Wednesday, December 12, 2018

NMRlipids IV: Toward submission of the manuscript about PS lipids

I believe that the manuscript about PS lipids contains now all the essential content (see also the supplementary information). The main results are:
  • None of the tested models reproduces the PS headgroup order parameters within experimental accuracy but the best models suggest that the carboxyl group in serine headgroup does not rotate freely.  
  • Cation binding to bilayers containing PS lipids is overestimated in all the tested simulations, except in MacRog with potassium and CHARMM36 with the NBfix for the calcium. However, the latter underestimates the binding affinity.
  • The qualitative response of PS lipid headgroups to the bound calcium and dilution with PC lipids do not agree with experiments in any of the tested models, indicating that the force field development is necessary for MD simulation studies of PS lipids and their interactions with other biomolecules.
There is a todo list in the manuscript, mainly containing missing details. Current contributors, please check the list (also in the supplementary information). One question regarding many points in the todo list is that how much simulation details we should put in the supplementary information when the data is shared in Zenodo? When the details are well described in Zenodo (for example here), it may not be necessary to repeat all the information in the SI. However, cases where the data is not much described in Zenodo but the repository contains all the files having all the information about simulations (for example this) are more complicated. Are there any opinions about this? 

In addition, all kind of comments are welcomed, also from the people who have not contributed to the manuscript. At this point, I hope to have comments from as many people as possible, including also critical comments. 

If you have major concerns about scientific content, I hope to hear about that before the end of January 2019. If there are no major concerns, I hope that we could submit the manuscript by the end of February 2019. 

Wednesday, December 5, 2018

Correlation times of C–H bond direction in different force fields

We are starting a small project to see how well the correlation times of C–H bond vectors are reproduced in different force fields. We aim to simply follow the analysis of the R1 and τeff times as was done earlier by Tiago, Samuli, et al. for the Berger force field, but for all the different force fields available at the Zenodo repository of NMRlipids.

The plan is to not make this an official NMRlipids project, but still to do it fully openly; to this end, we warmly welcome anyone interested to follow the progress and to participate on GitHub!

Leftin and Brown have published experimental Rtimes for many different lipids in various temperatures and at various field strengths. Concerning the experimental τeff times, in Tiago's paper they are reported for POPC at full hydration and 298 K, and they have also been published for DMPC at low hydration and 300 K. However, as many of the trajectories at the Zenodo repository have DPPC, the experimental τeff times for DPPC at full hydration would be warmly welcomed.

Beyond force field comparison, possibly interesting physical questions to look from the data would be the effects of hydration, salts, and cholesterol on C-H bond dynamics. A possible extension could also be to look at the temperature dependence.

Hanne Antila and Markus Miettinen

Thursday, September 20, 2018

NMRlipids III: Quantitative measure for the force field quality needed

Progress in the NMRlipids III project about lipid-cholesterol interactions has been slow because the focus has been recently in improving ion binding to PC lipid bilayers and charged membranes.

After revisiting the manuscript with a serious intention to finalize the project, I think that we need to define a quantitative measure for the force field quality to simplify the discussion. For example, Berger model gives the best agreement with form factor data with high cholesterol content, but too large order parameters. On the other hand, Slipids give better order parameters with and without cholesterol but the form factor with high cholesterol concentrations is less good, and so on. This kind of discussion could be significantly simplified with a quantitative quality measure for the force field quality.

The quality measure could be also used to rank the force field quality in the databank collected from the contributions to the NMRlipids project and in further automatic force field development. The simplest measure to start with could, for example, sum up the deviation from the experimental order parameters for different segments and the deviation from experimental form factor using equation (3) from the SIMtoEXP publication. Similar measure has been recently introduced for proteins in solution.

Any kind of ideas and contributions about measuring the force field quality are welcomed.

Tuesday, September 18, 2018

NMRlipids IV: Challenges in evaluating counterion binding affinity to PS bilayers

Counterion binding to POPS lipid bilayers is signifincantly different between simulation models, as seen from density profiles along membrane normal in figure 1.
Figure 1:    Counterion densities of POPS lipid bilayer along the membrane normal from simulations with different force fields.

In the NMRlipids IV manuscript, we are trying to figure out which one these is the most realistic. In the NMRlipids II project we evaluated the sodium binding affinity to PC lipid bilayers using the changes of headgroup order parameters and electrometer concept. Similar experimental data is available for POPC:POPS (5:1) mixtures as a function of KCl, NaCl and LiCl concentration. This data is compared with different simulations in figure 2.
Figure 2:     Changes of the PC (left) and PS (right) headgroup order parameters as a function of added NaCl, KCl and LiCl from POPC:POPS (5:1) mixture at 298 K (except Berger simulations are (4:1) mixture at 310 K).
Rough conclusions could be that the binding of sodium and potassium is overestimated in Berger and CHARMM36 models, while Lipid17 performs better. Changes in MacRog model with potassium are somewhat overestimated but not systematically. However, these results are more difficult to interpret than the corresponding figure for PC lipids in the NMRlipids II, because zero point of the x-axis (added salt concentration) is not free of ions in this case. The systems always contain counterions when PS lipids are present and the binding affinity of these counterions differ between simulation models. Therefore, the changes of order parameters in figure 2 cannot be plotted against completely ion free state and the interpretation is more complicated.

The changes of PC headgroup order parameters with increasing amount of PS lipids give additional information about counterion binding affinity. According to the electrometer concept, the headgroup order parameters increase with increasing amount of negatively charged PS lipids, as seen in experiments and MacRog simulations with potassium in figure 3.
Figure 3:  Changes of PC (left panel) and PS (right panel) headgroup order parameters from POPC:POPS mixtures with increasing amount of POPS.
However, POPC headgroup order parameters are unaffected by POPS in CHARMM36 simulations with potassium and sodium counterions, and decrease in Berger simulations with sodium. This can be explained by overestimated counter ion binding in these simulations, which cancel the negative charge by the PS headgroups.

In conclusion, the results would suggest that sodium clearly overbinds in Berger simulations, both sodium and potassium slightly overbind in CHARMM36 simulations and most realisistic binding to PS containing bilayers is observed in MacRog simulations with potassium. The case of Lipid17 is little bit unclear because the relatively strong counterion binding in Figure 1 is not seen in Figure 2. The Lipid17 results are yet missing from Figure 3 because we do not have pure POPC simulation with Lipid14 at 298 K.

ToDo: My current suggestion is to do following. We conclude from this data that the MacRog simulations with potassium probably represent the most realistic counterion binding affinity to PS containing bilayers and compare its density profile to the results from other simulations in Figure 1. To finalize this part we still need the following data:
  1. Pure POPS MacRog simulation with potassium counterions at 298 K
  2. Pure POPC Lipid14 simulation at 298 K
  3. Ion density profiles from Lipid17 POPC:POPS (5:1) simulations with different ion concentrations for figure: (we already have the data in Zenodo in Amber format but we need to calculate the density profiles.)
Maybe we should also move current figures 7 and 8 to the supplementary information.

Fortunately, the calcium binding results seems to be more clear because changes in both simulations and experiments are larger.

Wednesday, July 4, 2018

NMRlipids IV: First draft of the manuscript about PS lipids

Thank you again for the 168 comments in the two previous blog posts about NMRlipids IV project (NMRlipids IV: Headgroup & glycerol backbone structures, and cation binding in bilayers with PE, PG and PS lipids and NMRlipids IV: Current status and reorganization of the manuscript) and numerous contributions to the GitHub repository.

First draft of the manuscript summarizing the results related to PS lipids is now compiled. As discussed previously, the goal is to first finish the manuscript about PS headgroup and then progress with PG and PE results separately. In addition to the up-to-date manuscript (pdf, tex) and supplementary information (pdf, tex), also the figures and data are available in the GitHub repository.

I believe that the manuscript already contains most of our conclusions, but several issues need to be addressed before we can start to prepare the submission. Most important current issues are listed here, accompanied with the names of persons who can hopefully help me to resolve the issues.
  1. We need slightly more detailed description of the NMR experiments by Tiago Ferreira. Also the results and discussion about the experiments, including Figure 1, needs to be polished.
  2. Atom names and dihedral notations should be made consistent between the chemical structures in figure 2, and dihedral angles distributions in  figures S6 and S7 by Pavel Buslaev. I am not sure if we could just the atom names currently used in figure 2, or if we need more labels. Slight polishing of figures S6 and S7 is also needed.
  3. The subjective ranking criteria in figure 4 by Markus Miettinen should be probably improved, see issue #4 in the GitHub repository.
  4. CHARMM36 data from POPC:POPS (5:1) mixtures with added NaCl would be highly useful for figure 6. CHARMM36 gives two maxima in the counterion density profiles in figure 5, while for example Lipid17 gives only one. From the right column of figure 6 we could see if this difference is reflected to the order parameter response of PS headgroup to the excess counterion concentration. Data from Gromos-CKP would also be useful for this. If someone has such data or is willing to generate it, let us know. 
  5. CHARMM36 simulations with the added CaCl2, but without the new NBfix, mentioned by Jesper Madsen would be useful for figure 9
  6. Lipid17 simulations ran with the Amber package, mentioned by Batuhan Kav, would be useful for figure 9
  7. Details of simulations by Thomas Piggot, Jesper Madsen, Fernando Favela, Batuhan Kav, Markus Miettinen, Josef Melcr and Matti Javanainen should be added in section S1 in the supplementary information (tex file here).
In addition, all comments regarding the current manuscript are welcomed. I will continue working with the manuscript and keep the the list updated.  

Thursday, April 12, 2018

New NMRlipids-related publication: Accurate Binding of Sodium and Calcium to a POPC Bilayer by Effective Inclusion of Electronic Polarization

One of the original goals of the NMRlipids project was to find a MD model that correctly describes cation binding to zwitterionic PC lipid bilayers. In the NMRlipids II project, we concluded that the currently available MD models typically overestimate cation binding and that none of them was accurate enough to capture calcium binding details to PC bilayers. Furthermore, the improved ion models available at the time were not sufficient to reproduce the correct binding behavior.

In early 2017, we started to develop a new MD simulation model of POPC in the group of Pavel Jungwirth. The goal was to improve one of the existing lipid force fields to reproduce the experimental Na+ and Ca2+ ion binding affinities to PC bilayers without destructing the ion-free bilayer properties. We assumed that cation binding could be improved by implicitly including the electronic polarizability to an existing lipid model by using the electronic continuum correction (ECC) [Leontyev et al. PCCP 13, 2613 (2011)]. ECC had been previously applied to ion parameters in Jungwirth's group by simply scaling the ion charges [Kohagen et al. J. Phys. Chem. B 118, 7902 (2014), Kohagen et al. J. Phys. Chem. B 120, 1454 (2016)]; we decided to extend this approach also to lipid models. Our original plan was to proceed using the open collaboration approach, that is, turn our initial development efforts into a NMRlipids project. Due to the surprisingly rapid progress, we, however, decided to finish the manuscript in the traditional way: It is now accepted to be published in the Journal of Physical Chemistry B [J. Melcr, H. Martinez-Seara, R. Nencini, J. Kolafa, P. Jungwirth, and O.H.S. Ollila J. Phys. Chem. B, DOI: 10.1021/acs.jpcb.7b12510]. All our data and files are available in the GitHub repository and in Zenodo (see e.g. DOI: 10.5281/zenodo.1118265), similarly to the NMRlipids projects.

Shortly, we used the Lipid14 parameters of POPC as a starting point and applied ECC to the headgroup and glycerol backbone atoms. The scaling factors of 0.8 and 0.89 were used for partial charges and LJ radii, respectively. Numerical values of the scaling factors were tuned to reproduce the experimentally observed calcium binding affinities as well as structural details without additional ions. Figure 1 shows comparison between the developed ECC-POPC model, Lipid14, and experiments. For more details, further analysis, data, and parameter files see the manuscript in press, the GitHub repository, and Zenodo.

Figure 1: Response of the headgroup order parameters (which is a direct measure for ion binding affinity, see NMRlipids IIto added CaClin the ECC-POPC model, in the Lipid 14 model, and in experiments.

We are currently running further tests for the developed ECC-POPC model and extending the development to negatively charged phosphatidylserine (PS) lipids. The development of ECC-PS lipids will be done in parallel with the NMRlipids IV project. The GitHub repository for the ECC-PS model development will be publicly available, but at least for now it is separated from the NMRlipids project core.