The Impact of Dyskeratosis Congenita Mutations on the Structure and Dynamics of the Human Telomerase RNA Pseudoknot Domain

The Impact of Dyskeratosis Congenita Mutations on the Structure and Dynamics of the Human

Telomerase RNA Pseudoknot Domain

http://www.jbsdonline.com

Abstract

The pseudoknot domain is a functionally crucial part of telomerase RNA and influences the activity and stability of the ribonucleoprotein complex. Autosomal dominant dyskerato-sis congenita (DKC) is an inherited disease that is linked to mutations in telomerase RNA and impairs telomerase function. In this paper, we present a computational prediction of the influence of two base DKC mutations on the structure, dynamics, and stability of the pseudoknot domain. We use molecular dynamics simulations, MM-GBSA free energy cal-culations, static analysis, and melting simulations analysis. Our results show that the DKC mutations stabilize the hairpin form and destabilize the pseudoknot form of telomerase RNA. Moreover, the P3 region of the predicted DKC-mutated pseudoknot structure is unstable and fails to form as a defined helical stem. We directly compare our predictions with experimen-tal observations by calculating the enthalpy of folding and melting profiles for each structure. The enthalpy values are in very good agreement with values determined by thermal denatur-ation experiments. The melting simulations and simulations at elevated temperatures show the existence of an intermediate structure, which involves the formation of two UU base pairs observed in the hairpin form of the pseudoknot domain.

Keywords: Dyskeratosis congenita; Telomerase; RNA; Molecular modeling; and pseudoknot.

Introduction

Dyskeratosis congenita (DKC) is a rare human genetic disorder that is primarily characterized by abnormal skin pigmentation, nail dystrophy, leucoplakia (1), and anomalies in other tissues with rapid cell divisions. DKC causes premature mortal-ity mainly due to bone marrow failure. It can be found in three genetic subtypes, namely, X-linked recessive, autosomal dominant, and autosomal recessive (2, 3). The X-linked recessive type is associated with the DKC1 gene (4, 5) and the autoso-mal dominant type is associated with mutations in the RNA subunit of telomerase, hTR (6). The genetic basis for the autosomal recessive type is currently unknown (7). In both X-linked and autosomal dominant subtypes of DKC it has been shown that the patients have shorter than normal telomeres (6, 8-10), indicating that DKC is predominantly related to defective telomere maintenance.

Telomerase is an enzyme that is responsible for maintenance of the balance between telomere shortening and telomere elongation by adding telomeric DNA repeat se-quences to the ends of chromosomes. Most normal somatic cells do not express telomerase and, consequently, telomere length gradually decreases with age in near-ly all human tissues. However, telomerase is active in other proliferative tissues, as in early embryonic and fetal cells, stem cells, germline cells, inflammatory cells, and cells in other periodically or continuously renewing tissues. Moreover, telom-erase is consistently active in the majority of human cancer cells.

Journal of Biomolecular Structure &Dynamics, ISSN 0739-1102Volume 24, Issue Number 4, (2007)©Adenine Press (2007)

Yaroslava G. YinglingBruce A. Shapiro*

Center for Cancer Research Nanobiology ProgramNational Cancer Institute, NCI-Frederick National Institutes of HealthBuilding 469, Room 150Frederick, MD 21702, USA

303

*Phone: 1-301-846-5536Fax: 301-846-5598Email: [email protected]

Open Access ArticleThe authors, the publisher, and the right hold-ers grant the right to use, reproduce, and dis-seminate the work in digital form to all users.

304

Yingling and Shapiro

Human telomerase consists of an RNA (hTR), a telomerase reverse transcrip-tase (hTERT) protein, and a variety of other proteins. Mutations in hTR lead to shorter telomeres, faster aging, predisposition to cancer, and are linked not only to DKC but to other syndromes such as aplastic anaemia, myelodysplastic syn-dromes, paroxysmal nocturnal haemoglobinuria (1, 11-13). Thus, dyskeratosis congenita provides a test not only for the significance of telomerase defects but for the overall importance of telomerase in aging and cancer. Recent reviews have highlighted current achievements in DKC research and its relevance to other forms of bone marrow failure, cancer, and aging (1, 2, 7, 12, 14, 15). Greater understanding of the molecular, structural, and dynamical aspects of DKC should result in improvements in diagnostics and possible therapy for diseases character-ized by impaired telomerase function.

In this study, we concentrate on the effects of mutations in the pseudoknot do-main of hTR (Figure 1a), which are known to reduce telomerase functionality (14). The pseudoknot domain is highly conserved and has been found in all telomerase RNAs (16). It provides the template for telomeric repeat synthesis, enhances re-peat amplification processivity (17), and is required for telomerase activity and stability (18, 19). Studies using in vitro telomerase assays, NMR, and UV absor-bance melting analyses suggest that the pseudoknot domain can exist in two stable conformations (20, 21). The first conformation is a pseudoknot (Figure 1b), and the second conformation is a hairpin pentaloop alone (Figure 1c). It has been pro-posed that there is a dynamic biological switch between these two conformations, which is crucial for efficient telomerase functioning (20, 21). In the wild-type telomerase RNA the pseudoknot dominates by approximately 95% (21). A two-base mutation, GC → AG, at nucleiotides 107-108 in the telomerase RNA pseu-doknot domain (circled bases in Figure 1a) was found in patients with autosomal dominant DKC (6). These specific mutations in the pseudoknot domain, disrupt the base pairing associated with the P3 domain of the pseudoknot and prevent the stable assembly with the catalytic reverse transcriptase component of telomerase (20, 22). Moreover, these DKC mutations significantly shift the equilibrium be-tween the hairpin form and the pseudoknot form to a 50/50 conformation ratio (21) or completely favors the hairpin structure (20). These mutations are found to result in deleterious cellular growth, dramatic reduction of telomerase activity by 100-fold (20) or by 93% (23), or the formation of a weakly active telomerase enzyme defective in telomere elongation (24). Thermal denaturation experiments (21, 23) and free energy calculations (25) demonstrated that the DKC mutations significantly worsen the free energy of the pseudoknot and slightly improve the free energy of the hairpin. NMR (26) and our molecular dynamics (MD) study (27) showed that DKC mutations increase the stability of the hairpin via hydrogen bond interactions between residues in the pentaloop.

Figure 1: Human telomerase RNA pseudoknot domain (a) schematic diagram of pseudoknot domain second-ary structure (16). (b)-(c) Conformational switch be-tween pseudoknot and hairpin forms. Circled residues participate in DKC mutations. Bases are numbered ac-cording to the full structure.

305DKC Mutations and the

Human Telomerase RNA Pseudoknot

In our previous publication, we reported the molecular dynamics study of the RNA hairpin structure in the telomerase pseudoknot domain and the effect of DKC mu-tations on it (27). The atomic coordinates of the hairpin were taken from NMR solution structures (21, 26) and subjected to extensive explicit and implicit water simulations. Our results showed that the wild-type hairpin structure is very flexible and undergoes periodic structural flips represented by the opening of four consecu-tive base pairs. We found that the flips are preceded by the rotation of U105 in the pentaloop. DKC mutations stabilize the hairpin structure and reduce the occurrence of flips by engaging U105 in extra non-Watson-Crick base pairs with loop residues. We suggested that the significance of these flips is in the creation of a nucleation point for pseudoknot formation. In this paper, we continue this line of research by discussing the implications of the DKC mutations on the pseudoknot structure.

Recently, we employed a molecular modeling approach to predict the 3D solution structure of the wild-type telomerase RNA pseudoknot (28). In this structure, the loops and stems of the pseudoknot formed triple helices with the loops positioned either in the minor or major groove of the stems. The pseudoknot junction region showed unique stable triple base interactions. We found that the dynamics of the interactions in the junction region were greatly influenced by the interactions of the U177 bulge and the rotation of residue A174. Overall the whole structure ex-hibited high stability and rigidity over 40 ns of implicit solvent and 6 ns of explicit solvent molecular dynamics simulations.

Here, we report the predicted structure of the DKC-mutated pseudoknot that results from 56 ns of atomistic molecular dynamics simulations in implicit solvent. We use the same methodology as in the case of the wild-type pseudoknot to determine the structure of the DKC-mutated pseudoknot. Due to time-limitations of MD simulations we can not guarantee that the predicted structure is in its global energy minimum. However, we believe that our protocol is useful and promising in pre-dicting the main features of the RNA tertiary structures. Our simulations show that the wild-type pseudoknot structure is notably more stable than the DKC-mutated pseudoknot structure. Moreover, the DKC-mutations prevent the stable formation of the P3 helix of the pseudoknot. Using the MM-GBSA method we verify our predicted structures by calculating the enthalpy of folding for the wild-type and the DKC-mutated hairpins and the wild-type and the DKC-mutated pseudoknot structures. Our enthalpy values are in excellent agreement with experimentally determined enthalpies, which gives us confidence that our predicted structures are within a reasonable approximation to the real structures.

Folding/unfolding events take place on at least micro to millisecond timescales, which prevents us from observing the dynamic switch from the hairpin to the pseu-doknot form or vice versa in our simulations. In order to gain insights into the path-ways between the pseudoknot and hairpin forms of the telomerase RNA we subject the wild-type and the DKC-mutated pseudoknot structures to high temperature mo-lecular dynamics simulations. During these simulations we observe the unfolding of the pseudoknots in three stages and the temporary formation of two UU base pairs. These UU base pairs are found in the hairpin form of the pseudoknot domain in both the wild-type and the DKC-mutated pseudoknots. The formation of these base pairs supports the occurrence of the molecular switch behavior. Overall, the results of our study show the effect of the two base-pair mutation, found in the auto-somal dominant DKC, on the pseudoknot telomerase RNA structure and dynamics.

Methods

The three-dimensional structure of the DKC-mutated pseudoknot is produced by replacing G107 and C108 with A107 and G108 in the starting wild-type pseudoknot structure. The starting coordinates of the wild-type pseudoknot are generated from the secondary structure (Figure 1) using the program RNA2D3D (29). The ini-

306


tial standard base pair interactions in the DKC-mutated pseudoknot are retained according to the secondary structure with minimal tertiary interactions involved. RNA2D3D is software designed by Martinez et al. (29) and explained in more details in Ref. (28). RNA2D3D can generate, view, and compare 3D RNA mol-ecules. In this software the atomic coordinates of a nucleotide are generated from a reference triad of atoms. The stems are created from the reference triad of its 5’ nucleotide using helical coordinates taken from the Biosym® database. The unpaired nucleotides, bulges, hairpin loops, branching loops, and other non-helical motifs are generated by using the coordinates of their reference triad relative to the 5’ neighboring nucleotide. As a result, a first-order approximation of the actual 3D molecule is established. Structure refinement involves molecular modeling or interactive editing. The interactive editing involves a rotation and translation of a nucleotide or a group of nucleotides and is used for the removal of structural clashes, enforcing tertiary interactions, and modification of mutual stacking.

Molecular Dynamics Simulations

All simulations were performed using the ff99 force field for RNA (30) and the molecular dynamics software AMBER 7.0 (31).

Recent reviews discussed the use of MD simulations for investigations of the dy-namical characterization of a wide range of nucleic acid structures (32-37). The length of MD simulations and the size of the investigated molecules are usually hampered by the necessity of including thousands of explicit solvent molecules. The continuum dielectric methods evaluate only the intrasolute electrostatics and consequently reduce the number of interactions with respect to explicit solvent methods (38). These methods have proven to be reliable and able to provide crucial information for various biomolecules (39, 40). The generalized Born (GB) theory is one of the most successful approximations of the Poisson equation for continuum electrostatic solvation energy (41-43). It involves accurate evaluation of Born ra-dii, which characterize the average spherical distances of each atom to the solvent boundary. Consequently, the GB energy expression is very good at reproducing the Poisson energy with much smaller computational cost (44).

Molecular dynamics simulations were performed at 300 K temperature using the GB implicit solvent approach as implemented in the SANDER (Simulated Anneal-ing with NMR-derived Energy) module of AMBER. Each starting structure was subjected to minimization (10,000 steps), followed by slow 20 kcal/mol constrained heating to 300 K over a 200 ps time period, and several consecutive MD equilibra-tions with declining constraints from 2 kcal/mol to 0.1 kcal/mol over a total 500 ps time period. The temperature was maintained at 300 K using a Berendsen thermo-stat (45). The monovalent salt concentration was set to 0.5 mol/L. The production simulations were performed for 56 ns using 1 fs time step.

The simulations were carried out on SGI-Altix and SGI-Origin computers using eight processors. The analysis for all simulations was performed using the Ptraj modules on the production simulations excluding the initial equilibration stage.

Structural Analysis

The program CURVES 5.1 (46) was used to evaluate global axis changes and groove parameters.

Energetic Analysis

The MM-PB(GB)SA module in Amber 7.0 was used to calculate the contributions of gas-phase and solvation free energies for snapshots of the MD trajectories. The energies were calculated for configurations extracted every 100 ps from the stable



36 ns MD trajectory. Total molecular-mechanics energies (Egas), internal energies (Eint, i.e., bonds, angles, and dihedrals), van der Waals (Evdw), electrostatic (Eelec) components were determined. An infinite cutoff for evaluation of all interactions was used. The electrostatic contribution to the solvation free energy (Egb) involves using the GB equation to estimate the electrostatic contributions to the solvation free energy and leads to a 1% error compared to the Poisson-Boltzmann (PB) ap-proach (39). The non-polar contribution to the solvation free energy (Enp) was determined using the Linear Combinations of Pairwise Overlaps (LCPO) method where the hydrophobic contribution to the solvation free energy is determined by the Solvent-Accessible-Surface-Area (SASA) dependent term.

The calculations of the entropic contributions of an atomic-resolution structure in-volve normal mode analysis. Normal mode analysis requires the calculation and diagonalization of a mass-weighted second derivative matrix. These calculations scale quadratically with the size of the structure and are very computationally ex-pensive for our system’s size and therefore were omitted in our results.

Calculations of Enthalpy of Folding

We assume the two state model for calculation of the enthalpy. We evaluate the differences in energy using MM-GBSA module between folded and coiled RNA structures. Four coiled RNA structures with sequences corresponding to the wild-type and the DKC-mutated hairpin and pseudoknot structures were built using Insight II®. The coiled structures were then minimized, heated to 300 K, equili-brated with a slow reduction of constraints, and subjected to 200 ps of molecular dynamics simulations. The MD simulations were stopped before the folding of the RNA coil started. The energies were then calculated using MM-GBSA for configurations extracted every 2 ps from the 200 ps trajectories (Table II). The enthalpies for folding were then estimated as

ΔH = Etot folded – Etot coil

and presented in Table III.

Melting Analysis

Simulations of the melting curves are straightforward if only two-state transitions are assumed, where any single biomolecule is either single stranded or fully folded. To derive a melting/denaturation curve we follow the experimental procedure of heating each structure until they are completely unfolded while measuring the prop-erties that change as a result of unfolding. The final structures of the pseudoknots are subjected to a number of melting molecular dynamics simulations, where each structure is heated from 300 K to 500 K in 5K increments. Each heating increment takes 50 ps and a 0.1 molar salt concentration is used. For statistical purposes each heating simulation is repeated five times using a different random number seed.

The melting temperature of a nucleic acid is defined as the temperature at which half of the structure exists as a duplex and half is single stranded. An accurate measure of the melting temperature can be obtained by conducting a UV melting experiment. The UV absorbance of a single stranded nucleic acid differs from that of a duplex nucleic acid and is higher. For single strands, the absorbance slightly increases with increasing temperature due to decreased base stacking. For double stranded nucleic acids, the absorbance increases abruptly as single strands are pro-duced and involves the combination of both the loss of base pairing and a decrease in stacking. The relative contributions of base pairing and stacking to the melting transition are interrelated and quite complex (47). UV absorption melting curves can be directly compared with calculated curves, dθ/dT, where θ is the calculated fraction of base pairs remaining in the helical stem.

308


In addition, the melting curves for DNA and RNA can be calculated as the temporal average of base-pairing probabilities (48), or based upon an empirically parameter-ized nearest neighbor thermodynamic model (49), or using published extinction coefficients and computed base pair probabilities (50). Alternatively, the melting temperature of a duplex can be estimated from the base pair content of the duplex using a simple empirical formula (51).

A melting curve can be calculated as a first step in structure analysis to find the range of temperatures where interesting events takes place. Moreover, melting profiles can provide information on the relative stability of different motifs and stems. We use high temperature molecular dynamics simulations to estimate the melting profiles. We calculate a folded fraction function with respect to the temperature rather than UV absorbance and use the base pair bonding distance and hydrogen bond occupan-cy between two residues as a primary measure for the loss of helicity. We assume that the two originally bonded strands are melted when they are separated by more than the distance to participate in hydrogen bonding. In our simulations the distance between each bonded base was measured and classified as bonded or non-bonded. We assign three hydrogen bonds for each observed GC base pair and two hydrogen bonds for each observed AU base pair. Other numbers of hydrogen bonds are as-signed to other base pairs according to our observations in Figure 4b. When the distance between bases increase by more than 1 Å the interaction is considered lost and the base pair is “unfolded”. We also conducted visual inspection of all melting simulations trajectories to make sure that our assumptions were correct. Fraction of folded was computed as the average number of hydrogen bonds in the pseudoknot at a given temperature divided by the total maximum number of hydrogen bonds in the pseudoknot. This data was converted into a normalized fraction of folded, θ, versus temperature representation. The normalized fraction of folded is a number between 0 (unfolded) and 1 (folded), where θ = 0.5 at the melting temperature (52).

Because of the relatively long timescale that is required for energy transfer from the fast oscillation of the bond and angle potentials as a result of heating to slow overall structural changes, the melting protocol that was used in this work exaggerated the melting temperatures. For example, we conducted sample simulations at a constant temperature of 360 K and 400 K where the pseudoknot structure was melted after 5 ns and 2 ns, respectively. Therefore, the 50 ps heating time increment that was used in our melting protocol falls short for cooperative unwinding of the strands at each given temperature and results in overestimation of melting temperatures. However, an increase in the time intervals to more than 5 ns would lead to unreasonable computational requirements. Therefore, for comparative purposes with thermal de-naturation experiments the temperatures in Table III and Figure 6 were scaled by 1.37 times. The main goal of the melting simulations is to access the information on the relative stabilities of the stems and tertiary interactions and compare it to the relative stabilities determined by UV absorption experiments.

Results and Discussion

Starting Structure

The 48-nucleotide RNA pseudoknot in this study includes the vitally important and conserved regions of the pseudoknot domain including P3 (Stem 2), J2b/3 (Loop 1), part of P2b (Stem 1), and J2a/3 (Loop 2) domains (Figure 1). More-over, this pseudoknot and its mutated versions have been extensively investigated using NMR spectroscopy, telomerase activity assays, and thermal denaturation experiments (21, 23, 26), which allow us to directly compare our simulated struc-ture with experimental findings.

The starting 3D structure of the DKC-mutated pseudoknot was obtained by a two-base substitution in Stem 2 of the starting wild-type pseudoknot structure. The



method used to build the starting structure of the wild-type pseudoknot was de-scribed in detail in Ref. (28) and in Materials and Methods. The schematic rep-resentation of the DKC-mutated secondary structure and the stereo snapshot of the 3D starting structure are shown in Figure 4a. No tertiary interactions between loops and stems are initially included. Therefore, the initial pseudoknot is in a relatively open state and retains the rest of the initial structure and the original base-pairing which include six base pairs in Stem 1 and seven base pairs in Stem 2. After minimization, heating, and equilibration the starting structure is subjected to 56 ns of unconstrained MD simulations.

DKC-mutated Pseudoknot Structure of Telomerase RNA

The simulations of the DKC-mutated pseudoknot reveal a different dynamical be-havior and a pronounced change in the final structure of the pseudoknot compared to the wild-type pseudoknot structure (28). We find that the stabilization period, which is based on the total energy fluctuation (Figure 2a) is slightly shorter (13 ns) than in the case of the wild-type pseudoknot (16 ns). However, the total energy of the stabilized DKC-mutated pseudoknot is higher than that of the wild-type pseu-doknot by roughly 100 kcal/mol. The total RMSD plot with respect to the average frame exhibits high fluctuations indicating general flexibility and low stability of the DKC-mutated pseudoknot. To highlight the differences in stability between the wild-type and the DKC-mutated pseudoknots the RMSDs relative to the aver-age structure of the stable portion of the trajectory (last 40 ns of simulation time) were calculated and presented in Figure 3. It is clear that the wild-type structure is significantly more stable (Figure 3b) than the DKC-mutated structure (Figure 3a). Examination of the individual contributions to the RMSD from the stems and loops of the DKC-mutated pseudoknot (Figure 2b-e) reveal that Stem 1/Loop 2 form a stable rigid structure; however, Stem 2 and Loop 1 are highly flexible and unstable. The most probable base interactions in the DKC-mutated pseudoknot are presented in Figure 4 and in Table I. The stereo snapshot of the lowest energy DKC-mutated pseudoknot structure is shown in Figure 3b. The bases that were changed due to the DKC mutation, A107 and G108, are colored orange.

Similar to the wild-type pseudoknot (28) and the pseudoknot structure without the U177 bulge (ΔU177) (23), Stem 1 and Loop 2 of the DKC-mutated pseu-doknot form a well-defined and stable triple helix (Figure 2c,d and Figure 4b). All Stem 1 hydrogen bonds are highly stable (Table I). Loop 2 is positioned in the minor groove of Stem 1 and participates in triple base interactions with Stem

Figure 2: Total energy and RMSDs compared with the initial frame of the DKC-mutated pseudoknot struc-ture. The vertical dashed line represents the end of the equilibration period.

310


1. The nature of these triple interactions is slightly different from those found in the predicted wild-type telomerase pseudoknot (Figure 5b). The RMSDs between the average wild-type pseudoknot structure and the average DKC-mutated pseu-doknot structure is equal to 1.0 Å for Stem 1 and is equal to 4.1 Å for Loop 2. We presume that different dynamics and interactions in the Stem2/Loop1 region lead to different Stem 1/ Loop 2 interactions.

The major difference between the wild-type and the DKC-mutated structures, however, comes from the interaction between Loop 1 and Stem 2. There are no original base pairs retained in Stem 2. In fact, Stem 2 failed to form. However, there is a network of intermittent base pairs that form between Stem 2 and Loop 1. The base-pair stackings indicate the formation of a weak flexible helix that is comprised of the base pairs U99:A174, U100:A175, C104:A176, C112:U177, and U105:G178 (Figure 4b). This helix represents the combined interactions between Stem 2 and Loop 1. There are three triple base interactions between the Stem 2 and Loop 1 strands, U99:A174:U102, U113:G178:U105, and C106:G110:A181. The last triple base pair starts a small helix that is comprised of G110:A181, U109:G182, and G108:C183. Overall the Stem 2/Loop 1 part of the DKC-mutated struc-ture undergoes periodic reshuffling and fluctuations.

The Effect of the DKC Mutations on the Telomerase RNA Pseudoknot

Global Positioning: Comparison of the average structures of the DKC-mutated pseudoknot and the wild-type pseudoknot reveals that the global curvature and po-sition of these two structures are distinctly different. The superposition of the DKC-mutated and the wild-type pseudoknot structures using only the heavy atoms of the matching residues resulted in a large RMSD of 7 Å. Due to the dramatic change in the global appearance, the structures could not be well aligned but were superim-posed in Figure 5a for comparative purposes using the Stem 1/Loop 2 domain only. The Stem 1/Loop 2 interactions and positions are similar between the wild-type and the DKC-mutated pseudoknots with an RMSD of 3.7 Å. Consequently, the major disparity comes from the Loop 1/Stem 2 orientation and position. The Loop 1/Stem 2 part of the DKC-mutated pseudoknot is largely unstructured in the junction region and bends to the side. Therefore, the two-base mutations not only completely de-stroy the structure of the P3 helical region, but also change the global positioning of the pseudoknot structure. Experimental observations of, for example, the VS ribozyme show that the global architecture of the folded RNA must be correct for it to function (53). Therefore, the global positioning of the pseudoknot could be crucial for telomerase function and also for hTERT protein binding.

Energy Calculations: We computed the energetic cost of the DKC mutations on the wild-type pseudoknot. A comparison of the average energy components is giv-en in Table II. Even though the internal energies are roughly the same, the main difference comes from the electrostatic energies (Eele) and solvation energies (Ggb). The DKC mutations impair the electrostatic energy but improve the solvation free energy and the van der Waals energy.

Figure 3: RMSDs compared to the average stable structure of (a) the DKC pseudoknot structure and (b) the wild-type pseudoknot structure estimated during last 40 ns.



The relative favorability of the van der Waals (Evdw) energy for the DKC-mutated structure could mean that the structure is more compact. However, it is possible that the more compact structure will not have well-packed interiors as in the case of the DKC-mutated pseudoknot. The small advantage of the DKC-mutated structure in the van der Waals interactions is completely diminished by the large disadvan-tage in the gas phase electrostatic energy (Eele) as compared to the wild-type pseu-doknot. The significant advantage of the wild-type telomerase due to gas-phase

Table I Hydrogen bond occupancy in percentage, average distance while occupied, and average angles while occupied computed over 40 ns for the DKC-mutated pseudoknot structure. The maximum allowable hydrogen bond length is 3.8 Å. Number in parenthesis indicates the standard deviation.

Base-pair Annotation Hydrogen bond Occupancy(%)

Average distance

Average angle

Stem 1/ Loop 2H1(G93)…N3(C121) 99.8 3.1(0.1) 17.8(9.1) H21(G93)…O2(C121) 99.8 3.2(0.1) 18.0(8.1)

G93:C121 GC cis W.-C.

H41(C121)…O6(G93) 91.1 3.2(0.3) 19.6(9.4) H1(G94)…N3(C120) 98.8 3.1(0.1) 20.2(9.1) H21(G94)…O2(C120) 99.8 2.9(0.1) 21.4(10.1)


H41(C120)…O6(G94) 86.3 3.1(0.3) 28.0(13.9) triple G94:A167 GA trans S./S. H22(G94)…N3(A167) 97.6 3.1(0.2) 20.2(10.9)

H1(G95)…N3(C119) 100.0 3.0(0.1) 16.8(8.5) H21(G95)…O2(C119) 100.0 2.1(0.1) 16.4(8.4)


H41(C119)…O6(G95) 98.8 3.1(0.2) 17.9(9.9) H22(G95)…N3(A168) 68.5 3.3(0.3) 24.5(14.0) triple G95:A168 GA trans S./S. H8(A168)…N3(G95) 30.0 3.6(0.1) 48.6(8.2) H1(G118)…N3(C96) 100.0 3.0(0.1) 16.9(8.3) H21(G118)…O2(C96) 100.0 2.9(0.1) 15.7(8.3)

C96:G118 CG cis W.-C.

H41(C96)…O6(G118) 99.9 3.0(0.2) 16.8(9.1) triple G118:A169 GA trans S./S. H22(G118)…N3(A169) 61.0 3.4(0.3) 53.1(5.7)

H3(U97)…N1(A117) 99.7 3.0(0.1) 18.9(10.6) U97:A117 UA cis W.-C. H61(A117)…O4(U97) 99.2 3.1(0.2) 16.9(9.5)

triple U97:A171 UA cis W.-C./H. H62(A171)…O2(U97) 61.5 3.2(0.3) 42.9(11.1) H1(G95)…N3(C119) 99.2 3.1(0.2) 19.5(10.4) H21(G95)…O2(C119) 100.0 3.0(0.1) 16.2(8.2)


H41(C119)…O6(G95) 92.4 3.2(0.3) 18.7(10.3) H22(G98)…N7(A173) 99.9 3.1(0.2) 25.5(9.4) triple G98:A173 GA trans S./H. H62(A173)…N3(G98) 72.9 3.1(0.2) 38.3(12.9)

Stem 2 / Loop 1U99-A174 UA trans S./W.-C. H61(A174)…O2(U99) 29.9 3.1(0.3) 25.1(13.4)

H3(U99)…O4(U114) 13.6 3.1(0.3) 25.9(13.7) U99-U114 UU cis W.-C. H3(U114)…O4(U99) 13.9 3.5(0.3) 45.2(12.5)

U102-A174 UA trans W.-C./H. H8(A174)…O4(U102) 66.4 3.5(0.2) 39.1(12.2) H61(A175)…O2(U100) 69.8 3.0(0.2) 22.9(12.4) U100-A175 UA trans W.-C. H3(U100)…N1(A175) 54.2 3.1(0.2) 20.1(12.5) H3(U179)…O4(U101) 29.5 3.2(0.3) 24.0(11.2) U101-U179 UU cis W.-C. H3(U101)…O4(U179) 24.1 3.1(0.3) 32.3(14.6)

C104-A176 UA trans W.-C./S. H41(U104)…N3(A176) 33.4 3.2(0.2) 41.7(11.9) H41(C112)…O4(U177) 27.8 3.3(0.3) 24.5 (14.2) C112-U177 CU cis W.-C./H. H5(U177)…N3(C112) 22.8 3.7(0.2) 38.9 (10.4) H3(U105)…O6(G178) 29.7 3.4(0.3) 35.8(13.9) U105-G178 UG trans W.-C. H1(G178)…O2(U105) 19.9 3.1(0.3) 36.1(14.8)

U113-G42 UG trans W.-C./H. H8(G178)…O4(U113) 61.9 3.5(0.2) 30.3(14.1) H1(G110)…N3(C106) 47.0 3.2(0.2) 26.9(12.8) H21(G110)…O2(C106) 66.3 3.1(0.3) 24.3(12.1)

C106-G110 CG trans W.-C.

H41(C106)…O6(G110) 52.9 3.0(0.2) 26.0(14.6) H61(A181)…N3(G110) 13.3 3.4(0.3) 45.4(11.9) G110-A181 GA trans S./H. H22(G110)…N7(A181) 2.4 3.2(0.4) 42.3(11.7) H3(U109)…N7(G182) 89.9 3.2(0.3) 25.5(13.6) U109-G182 UG trans W.-C./H. H1(G182)…O2(U109) 95.5 3.0(0.2) 22.9(11.6) H6(C183)…O6(G108) 88.9 3.4(0.2) 49.7(8.2) G108-C183 GC trans H./H. H5(C183)…N7(G108) 20.8 3.6(0.2) 48.6(9.2)

312


electrostatics means that the DKC mutations lead to unfavorable arrangements of the phosphate atoms. As expected the large electrostatic advantage is reduced by the solvation terms, since solvent mainly screens electrostatic interactions in RNA structures (54). Since in the DKC-mutated pseudoknot the bases are splayed out into the solvent and the phosphate backbone is more exposed, there are ad-ditional interactions with the solvent and consequently a more favorable solvation energy term. The non-polar solvation term (Gnp) is solvent-accessible surface area dependent and, therefore, also favors the DKC-mutated pseudoknot. Ultimately, the electrostatic contributions make the wild-type pseudoknot energetically more favorable than the DKC-mutated pseudoknot.

More favorable electrostatic energy leads to the increased rigidity of the wild-type telomerase, which is indicated by the stable hydrogen bonding network and lower RMSD fluctuations. It has been shown that the favorable electrostatic solvation of an unbound RNA dominates and opposes binding to proteins (55). Theoretical studies of DNA-protein complexes have also found that unfavorable electrostatic contributions to binding are due to favorable solvation of the unbound nucleic acids (56). The better electrostatic contribution to solvation free energy of the DKC-mu-tated pseudoknot suggests that it could bind less easily to, for example, hTERT.

Melting Simulations: Melting simulations can provide important information on structural stability and folding pathways, and also can be used to directly compare the results from molecular dynamics simulations to thermal denaturation experi-ments. Moreover, individual contributions to the stability and the unfolding process from parts of the pseudoknot can be extracted. Melting profiles (see Methods) for the wild-type and the DKC-mutated pseudoknots are shown in Figure 6. The plots show a structural transition with a midpoint corresponding to melting temperatures which are extracted from this profile and presented in Table III.

The melting of RNA is expected to be hierarchical, with tertiary structure unfold-ing first, followed by secondary structural elements. The wild-type pseudoknot profiles indeed show that the tertiary interactions are unfolding first, followed by Stem 2, and then Stem 1. These results are consistent with the unfolding pathway for the wild-type pseudoknot determined by analysis of the experimental melting profiles (21, 23). The melting plot of the DKC-mutated pseudoknot shows that

Table IICalculated energy values for the wild-type, DKC-mutated hairpin, pseudoknot, and coil structures. Energies are in kcal/mol. Number in parenthesis indicates the standard deviation.

RNA coil

RNA Wild-type pseudoknot

DKC-mutated pseudoknot

Wild-type hairpin

DKC-mutated hairpin

Wild-type pseudoknot

coil


coil

Wild-type hairpin coil

DKC-mutated hairpin coil

Eele10581.6 (244.6)

11356.1 (329.1)

2367.7 (148.5)

2607.1 (140.1)

4879.0 (292.0)

4734.7 (130.0)

1309.8 (114.6)

1565.6 (215.1)

Evdw-567.3 (15.8)

-600.5 (23.7)

-279.5 (9.6)

-281.7 (9.5)

-327.4 (15.0)

-322.8 (13.5)

-190.1 (11.8)

-206.2 (13.0)

Eint2568.4 (25.9)

2568.2 (25.2)

1601.9 (25.0)

1596.8 (18.3)

2567.1 (23.2)

2566.7 (25.3)

1596.6 (19.3)

1603.6 (20.5)

Egas12582.7 (242.3)

13323.9 (314.9)

3690.0 (140.6)

3922.2 (131.4)

7118.7 (285.9)

6978.6 (128.9)

2716.3 (117.4)

2963.0 (214.0)

Enp48.0 (0.9)

43.8 (1.6)

33.4 (0.4)

33.9 (0.5)

62.4 (0.9)

63.0 (0.7)

38.9 (0.5)

38.4 (0.7)

Egb-22076.7 (240.5)

-22680.6 (311.4)

-9744.8 (136.1)

-9898.7 (131.8)

-16450.2 (284.4)

-16227.5 (126.4)

-8677.7 (112.6)

-8846.2 (209.0)

Esol+gb-22028.7 (241.0)

-22636.8 (312.5)

-9711.4 (135.9)

-9864.8 (131.8)

-16387.7 (285.0)

-16164.5 (126.6)

-8638.7 (112.6)

-8807.9 (209.4)

Eelec_tot -11495.1

(20.7) -11324.5

(28.1) -7377.1 (15.6)

-7291.5 (14.0)

-11571.1 (18.8)

-11492.8 (16.4)

-7367.9 (14.3)

-7280.7 (13.8)

Etot-9446.1 (21.8)

-9312.9 (23.0)

-6021.3 (14.3)

-5942.6 (18.5)

-9269.0 (21.1)

-9185.9 (21.5)

-5922.5 (14.7)

-5844.9 (17.3)



the tertiary interactions and Stem 2 unfold almost at the same time with a small difference in the sharpness of the transition and melting temperatures. Clearly, the DKC mutated pseudoknot does not have a well defined Stem 2 (Figure 4b). Generally, a sharp transition in the melting profile indicates that the affinity con-stant and consequently the Gibbs free energy are strongly temperature dependent, thus exhibiting more stability (52). Stem 2 shows a sharper transition than the tertiary interactions in the DKC-mutated pseudoknot and, therefore, is a little bit more stable than the tertiary interactions, which could be attributed to the exis-tence of the small stem comprised of the base pairs G110:A181, U109:G182, and G108:C183. It is clear from these melting profiles that Stem 2 is responsible for lowering the stability of the DKC-mutated pseudoknot.

The melting profiles of Stem 1 for both structures are comparable in stability. In-terestingly, close to its melting temperature Stem 1 exhibits a temporal increase in its melting profile (around 70 ºC for the wild-type pseudoknot and around 74 ºC for the DKC-mutated pseudoknot). We observe that this increase is attributed to the temporary stabilization of Stem 1 due to the formation of two UU base pairs, U99:U115 and U100:U114. These base pairs are observed in the hairpin form of the pseudoknot. We have conducted several high temperature simulations (360K, 380K, 390 K, 400K) and observed the formation of the same intermediate structure.

Table IIIThermodynamic values for the wild-type and the DKC-mutated hairpin and pseudoknot structures and the coil structures. Energies and enthalpies are in kcal/mol, entropies are in cal/K mol, and temperatures are in ºC. 3º denotes tertiary interactions. Stem 1 are residues 93-121 and Stem 2 are residues 107-184.

RNA Wild-type pseudoknot


Wild-typehairpin

DKC-mutated hairpin

Eele 5702.6 6621.4 1057.9 1041.5 Evdw -239.9 -277.7 -89.4 -75.5 Eint 1.3 1.5 5.3 -6.8 Egas 5464.0 6345.3 973.7 959.2 Enp -14.4 -19.2 -5.5 -4.5 Egb -5626.5 -6453.1 -1067.1 -1052.2

Esol+gb -5641.0 -6472.3 -1072.7 -1056.9 Eelec_tot 76.0 168.3 -9.2 -10.8

Htot Etot -177.7 -127.0 -98.8 -97.7 Tm (Stem 1) 72 76 Tm (Stem 2) 62 54

Tm (3º) 54 52

Experimental Observations (Theimer et al)a,b

Hexp -164.7 -124.0 -92.0 -99.0 Gexp -18.7 (0.2)a, -17.8b -12.1 (0.3)a, -11.2b -9.8 -10.5

Tm (3º) 49.5 35.1 Tm (Stem 1)c 71 72 Tm (Stem 2)c 59 56 Activity, % 100 (6.9) 7 (0.8)

Free Energy Calculations (Dirks et al)d

Gcalc -18.5 -11.3 -11.5 -11.5

Free Energy Calculations (efn server)e

Gcalc(3.2) Stem 1 Stem 2 Stem 1 Stem 2 -11.5 -11.5 Gcalc(2.3) -8.3 -7.9 -8.3 -1.8 -10.1 -10.1 H (2.3) -66.5 -89.9 -66.5 -65.2 -90.8 -90.8 S(2.3) -187.6 -264.3 -187.6 -204.4 -260.1 -260.1

Tm (2.3) 81.2 66.8 81.2 45.8 75.8 75.8 a Theimer, C. A., Blois, C. A., Feigon, J. Mol Cell 17, 671-682 (2005). b Theimer, C. A., Finger, L. D., Trantirek, L., Feigon, J. PNAS 100, 449 (2003). c Estimated from melting profiles published in Theimer, C. A., Finger, L. D., Trantirek, L., Feigon, J. PNAS 100, 449 (2003). c Dirks, R. M., Pierce, N. A. J Comp Chem 25, 1295 (2004). e Zuker, M. efn server at http://www.bioinfo.rpi.edu/applications/mfold/old/rna/energy, 2.3 and 3.2 denotes the energy rules.

314

Yingling and Shapiro(a)

(b)

G CG CG CC GU AG C

U AU AU AC GA UG CU AG GA C A

C AAA

CA

AA

U

UU

UU

UC

UC

G CG CG CC GU AG C

U AU AU AC GA UG CU AG GA C A

C AAA

CA

AA

U

UU

UU

UC

UC

5ʹ166

170

120

95

100

110 180

175

3ʹ

105

Ste

m 1

Lo

op

2

Lo

op

1

Ste

m 2

5ʹ166

170

12095

100

110 180

175

3ʹ

105

GCcisWatson-Crick

AUcisWatson-Crick

transWatson-Crick

transHoogsteen/Hoogsteen

transWatson-Crick/SugarEdge

transSugarEdge/SugarEdge

transWatson-Crick/Hoogsteen

transHoogsteen/SugarEdge

cisWatson-Crick/Hoogsteen

Figure 4: (a) Starting and (b) final structure of the DKC-mutated pseudoknot which resulted from 56 ns of molecular dynamics simulations. Shown are the secondary structures with tertiary interactions and the stereo snapshots of the starting structure and the lowest energy final structure. The coloring of the nucleotides is based on their association with Stem 1 (red), Loop 1 (cyan), Loop 2 (magenta), and Stem 2 (red). The highest occupancy tertiary interactions are marked ac-cording to the proposed geometric nomenclature (63). The bases participating in DKC mutations are circled and highlighted in orange.

(a) (b)

5’

3’

166

120

17095

100 175

177

180110

105

C AAA

CA

AA

AAA

UGUCAGCA

CU

C

U

U

UU

UGUCGGG C

CCGACUUUCAGU

CG

ACCCGACU

GGGCUG

U

CAACA

AA

A

166

170

120

5’

95

(c)

UU

UU

UUCAG

A 175100

110 180

AUGUC

CU

C105

U AGCA

GA

3’

Figure 5: (a) Superposition of the average wild-type (red) and the average DKC-mutated (grey) pseudoknots. The bases participating in the mutations are highlighted in purple. In this figure only Stem 1/Loop 2 regions are superimposed. Secondary structures with observed tertiary interactions of (b) the wild-type pseudoknot and (c) the DKC-mutated pseudoknot.



This intermediate structure may be indicative of the possible transition from the pseudoknot to the hairpin form, thus, supporting the molecular switch.

Our scaled melting temperatures and profiles for both structures are in agree-ment with experimental observations (21) and calculated predictions (Table III). Albeit our theoretical melting calculations are approximate, they appear to give qualitatively the same peaks and shoulders and show similar relative stability in the melting profiles.

The Effects of DKC Mutations on the Hairpin

We have previously showed that the telomerase hairpin undergoes periodic struc-tural flips represented by the opening of base pairs in the helix (27). The DKC mutations in the hairpin, which are located in the pentaloop, reduce the number of structural flips by 80%. The RMSD plot of the DKC-mutated pseudoknot shows one structural fluctuation per 20 ns (Figure 7a) to five structural fluctuations per 20 ns in the wild-type hairpin (Figure 7b). Figure 7c illustrates the change in the hairpin structure at the flip (point A). The structural changes at the flip include the loss of hydrogen bonds for the U100:U114, U101:U113, U102:C112, U103:A111 base pairs, the rotation of A111-U114 outside of the helix, and the shortening of the helical axis curvature by 15% (Figure 7c). To assess the bend during the flip we calculated the global curvature and axis shortening using the program CURVES5.1 (46). The global curvature is the angle between the local helical axes of the second and n-1 base pairs of the helical region. The axis shortening is defined as one minus the ratio between the end-to-end distance of the helix and the axis path length and is presented in percentage. In the average wild-type hairpin structure the global curvature is 23.4 degrees and the axis shortening is 36.3%. In the hairpin structure at the flip the global curvature is 66.1 degrees and the axis shortening is 51.7%. The numbers represent a significant overall helical bend during the flip.

We discovered earlier that the flips in the wild-type hairpin structure are directly related to the rotation of residue U105 (27). We also noticed that in the DKC-mu-tated hairpin U105 is hydrogen bonded with U109 or G108, consequently reducing the conformational freedom of U105 and reducing the number of structural flips. The U105:U109 base pair can adopt two alternative conformations and sporadi-

Figure 7: RMSDs relative to the average structure of (a) the DKC-mutated hairpin structure and (b) the wild-type hairpin structure. Letters A, B, C, D, E, and F denote the approximate locations of the flips in the trajectory. (c) Superposition between the backbone atoms of residues 93-98 and 116-121 of the average wild-type hairpin structure (blue) and the structure at the 3.9 ns flip (red). Residues 111-114 are represented as sticks with U113-114 in yellow, C112 in magenta, and A111 in green. This represents the helical twist, opening of the base pairs, and the rotation of 112-114 onto the side of the structure.

Figure 6: Melting diagrams for (a) the wild-type pseu-doknot and (b) the DKC-mutated pseudoknot structures. Stem 1 profile is a solid line, Stem 2 profile is a dashed line, and a dotted line depicts the tertiary interactions. The horizontal dashed line represents the halfpoint cor-responding to the melting temperature evaluation. The temperature on x-axis was scaled by 1.37 times.

(a)

(b)

(c)

A B C D E

F

RM

SD, Å

Time, ns

54321543210 5 10 15 20

(a) (b)

(c)

C G CU UCU

GA

UUUU U

UUC

GUCG C

GAC

GG

CCC

121

5’3’

98

112

105

CU105

98

112

CU

UUUUGUCGG C

CGACUUUC

AGU

GA

C121

3’5’G

C

Figure 8: Superposition of the average wild-type (red) and the average DKC-mutated (grey) hairpins. Secondary structures with predicted tertiary inter-actions of (b) the wild-type pseudoknot and (c) the DKC-mutated hairpin.

316


cally switch from one to another (27). The hydrogen bond occupancies in the first conformation are 48.5% for O4(U109)…H3(U105) and 43.8% for O2(U105)…H3(U109). The hydrogen bond occupancies for the second conformation are 11.4% for O4(U105)…H3(U109) and 10.8% for O2(U109)…H3(U105). The hydrogen bond occupancy for U105:G108 is about 10.7%. Therefore, the rotation of U105 that leads to a structural flip is significantly reduced by the DKC mutations, but is still possible during the switch from one base pair conformation to another.

The average structures of the wild-type and DKC-mutated hairpins are very simi-lar, with an RMSD of 0.5 Å in the helix. The only visible change is in the pen-taloop (Figure 8). This indicates that the only structural change that the DKC mutations induce on the hairpin structure is the change of the pentaloop structure and not the helix itself. Yet dynamically the DKC mutation brings more stability to the hairpin structure.

Enthalpy of Folding

Free energy has two components, enthalpy and entropy. Entropy is a measure of disorder and enthalpy (ΔH) is the measure of the internal energy (ΔE) of a biological system. All biological reactions take place at constant pressure (P) and temperature. Therefore, ΔH = ΔE + PΔV. Since biological reactions occur in a large excess of liquid, volume changes (ΔV) are extremely small, hence, PΔV is very small as well. Therefore, enthalpy and internal energy values of biological reactions are approxi-mately the same and are referred to as the energy change of a reaction. Since the energy change in a chemical reaction comes from making and breaking of bonds, the value of ΔH can be calculated from the energy of the bonds. Since in our case no breaking or making of covalent bonds occur, the enthalpy change depends heavily on noncovalent bond energies, such as hydrogen bonds and van der Waals contacts.

Free energy and enthalpy calculations by molecular dynamics simulations can pro-vide direct feedback between our findings on change in structural interactions and macroscopic thermodynamics. Moreover, free energy and enthalpy can also be measured experimentally. Furthermore, MM-PB(GB)SA calculations of binding free energy, absolute free energy, entropy, and enthalpy have been shown to cor-relate well with experimental observation for a number of molecular complexes (57-61). The entropy calculations are very computationally expensive for our sys-tem’s size and, therefore, were omitted in our results. We computed the enthalpy of folding (ΔH) for all four molecules, the wild-type hairpin and pseudoknot, and the DKC-mutated hairpin and pseudoknot. The folding process generally involves going from an RNA coil conformation to a folded structure. We have thus calcu-lated the energetic differences between the folded structure and coiled sequence as described in the Methods section and compared it to the enthalpy obtained from the optical melting data analysis (21) and the enthalpy obtained from the efn server (62) (Table III). Since the efn server can not calculate the energy of the pseudoknot, pseudoknot structures were separated into Stem 1 (residues 93-121) and Stem 2 (residues 107-184). The Efn server 2.3 energy rules provide the free energy, en-tropy, enthalpy, and melting temperature values. First, we will discuss the hairpin structures followed by the pseudoknot structures.

For the hairpin structures, the secondary structure free energy calculations (62), which are based on the stability of the helix and loop size (independent of loop con-text), obtained identical thermodynamic parameters as expected. The experimen-tally observed enthalpy favors the DKC-mutated hairpin structure by 7 kcal/mol (21). Our MM-GBSA calculations show a similar enthalpy value for the hairpin with or without the DKC mutations.

For the pseudoknot structures, secondary structure free energy calculations, ex-perimental results, and our simulation analysis agree that the DKC mutation sig-



nificantly destabilizes the pseudoknot structure. Secondary structure free energy calculations (efn server) show an increase in the enthalpy due to the DKC mutations in the pseudoknot by 21 kcal/mol, which is attributed to the loss of two base pairs in Stem 2. Thermal denaturation experiments and our simulations show enthalpy gains of 41 to 51 kcal/mol, which indicate that there is a significant disruption of the bonding in the P3 helix involving more than two base pairs. Moreover, the differ-ence between our simulated enthalpy and the experimentally determined enthalpy is reasonably small, with our values being slightly lower than the experimentally determined ones. The enthalpy agreement indicates that the noncovalent bonds, in-cluding H-bonds and van der Waals contacts in our predicted pseudoknot structures are within a reasonable proximity of the experimental structures.

Experimental Observations

The pseudoknot structure of the telomerase RNA is important for the catalytic activ-ity of telomerase and TERT binding. Experimental studies of a two base mutation in DKC (GC107/108 → AG107/108) show significant reduction of telomerase activity and possible structural changes. Comolli et al. (20) showed that the DKC-mutations abrogated in vitro telomerase activity and hyperstabilized the hairpin conformation, blocking pseudoknot formation. Moreover, nondenaturating gel electrophoresis confirmed by NMR indicates that the DKC-mutations prevent formation of the P3 helix (20), in agreement with our results. Theimer et al. (21, 23, 26) reached similar conclusions about the effect of the DKC-mutation on telomerase and indicated that the DKC mutations significantly destabilized the pseudoknot conformation, result-ing in the favorability of the hairpin structure and a reduction of activity by 93%. DKC mutations destabilize the pseudoknot by 6.6 kcal/mol and lower the tertiary melting temperature by 12-14 ºC. Fu and Collins (22) found that the DKC-mutation and mutations in the P3 helix of the pseudoknot domain led to a strong decrease in telomerase activity in vivo and in vitro. Cerone et al. (24) showed that the DKC-mu-tated telomerase produced a weakly active telomerase enzyme defective in telomere elongation. Our results agree with the experiments that the DKC-mutations over-stabilize the hairpin structure and disfavor pseudoknot formation. Moreover, we have suggested potential structural implications in the formation of the pseudoknot and its binding to hTERT, which can explain the reason for abrogated functionality.

Conclusions

In this paper, detailed analyses of the effects of the DKC mutations on the wild-type pseudoknot structure and dynamics were carried out. The DKC-mutated pseudoknot structure was predicted using molecular modeling. We show that DKC-mutations abolish the formation of the P3 helix, change the global orientation and appearance of the pseudoknot, and overall destabilize the structure. Energetic analysis reveals that the lower free energy and the higher stability of the wild-type telomerase pseu-doknot are associated with a favorable electrostatic energy. The 3D prediction of the wild-type and the DKC-mutated pseudoknot structures is reinforced by com-parison of the enthalpy of folding calculations.

In the hairpin form, the DKC mutations significantly improve stability and reduce the number of structural flips by roughly 80%. These flips were previously sug-gested to represent a nucleation point for pseudoknot formation or initiation of a molecular switch. Thus, we presume that the DKC mutations reduce the probability of a molecular switch by at least 80%.

We also conducted melting experiments that exhibit good agreement with UV denaturation experiments. Melting profiles show a temporary stabilization of Stem 1 just below the melting temperature due to the formation of U99:U115 and U100:U114 base pairs, as is found in the hairpin form of the pseudoknot telomerase RNA domain.

318


Overall, our study shows that the DKC mutations stabilize the hairpin form by ex-tra non-Watson-Crick interactions in the pentaloop and destabilize the pseudoknot form by preventing the stable formation of the P3 helix. Our results are consistent with experimental observations and support published biochemical data. A poten-tial drug can be designed to suppress the wild-type telomerase activity in cancerous cells by attacking the same bases that participate in the DKC mutation.

The predicted pseudoknot structure in this study was determined purely by molecu-lar modeling. The accuracy of this structure is heavily dependent on accuracy of the force field, efficient conformational sampling, and the global energy minimum. Since the wild-type and the DKC-mutated pseudoknot structures have not been experimentally determined, presumably due to their complexity and lack of stabil-ity, our method shows a possible way to predict the main features of tertiary struc-tures from a known secondary structure via molecular modeling and to examine the structural effects of mutations. The predicted structures can then be fitted into NMR NOESY data to determine their compatibility.

Acknowledgments

This research was supported by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research. The computational sup-port was partly provided by the National Cancer Institute’s Advanced Biomedical Computing Center.

References and Footnotes1.2.3.

4.

5.6.

7.8.

9.10.11.

12.13.14.15.16.17.

18.19.20.

21.22.23.24.25.26.27.28.29.30.31.

A. Marrone, A. Walne, I. Dokal. Curr Opin Genet Dev 15, 249-257 (2005).M. Bessler, D. B. Wilson, P. J. Mason. Curr Opin Pediatr 16, 23-28 (2004).S. Knight, T. Vulliamy, A. Copplestone, E. Gluckman, P. Mason, I. Dokal. Br J Haematol 103, 990-996 (1998).N. S. Heiss, S. W. Knight, T. J. Vulliamy, S. M. Klauck, S. Wiemann, P. J. Mason, A. Poustka, I. Dokal. Nat Genet 19, 32-38 (1998).S. Hassock, D. Vetrie, F. Giannelli. Genomics 55, 21-27 (1999).T. Vulliamy, A. Marrone, F. Goldman, A. Dearlove, M. Bessler, P. J. Mason, I. Dokal. Nature 413, 432-435 (2001).T. P. Handley, J. A. Mccaul, G. R. Ogden. Oral Oncol (2005).F. Goldman, R. Bouarich, S. Kulkarni, S. Freeman, H. Y. Du, L. Harrington, P. J. Mason, A. Londono-Vallejo, M. Bessler. Proc Natl Acad Sci USA 102, 17119-17124 (2005).T. J. Vulliamy, S. W. Knight, P. J. Mason, I. Dokal. Blood Cells Mol Dis 27, 353-357 (2001).J. W. Shay, W. E. Wright. Nat Genet 36, 437-438 (2004).H. Yamaguchi, R. T. Calado, H. Ly, S. Kajigaya, G. M. Baerlocher, S. J. Chanock, P. M. Lansdorp, N. S. Young. N Engl J Med 352, 1413-1424 (2005).P. J. Mason, D. B. Wilson, M. Bessler. Curr Mol Med 5, 159-170 (2005).T. Vulliamy, A. Marrone, I. Dokal, P. J. Mason. Lancet 359, 2168-2170 (2002).R. Marciniak, L. Guarente. Nature 413, 370-371, 373 (2001).J. L. Chen, C. W. Greider. Trends Biochem Sci 29, 183-192 (2004).J. L. Chen, M. A. Blasco, C. W. Greider. Cell 100, 503-514 (2000).J. Feng, W. D. Funk, S. S. Wang, S. L. Weinrich, A. A. Avilion, C. P. Chiu, R. R. Adams, E. Chang, R. C. Allsopp, J. Yu, et al. Science 269, 1236-1241 (1995).L. Martin-Rivera, M. A. Blasco. J Biol Chem 276, 5856-5865 (2001).F. Bachand, C. Autexier. Mol Cell Biol 21, 1888-1897 (2001).L. R. Comolli, I. Smirnov, L. Xu, E. H. Blackburn, T. L. James. Proc Natl Acad Sci USA 99, 16998-17003 (2002).C. A. Theimer, L. D. Finger, L. Trantirek, J. Feigon. PNAS 100, 449-454 (2003).D. Fu, K. Collins. In Mol Cell, pp 1361-1372 (2003).C. A. Theimer, C. A. Blois, J. Feigon. Molecular Cell 17, 671-682 (2005).M. A. Cerone, R. J. Ward, J. A. Londono-Vallejo, C. Autexier. Cell Cycle 4, 585-589 (2005).R. M. Dirks, N. A. Pierce. J Comput Chem 25, 1295-1304 (2004).C. A. Theimer, L. D. Finger, J. Feigon. RNA 9, 1446-1455 (2003).Y. G. Yingling, B. A. Shapiro. Journal of Molecular Biology 348, 27-42 (2005).Y. G. Yingling, B. A. Shapiro. J Mol Graph Model 25, 261-274 (2006).H. M. Matinez, J. Maizel, Jr., B. A. Shapiro. (in preparation).J. M. Wang, P. Cieplak, P. A. Kollman. J Comp Chem 21, 1049-1074 (2000).D. A. Case, D. A. Pearlman, J. W. Caldwell, T. E. Cheatham, III, J. Wang, W. S. Ross, C. L. Simmerling, T. A. Darden, K. M. Merz, R. V. Stanton, A. L. Cheng, J. J. Vincent, M. Crowley, V. Tsui, H. Gohlke, R. J. Radmer, Y. Duan, J. Pitera, I. Massova, G. L. Seibel, U. C. Singh, P. K. Weiner, P. A. Kollman. In San Francisco: University of California (2002).



32.33.34.35.36.37.

38.39.

40.41.42.43.44.45.

46.47.

48.49.

50.51.

52.53.54.55.56.

57.58.59.60.61.62.63.

M. Zacharias. Curr Opin Struct Biol 10, 311-317 (2000).E. Giudice, R. Lavery. Acc Chem Res 35, 350-357 (2002).P. Auffinger, E. Westhof. Curr Opin Struct Biol 8, 227-236 (1998).T. E. Cheatham, 3rd, P. A. Kollman. Annu Rev Phys Chem 51, 435-471 (2000).T. E. I. Cheatham, Young, M. A. Biopolymers 56, 232-256 (2001).P. Auffinger, A. C. Vaiana. Molecular Dynamics of RNA Systems. Weinheim: Wiley-VCH Verlag GmbH & Co. KGaA (2005).M. S. Lee, F. R. Salsbury, Jr., C. L. Brooks, III. J Chem Phys 116, 10606-10614 (2002).M. Feig, A. Onufriev, M. S. Lee, W. Im, D. A. Case, C. L. Brooks, 3rd. J Comput Chem 25, 265-284 (2004).L. Y. Zhang, E. Gallicchio, R. A. Friesner, R. M. Levy. J Comp Chem 22, 591-607 (2001).D. Bashford, D. A. Case. Annu Rev Phys Chem 51, 129-152 (2000).V. Tsui, D. A. Case. Biopolymers 56, 275-291 (2000).J. Weiser, P. S. Shenkin, W. C. Still. J Comp Chem 20, 217-230 (1999).M. Scarsi, J. Apostolakis, A. Caflisch. J Phys Chem B 102, 3637-3641 (1998).H. J. C. Berendsen, J. P. M. Postma, W. F. Van Gunsteren, A. Dinola, J. R. J. Haak. Chem Phys 81, 3684-3690 (1984).R. Lavery, H. Sklenar. J Biomol Struct Dyn 6, 63-91 (1988).V. A. Bloomfield, D. M. Crothers, I. Tinoco, Jr. Nucleic Acids. Sausalito, CA, USA: Univer-sity Science Books (2000)E. Tostesen, G. I. Jerstad, E. Hovig. Nucleic Acids Res 33, W573-576 (2005).R. D. Blake, J. W. Bizzaro, J. D. Blake, G. R. Day, S. G. Delcourt, J. Knowles, K. A. Marx, J. Santalucia, Jr. Bioinformatics 15, 370-375 (1999).N. R. Markham, M. Zuker. Nucleic Acids Res 33, W577-581 (2005).R. Deaton, M. Garzon, R. C. Murphy, J. A. Rose, D. R. Franceschetti, S. E. Stevens, Jr. Physical Review Letters 80, 417-420 (1998).J. L. Mergny, L. Lacroix. Oligonucleotides 13, 515-537 (2003).D. A. Lafontaine, D. G. Norman, D. M. Lilley. Embo J 20, 1415-1424 (2001).J. Srinivasan, J. Miller, P. A. Kollman, D. A. Case. J Biomol Struct Dyn 16, 671-682 (1998).C. M. Reyes, P. A. Kollman. J Mol Biol 297, 1145-1158 (2000).B. Jayaram, K. Mcconnell, S. B. Dixit, A. Das, D. L. Beveridge. J Comput Chem 23, 1-14 (2002).H. Gohlke, C. Kiel, D. A. Case. J Mol Biol 330, 891-913 (2003).I. Bea, M. G. Gotsev, P. M. Ivanov, C. Jaime, P. A. Kollman. J Org Chem 71, 2056-2063 (2006).B. Kuhn, P. A. Kollman, M. Stahl. J Comput Chem 25, 1865-1872 (2004).H. Zhong, H. A. Carlson. Proteins 58, 222-234 (2005).S. Yan, M. Wu, D. J. Patel, N. E. Geacintov, S. Broyde. Biophys J 84, 2137-2148 (2003).M. Zuker. In Rensselaer Polytechnic Institute (1996-2005).N. B. Leontis, E. Westhof. RNA 7, 499-512 (2001).

Date Received: September 1, 2006

Communicated by the Editor Ramaswamy H. Sarma

The Impact of Dyskeratosis Congenita Mutations on the Structure and Dynamics of the Human Telomerase RNA Pseudoknot Domain

Documents