Article Information

• PDF (260 kb)
• Full Text with Large Figures
• Supporting Material
• Cited by in Scopus (1)

PubMed

• Articles by Philip Shemella
• Articles by Saroj K. Nayak

Related Articles

• Structure and dynamics of hydronium in the ion channel grami...

  Structure and dynamics of hydronium in the ion channel gramicidin A Biophysical Journal, Volume 70, Issue 5, 1 May 1996, Pages 2043-2051 D.E. Sagnella and G.A. Voth Abstract The effects of the hydronium ion, H(3)0+, on the structure of the ion channel gramicidin A and the hydrogen-bonded network of waters within the channel were studied to help elucidate a possible mechanism for proton transport through the channel. Several classical molecular dynamics studies were carried out with the hydronium in either the center of a gramicidin monomer or in the dimer junction. Structural reorganization of the channel backbone was observed for different hydronium positions, which were most apparent when the hydronium was within the monomer. In both cases the average O-O distance between the hydronium ion and its nearest neighbor water molecule was found to be approximately 2.55 A, indicating a rather strong hydrogen bond. Importantly, a subsequent break in the hydrogen-bonded network between the nearest neighbor and the next-nearest neighbor(approximately 2.7 -3.0 A) was repeatedly observed. Moreover, the carbonyl groups of gramicidin A were found to interact with the charge on the hydronium ion, helping in its stabilization. These facts may have significant implications for the proton hopping mechanism. The presence of the hydronium ion in the channel also inhibits to some degree the reorientational motions of the channel water molecules. Abstract | PDF (1029 kb)

• Identification of a virus intein and a possible variation in...

  Identification of a virus intein and a possible variation in the protein-splicing reaction Current Biology, Volume 8, Issue 18, 10 September 1998, Pages R634-R638 Shmuel Pietrokovski Full Text | PDF (263 kb)

• A natural example of protein trans-splicing

  A natural example of protein trans-splicing Trends in Biochemical Sciences, Volume 24, Issue 6, 1 June 1999, Pages 209-211 Francine B. Perler Full Text | PDF (206 kb)

• …more

Copyright © 2007 The Biophysical Society. All rights reserved.
Biophysical Journal, Volume 92, Issue 3, 847-853, 1 February 2007

doi:10.1529/biophysj.106.092049

Biophysical Theory and Modeling

Mechanism for Intein C-Terminal Cleavage: A Proposal from Quantum Mechanical Calculations

Philip Shemella^*, ,  , Brian Pereira^†, Yiming Zhang^*, Patrick Van Roey^‡, Georges Belfort^†, Shekhar Garde^† and Saroj K. Nayak^*

^* Department of Physics, Applied Physics and Astronomy, Rensselaer Polytechnic Institute, Troy, New York 12180
^† Howard P. Isermann Department of Chemical and Biological Engineering, and Center for Biotechnology and Interdisciplinary Studies, Rensselaer Polytechnic Institute, Troy, New York 12180
^‡ Wadsworth Center, New York State Department of Health, Albany, New York 12201

Address reprint requests to Philip Shemella.

Found in more than 250 proteins, inteins are autocatalytic protein elements that undergo N- and C-terminal cleavage reactions as well as a splicing reaction where the cleaved N- and C-exteins are ligated (Fig. 1) 1,2. These reactions are largely dependent on the presence of conserved amino acids mostly located near the N- and C-termini 3,4,5. The mechanism for intein splicing (or cleavage) is proposed based on mutational analyses and structural information obtained from x-ray crystallography 6,7,8,9,10. Mutational analyses further show that certain highly conserved residues are critical (i.e., their mutation leads to a complete loss of activity), whereas others can accommodate alternative amino acids that change the efficiency or rate of reaction without a complete loss of activity 11,12,13,14,15.

Display large version of this figure

Figure 1

Intein reactions—splicing and cleavage (after C1A mutation). C, A, H, N, and N* represent cysteine, alanine, histidine, asparagine, and succinimide, respectively.

The cysteine (or serine) at the N-terminus of the intein, when mutated to alanine, eliminates N-terminal cleavage activity, and therefore also inhibits splicing (Fig. 1, right). With this mutation, C-terminal cleavage is isolated, and the peptide bond between the C-terminal of the intein (asparagine) and the N-terminal of the C-extein is cleaved.

Despite significant insights into the intein splicing and cleavage mechanisms obtained from mutational analyses, the details of the component reactions remain to be worked out. It is not even clear what step initiates the splicing or cleavage reactions. Specifically, the origins and movement of protons 16 and the roles of vicinal amino acids or water molecules as proton donors or acceptors during the cleavage reaction are not fully known or understood as illustrated below.

In the experimental study by Wood et al. on the Mycobacterium tuberculosis RecA cleavage mutant mini-intein (ΔI-CM), a decrease in solution pH from 7.5 to 6.0 was found to lead to a significant increase in the rate of C-terminal cleavage 17,18. The higher C-terminal cleavage activity at lower pH for this intein as well as for the Ssp DnaB 19 intein are, however, inconsistent with the currently available details of the mechanism proposed for intein cleavage. For example, Ding et al. have suggested that for the reaction of the Ssp DnaB mini-intein, the F-block histidine acts as a base, deprotonating the nitrogen of the C-terminal asparagine side chain via a vicinal water molecule 20. Experiments on short peptides in solution also show an increased tendency of asparagine to be ionized and to cyclize (leading to succinimide formation) over the pH range of 7.4–13.8 21,22,23. For a histidine to act as a base and accept protons, one of its two imidazole side-chain nitrogens must be deprotonated on average. At lower pH, especially below the pK_a, histidine is more likely to be present in the doubly protonated state, diminishing its ability to accept protons. Thus, the mechanistic details underlying the increased C-terminal cleavage activity observed in experiments of Wood et al. 17,18 are expected to be different from those proposed by Ding et al. 20.

Based on a combination of semiempirical and first principles computational analyses, here we propose what we believe are new details of the mechanism of asparagine cyclization catalyzed by the cleavage mutant of the M. tuberculosis RecA mini-intein that accounts for an increase in activity of C-terminal cleavage at low pH. These details involve protonation of the nitrogen of the scissile peptide bond by a vicinal hydronium ion (). This leads to stretching of that peptide bond due to the loss of π-bond resonance and consequent increase in carbon electrophilicity. Asparagine cyclization and subsequent succinimide formation occur 24, resulting in peptide bond cleavage. Our results are consistent with the experimental observations of Wood et al. that indicate a simple proton-catalyzed reaction 18.

Below we present the details of computational methodology that utilizes available experimental crystal structure information, experimental kinetic data, and mutational analyses, to provide a proposal for mechanistic details of the C-terminal reaction of the cleavage mutant intein. Our proposal highlights the need to account for the local structural information as well as solution conditions (e.g., pH) in formulating details of enzymatic catalysis.

Given current computational resources, detailed quantum mechanical calculations are limited to small systems 25,26,27,28,29. In the present context, where the mechanism of enzymatic catalysis is of interest, one may choose a large number of atoms (50–100) and perform only a few calculations, or choose a smaller but relevant part of the system, and explore a variety of possible reaction pathways. Inclusion of a larger number of atoms not only adds to the computational expense, but could allow conformational rearrangements that are inconsistent with protein structural context being studied. As a compromise, and based on available crystal structure knowledge, our system includes 25 atoms—the C-terminal asparagine side chain, the backbone atoms of the penultimate histidine and of the cysteine of the C-extein (the dangling bonds are passivated with hydrogen atoms), and one explicit water molecule (see Fig. 2). Additional calculations (not shown) with larger systems containing 45 atoms for limited cases, and a full protein quantum mechanics/molecular mechanics (QM/MM) treatment indicated energetic results similar to the ones presented here with a 25-atom system.

Display large version of this figure

Figure 2

Computational system containing 25 atoms includes C-terminal asparagine of the intein, backbone atoms of the adjacent amino acids, and a protonated water molecule (panel A). Panel B identifies x, y, and z coordinates used in the exploration of reaction pathways.

Locating the transition state for biomolecular reactions is a complex problem that requires a search in a multi-dimensional local conformation space using methods such as transition path sampling 30. The computational expense of quantum mechanical calculations, however, prohibits such investigations. Instead, we heuristically choose the constraint minimization procedure based on a two-dimensional space defined by the distances x and y indicated in Fig. 2 (see below and figure captions for details) to locate the transition state region. The precise energy barrier is found by constraining only the y coordinate and relaxing all other internal coordinates.

Semiempirical PM3 31 calculations were performed using Gaussian code 32 to obtain a two-dimensional potential energy surface in the x-y plane. Geometry optimizations with PM3 are computationally efficient and sufficiently accurate 33. Deficiencies of PM3 with respect to chemical structures, such as the “flattening” of small and medium-sized rings 34, nitrogen atoms with a lone pair having pyramidal geometries 35, and inaccurate hydrogen bonding distances are well known. Nevertheless, PM3 is useful for efficiently scanning many geometries and locating general locations of transition states, as is done here.

The PM3 calculations described above were followed by detailed quantum mechanical calculations along selected values of relevant reaction coordinates. Specifically, we employed first principles density functional theory based on the B3LYP gradient corrected functional 36. We used 6-31G(d,p) basis sets to calculate energies of optimized geometries and compared them with values obtained using larger basis sets (6–311++g(2d,p)) to test for convergence 37. MP2 calculations were conducted to test the accuracy of the B3LYP method for this system, and the energy barrier calculations were consistent. In calculations with a hydronium ion, the three O-H bond distances were constrained to 0.98Å to avoid spontaneous proton donation observed otherwise.

The polarizable continuum model was used to simulate solvent effects in the detailed calculations 38. Geometry optimizations performed in implicit solvent were compared with similar calculations in the gas phase. The numerical integral equation formalism was used because it allows defining interlocking atomic spheres to represent the extent of the system in solution (a single sphere was put on the hydrogen atoms belonging to the asparagine side-chain nitrogen and to hydrogen atoms of the water molecule) 39. Nondimensional dielectric constants are defined by ɛ_r≡ɛ_s/ɛ_o, where ɛ_o is the vacuum permittivity and ɛ_s is the static dielectric constant for the dielectric. For the gas phase, ɛ_r=1. For water, ɛ_r=78.39.

Thermal and entropic contributions calculated with a harmonic approximation for the optimized geometries at the B3LYP/6-31G(d,p) level were combined with the electronic energy to obtain free-energy profiles in the gas phase and in the implicit solvent. Zero point energies were found to differ by between 0.04 and 1.33kcal/mol, which are within the expected error for the calculation. The approximate entropic components of the free energy include contributions from translational, electronic, rotational, and vibrational degrees of freedom 40,41 and were obtained from frequency calculations at room temperature. Thermal corrections do not include imaginary frequencies of vibrational modes for transition states.

Fig. 2 shows the system considered in our calculations and includes the C-terminal asparagine residue of the intein, and the backbone atoms of the adjacent amino acids, including the scissile peptide bond. In addition, we include a water molecule in the form of a hydronium ion to mimic a low pH environment. Coordinate x in Figure 2B represents the distance between the oxygen atom of the water molecule and the hydrogen atom of the asparagine side chain, referred to as the asparagine ionization distance. Coordinate y, the asparagine cyclization distance, is the separation between the asparagine side-chain nitrogen and the peptide carbonyl carbon. Coordinate z represents the length of the scissile peptide bond.

As will be apparent in the discussion below, protonation of the backbone nitrogen of the scissile peptide bond is a necessary first step in the reaction at low pH. To this end, Fig. 3 considers three scenarios corresponding to i), a neutral peptide case in which neither of the atoms are protonated, ii), O-protonation of the carbonyl oxygen, and iii), the N-protonation of the peptide nitrogen 42. Specifically, we study consequences of these three scenarios on the system energy and the equilibrium length of the scissile peptide bond.

Display large version of this figure

Figure 3

Schematic of the three scenarios considered for asparagine cyclization: i), normal peptide, ii), O-protonation, and iii), N-protonation. Asparagine cyclization energies (panel A) and peptide bond stretching (panel B) versus the asparagine cyclization distance, y, for the neutral peptide bond system (□), for the system where the carbonyl oxygen is protonated (▵), and where the peptide nitrogen atom is protonated (■). Gas phase structures are optimized at the B3LYP/6-31G(d,p) level.

In a neutral peptide (scenario i above), the relaxed peptide bond length is 1.35Å. As asparagine cyclization proceeds (i.e., as y is reduced), the system energy increases significantly (Figure 3A). Correspondingly, there is only a slight increase in the peptide bond length, indicating that it remains essentially intact. When the carbonyl oxygen atom is protonated (scenario ii), the relaxed peptide bond length in fact decreases to 1.32Å, as expected from the increased π-conjugation (or the double bond character of the bond) between C and N. Asparagine cyclization energy (Figure 3A) in this case is lower than that for the neutral peptide case; however, the peptide bond is significantly more stable and remains essentially intact as asparagine cyclization proceeds. In contrast, when the peptide nitrogen atom is protonated (scenario iii), the relaxed peptide bond length increases to 1.51Å, indicating the weakening of that bond. As asparagine cyclization proceeds, that distance increases further and leads to breaking of that bond, resulting from the fact that a doubly protonated nitrogen makes a good leaving group (Figure 3B). The cyclization energy (Figure 3A) in this case is lower than that for the neutral peptide case and similar to that for oxygen protonation, which does not lead to peptide bond cleavage (see above). Collectively, these preliminary calculations indicate that protonation of the amide nitrogen is an important first step for C-terminal cleavage in low pH environments.

The protonation of the nitrogen of the scissile peptide bond proposed above makes that nitrogen atom transiently doubly protonated. In the broader context of enzymatic catalysis, this proposal is not new. Indeed, in the hydrolysis of a peptide bond by serine proteases, the nitrogen of the scissile peptide bond accepts a proton from the histidine of the catalytic triad 43. In their study of enzymatic reaction catalyzed by the HIV-1 protease, Trylska et al. found that protonation of the amide nitrogen was essential for peptide bond cleavage 44. Similarly, the protonation of the amide nitrogen was found to be the essential step in the hydrolysis of a formamide molecule, which was used as a computational model for peptide bond hydrolysis 45.

Given that the doubly protonated state of the amide nitrogen is a likely starting point of the C-terminal cleavage reaction at low pH, we explored further reaction pathways using semiempirical PM3 calculations, which guided additional high-level quantum calculations. We performed geometry optimizations using the PM3 method for 428 independent points in the two-dimensional space (shown in Fig. 4) based on reaction coordinates x and y (see Figure 2B).

Display large version of this figure

Figure 4

Potential energy surface obtained from semiempirical PM3 calculations for the N-protonated system. The initial state is located at the top right and the final one at the bottom left. The larger region on the right is relatively featureless and can be sampled by multiple possible paths to the barrier region (dashed ellipse) that all lead to cyclization of asparagine, which is followed by ionization of the asparagine side-chain nitrogen.

The space of reaction coordinates considered in Fig. 4 is broad. The asparagine ionization distance, x, ranges from 0.9 to 3.0Å (x=2.0Å corresponds to a typical hydrogen bond, whereas x=1.0Å indicates the deprotonated asparagine side chain and a re-formed hydronium ion). The asparagine cyclization distance, y, ranges from 1.4 to 3.5Å (y=3.5Å is the relaxed distance in the initial state, whereas y=1.5Å indicates fully cyclized asparagine). We note that in these calculations the coordinate z (Figure 2B) is relaxed and the two original O-H distances in the water molecule were frozen as discussed in the Methods section.

The initial state (x=3.0Å, y=3.5Å) located on the top right of the plot (Fig. 4) is chosen to be zero for the relative reaction energy. The final products state is located in the bottom left corner (x=1.0Å, y=1.5Å) and corresponds to cyclized asparagine (succinimide), a re-formed hydronium ion, and a cleaved peptide bond. The path marked by arrows on Fig. 4 indicates the likely path followed by the C-terminal cleavage reaction. Along that path, y is reduced significantly first, and is then followed by a reduction in the value of x. The reduction in y can happen along combinations of paths shown in Fig. 4 because that region of the energy landscape is relatively featureless. In any case, cyclization of asparagine appears to be almost complete before the ionization of the side-chain nitrogen takes place. The barrier region is located near x=1.6Å, y=1.6Å indicated by the ellipse in Fig. 4 and has energy of ∼25kcal/mol higher than the reference state. Fig. 4 shows that alternate paths, e.g., in which asparagine ionizes before its cyclization, are highly unlikely as they sample regions of considerably high energies.

We note that PM3 calculations for geometry optimizations do not converge for certain choices of x and y (open squares in Fig. 4). Most of these points are neighbored by points of higher energies, and therefore, should not affect the general conclusions drawn above.

The observation made above that the asparagine cyclization proceeds before the ionization of its side chain is supported independently by high-level quantum calculations at the B3LYP/6-31G(d,p) level. Specifically, we followed the ionization of the asparagine by gradually transferring the proton from the side-chain nitrogen to the vicinal water molecule (Figure 5A). These high-level calculations also included effects of the dielectric constant of the local environment, which was assumed to be equal to 1 in the (gas phase) PM3 calculations.

Display large version of this figure

Figure 5

Energy of ionization of asparagine side-chain nitrogen (panel A) calculated in solvents of different dielectric constants: vacuum (ɛ_r=1), argon (ɛ_r=1.43), benzene (ɛ_r=2.247), chlorobenzene (ɛ_r=5.621), dichloroethane (ɛ_r=10.36), ethanol (ɛ_r=24.55), methanol (ɛ_r=32.63), and water (ɛ_r=78.39). Each implicit solvent has unique parameters such as radius and density. Gibbs energy is shown by blue triangles. Note that the peptide nitrogen has only one proton in this case. Gas phase coordinates, energy in solvent (○); solvent optimized coordinates, energy (□); and Gibbs free energy (▵).

Figure 5B shows that both the electronic energy (E) and Gibbs free energy (G) for the ionization of the asparagine side chain are rather high, equal to ∼31kcal/mol and ∼35kcal/mol, respectively, even in the highly polar medium such as water. The intein active site is expected to have a dielectric constant lower than that of water, and therefore, the relative energy of ionization will be even higher. We note that the asparagine cyclization distance, y, is relaxed in these calculations and does not reduce significantly. Thus, asparagine cyclization will require additional energy. In contrast, as shown later, the side-chain ionization is almost spontaneous once the asparagine side chain has undergone cyclization and formed a succinimide, consistent with the experimental observations of enhanced cleavage at low pH by Wood et al. 17,18.

Collectively, the above calculations allow us to propose a somewhat detailed C-terminal cleavage reaction mechanism at low pH, in which six states shown in Fig. 6 are particularly important. Figure 6A shows the hydronium ion in the context of the relevant part of our intein system. The second state involves the donation of a proton by the hydronium ion to the peptide nitrogen, resulting in water and N-protonated state (Figure 6B). Asparagine cyclization is shown in Figure 6CD, where the asparagine side chain still has two protons. The explicit water molecule is adjacent to the peptide nitrogen in one case (Figure 6C) and moves to accept a proton from the forming succinimide in another (Figure 6D). The formed succinimide with the proton passed back to the hydronium ion is shown in Figure 6E, whereas the final product is shown in Figure 6F. Water is re-formed and the extein segment leaves with an NH₃ group.

Display large version of this figure

Figure 6

Important states in the proposed hydronium ion catalyzed asparagine cyclization and peptide bond cleavage mechanism (see text for details).

The vicinal water molecule plays an important role in this mechanism and is used both as an acid (state A→B) and a base (state D→E). Indeed, succinimide with is highly acidic due to the resonance effect of amide bonds on either side of that nitrogen. As a result, the nitrogen readily gives a proton to a nearby water molecule (state F).

The energies corresponding to the various states (from A to F) presented in Fig. 6 are shown in Fig. 7. State Z in Fig. 7 is shown as a reference. The transition from Z to A corresponds to a positively charged histidine and a water molecule (Z) forming a neutral histidine and a hydronium ion (state A). The energy barrier from B to C/D corresponds to the asparagine cyclization, where the second proton is still attached to the asparagine side chain. Indeed the normal mode corresponding to the single imaginary frequency in state C was in the direction of bond formation, as expected. For this system, the energetics suggest that the nitrogen will give its second proton to water to reform the hydronium ion (Figure 6E). As discussed above, in the absence of protonation of the peptide nitrogen (e.g., in the case of a neutral peptide), the asparagine cyclization has a higher barrier.

Display large version of this figure

Figure 7

Electronic energies and approximate Gibbs free energies for states A through F in Fig. 6 (energies are relative to state B). State Z is shown as a reference and corresponds to a positively charged histidine and a water molecule (Z). Figure key: vacuum optimized coordinates energy (□) and Gibbs free energy (○); water optimized coordinates, energy (▵); and Gibbs free energy (▿). Single point energies calculated with B3LYP/6-311++g(2d,p) using the coordinates optimized in implicit water without the diffuse functions (♢). Energies of intermediate structures between the states in Fig. 6 optimized in the gas phase (■) are also shown.

Fig. 7 also highlights the effects of taking into account the dielectric constant of the environment of the active site on intein. As expected, the energy barrier (∼33kcal/mol) in the gas phase is reduced to ∼25kcal/mol in implicit solvent having a high dielectric constant. When tested with MP2, the energy barrier was found to be 29.4kcal/mol at the MP2/aug-cc-pVDZ level. This mechanism was also tested with the 45-atom tripeptide system (histidine-asparagine-cysteine), and the gas phase energy barrier was found to be 28kcal/mol, with B3LYP/6-311++G(d,p). QM/MM calculations that include more explicit water molecules suggest that the computational energy barrier calculated with B3LYP/6-31G(d,p) is 3–4kcal/mol lower than the 33kcal/mol found by the gas phase calculation. Since the crystal structures of active inteins do not include exteins, the conformation of the exteins and N- and C-terminal active sites is unknown. Hence, the intein plus extein system used in the QM/MM calculation will require additional verification to be comparable to precursor inteins used in experiment. The actual dielectric constant of the protein interior is between that of bulk water and vacuum, and hence, our calculations suggest that the energy barrier for C-terminal cleavage lies between 25 and 33kcal/mol. The experimental value is ∼21kcal/mol at pH 6.0 17.

We note that our calculations have several limitations. In the actual intein system, the overall protein structure (including both the intein and exteins) that surrounds the active site provides a significantly greater structural as well as chemical context for the reaction to occur. Also, there will likely be more than one water molecule in the vicinity of the active site that could mitigate the C-terminal cleavage reaction. Our system, in contrast, is significantly smaller due to computational limitations. In addition, our calculations are by necessity static in nature, and ignore the conformational and water exchange dynamics that are important in enzymatic catalysis 46. These types of concerns are shared by most (if not all) quantum calculations of enzymatic reactions. Nevertheless, our calculations provide a plausible mechanism for C-terminal cleavage that could be tested in the future by application of better multi-scale methods and/or detailed experimentation.

In summary, we present computational analysis of the intein C-terminal cleavage reaction. In particular, we have studied the cleavage mechanism of the M. tuberculosis RecA intein using information from available intein crystal structures and mutagenesis experiments. Part of the intein structure close to the C-terminal reaction site was considered in both gas phase as well as implicit solvent models calculations. Specifically, we performed computationally efficient PM3 calculations to obtain a qualitative understanding of the reaction pathway in the two-dimensional space of reaction coordinates chosen heuristically. More accurate high-level density functional calculations after the PM3 calculations allowed us to propose a detailed mechanism of the C-terminal cleavage at low pH.

Our calculations suggest that N-protonation, protonation of the nitrogen atom of the scissile peptide bond, is the likely starting point of the C-terminal cleavage reaction. A hydronium ion, included in our calculations to mimic a low pH environment, protonates the peptide nitrogen. After N-protonation, the reaction proceeds through asparagine cyclization followed by succinimide deprotonation. The energy barrier of ∼33kcal/mol for the rate determining step in gas phase reduces to ∼25kcal/mol in solution, and is close to the value of ∼21kcal/mol obtained from experiments 18. Our proposed mechanism uses one water molecule, which acts both as an acid and a base, and plays a critical role in the cleavage reaction.

Further work will investigate the importance of the penultimate histidine and first C-extein residue, cysteine, or serine by increasing the system size. The possibility of protonation by histidine is not excluded by the mechanism proposed here. For example, positively charged histidine could donate a proton to the peptide nitrogen via a water molecule; the reaction can then follow steps similar to the ones outlined here.