Min Distance of PLM 210 Carboxy from Long TIM44. 220. 245. 270. 295 ... to
better illustrate the absorption of PLM210 at the base of A1\A2. Thus it is
apparent ...
Binding of Palmitic Acid to the C-TerminalDomain of Mitochondrial TIM44 Protein in Yeast
Molecular Dynamics Simulation and Analysis
Under supervision of Dr. Yossi Tsfadia and Mr. Avneon Yoav with Prof. Menachem Gutman research group
Amram Shay 27/11/2008
1
A. Introduction Mitochondria play a critical role in cell physiology and in cellular respiration. According to the endosymbiotic theory, the mitochondria originated from a bacterial endosymbiotic event approximately two billion years ago and the course of evolution has displaced most of the mitochondria native DNA so that large proportion of those genes is encoded by the host. Thus, the majority of mitochondrial proteins is synthesized on cytosolic ribosomes and must be imported into the mitochondria. Mitochondria are divided into four major compartments – outer membrane (OM), intermembrane space (IMS), inner membrane (IM) and matrix, so cytosol proteins should be imported and sorted into each of these compartments. The proteins are imported in their preprotein state, of which two main classes can be distinguished: amino-terminal, cleavable presequence and internal signals that are distributed throughout the protein (1). Those characteristics signal for recognition by the mitochondria and passage across the membranes. Translocation of pre-proteins into the mitochondria is mediated by a large number of components, one of which is the TIM44 (Tranlocase of the Inner Membrane 44) protein in a mechanism that has yet to be understood. Tim44 is functioning at the junction between the membrane and the mitochondrial matrix, and is capable to adsorb to a phospholipid membrane containing acidic residues. Azem et al. showed increase in Tim44 binding to vesicles containing increasing cardiolipin concentrations (2), which was abolished only at high concentrations of salt (0.5M). The necessity of cardiolipin presence in vesicles suggests electrostatic manner of binding on one hand. On the other hand, the high concentration of salt required to disrupt biding, suggests also involvement of hydrophobic interactions, beside the electrostatic one. Previous MD work was done by Avneon et. al with the human TIM44 C-Domain (5) with both NH 3+ and NH 2 N-terminals. The work regarding the human TIM44 homolog consisted of the published crystal structure (13, PDB codename 2CW9) with no modifications, as well as yeast TIM44 C-Domain (4, PDB codename 2FXT) and modified TIM44. Modification done by Avneon et. al consist of elongation by 14 a.a. of TIM44 in the same method as explained in section B in this work, and also mutations of the elongated version. This work is based on the aforementioned work done by Avneon et. al and includes several repeats of the same composition of simulations . The work of Avneon et al has been performed in collaboration with Dr. Azem's research group where in vitro experiments (17) revealed that the area of A1 helix may associate with TIM44 ability to bind to membrane. 2
Elongation was facilitated in light of some preliminary results of unmodified PDB of TIM44-C-Domain simulations by Avneon et al which suggested A1 may play a role in membrane binding. Thus, elongation of TIM44 serves to asses whether these results ensue from an "end" effect of A1 being incomplete, or if other areas in A1 may be also important for interaction with the membrane. Evaluating the working mechanism of proteins is a common problem in biochemistry. Understanding the mechanism is important to the complete understanding of the process, and thus making human intervention in the process possible (creation of medicine, drugs etc). It is the objective of this study to try and shed some light on the mechanism that leads to the binding of TIM44 C-Domain to the matrix-side of the membrane. To this end we employ the use of Molecular Dynamics (MD) tools. As to this date, cardiolipin parameterization for the GROMACS 53A6 force field is unavailable. Instead, we simulated TIM44 C-terminal domain using palmitate as a model ligand in order to detect areas that may interact with negatively-charged molecules such as cardiolipin. The choice of the palmitate was due to its simple structure, its negatively charged head and its known parameterization from previous works (12).
3
B. Methods Simulations were conducted using GROMACS software, ver. 3.3.1, and force-field G53A6 (3) and its package of programs. Protein structures of C-Domain of both human and yeast TIM44 has been determined by methods of X-ray crystallography (13 ,4) with resolutions of 1.9Å and 3.2Å respectively. Structures were acquired from the Protein Data Bank (14) by code names 2CW9 (human) and 2FXT (yeast). Parameterization for palmitic acid has been determined elsewhere (12) and simulations used the SPC water model (11). Box shape was set to dodecahedron in order to save computer power spent on solvent (3) (water molecules extended at least 12Å from protein's surface area). Charge neutralization and ionic strength of approximately 100mM were obtained with appropriate numbers of Na+ and Cl- ions. Protein C/N Terminal sites were selected COO − and NH 2 respectively. The choice of COO − was due to the common state of the carboxyl group at the pH level in the matrix side of the mitochondria (pH ~8.00). The N-terminal of the C-domain in vivo is linked to the N-domain (which was not crystallized) through peptide bond. Thus, we determined the N-terminal of the C-domain to be NH 2 in silica to better emulate the native protein and avoid bogus positive charge. Simulations were preceded by energy minimization (EM) by method of conjugate gradient in order to relax the system (7), followed by position restraint (at least 40ps), and equilibrium (at least 200ps) (6). Full MD simulations were run with timesteps of 2fs, under NPT conditions (P=1 bar; τ P = 0.5 ps ; T=300ºK; τ T = 0.1 ps ) using Berendsen's coupling algorithm (8). VDW forces were treated with a cutoff at 12Å and long range electrostatic forces were treated using Particle Mesh Ewald (PME) (9). Data analysis and graphs were generated by the GROMACS package ver. 3.3.1 (3, 7), all protein images were generated using VMD software (10). 1.A
1.B A1 helix 43.19Å
A1 helix 21.38Å
Fig 1. Illustration of the Short (A) and Long TIM44 C-terminal domain (B) and the corresponding length of A1 helix. Number of
molecules in 1.A. ~18,000 and in 1.B. ~30,000. B.solvent 1. Elongation of TIM44
4
Residues SER234-GLY425 of the yeast TIM44 C-domain were crystallized though several residues at the C-terminal end of the C-domain were not visible due to their flexible nature (4). In addition to the simulations of published structure of yeast TIM44 C-Domain we have also included several simulations elongated by 14 amino acids (LYS220-ARG233). Elongation was facilitated using Swiss-PDB Viewer (15). The 14 a.a. were added to A1 as an extension of the existing alpha-helix (Figure 1). In this regard it is important to note that elongation of TIM44 was done in order to inspect the possible contribution of different areas of helix A1 (Figure 3) to the absorption of palmitate molecules. The simulation box size of the short protein was determined by extending the box 15Å from protein surface that consisted of approximately 18,000 solvent molecules. Since A1 helix sticks out of the main body of the protein (Figure 1) the corresponding box of the elongated protein would have included approximately 36,000 solvent molecules. In order to minimize computation power simulating solvent, box size for the elongated protein was adjusted to extend 12Å instead of 15Å from protein surface, thus including only ~30,000 solvent molecules. In both cases (Short and Long), shape was set to dodecahedron (3); for all simulations cutoff for both coulomb and VDW were left at 12Å. Table 1. Description of MD simulated systems
Ref. Name A. B. C. D. E. F. G. H.
Short I Short II Short PLM Long I Long II Long PLM I Long PLM II Human TIM I
TIM44 Type*
#PLM**
Distance***
Simulation Time
Short Short Short Elongated Elongated Elongated Elongated Human TIM44
------4 ------4 4 ----
15Å 15Å 15Å 12Å 12Å 12Å 12Å 15Å
12ns 14ns 20ns 14ns 14ns 20ns 20ns 10ns
*
The Short type is the original PDB file downloaded from Protein Databank. Elongation of TIM44 was by 14 AA (220KKVEDFKEKTVVGR 234SIQ ) of the N-Terminus side of the C-Domain as specified by Josyula et al. (4). The addition of a.a. was generated using SPDBView (15). ** Number of palmitate molecules in simulation box. *** Box size was established by taking 15/12Å from TIM44 surface area and filling it with water molecules. The simulations shown in gray will not be discussed in this work.
In order to facilitate discussion in this study, we refer to SER234-ASP244 as helix A1 of the Short protein, and LYS220-ASP244 as helix A1 of the Long protein. Unless stated otherwise, references to TIM44 throughout this work will always refer to TIM44 C-Terminal Domain.
5
(A)
(B) A6
B2 B1
"Base" of A1/A2 helices
A6 A5 A1
A4.2 A2
B4 A4.1 A3
A3
Figure 2.A, 2.B: Schematic structure of TIM44 C-Domain will be used to In the elongated version A1 is longer.
In this study we will analyze and discuss the binding of palmitate to the various sites of TIM44 while simulations in physiological water without palmitate (B, D, E) were used as the base for comparison (both for Long and Short TIM44). Figure. 2.A, B. shows structural elements (α-helix, β-strand) of TIM44, though some may vary during simulation. Coloration was applied as to represent types of secondary structures (purple=α-helix, yellow=β-strand/sheet, cyan/white=extended strands and turns).
B. 2. Methods of Analysis RMSD (Root-Mean-Square-Deviations) RMSD computes the root mean square deviation of atom distances using the g_rms algorithm in GROMACS package. This analysis was used to measure overall change of atoms' position. In this type of analysis each timeframe t2 of the MD simulation was compared to a reference frame at time t1 by generating root-min-square-distances for atoms of interest. The General formula for the calculation is: 1 RMSD(t1 , t 2 ) = M
(B.2.1)
N
∑ m r (t ) − r (t ) i
i =1
i
1
i
2
2
1
2
Where N is total number of atoms, mi is the mass of atom i, ri (t ) is the position of atom i at N
time t, and M = ∑ mi . Note that the RSMD calculation here gives greater emphasis to large i =1
atoms by multiplying it by mi . All RMSD calculations throughout this study refer to the protein backbone and using the ending frame of the equilibrium run as reference. By using RMSD for TIM44 backbone we were able to detect simulation timeframes where major changes in protein conformation 6
manifested, or alternatively, indicate when protein achieves stability and no further conformational changes are manifested. RMSF (Root-Min-Square-Fluctuations) RMSF calculates the standard deviation of atomic positions. This analysis typically shows the tendency of a specific group to move (residue, by choosing each a.a.'s α-carbon). RMSF is plotted as function of α-carbon, thus it is possible to infer the behavior of the corresponding amino acid. S.A.A.P. Analysis S.A.A.P. is method for fast reliable tracking and viewing the trajectory of each ligand in the simulation which was developed for the purpose of this study. The method consists of finding the minimum distance of each ligand from the protein thus creating a list for each frame of the simulation that includes amino-acid/ligand numbers and the corresponding minimum distance between them. The list is then divided to several distance-groups and each group is assigned a specific color to reflect that distance group. It is then displayed on a scatter graph with X-axis as time, Y-axis as amino-acid-number and each groups' points are colored with their predefined distance color. In this manner we searched the minimum distance of TIM44 from each atom of the Palmitate COO − groups throughout the simulation and for each frame, the best minimum of the three atoms was taken. Distance groups and colors were assigned as following: 0-0.3nm red; 0.3-0.6nm orange; greater then 0.6nm in purple. A long stretch of colored points implies that the palmitate molecule had resided for a long period of time in the vicinity of the specified amino acid. This view analysis enabled us to compare the paths of each COO − group throughout the simulation and compare common trends and, specifically, areas of electrostatic interaction of the TIM44 with the COO − group .Note should be made that a similar analysis for the aliphatic tail is still in progress and would not be included in this work. Ligand Contact to the Protein In order to illustrate the regions in which TIM44-palmitate- COO − interacts we include images of TIM44 colored as function of number of contacts throughout the simulations. Contacts were defined as the total time (per simulation) that any palmitateCOO − spends in a distance lower than 0.6nm from a given amino acid. This analysis helps to
visualize and identify "hotspots" for the interaction with COO − on TIM44 surface.
7
C. Results and Discussion Simulations of TIM44 in Physiological Solution Three simulations of TIM44 in physiological solution were initiated: Short II, Long I and Long II. We have calculated RMSD values for TIM44 backbone, and RMSF for the 3.A.
RMSD of Yeast TIM44 backbone C-Domain, Physiological Water Short Vs. Elongated
1.2 1.1 1 0.9
RMSD [nm]
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0
2000
4000
6000
8000
10000
12000
14000
frame [ps] Short II
Long I
Long II
RMS Fluctuations of Yeast TIM44 α-carbon C-Domain in Physiological Water: Short Vs. Long II
3.B. 1.2 1.1 1 0.9
RMSD [nm]
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 220
234
244
260
300 Short II
# frame α carbone [ps] 340
Long I
380
420
Long II
Fig. 3.A RMSD of protein backbone in Short II and Long I/II . Fig. 3.B RMS fluctuations of α-carbons in Short II and Long I/II . Residues of Short A1 are marked with purple arrow; Long A1 residues with a blue arrow.
8
4.A.
RMSD of Yeast TIM44 backbone C-Domain in Physiological Water, Short II A1 Vs. Protein Body
4.B.
RMSD of Yeast TIM44 backbone C-Domain in Physiological Water, Long I A1 Vs. Protein Body
4.C. 1
0.9
0.9
0.9
0.8
0.8
0.8
0.7
0.7
0.7
0.6
0.6
0.6
R M S D [n m ]
R M S D [n m ]
1
R M S D [n m ]
1
0.5
0.5
0.5
0.4
0.4
0.4
0.3
0.3
0.3
0.2
0.2
0.2
0.1
0.1
0.1
0
0 0
2000
4000
6000
8000
frame [ps] Short A1
Short Body
10000
12000
14000
RMSD of Yeast TIM44 backbone C-Domain in Physiological Water, Long II A1 Vs. Protein Body
0 0
2000
4000
6000
8000
10000
12000
14000
0
2000
frame [ps] Long I A1
Long I Body
4000
6000
8000
10000
12000
frame [ps] Long II A1
Long II Body
Fig. 4. Comparison of RMSD of each simulation – Backbone of A1 Vs. Backbone of protein body. The body was defined as all backbone amino-acids except A1's. (4.A.) Short II. (4.B.) Long I (4.C.) Long II.
α-carbon RMSF values can easily reflect the behavior of the corresponding amino acid (Figure 3.A,B). All simulations show that soon after the initiation of the simulation that RMSD values reach a plateau (Figure 3.A, 2-4ns) thus it can be seen that overall structure of TIM44 main body (GLU245-GLU425) remains relatively stable with no major changes of conformation during the simulation time. Helix A1 sticks outside the globular body of the protein and retains some rotational degree of freedom, consequently contributing significantly to the RMSD values. Figures 4.A-C treat separately the RMSD of the main body of the protein and of helix A1 to emphasize the difference of behaviors between A1 helix and the protein main body. In this analysis it is apparent that RMSD values reach a plateau after ~2ns of the simulation time. Long-I also exhibits partial folding of A1 upon itself next to the point of elongation (Figure 5.B) which is near VAL230-VAL231. In Long-II, A1 remains stable throughout the entire simulation though it bends at the end of the elongation of A1. By utilizing the RMSF analysis (Fig 3.B.) it can be seen that residues belonging to A1 in the Short/Long protein show considerably higher RMSF values than the rest of the protein (purple and blue arrows, respectively). This suggests that A1 retains high degree of freedom in its movement throughout the simulations. This type of behavior in TIM44 was observed in the works of others in the yeast TIM44 (4) where it was suggested as the reason these residues were not visible through crystallography. In Short-II (Figure 4.A.) it seems that RMSD values for A1 are lower when comparing to those of the whole protein. While this may seem contradictory to the above statement regarding A1 degrees of freedom, the included RMSF analysis (Figure 3.B.) shows that residues of A1 indeed retain high degree of freedom in its movement throughout the simulation along ASP244-SER234. 9
14000
Short-II
Long-I
5.A.
Long-II
5.B.
5.C.
A1
A1
A1
A1
A1
A1
Fig. 5. Series of snapshot from times 0ns (top) and 14ns (bottom) of each simulation. (5.A.) Short-II, (5.B.) Long-I, (5.C.) Long-II. Colored by secondary structures.
Comparison of RMSF analysis in all three simulations reveals that elongation of TIM44 has stabilizing effect on the A1 helix; residues SER234-LEU242 in Long I/II has low RMSF values, while residues LYS220-GLY232 remain flexible with high RMSF values. Other areas of TIM44 appear to be flexible – LYS380-ASP390; ASP405-GLU415. Both these areas contain long extended strands, thus accounting for the higher RMSF values. The same trends were observed in the Long PLM I/II simulations as well (Figures 6, 7). We include snapshots of TIM44 to show the initial and final positions of A1 and TIM44 structure. It is seen that A1 folds, partially or fully, upon itself (Figure 5). Thus it is concluded that the main body of TIM44's C-Domain remain relatively unchanged, while the A1 helix seems to retain flexibility of movement.
10
C.1. Adsorption of Palmitate molecules to the TIM44 Three simulations of TIM44 in physiological solution were initiated, each simulation including four palmitate molecules. Comparison of RMSD analysis of TIM44 in water and TIM44 with palmitate simulations shows great similarity; even though some fluctuations can be observed the range of RMSD values (Figure 6.A) and overall behavior is still maintained. The separate RMSD of A1 and TIM44's main body of these simulations (Figure 6.B1-3) also maintains the same similarity to the TIM44 in physiological water (Figure 4.A-C), thus it is inferred that in these simulations the palmitate molecules do not induce major conformational changes in TIM4. In both Short and Elongated proteins it can be seen RMSD values of the main body reaches a plateau after ~2ns of the simulation. RMSD of Yeast TIM44 backbone C-Domain, PLM simulations :Short, Long I, Long II PLM
6.A 1.2 1.1 1 0.9
RMSD [nm]
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
frame [ps] Short PLM
Long I PLM
Long II PLM
Fig. 6.A. Comparison of RMSD of each simulation – Backbone of TIM44 and simulations Short-PLM, Long I PLM, Long II PLM. 6.B.1
RMSD of Yeast TIM44 backbone C-Domain, Short PLM A1 Vs. Protein Body
RMSD of Yeast TIM44 backbone C-Domain, Long I PLM A1 Vs. Protein Body
6.B.2
0.9
0.9
0.9
0.8
0.8
0.8
0.7
0.7
0.7
0.6
0.6
0.6
R M S D [n m ]
1
R M S D [n m ]
1
R M S D [n m ]
1
0.5
0.5
0.4
0.5
0.4
0.4
0.3
0.3
0.3
0.2
0.2
0.2
0.1
0.1
0.1
0
0 0
2000
4000
6000
8000
10000
frame [ps] Short PLM A1
Short PLM Body
12000
14000
16000
18000
20000
RMSD of Yeast TIM44 backbone C-Domain, Long II PLM A1 Vs. Protein Body
6.B.3
0 0
2000
4000
6000
8000
10000
12000
frame [ps] Long I PLM A1
Long I PLM Body
14000
16000
18000
20000
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
frame [ps] Long II PLM A1
Long II PLM Body
Fig. 6.B. Comparison of RMSD of each simulation – Backbone of A1 Vs. Backbone of protein body. The body was defined as all backbone amino-acids except A1's. (7.B.1.) Short PLM. (7.B.2.) Long I PLM (7.B.3.) Long II PLM
11
20000
Comparison of RMSF (Figure 7.) of the palmitate simulations shows that, like in the water simulations, elongation of TIM44 by 14 amino acids contributes to the stability of A1 in similar areas (SER246-GLN236). Another common behavior to the Long I and Long I PLM simulations is the folding of A1 upon itself. RMS Fluctuations of Yeast TIM44 α-carbon C-Domain, PLM simulations: Short Vs. Long II
7. 1.5 1.4 1.3 1.2 1.1
RMSD [nm]
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 220
234
244
260
300 Short PLM
α-carbon frame [ps] 340
Long I PLM
380
420
Long II PLM
Fig. 7. Comparison of RMSF of each simulation – Backbone of TIM44 and simulations Short-PLM, Long I, Long II. Residues of Short A1 are marked with purple arrow; Long A1 residues with a blue arrow.
Ligand Contact map is presented to illustrate "hotspots" of palmitate contact with TIM44. The list of contacts revealed that the carboxyl moiety has a tendency to occur mostly near arginine, lysine, and to some extent serine/glutamic acid and several other less frequently occurring a.a. (amino acids). In spite of being purely statistical, several specific a.a. stand out in a manner that should require a more rigorous analysis (Table 2, Figure 8). Table 2. Each column includes the total time carboxylate moiety was less than 0.6nm from any given amino acid in either Short-PLM, Long PLM I, Long PLM II simulations. Last column includes the Total Time. Note that this is only partial list, as the carboxylate moiety is found near other residues as well but to a marginal extent. AA are ordered in descending order, starting from the left. Amino acids that stood out in further analysis were marked in bold. AA ASN259 ASN323 ARG255 SER237 SER234 LYSH260 ARG342 LYSH307 ARG272 SER275 THR268 GLN236 LYSH241
Short PLM [ps] 2003 1825 3179 1858 3072 60 2411 2252 438 500 0 1842 954
Long I PLM [ps] 2 3202 0 1721 74 1192 258 0 0 0 1858 4 808
Long II PLM [ps] 3261 14 470 0 0 1428 0 0 1757 1477 1 0 0
Total time [ps]
AA
5266 5041 3649 3579 3146 2680 2669 2252 2195 1977 1859 1846 1762
LYSH330 LYSH333 SER288 LYSH256 GLU247 VAL324 TYR377 LYSH239 LYSH355 SER285 GLY308 ASN248 ILE235
12
Short PLM [ps] 331 785 0 0 1084 8 0 991 8 0 928 113 908
Long I PLM [ps] 899 395 1166 660 8 992 4 0 932 936 0 808 0
Long II PLM [ps] 1 5 0 493 5 68 1025 0 0 1 0 0 0
Total time [ps] 1231 1185 1166 1153 1097 1068 1029 991 940 937 928 921 908
A3 A3 A1
A2 A1
A2 A4
A5 A6
Figure 8. TIM44 was colored as function of number of contacts (Short PLM, Long PLM I/II simulations) near different amino acids. Top: Cartoon representation Bottom: SURF representation. Coloring goes from white (marginal/no occurrence) to dark-red which represents high occurrence. Ligand Contact Legend [#contacts] 0
1-250
251-500
501-1500
1501-2500
> 2500
13
C.1. Long PLM I 9.A
Min Distance of PLM 210 Carboxy from Long TIM44
420 395 370
[#AA]
345 320 295 270 245 220 0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
frame [ps]
9.B
Elongated Yeast TIM44 with PLM, 1st run, Changes in Coulomb and LJ againt time of Protein to PLM1-residue210
20 -30
Energy [KJ/mol]
-80 -130 -180 -230 -280 -330 0
2000
4000
Coul-SR:Protein-PLM210
6000
8000
LJ-SR:Protein-PLM210
10000
12000
14000
16000
18000
20000
tim e [ps]
Figure 9.A S.A.A.P. graph for palmitate molecule in Long I PLM TIM44 simulation. Distance Legend: red: 0-0.3nm, orange: 0.3-0.6nm, purple > 0.6nm. Figure 9.B Energy profile of Coulomb (SR) and Lennard-Jones potential (SR), both generated using GROMACS g_energy algorithm.
Figures 9.A, B. show S.A.A.P. representation of one palmitate molecule complemented by its energy profile throughout the simulations. The energy profile shows analysis of Coulomb and LJ potentials between TIM44 and a given palmitate molecule during the simulation. From these figures it can be seen that once a carboxylic head reached contact-range ( 0.6nm.
theMin whole Distance of PLM207 Carboxy from Long TIM44 - 2nd run
420
395
370
Figure 14.B Energy profile of Coulomb (SR) and LennardJones potential (SR), both generated using GROMACS g_energy.
[#AA]
345
320
295
270
245
220
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
frame [ps]
Figure 13.B
Elongated Yeast TIM44 with PLM, 2nd run, Changes in Coulomb and LJ againt time of Protein to PLM207
20
-30
Energy[KJ/m ol]
-80
-130
-180
-230
-280 0
2000
4000
Coul-SR:Protein-PLM207
6000
8000
LJ-SR:Protein-PLM207
10000
12000
time [ps]
19
14000
16000
18000
20000
2726ps
2726ps
A2
A1
A1
N259
N259 R255
R255
T258
K256
A2
K260
T258
K256
K260
Figure 14. Snapshot of Long PLM II simulation at 2726ps. PLM207 in cyan, interacting exclusively with A2 helix (facing parallel to A2) and with the conserved R255, K260 amino acids. The snapshot is repeated using both the Cartoon and the SURF representations. Ligand Contact Legend [#contacts] 0 1-250 251-500 501-1500 1501-2500 > 2500
simulation (~2-14ns), it is worth noting due to the involvement of the preserved R255 and K260. The above snapshots (Figure 14) depict PLM207 of Long-PLM-II simulation. It can be seen that the A2 helix adsorbed the palmitate molecule quite effectively, and five a.a. appear to be the main players of this event – R255, T258, K256, K260 and N259. The palmitate molecule interacts almost exclusively with A2 for most of the simulation (Figure 13.A, ~1.7ns-14ns). The energy profile (Figure 13.B) shows correlation with the S.A.A.P. analysis where interaction with TIM44 dissipates after the carboxylate moiety is released (Figure 13.A,B). Similarly to the Short PLM and Long PLM I, the interaction is also in the A1/A2 vicinity but limited to the A2 helix, which again points to the A1/A2 vicinity as the area responsible for interacting with negatively charge lipids.
20
Discussion Several simulations of TIM44 were initiated with and without palmitate molecules. All simulations supports the C-Terminal domain of TIM44 is stable and undergoes no major conformational changes throughout any of the simulations. The stability of the main body of the C-terminal is demonstrated in Figures 4 and 6, where it is seen that RMSD values for the main body in all simulations converge to the same low range of values (0.3-0.4nm). This conclusion is also supported by experimental results by Weiss et al., (2) through the use of limited proteolysis to establish stable folding. The A1 helix area shows to be of some rotational degrees of freedom in both Short and Long protein simulations. Long simulations also show folding of A1 towards its base. With little information about the N-terminal domain structure of TIM44 it is unclear whether this behavior can take place with the whole protein, though flexibility of A1 is consistent with the work done by Josyula et al. (4). Thus, so far these studies do not reveal major contradictions between our simulated model and S237 data available as to the publication date of this work. experimental 6.6Å Palmitate molecules showed preference to the A1/A2 base area in all simulations with 3.6Å residues involved and analysis of Long I PLM suggests that the A1/A2 area several 4.4Å specific
might adsorb K355 to a negatively charge molecule such as cardiolipin. The interaction shown (Figure 9.A,B TIM44\PLM210) utilizes a large number of amino acids with weak positive electrostatic charge that directs the palmitate molecule to its final location at the base of A1/A2, where it is yet surrounded by a large number of positively charged amino acids to facilitate strong binding of the molecule. This may serve to direct a negatively charged molecule to bind at a certain site of TIM44 and also determine the orientation in which TIM44 may be bound to membrane or ensure a specific positioning of the C and N domains. In addition, the Short PLM simulation shows a bilayer-like formation of palmitic acids in the A1/A2 base (Figure 11) that stays stable throughout the rest of the simulations and interacts with the aforementioned helices. The Long PLM II simulation shows a palmitate molecule to interact almost exclusively with A2 helix for much of the simulation time (Figures 13, 14). Thus it is concluded that all three PLM simulations suggest that the area of A1/A2 may stably interact with parts of a negatively charged membrane even while it is yet unclear as to the mechanism by which this interaction occurs. These results appear consistent with the mentioned preliminary results by Avneon et al. Also, all three simulations show that in each case of palmitate binding there are a number of amino acids involved in varying degrees of intensity, and therefore we suggest 21
that the mechanism of binding utilizes a large number of weak interactions with several amino acids to achieve stable binding to the membrane. Suggestions for further studies to unravel the role of A1\A2 in this mechanism may include taking the A1/A2 helices sequence and incorporate them in vitro into the sequence of a known non-membrane-binding protein, thus testing if the A1/A2 element introduces membrane binding capability to the protein. Also possible are mutations of in the A1/A2 sequence, where each mutant consists of several mutations to key amino acids presented in this work (R233, S234, S237, K241, T268, K355). Since several amino acids may be involved, it would be necessary to mutate several of them if any change in binding strength to membrane is to be observed. This procedure however should be carefully designed with proper controls as multiply mutations may induce conformational change in the protein, thus it may be the change to A1/A2 structure, rather than change in electrostatic charge of the helices that would disrupt TIM44's binding to the membrane. This may also serve to demonstrate the importance of the A1/A2 element in membrane-binding rather than the caveola. In the theoretical arena a better analysis of the actual mechanism and key amino acids involved may be achieved through simulations of TIM44 with negatively charged membranes such as cardiolipin, phosphatidylserine or POPG (palmitoyloleoylphosphatidylglycerol). These simulations may provide valuable data that would enable the design of other, and more specific, experiments to unravel the actual mechanism of TIM44 binding to the matrix side of the inner-membrane. Of the aforementioned membranes PDB structure is available only for palmitoyloleoylphosphatidylglycerol (16) by means of replacing certain elements of a preequilibrated POPC bilayer. Further MD effort should be directed either for simulations of TIM44 with POPG membrane or modeling in similar manner a workable cardiolipin membrane. We therefore conclude that while simulations with palmitate molecules have yielded some valuable information, the next step should be MD of negatively charged membrane which may allow new insights to the mechanism.
2606ps
22
References (1) Van der Laan M., Rissler M and Rhling P., Mitochondrial preprotein tranlocases as dynamic molecular machines, FEMS Yeast Res 6 (2006) 849-861 (2) Celeste Weiss, Wolfgang Oppliger, Guy Vergeres, Rudy Demel, Paul Jeno, Martin Horst, Ben de Kruiff, Gottfried Schatz and Abdussalam Azem, Domain structure and lipid interaction of recombinant yeast Tim44, Proc. Natl. Acad. Sci. USA Vol. 96, pp. 8890–8894, August 1999. (3) David van der Spoel,Erik Lindhal, Berk Hess, Gerrit Groenhof, Alan E. Mark, Herman J. C. Berendsen, GROMACS: Fast, Flexible, and Free, J Comput Chem 26: 1701–1718, 2005. (4) Ratnakar Josyula, Zhongmin Jin, Zhengqing Fu and Bingdong Sha, Crystal Structure of Yeast Mitochondrial Peripheral Membrane Protein Tim44p C-terminal Domain, J. Mol. Biol. (2006) 359, 798–804. (5) Avneon Y., Tsfadia Y. and Gutman M. Molecular Dynamics Simulations of Palmitate Anion Binding to the Human TIM44 C-Terminal Domain (yet unpublished as to this date). (6) Stewart A. Adcock and J. Andrew McCammon, Molecular Dynamics: Survey of Methods for Simulating the Activity of Proteins, Chem. Rev. 2006, 106, 1589-1615. (7) More data can be found in the GROMACS Manual at http://www.gromacs.org/ (8) H. J. C. Berendsen, J. P. M. Postma, W.F. van Gunsteren, A. DiNola and J.R. Haak, Molecular Dynamics with Coupling to an External Bath, J. Chemical Physics, 81 (8), 15 October 1984. (9) Essman U., et al., A Smooth Particle Mesh Ewald, J. Chemical Physics 1995 103: p. 8577-8592. (10) Software can be found at http://www.ks.uiuc.edu/Research/vmd/ (11) Berendsen, H.J.C, et al., Interaction Models for Water in Relation to Protein Hydration. Nature 1969. 224: p. 175-177. (12) Tsafadia Y., Friedman R., Kadmon J., Selzer A., Nachliel E. and Gutman M., Molecular Dynamics Simulation of Palmitate Entry into the Hydrophobic Pocket of the Fatty Acid Binding Protein, FEBS Letters 581 (2007), 1243-1247. (13) Yokoyama S. et al., Structure of the Human Tim44 C-terminal Domain in Complex with Pentaethylene Glycol: Ligand-Bound From, Biological Crystallography, Acta Cryst. (2007). D63, 1225-1234. (14) Protein DataBank at http://www.rcsb.org/pdb/ (15) SPDBView Software can be downloaded from Expasy - http://www.expasy.org/ (16) Wei Zhao, Tomasz Róg, Andrey A. Gurtovenko, Ilpo Vattulainen and Mikko Karttunen , AtomicScale Structure and Electrostatics of Anionic Palmitoyloleoylphosphatidylglycerol Lipid Bilayers with Na1 Counterions, Biophysical Journal 92:1114-1124 (2007). (17) Roman Safonoff, Abdussalam Azem, Analysis of the Interaction of the Mitochondrial Protein TIM44 with Model Membranes, 2007 (Thesis work, yet unpublished).
23
