Correction to Molecular Dynamics Simulations of Phosphatidylcholine

Correction to Molecular Dynamics Simulations of Phosphatidylcholine Membranes: A Comparative Force Field Study. Thomas J. Piggot, Ángel Piñeiro*, an...
0 downloads 15 Views 1015KB Size
Download PDF
Erratum pubs.acs.org/JCTC

Correction to Molecular Dynamics Simulations of Phosphatidylcholine Membranes: A Comparative Force Field Study Thomas J. Piggot, Á ngel Piñeiro,* and Syma Khalid* J. Chem. Theory Comput. 2012, 8 (11), 4593−4609. DOI: 10.1021/ct3003157 S Supporting Information *

I

t has recently come to our attention that the analysis tool used to calculate the deuterium order parameters in our original work1 produced incorrect order parameters for a small part of the analysis that was performed. In particular, it did not calculate the correct order parameters for unsaturated carbons (C9 and C10) within the oleoyl tail of POPC lipids for the united-atom force fields. We note here that this is not just an issue within our tool but a more widespread problem among many different tools available for analyzing united-atom lipid membrane simulations. A shortly forthcoming work will describe this in greater detail, including the validation of a corrected version of the GROMACS g_order analysis tool used herein (modified from a version kindly provided by Reid Van Lehn2). This tool performs the analysis using the methods as detailed by Douliez et al.3 It is worth noting that this analysis approach assumes idealized 120° angles around the double bond. This assumption makes no noticeable impact upon the order parameter results for any of the united-atom force fields studied in the original work apart from the GROMOS 43A1-S3 force field. However, as will be described in further detail in our forthcoming work, we believe that the corrected results reported herein determined using these assumptions produce a reasonable representation of the order parameters for this force field. In addition, this approach exactly matches the method used in the original POPC order parameter analysis with this force field.4 Repeating the order parameters analysis of the united-atom POPC simulations using this validated analysis tool results in substantially different order parameters for C9 and C10 of the oleoyl tail in several of the original simulations. Within this work we provide corrections to both the order parameters reported in the original work and to any incorrect conclusions and recommendations made from these erroneous results. Furthermore, we wish to apologize for these mistakes within the original analysis, particularly regarding the GROMOS 43A1-S3 POPC force field, which we would now recommend for use. Finally we wish to stress that the overwhelming majority of the results and conclusions originally presented are still correct and remain valid.

Figure 2. (Center) Deuterium order parameters for the sn-1 (*) and sn-2 (Δ) chains of the lipids. Experimental deuterium order parameters are also shown in the plot for both chains of POPC at 300 K (◊) and (○) [Seelig et al.5] and only for the oleoyl chain at 296 K (•) [Warschawski et al. Figure 36] and at 310 K (+) [Warschawski et al. Figure 26].

Furthermore, the description of the order parameters for this force field on Page 4598 “The largest disagreement with the experimental properties is the deuterium order parameters of carbons 9, 10, and 11 of the sn-2 chain (Figure 2B). The order parameters of these carbon atoms, which are located next to or within the double bond of the oleoyl chain, were substantially larger (carbon 10 is ∼5 times larger) than the experimentally determined values. These results, interestingly, are in contrast to the previously published order parameters for POPC with this force f ield. However, we believe that this dif ference is due to the problems associated with using the GROMACS program g_order to calculate the order parameters around the double bond (see the Methods for f urther details).” should now be replaced with “The experimental deuterium order parameters for this force f ield are also in good general agreement with the experimentally determined values (Figure 2). However, there are some discrepancies, particularly with carbons 10−12 of the sn-2 chain being somewhat larger than the experimental values. These results are also in good agreement with previously published order parameters for this force f ield.4” GROMOS 53A6L Force Field. The second set of results presented in the original work are for the GROMOS 53A6L force field simulations. In the Results text on Page 4598, the sentence

■

CORRECTION TO THE RESULTS GROMOS 43A1-S3 Force Field. The first force field described within the results of the original work is the GROMOS 43A1-S3 force field and is where the most dramatic of differences arise using the corrected analysis tool. Figure 2, presented below, now shows the replacement center panel of Figure 2 from the original work. © 2012 American Chemical Society

A

DOI: 10.1021/acs.jctc.7b00244 J. Chem. Theory Comput. XXXX, XXX, XXX−XXX

Erratum

Journal of Chemical Theory and Computation “The order parameters of carbons 9, 10, and 11 around the double bond of the sn-2 chain of POPC are closer to the experimental values than for the GROMOS 43A1-S3 force f ield.” should now be replaced with “The order parameters of carbons 9, 10, and 11 around the double bond of the sn-2 chain of POPC are slightly closer to the experimental values than for the GROMOS 43A1-S3 force field. The presented order parameters for carbon 10 are either very close to zero or negative in the plots (Figure S2; note that the f igures show − SCD), which is in contrast to the originally published order parameters with this force f ield.7 We believe that this dif ference is due to the problems associated with methods used to previously calculate these order parameters (Piggot et al., ‘On the Calculation of Acyl Chain Order Parameters f rom Lipid Simulations’, manuscript in preparation).” We note here that a corrected version of Figure S2 is provided in the Supporting Information of this work. Berger Force Field. The next results described in the original work are for the Berger force field. Figure 3 of the original work, which demonstrates the order parameters using different force field parameters, should now be replaced with Figure 3 presented below.

GROMOS 53A6 Kukol (and GROMOS-CKP) Force Field. The final united-atom POPC results presented in the original work were for the GROMOS 53A6 Kukol force field and modifications to the force field, including the so-called GROMOS-CKP (which stands for Chandrasekhar-KukolPiggot) POPC parameters. Figure 4 (bottom) of the original work, should now be replaced with Figure 4 presented below.

Figure 4. (Bottom) Deuterium order parameters for the palmitoyl (circles) and for the oleoyl (triangles) of three POPC bilayers that were obtained using the original GROMOS53A6 Kukol force field (red), the same force field but without the extraneous dihedrals in the glycerol region (blue), and with the GROMOS-CKP parameters (black).

In the original work it was demonstrated that using the original Kukol parameters, both with and without extraneous dihedrals in the glycerol region, results in order parameters that are in substantial disagreement with experiment, particularly for carbons 11 and 12 of the oleoyl tail. This is still the case upon reanalysis of the order parameters, which is not surprising given the fact that the original tool only calculated the order parameters of the unsaturated carbons incorrectly. Removal of the extraneous dihedrals in the glycerol region and reversion of the double-bond dihedrals to use the standard GROMOS parameters (i.e., the GROMOS-CKP parameters) continues to result in substantial improvements over the 53A6 Kukol POPC parameters. Therefore, only the following minor changes to the results need to be made: On Page 4600 the text “In particular, the largest deviations were for carbons 11 and 12, located just af ter the double bond in the sn-2 chain, where the deuterium order parameters were negative (∼−0.05 and −0.04 for carbons 11 and 12, respectively). Simulations were also performed in which, as for the Berger force f ield, extraneous dihedrals in the glycerol region of the POPC topology were removed (simulations KP2a and KP2b) (see the Methods for more details). However, unlike the Berger force f ield (see above), these modif ications did not substantially impact upon the deuterium order parameters of either the sn-1 or sn-2 chains. The modifications did, however, impact upon AL, VL, and DHH for this force field (Figure 4).” should be replaced with “In particular, the largest deviations were for carbons 10, 11, and 12, located in and just af ter the double bond in the sn-2 chain, where the deuterium order parameters were negative (∼−0.04 for all three of these carbons). As per the Berger force f ield, simulations in which the extraneous glycerol region dihedrals were removed resulted in a more negative order parameter for C10. These modif ications also impacted upon AL, VL, and DHH for this force field (Figure 4).”

Figure 3. Deuterium order parameters obtained for the palmitoyl (circles) and oleoyl (triangles) chains of POPC calculated from three trajectories that were obtained using the original Berger force field (red), the Berger force field with the extraneous dihedrals in the glycerol region were removed (blue), and, additionally, with the van der Waals parameters of the LOS oxygen atoms were modified (black) as described in the Methods section.

As demonstrated in this new figure, the same pattern of results remains as per the original work. In particular the extra glycerol dihedrals improve the double bond order parameters. However, the order parameters of C10 with this force field are now in further disagreement with the experimental values than previously reported. Therefore, only the following minor change needs to be made to the Results section for this force field. On Page 4599, the text “the only substantial dif ferences were observed in the deuterium order parameters (Figure 3), in particular for carbon 10, located in the double bond of the oleoyl chain, where the deuterium order parameters were negative (∼−0.03) using the corrected topology.” should be replaced with “the only substantial dif ferences were observed in the deuterium order parameters (Figure 3), in particular for carbon 10, located in the double bond of the oleoyl chain, where the deuterium order parameters were more negative (∼−0.06) using the corrected topology.” B

DOI: 10.1021/acs.jctc.7b00244 J. Chem. Theory Comput. XXXX, XXX, XXX−XXX

Erratum

Journal of Chemical Theory and Computation

f ields were closer to the experimentally determined values for the deuterium order parameter of this carbon.” should now be replaced with “Experimentally there is a characteristic drop in the deuterium order parameter for carbon 10 of the oleoyl tails. This drop in the deuterium order parameter at carbon 10 was reasonably well reproduced for all of the force f ields, despite some dif ferences being observed (Figure 6). In particular, the size and sign of the order parameter for carbon 10 dif fered between force f ields. The greatest drop was with the Berger force f ield (to a value of ∼−0.03) and resulted in the largest disagreement with the experimentally determined values for the deuterium order parameter of this carbon.”

and on Page 4601 “apart f rom carbons 9−11 of the sn-2 chain. As mentioned above, the overall trends in the order parameters were not substantially altered” should be replaced with “apart f rom carbons 10−12 of the sn-2 chain. The overall trends in the order parameters were not substantially altered, despite the worsening of the experimental agreement for C10.” Simulations with Optimal Parameters. Finally, for the Results section of the original work, Figure 6 showing the order parameters from the simulations with optimal parameters should be replaced with Figure 6 presented below.

■

CORRECTION TO THE DISCUSSION On Page 4604, in the section regarding the order parameters with the Berger force field “Removal of the dihedrals in the glycerol region resulted in a negative order parameter for carbon 10, located in the double bond of the oleoyl chain. This, somewhat unexpected, impact meant that the order parameters for this carbon atom were f urther away f rom the experimental values.” should be very slightly modified to “Removal of the dihedrals in the glycerol region resulted in a more negative order parameter for carbon 10, located in the double bond of the oleoyl chain. This, somewhat unexpected, impact meant that the order parameters for this carbon atom were even f urther away f rom the experimental values.” On Page 4606, in the section discussing order parameters with the GROMOS 43A1-S3 “For POPC membranes and the GROMOS 43A1-S3 force f ield, the deuterium order parameters of the carbons in the double bond of the oleoyl chain were in substantial disagreement with the experimentally determined values. Therefore, despite the other properties of the membrane being in a good agreement with the experimentally determined values, we would recommend caution in using this force f ield when simulating POPC membranes.” should be now changed to “For united-atom POPC membranes, despite some dif ferences in order parameters between force f ields, most of the force f ields produced reasonable order parameters, and therefore we consider most of these force f ields acceptable for use with respect to this experimental property. The primary exception to this is the GROMOS 53A6 Kukol parameters, which should be avoided. The standard Berger POPC parameters should also be used with caution due to the relatively large negative order parameters for C10 of the double bond.”

Figure 6. Deuterium order parameters obtained for the sn-1 (*) and sn-2 (△) chains of the lipids from one of the replicas of the simulations with optimal parameters for the five force fields.

■

Furthermore, the following Results section regarding these order parameters described on Pages 4603−4604 “Experimentally there is a characteristic drop in the deuterium order parameter for carbon 10 of the oleoyl tails. This drop in the deuterium order parameter at carbon 10 was reasonably well reproduced for all of the force f ields apart f rom the GROMOS 43A1-S3 force f ield (Figure 6). For this force f ield, there were lower order parameters for carbons 8 and 11, with slightly higher order parameters for carbons 9 and 10. Slight dif ferences in the deuterium order parameters for these carbons with the other force f ields were also observed. In particular, the size of the order parameter for carbon 10 dif fered between force f ields. The smallest value was with the Berger force field (to ∼0.01), with the largest for the GROMOS 53A6L and GROMOS 53A6 Kukol/GROMOSCKP force f ields (to ∼0.06). The CHARMM36 and Berger force

CORRECTION TO THE CONCLUSIONS

On Page 4606, the sentence “In particular, this applies to the GROMOS 53A6 POPC parameters of Kukol, 1.0 nm cut-of fs (without dispersion correction) for DPPC with the Berger force f ield, the GROMOS43A1-S3 POPC parameters, and the use of the standard TIP3P water model with the CHARMM36 force f ield in GROMACS.” should now be replaced with “In particular, this applies to the GROMOS 53A6 POPC parameters of Kukol, 1.0 nm cut-of fs (without dispersion correction) for DPPC with the Berger force f ield, the standard Berger POPC parameters, and the use of the standard TIP3P water model with the CHARMM36 force field in GROMACS.” C

DOI: 10.1021/acs.jctc.7b00244 J. Chem. Theory Comput. XXXX, XXX, XXX−XXX

Erratum

Journal of Chemical Theory and Computation

■

ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jctc.7b00244. Figures S2 and Figures S5 in the Supporting Information should replace the corresponding figures presented in the original Supporting Information of this work. We note that only the order parameters for C9 and C10 in the oleoyl tail of the united-atom POPC simulations have changed in all of these plots (PDF)

■

REFERENCES

(1) Piggot, T. J.; Piñeiro, Á .; Khalid, S. J. Chem. Theory Comput. 2012, 8, 4593−4609. (2) Van Lehn, R. C.; Alexander-Katz, A. J. Phys. Chem. B 2014, 118, 12586−12598. (3) Douliez, J.-P.; Ferrarini, A.; Dufourc, E.-J. J. Chem. Phys. 1998, 109, 2513−2518. (4) Pandit, S. A.; Chiu, S.-W.; Jakobsson, E.; Grama, A.; Scott, H. L. Langmuir 2008, 24, 6858−6865. (5) Seelig, J.; Waespe-Šarčević, N. Biochemistry 1978, 17, 3310−3315. (6) Warschawski, D.; Devaux, P. Eur. Biophys. J. 2005, 34, 987−996. (7) Poger, D.; Mark, A. E. J. Chem. Theory Comput. 2010, 6, 325− 336.

D

DOI: 10.1021/acs.jctc.7b00244 J. Chem. Theory Comput. XXXX, XXX, XXX−XXX