation of CMP at position +25
(two base pairs upstream from the mutation of interest) is unaffected by the
change in the DNA base (data not shown). Only the rate of addition of the base
immediately upstream from the DNA mutation is affected in these experiments
using this sequence.
Each of the mutant DE13 templates contains several pyrimidines in the proximal downstream region of the non-template strand of the DNA. At most intrinsic termination sites, the nascent RNA is shown to fold into a GC-rich hairpin followed by a stretch of U residues (11, 36–39). The presence of the U residues causes RNAP to slow down prior to the dissociation of the ternary complex (5, 6, 38, 39). Perhaps the series of pyrimidines in these mutant templates causes RNAP to decrease the rate of incorporation, similar to a pause or termination site. To determine if the reduced rate of synthesis at +26 is specifically due to the presence of a single pyrimidine in the non-template strand at position +27, or to the presence of two contiguous pyrimidines at +27 and +28, or simply due to having many pyrimidines in the vicinity of this position, we constructed mutants on the A27c template in which the non-template strand downstream pyrimidines were changed to purines (Fig. 1B) to see if the higher rate of synthesis could be recovered. Fig. 3B shows the rates of incorporation of AMP at +26 for these templates as compared with the wild type and A27c templates at 5 µM ATP. Templates in which the non-template strand pyrimidines are changed to purines at +28, +30, and +36 still exhibit the decreased rate of AMP incorporation at position +26 as with the A27c template relative to the wild type. These results suggest that the observed decrease in AMP incorporation at +26 is due solely to the pyrimidine on the non-template strand (or a purine in the template strand) at position +27 in this sequence context. (We did not examine the effect of changing the upstream sequence, which may play a role in the observed results.)
The reduced rate of AMP incorporation could result from a decreased rate of NTP binding to the allosteric or catalytic sites, from a reduced rate of catalysis, or from a combination of effects. To understand the origin of this decreased rate, we have characterized the kinetics of AMP incorporation at position +26 using the A27c template and fit the data to the mechanisms described by Fig. 2 (A and B). To obtain constants that would fit the data for the A27c template using the mechanisms shown in Fig. 2 (A and B), the constants for the wild type template (position +26) were used as starting values. In addition, several other starting values for the rate and binding constants that were significantly different were used. As with other data sets, independent of the starting values, the constants always converged to the same final values.
Inspection of Fig. 4C shows that the kinetic mechanisms can fit the data for all [ATP] with a single set of rate and binding constants (Table I). The most significant difference between the wild type template and the A27c template is given in boldface in Table I. Specifically, the rates of NTP binding and dissociation to the catalytic site in the activated state (kact and k–act) are 1000 times faster for the wild type template than the A27c template. Perhaps the catalytic site in the activated state becomes difficult for NTPs to access when the A27c template is used. As discussed above, the rate of synthesis in the unactivated state (kslow) can be zero for this template position, indicating that this state is not a productive synthesis state. Consequently, at this template position, synthesis is forced to occur on the activated pathway, where the catalytic site in this state becomes less accessible to NTPs, thus explaining the reduced rate of synthesis observed when using the A27c template. The structural implications of this idea will be discussed further below.
Structural and Functional Implications—The observed
downstream sequence effects prompted us to examine which amino acid residues
on the RNAP could be responsible for distinguishing between a pyrimidine and a
purine base in the +2 downstream DNA. We examined the crystal structure of the
RNA polymerase II elongation complex and discovered that the residues nearest
to this base belong to a loop (fork loop 2 in Rpb2) that is disordered and
hence absent in the crystal structure
(28,
40). To further investigate
the potential role of this loop, we made models (see "Materials and
Methods") of elongation complexes of the T. aquaticus and
T. thermophilus RNAPs, because the loop of interest is ordered in
these crystal structures
(41–43).
Inspection of these models reveals that the homologous loop in the bacterial
RNAPs, which comprises residues 413–431 (D loop I) in T.
thermophilus and T. aquaticus RNAP (residues 533–541 in
E. coli), is within 5–6 Å of the downstream
template base. Sequence alignments indicate that this loop is conserved among
multisubunit prokaryotic RNA polymerases
(Fig. 5).
A comparison of this loop in the T. aquaticus structure (42) with the corresponding loop in the T. thermophilus structure (43) shows that the loop has a different conformation in each of the structures (Fig. 6). In particular, several of the conserved residues are in different conformations in the two different structures (Fig. 6). These data along with the fact that the loop is disordered in the RNA polymerase II structures suggest that the loop is flexible and can easily change conformations. Consequently, it could make unique interactions with different template bases, and it could play an important role in elongation in all multisubunit polymerases.
|
Interestingly, mutations in this loop have been shown to be associated with
streptolydigin resistance in E. coli
(44,
45) and in Bacillus
subtilis (46).
Streptolydigin is an antibiotic that inhibits transcription in E.
coli during initiation and elongation
(44,
47,
48). In particular,
-subunit residues 543–546 in E. coli
(44,
45) and 499, 500, and 502 in
B. subtilis (46)
confer streptolydigin resistance upon mutation (see
Fig. 5). These residues
correspond to 423–426 in T. thermophilus and T.
aquaticus RNAPs. The rifampicin binding region, which has been shown to
be important in elongation and nucleotide binding
(44,
45,
49), is nearby. This loop is
also very close to the bridge helix (' F helix)
(Fig. 7, A and
B) and could potentially interact with it upon a
conformational change in the enzyme.
The Streptolydigin Binding Domain May Comprise an NTP Binding Site—A deeper inspection of the streptolydigin-binding loop suggests that it may play a more important role in transcription and translocation than previously thought. A thorough investigation of the region of RNAP surrounding this loop (Fig. 7, A and B) reveals that it could comprise a nucleotide binding site. This site is on the opposite side of RNAP from the secondary tunnel and is solvent-accessible. We propose that this region of the protein contains the putative allosteric NTP binding site.
To assess whether this loop could comprise the allosteric NTP binding site,
we have examined the sequence and structure in the regions surrounding this
loop. Notably, the overall fold of this region is similar to nucleotide
binding domains. Specifically, it contains a flexible loop (D loop I)
flanked by a -sheet on one side and -helices on the other
(50–52)
(Fig. 7A). To
investigate the sequence conservation in these regions, we aligned the
sequences based on the crystal structures of T. aquaticus, T.
thermophilus, and S. cerevisiae RNAPs. This structural alignment
was necessary, because the published alignments based only on sequence were
misaligned, especially the -sheet region, between prokaryotic and
eukaryotic RNAPs (28). The
structure of the -sheet region is conserved between T. aquaticus, T.
thermophilus, and S. cerevisiae RNAPs, and its sequence shows
moderate conservation with a conserved hydrophobic region found in the
-sheets of NTP binding domains (Fig.
5).
The loop is glycine-rich, in both eukaryotes and prokaryotes, which is an
evolutionarily conserved feature found in "P-loops" responsible
for binding NTPs (50,
53). The sequence of the loop
is moderately conserved between prokaryotes and eukaryotes and is highly
conserved in each class. Significantly, the eukaryotic RNAPs contain a
conserved GK sequence motif found in the P-loops of NTP binding domains
(50,
53). It has been suggested
that the function of this lysine in P-loops is to interact with the
-phosphate of the NTP
(50,
53). Although the bacterial
RNAPs do not contain the conserved GK motif, they are glycine-rich and contain
three conserved arginine residues that could perform the same function.
Finally, in addition to the above conserved features, a totally conserved
Walker B motif (DXXG) is found at the rear of the loop, with the
position of the aspartic acid residue conserved in all of the RNAP crystal
structures (28,
40,
42,
43). In NTP-binding proteins,
the aspartic acid residue of this motif binds a magnesium ion, which, in turn,
binds to the 
-phosphate of the NTP
(50,
51,
53). Significantly, in the
structure of T. thermophilus holoenzyme, a magnesium ion is bound to
this aspartic acid residue
(43).
Further evidence that the loop may comprise an NTP binding site comes from
studies of a kinase. A large conformational change is required for binding of
an inhibitor near the ATP binding site of p38 mitogen-activated protein kinase
(54). In this conformational
change, a Phe side chain moves by about 10.0 Å
(54). Similarly, the Phe side
chain of residue 425 in the T. aquaticus RNAP structure has
moved about 14.0 Å relative to its conformation in the T.
thermophilus RNAP structure (Fig.
6). Taken together, these sequence and structural analyses
strongly support our suggestion that the conserved flexible loop comprises an
NTP binding site.
To investigate this idea further, we have modeled a UTP bound in the loop
of the models of the T. thermophilus
(Fig. 7, A and
B) and T. aquaticus (not shown) elongation
complexes. These models were made by docking a UTP into the loop followed by a
molecular dynamics minimization to remove bad contacts and partially optimize
contacts (Fig. 7, A and
B). This minimization only slightly altered the
conformation of the loop. Root mean square bond deviations were 0.01 Å
for both proteins, and root mean square angle deviations were 1.39 Å and
1.51 Å for T. Thermophilus and T. aquaticus,
respectively. Although this model is probably not entirely correct in the
details, it provides insight into how an NTP may bind in this loop and how an
NTP could act as an allosteric effector. Inspection of the model of T.
thermophilus RNAP (Fig.
7A) reveals that the magnesium ion, which is chelated to
the conserved Walker B motif aspartic acid residue is poised to interact with
the -phosphate of the UTP. The NTP lies on top of the loop in a manner
similar to that found in the P-loops of NTP-binding proteins
(52,
53). The conserved glutamate
residue at position 421 in T. thermophilus is in hydrogen bonding
distance from the 2'-OH of the ribose, thereby providing specificity for
ribose versus deoxyribose NTPs. In addition, there is a nearby
threonine residue (419) that could provide specificity for ribose
versus deoxyribose, as is often seen in the P-loops of NTP-binding
proteins (50,
51,
53).
Structural Model for Allosteric Activation: Binding of the NTP to the Allosteric Site Induces Translocation Via A Ratchet Motion—With this structural model in hand, it still remains to be explained how binding of an NTP to this loop could provide allosteric activation of RNAP. Inspection of Fig. 7 (A and B) reveals that an NTP bound in the loop is poised to interact with the +1 base in the downstream DNA, if RNAP is in the pre-translocated state, as it is in this model. This interaction would provide specificity in the allosteric site for the base that is to be incorporated, giving us a structural explanation for our kinetic observation that the allosteric site is specific for the incoming NTP (1).
We propose that in stalled complexes, RNAP resides primarily in the pre-translocated state. In support of this idea, the structures of the yeast RNAP II elongation complex (40) and the T7 RNAP elongation complex (55) are both in the pre-translocated state. We propose that productive binding (i.e. correct pairing with the +1 DNA base) of an NTP to the allosteric site induces a conformational change to the post-translocated state. Inspection of the structure supports this idea (Fig. 7, A and B). First, the loop interacts with the F-helix; therefore, a conformational change in the loop could lead to a conformational change in the F-helix, which interacts with the downstream end of the RNA-DNA hybrid. Second, and perhaps more importantly, the putative P-loop is directly linked, via primary sequence, to the rifampicin binding domain, which interacts extensively with the RNA in the RNA-DNA hybrid (Figs. 5 and 7B). Consequently, a conformational change in the loop upon productive NTP binding could lead to a cooperative movement of the F-helix and the rifampicin region, which, in turn, would lead to the translocation of the DNA and the RNA to the post-translocated state. This idea is best envisioned as a ratchet model, shown in Fig. 7B. Notably, McClure postulated over two decades ago that the streptolydigin-binding domain might overlap with an NTP binding domain and that NTP binding might provide some of the driving force required for translocation (47).
Once NTP binding to the allosteric site facilitates RNAP translocation to
the post-translocated state, the catalytic NTP binding site becomes available
for a second NTP to bind via the secondary channel, with binding resulting in
rapid catalysis in the activated (fast) state. There are two possible
structural explanations, which correspond to the two kinetic mechanisms shown
in Fig. 2, that could describe
synthesis in the unactivated (slow) state. First, if RNAP in a stalled complex
is in equilibrium between the pre- and post-translocated states, as has been
suggested previously (47), an
NTP could bind via the secondary tunnel when RNAP is in the post-translocated
state and be incorporated (kinetic mechanism in
Fig. 2A). The slower
rate of catalysis (kslow) in this state could result from
suboptimal alignment of the -phosphate with the 3'-OH of the RNA
chain (1). Alternatively,
synthesis in the unactivated state may result from the NTP first binding in
the allosteric site and inducing translocation, and subsequently being
transferred over to the catalytic site (kinetic mechanism in
Fig. 2B). Such a
transfer may be rate-limiting and result in a slow rate of synthesis. Although
it has been suggested that the NTPs must enter via the secondary channel,
there is no direct evidence to support this hypothesis; therefore, this model
is a reasonable explanation for the slow rate of synthesis in the unactivated
state. This model can also be used to explain why the unactivated state is a
higher fidelity state (1). The
allosteric site first provides the specificity for the correct NTP, and after
it is transferred over to the catalytic site, the catalytic site provides
additional specificity. The model shown in
Fig. 8 is based on this latter
model; however, we do not have enough evidence to discriminate between the two
models, which differ only in the details. It is likely that RNAP may follow
either kinetic mechanism (Fig.
2) depending on the sequence context.
The location of the allosteric NTP binding site being on the solvent-accessible side of the enzyme can explain our observation that the binding and dissociation of NTPs to the allosteric site is in rapid equilibrium. Because entrance to the catalytic site is probably through the secondary channel, a trafficking problem between NTPs might occur if both the catalytic and allosteric NTPs had to enter via the secondary channel. With our model, this trafficking problem is overcome by having the allosteric site (which is easily accessible to NTPs) provide the initial specificity for the correct NTP instead of the catalytic site. Once the correct NTP binds to the allosteric site, translocation is induced, and the catalytic site becomes accessible. In the fast state of synthesis, a second NTP can then bind to the catalytic site. The catalytic site now provides additional specificity for the correct NTP.
Using this model, we can now explain the effects that we observe with the change in the downstream DNA. The A27c template kinetic data indicate that the reduced rate of synthesis is primarily due to a reduced rate of NTP binding to the catalytic site in the activated state (Fig. 2, A and B, and Table I). Specifically having a purine in the +2 downstream DNA base in this sequence context slows down the rate of synthesis. Because of the proximity of this base to the loop, the allosteric NTP could be binding to the loop suboptimally in this case and may not be able to induce translocation efficiently as a result. If translocation cannot be induced, the catalytic site cannot become accessible to the incoming NTP. The inaccessibility to the catalytic site could be observed as a reduction in the rate of NTP binding. The fact that kslow can be zero for some template positions (Table I) provides evidence for an improper alignment as a result of an incomplete or improper translocation event. The kinetic data may provide some of the strongest evidence for our translocation model, because the only structural element of the enzyme near the downstream base is the loop. The reduction in rate of NTP binding to the catalytic site may result from an interaction of the downstream DNA with the loop. Alternatively, the downstream DNA may be in an improper alignment relative to the loop such that correct NTP binding to the loop does not induce proper translocation. Such states may be present at pause sites, providing an alternative structural explanation for pausing other than backtracking or hypertranslocation (10, 32, 33).
It has been demonstrated previously that synthesis leaves RNAP in a long-lived activated state (2, 10) suggesting that binding to the allosteric site is no longer necessary once the RNAP has entered the activated (fast) state. Once RNAP has entered this state, the loop may serve as a site where NTPs can be pre-loaded, ready to become the allosteric effector for the next round of synthesis. Consistent with this idea, studies of RNAP II elongation kinetics indicate that the NTP that is complementary to the +2 DNA base, instead of the +1 base, is an allosteric effector for elongation (56). Specifically, it was found that the +2 NTP enhances the rate of incorporation at +1 if RNAP is already on the activated pathway. It was suggested that binding of the complementary NTP to the +2 site facilitates translocation and plays an important role in processive synthesis (56). Interestingly, Shimamoto and Wu (57) discovered over 20 years ago that the kinetics of initiation is affected by the concentration of the subsequent NTP to be added. Our ratchet model for translocation provides a structural explanation for the results with RNAP II as well as E. coli RNAP. Specifically, if RNAP in the activated state resides primarily in a post-translocated state, as opposed to a pre-translocated state of stalled complexes, the +2 base of the template strand would be in the same position as the +1 base in our model (Figs. 7A,7B, and 8). Consequently, the specificity in the allosteric site would be for the +2 NTP in the post-translocated state and for the +1 NTP in the pre-translocated state. Binding of the +2 NTP to the allosteric site could help the RNAP to maintain the post-translocated state and thereby facilitate elongation.
Our kinetic data and structural analysis strongly suggest that binding of the allosteric NTP to the streptolydigin-binding loop facilitates translocation via a ratchet motion. This ratchet motion most likely involves a concerted movement of the rifampicin domain and F helix. We suggest that this loop also helps to maintain processive elongation by keeping RNAP in a post-translocated state. This model indicates that the allosteric NTP induces the transition from the unactivated to the activated state by playing a role in translocation. It is likely that further investigation of the allosteric NTP binding site will provide more evidence for this important function.
|
FOOTNOTES |
|---|
Present address: Laboratory of Molecular Genetics, National Institutes of
Health, Research Triangle Park, NC 27709.
|| To whom correspondence should be addressed. Tel.: 919-962-6370; Fax: 919-966-3675; E-mail: derie{at}email.unc.edu.
1 The abbreviations used are: RNAP, RNA polymerase; NTP, nucleoside
triphosphate.
|
ACKNOWLEDGMENTS |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
E*25:SA:SC,
shown in gray, was present and NTP binding to the catalytic site in
both pathways was assumed to consist of rapid equilibrium steps. The
pre-translocated state is shown in gray. Although this state is not
used in the fits, the equilibrium constant Ktrans is
contained within Kallos and kunact.
B, kinetic mechanism in which NTP binding to the allosteric site
induces translocation. The pre-translocated state is shown at the
bottom in the gray-colored schematic. A structural schematic
of this mechanism is also given in 
and
(see 
