| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Physics Department, National Taiwan Normal University, Taipei, Taiwan, Republic of China
Correspondence: Address reprint requests to Chi-Ming Chen, 88 Sec. 4 Ting-Chou Rd., Taipei, Taiwan, 11718. Tel.: 886-2-86638109; E-mail: cchen{at}phy.ntnu.edu.tw.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMODELALGORITHM OF SIMULATIONSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
helices for both random folding and assisted folding. The chain length dependence of the folding time of a hydrophobic segment to a helical state is studied for both free and anchored chains. An unusual length dependence in the folding time of anchored chains is observed. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMODELALGORITHM OF SIMULATIONSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
; Leopold et al., 1992; Wolynes et al., 1995; Gutin et al., 1996; Li et al., 1996; Chan and Dill, 1997; Onuchic et al., 1997; Duan and Kollman, 1998). Recently, considerable attention has been focused on the cooperatively kinetic behavior of protein folding (Chan, 2000; Kaya and Chan, 2000; Fan et al., 2001). Kaya and Chan (2000) have shown the lack of cooperativity in popular lattice models, such as the two-letter HP model or the twenty-letter model, by using the calorimetric criterion. This may be due to oversimplification of these lattice models in both chain representations and intrachain interactions. Although these coarse-grained lattice models have the advantage of cutting computational efforts in studying protein folding, the kinetic information on the evolution of a protein chain from one coarse-grained structure to another is also lost. It is quite possible that some kinetic pathway is overwhelmingly enhanced or entirely blocked due to intermolecular interactions or molecular packing. A possible origin of this effect is the dipole-dipole interactions between amide groups in real helices (Cantor and Schimmel, 1980). However the resolution of coarse-grained lattice models is not enough to distinguish these pathways. Lattice models also tend to use statistical contact potentials extracted from the Protein Data Bank (PDB), which provide little information about how these potentials arise from realistic physical interactions. It is also unclear how these potentials would evolve when the protein structure deviates from its native structure.
Membrane proteins (MPs) perform important and diverse functions in living cells, such as regulation, communication, and assisting the folding of other MPs (for a review, see White and Wimley, 1999). They are partially buried in the nonpolar environment of a lipid bilayer, where the hydrophobic effect is absent. Because lipid tails are unable to form hydrogen bonds with proteins, the intrachain hydrogen bonding along the backbone of proteins in a membrane plays a significant role in forming their native structure. According to the structure of transmembrane segments, there are two known classes of MPs. The first class contains MPs whose transmembrane segments all form an -helical structure with lengths (Nc) of 17 to 25 amino acids (aa). In the second class, on the other hand, those MPs usually have a ß-barrel structure. However, due to difficulties in crystallizing MPs, only a dozen or so MPs have known crystallographic structures so far. Among them, helix bundles are much more abundant than ß barrels.
A previous model using a full-backbone atom representation in a diamond lattice initiates an interesting study on the insertion of polypeptides into a membrane (Milik and Skolnick, 1992). This model explicitly specifies that hydrogen bonds can form for only (i, i ± 4) pairs, where i labels amino acids in the chain. This can be considered as an extreme case in emphasizing the (i, i ± 4) hydrogen bonding state because there is no reason to forbid hydrogen bonding between (i, i ± n) residues for n > 4. Furthermore, this restriction also excludes the possibility to form ß strands. Our previous paper (Chen, 2001) proposes a lattice model for the folding of transmembrane polypeptides, in which the backbone hydrogen bonding of polypeptides can occur between i and i ± n for n 
4. Our model predicts two possible stable structures of transmembrane polypeptides, including helix and double helix structures, which have been observed for gramicidin dimers (Arumugam et al., 1996). However, the folding time of a polypeptide chain in this simple model is unexpectedly long, which might result from its incapability to distinguish the differences among various hydrogen bonding states. In this paper, we propose a lattice model of MPs to study their native structures and cooperative folding. The main goal of our model is to predict the native structures of membrane proteins by their relevant physical interactions alone, and this model is then reliable for us to study the folding kinetics of a transmembrane peptide with minimal artifacts. Predicted structures of MPs from our model can be refined by all-atom models and will be useful in studying their biological functions by docking studies. This model has a composite energy function to describe interactions among amino acids, which uses realistic interactions when residues are in a membrane but statistical potentials when they are in water. We show that this model predicts a reasonably good native structure of sensory rhodopsin I (SRI), which is a phototaxis receptor in Halobacterium salinarum and consists of 239 amino acids. In addition, we study the cooperative effect on the folding time of MPs by introducing a cooperative factor to favor the (i, i ± 4) hydrogen bonding state. Our chain representation is based on the bond-fluctuation model (Carmesin and Kremer, 1988; Chen and Fwu, 2001; Chen, 2001), which has advantages of giving reasonably good secondary structures and of simulating a more realistic diffusive kinetics than regular lattice models, while the computational cost is still quite limited compared to that of off-lattice models. We note that, although a helix bundle structure is studied in this paper, our model can also be used to predict ß-barrel structures because backbone hydrogen bonding is possible for amino acids far apart from each other in the sequence.
|
MODEL |
|---|
|
TOPABSTRACTINTRODUCTIONMODELALGORITHM OF SIMULATIONSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
H and O
C bonds are co-linear. Therefore we model the hydrogen bonding energy by EH-bond = 
i,j
|(ni • rij) (nj • rij)| 
r(i,j),4, where ni is the NH (or OC) bond orientation of the i-th amino acid, while r(i,j) and rij are the distance and its unit vector between amino acids i and j. Because the backbone hydrogen bonding is the dominant interaction for the formation of secondary structures of MPs, its energy strength is set to unity. Furthermore, we have explicitly excluded the possibility of forming 27 ribbons and 310 helices due to steric hindering by disallowing the hydrogen bonding between (i, i ± 2) and (i, i ± 3) pairs. The bending energy of the chain is assumed to be e1 i (1 - cos
i), where e1 is the bending rigidity and i is the angle between two consecutive bonds i and i + 1. The vdW interaction between amino acids is modeled by Evdw = e2 i,j {[1.78/r(i,j)]12 - [1.78/r(i,j)]6}, where e2 is its strength relative to hydrogen bonding. This vdW term has a minimum if two amino acids are next to each other in a cubic lattice model. For amino acids in water, their interactions are modeled by a residue-residue contact potential (Econtact) and the hydropathical interaction (Ehydropathy), i.e., Uwater = Econtact + Ehydropathy. The interactions between the exposed residues and the lipid bilayer are ignored. Here we use the Thomas-Dill contact potential with strength e3 to model the residue-residue interaction in water when residues are in contact (Thomas and Dill, 1996). Because MP residues exposed to water are mostly exterior residues, the positive contact energies between exterior residues in this potential imply highly dynamical loops of MPs in the water phase. The hydropathical interaction of amino acids in water can be modeled by using a rescaled Kyte-Doolittle hydrophathy index (spread between -1 and 1) with strength e4, which is mainly determined by the Gibbs free energy change for transferring amino acids from water into condensed vapor (Kyte and Doolittle, 1982). In addition, the insertion of a polypeptide chain into a membrane will disturb the integrity of the membrane and local lipid density around the chain, which increases the energy of the membrane (White and Wimley, 1999). We model this effect by introducing an effective lateral pressure (P) applied to the polypeptide chain to minimize its lateral area (A = i li2) in the membrane, where li is the projected length of the i-th transmembrane segment on the membrane surface (Chen, 2001). Therefore, to find the ground state structure of MPs, the relevant physical quantity to be minimized in our model is the enthalpy H = U + PA. |
ALGORITHM OF SIMULATIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMODELALGORITHM OF SIMULATIONSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
for studying dynamics of polymer chains in various spatial dimensions. Since then it has been used for investigation of the crossover between Rouse and reptation dynamics (Gerroff et al., 1993), for studying interdiffusion of polymer blends (Deutsch and Binder, 1991), the dynamics of polymer melts near glass transition (Ray et al., 1993), and polymer crystallization in dilute solution (Chen and Higgs, 1998).
Each monomer in the model is a cube of length 1 (lattice spacing) on a cubic lattice as shown in Fig. 1. The set of allowed bond vectors is B = P(2,0,0) 
P(2,1,0) P(2,1,1) P(2,2,1) P(3,0,0) P(3,1,0), where P(a,b,c) stands for the set of all permutations and sign combinations of ±a, ±b, ±c. The number of configurations per bond is z = 108. The length of one bond can take any one of the five values 2, 51/2, 61/2, 3, 101/2 (in units of lattice spacing). Chains satisfy the excluded volume constraint: no lattice site may be occupied by more than one monomer. The set B is chosen to satisfy the constraints of both excluded volume between monomers and topological entanglement between chains (i.e., two chains cannot pass through each other). If any other bond vectors were added to this set, some chains would become "phantom" chains.
|
H/T)], where H is the enthalpy change of the system. |
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMODELALGORITHM OF SIMULATIONSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
), and 2) the packing of transmembrane segments is much slower than the formation of each individual helix. Therefore we simulate the folding of SRI according to a three-stage model, which is an extension of the two-stage model (Popot and Engelman, 1990). This model is consistent with observed facts, including the nature of transmembrane segments in known structures, refolding experiments, the assembly of integral membrane from fragments, and the existence of very small integral membrane protein subunits. The physical picture of membrane protein folding proposed here is that a dominant hydrophobic interaction at early times hides the hydrophobic segments from water, followed by backbone hydrogen bonding assisted helix formation (a local interaction in sequence), and finally the packing of helices due to the vdW interaction (a nonlocal interaction in sequence). At the first stage, the chain is partitioned into several transmembrane segments by its hydropathy. Each segment then folds to form a secondary structure to optimize the hydrogen bonding along the backbone during the second stage. Finally, these autonomous folding domains will aggregate to form a compact structure due to the vdW interaction.
During the first stage, as shown in Fig. 2, the average hydropathy index of SRI using a window of 20 amino acids is calculated. To optimize the hydropathical interaction, the center of a transmembrane segment of 20 amino acids is located at those higher peaks of the hydropathy profile. Because no overlap is allowed for two segments, seven transmembrane segments are predicted for SRI from the first stage. The inset of Fig. 2 shows the reduction in hydrophobic energy when those transmembrane segments are placed in the membrane phase for various window sizes. The hydrophobic energy can be further reduced if the window size of each transmembrane segment is variable. This prediction is used to produce an initial configuration of SRI to perform MC simulations of its folding at the second stage: seven transmembrane segments are randomly distributed in the membrane phase, and two neighboring segments are connected by a random coil. The folding of the entire chain is simulated at this stage to optimize the enthalpy of the system. The result of the second stage is not sensitive to the window size used at the first stage. In fact, it remains the same even when only thirteen residues of each segment are placed in the membrane initially. Here we choose e1 = 0.3, e2 = 0, e3 = 0.3, e4 = 1.5, P = 0.1, and L = 23 lattice spacing (
31 Å, White and Wimley, 1999). The choice of these parameters is not unique. Because all interactions in our model are weak forces, we expect these energetic parameters to be of the same order of magnitude (the strength of hydrogen bonding is taken to be of order 1). No drastic changes in the secondary structure of SRI are observed if a slightly different set of parameters is used. We note that, to obtain a better representation of helices, the values of bending rigidity and lateral pressure adopted here are different from those in our previous study (Chen, 2001). However the general features of folding dynamics and structure in this case are not affected. During this stage, the vdW interaction is switched off and the secondary structure formation is dominated by the hydrogen bonding and the hydropathical interaction. In Fig. 3, we show the comparison of the PDB secondary structure of SRI (A) (Berman et al., 2000) to our predicted structures for L = 23 (B) and L = 24 (C). Both simulations predict a seven-helix structure of SRI. The average helix length of SRI is 22.4 aa in the PDB structure, and is 22.3 aa for L = 23 and 25.9 aa for L = 24. For L = 23, the prediction error in the average helix length is 0.6%, and the secondary structure alignment error (mismatch between Fig. 3 A and 3 B) is 11.3%. When the membrane thickness varies slightly, the seven-helix structure of SRI is still stable, but the helix regions will change correspondingly. For L = 24, its deviation from the PDB structure is 15.3% in the average helix length and is 17.6% in their mismatch. After the formation of all autonomous folding domains, we switch on the vdW interaction and switch off all other interactions. During this association stage, the initial configuration of SRI is taken from results at the second stage and each helix can only diffuse within the membrane (neighboring helices are constrained by their connecting loop). Packing of helices is a result of the vdW interaction alone. The external work done by the effective lateral pressure in our model does not drive helices to pack, because the helical structure (or projected area) of each transmembrane segment is fixed at this stage. Fig. 4 shows a comparison of our predicted tertiary structure (A) (only helical regions are shown) with the PDB structure (B). The resemblance of these two structures demonstrates the validity of our model in predicting the native structure of MPs. We note that tilting and distortion of transmembrane helices can be better studied in an off-lattice model, and the results will appear elsewhere (Chen and Chen, in preparation). Such a coarse-grained structure can be used as the initial configuration in an all-atom simulation or minimization to obtain a good native structure.
|
|
|
, 1993, 1995). An important question in studying protein folding dynamics concerns the cooperative effect in -helix formation. As indicated from the Ramanchandran plot, the formation of idealized helices strongly depends on the backbone structures (Mathews and van Holde, 1996); if the number of amino acids per turn is greater than four, no ideal helical structure can form. Therefore, the cooperative effect of helices resulting from the steric hindrance of the backbone structure would favor the formation of (i, i ± 4) hydrogen bonding over the others. For MPs, another feasible origin of cooperativity is membrane-promoting -helix formation, which has been studied by both experiments and simulations (Deber and Li, 1995; Deber and Goto, 1996; Efremov et al., 1999). Because our previous study (Chen, 2001) allows all (i, i ± n) hydrogen bonding for n 4, our previous model has no cooperative effect and leads to an exponential growth of the folding time for helix formation as the helix length increases. To properly include the cooperative effect of (i, i ± 4) hydrogen bonding, here we add an extra favorable factor exp(h) in the moving probability of each residue to enhance the cooperative helix formation, where h is the change of (i, i ± 4) hydrogen bonding pairs and is the cooperative factor. This cooperative effect on the folding time of a single transmembrane helix (AVATAYLGGAVALIVGVAFVWLLY, a transmembrane helix of SRI) has been studied for both random folding (with a random initial configuration) and assisted folding (with a parallel initial configuration to the membrane normal) using e1 = 0.3, e2 = 0, e3 = 0.3, e4 = 1.5, P = 0.1, T = 0.31 (the optimal folding temperature), and L = 24. The assisted folding is to mimic the helix formation of a hydrophobic segment assisted by a hydrophilic channel. This particular initial configuration selects a folding pathway with a smaller activated energy barrier than a random initial configuration, as described in our previous work (Chen, 2001). It is found that the mean first passage time (MFPT) to a helical state is minimized at 
1.2 for both cases, as shown in Fig. 5. The folding time to a helical state increases exponentially if deviates from this optimized value. For smaller values of , the chain is easily trapped at wrongly folded states. If is too large, the partially folded helix is often trapped at wrong positions in the membrane and the helix formation cannot continue to the rest of the peptide chain due to the presence of the water-membrane interface. These misfolded structures of a frustrated partial helix must be unfolded first before the chain can reach its ground state, which drastically increases the folding time. Note that, however, we do not know whether is optimized for membrane protein folding or, if it is optimized, why.
|
). This difference is clearly due to the cooperative helix formation. Because the configuration space of a hydrophobic peptide increases drastically with its chain length, the MFPT to search for helical states also increases drastically at zero cooperativity. At the optimal cooperativity, only part of the configuration space near helical states is focused in searching, which leads to a near linear MFPT. For anchored chains, the MFPT decreases slightly for Nc < 22 but increases linearly for Nc 22. The initial drop in the MFPT of anchored chains is due to the fact that shorter chains have a smaller cooperative effect and thus they might take longer times to fold even when their configuration space is also smaller. Because anchored chains have a smaller configuration space and a larger cooperative effect than free chains, a smaller slope of the MFPT is observed for anchored chains. We note that anchored chains have a larger MFPT than free chains at short chain lengths due to the anchored restriction in their folding pathways.
|
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMODELALGORITHM OF SIMULATIONSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
h) to -helical states. We find that the folding time of transmembrane helices is optimized for = 1.2. The polypeptide chains are usually trapped at wrong configurations for small values of cooperativity, whereas they tend to be trapped at wrong positions in the membrane at large cooperativity. The dependence of folding time on chain length is nearly linear for free chains but exhibits an unusual behavior for anchored chains, which might result from the competition between cooperativity and the number of configurations as chain length varies. |
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMODELALGORITHM OF SIMULATIONSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
This work is supported, in part, by the National Science Council of Taiwan under grant no. NSC 90-2112-M-003-024. C.M.C. is grateful for the hospitality at the Department of Pharmaceutical Chemistry, University of California at San Francisco.
Submitted on August 23, 2002; accepted for publication October 31, 2002.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMODELALGORITHM OF SIMULATIONSRESULTS AND DISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Berman, H. M., J. Westbrook, Z. Feng, G. Gilliland, T. N. Bhat, H. Weissig, I. N. Shindyalov, and P. E. Bourne. 2000. The protein data bank. Nucleic Acids Res. 28:235–242.
Bryngelson, J. D., and P. G. Wolynes. 1989. Intermediates and barrier crossing in a random energy model (with applications to protein folding). J. Phys. Chem. 93:6902–6915.
Cantor, C. R., and P. R. Schimmel. 1980. Biophysical Chemistry. W. H. Freeman Company, San Francisco. 862–863.
Carmesin, I., and K. Kremer. 1988. The bond fluctuation method: a new effective algorithm for the dynamics of polymers in all spatial dimensions. Macromolecules. 21:2819–2823.
Chan, H. S., and K. A. Dill. 1997. Solvation: how to obtain microscopic energies from partitioning and solvation experiments. Annu. Rev. Biophys. Biomol. Struct. 26:425–459.[Medline]
Chan, H. S. 2000. Modeling protein density of states: additive hydrophobic effects are insufficient for calorimetric two-state cooperativity. Proteins. 40:543–571.[Medline]
Chen, C.-M., and P. G. Higgs. 1998. Monte Carlo simulations of polymer crystallization in dilute solution. J. Chem. Phys. 108:4305–4314.
Chen, C.-M. 2001. Lattice model of transmembrane polypeptide folding. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 63:010901.
Chen, C.-M., and Y.-A. Fwu. 2001. Monte Carlo simulations of polymer brushes. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 63:011506.[Medline]
Deber, C. M., and S.-C. Li. 1995. Peptides in membranes: helicity and hydrophobicity. Biopolymers. 37:295–318.[Medline]
Deber, C. M., and N. K. Goto. 1996. Folding proteins into membranes. Nat. Struct. Biol. 3:815–818.[Medline]
Deutsch, H. P., and K. Binder. 1991. Interdiffusion and self-diffusion in polymer mixtures—a Monte Carlo study. J. Chem. Phys. 94:2294–2304.
Duan, Y., and P. A. Kollman. 1998. Pathways to a protein folding intermediate observed in a 1-microsecond simulation in aqueous solution. Science. 282:740–744.
Efremov, R. G., D. E. Nolde, G. Vergoten, and A. S. Arseniev. 1999. A solvent model for simulations of peptides in bilayers. I. Membrane-promoting -helix formation. Biophys. J. 76:2448–2459.
Fan, K., J. Wang, and W. Wang. 2001. Modeling two-state cooperativity in protein folding. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 64:041907.[Medline]
Gerroff, I., A. Milchev, K. Binder, and W. Paul. 1993. A new off-lattice Monte-Carlo model for polymers—a comparison of static and dynamic properties with the bond-fluctuation model and application to random media. J. Chem. Phys. 98:6526–6539.
Gutin, A. M., V. I. Abkevich, and E. I. Shakhnovich. 1996. Chain length scaling of protein folding time. Phys. Rev. Lett. 77:5433–5436.[Medline]
Kaya, H., and H. S. Chan. 2000. Energetic components of cooperative protein folding. Phys. Rev. Lett. 85:4823–4826.[Medline]
Kyte, J., and R. F. Doolittle. 1982. A simple method for displaying the hydrophobic character of a protein. J. Mol. Biol. 157:105–132.[Medline]
Leopold, P. E., M. Montal, and J. N. Onuchic. 1992. Protein folding funnels—a kinetic approach to the sequence structure relationship. Proc. Natl. Acad. Sci. USA. 89:8721–8725.
Li, H., R. Helling, C. Tang, and N. Wingreen. 1996. Emergence of preferred structures in a simple model of protein folding. Science. 273:666–669.[Abstract]
Mathews, C. K., and K. E. van Holde. 1996. Biochemistry. Benjamin/Cummings Publishing Co., Menlo Park. 165–174.
Milik, M., and J. Skolnick. 1992. Spontaneous insertion of polypeptide-chains into membranes: a Monte Carlo model. Proc. Natl. Acad. Sci. USA. 89:9391–9395.
Milik, M., and J. Skolnick. 1993. Insertion of peptide chains into lipid-membranes: an off-lattice Monte Carlo dynamics model. Proteins. 15:10–25.[Medline]
Milik, M., and J. Skolnick. 1995. A Monte Carlo model of fd and Pfl proteins in lipid membranes. Biophys. J. 69:1382–1386.
Onuchic, J. N., Z. Luthey-Schulten, and P. G. Wolynes. 1997. Theory of protein folding: the energy landscape perspective. Annu. Rev. Phys. Chem. 48:545–600.[Medline]
Popot, J. L., and D. M. Engelman. 1990. Membrane protein folding and oligomerization: the two-stage model. Biochemistry. 29:4031–4037.[Medline]
Ray, P., J. Baschnagel, and K. Binder. 1993. Dynamics near the glass-transition in two-dimensional polymer melts: a Monte Carlo simulation study. J. Phys.: Condens. Matter. 5:5731–5742.
Thomas, P. D., and K. A. Dill. 1996. An iterative method for extracting energy-like quantities from protein structures. Proc. Natl. Acad. Sci. USA. 93:11628–11633.
White, S. H., and W. C. Wimley. 1999. Membrane protein folding and stability: physical principles. Annu. Rev. Biophys. Biomol. Struct. 28:319–365.[Medline]
Wolynes, P. G., J. N. Onuchic, and D. Thirumalai. 1995. Navigating the folding routes. Science. 267:1619–1620.
This article has been cited by other articles:
 |
|
 |
|
  S. R. McAllister and C. A. Floudas {alpha}-Helical Topology Prediction and Generation of Distance Restraints in Membrane Proteins Biophys. J., December 1, 2008; 95(11): 5281 - 5295. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D.-M. Ou, C.-C. Chen, and C.-M. Chen Contact-Induced Structure Transformation in Transmembrane Prion Propagation Biophys. J., April 15, 2007; 92(8): 2704 - 2710. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
