normal complex membrane simulation - THICKNESS PROBLEM

  • audagnotto
  • audagnotto's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
9 years 2 months ago #4542 by audagnotto
Dear all,

I am new in CG simulations and I am trying to simulate a complex membrane composed by several lipids characterized by high number of insaturation (like in the Plasma Membrane example present in your web site).
I ran 4 equilibrations, decreasing restraints on phosphates at each stages (1000, 500, 250 and 2) but when I checked the thickness (distance between phosphates) of the membrane I observed that is around 36 A. From my atomistic simulations the mebrane thickness is around 38 A.
This is a problem when I inserted my TM-protein in the membrane in order to study its stability.

For what concerns the parameters I followed the Plasma Membrane example in the web site (md-production.mdp) and I tried to change both the lincs and the tau_p but without any results.

I will really appreciate any help
Many Thanks
Martina

Please Log in or Create an account to join the conversation.

More
9 years 2 months ago #4544 by mnmelo
Hi Martina,

You're talking about a 0.2 nm difference in a total 3.8 nm, or 0.1 nm per leaflet. At a first glance I'd say this is well within the coarseness scale of Martini. Your protein must be exquisitely structured to feel such a difference.

You don't mention what the problem was when your protein was inserted. Do you get system instabilities? Does the protein pop out of the membrane? Maybe your system will perform satisfactorily despite that small thickness difference.

Let us know,
Manel

Please Log in or Create an account to join the conversation.

  • audagnotto
  • audagnotto's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
9 years 2 months ago #4552 by audagnotto
Replied by audagnotto on topic complex membrane simulation - THICKNESS PROBLEM
Hello,

thanks for your reply.
My system is composed by two alpha-helices embedded in this complex membrane.
These 2 helices form a dimeric structure. What I observed during the CG simulation was a bending of the helix and I am wondering that could be due to the thickness of the membrane.
Indeed when I compared the thickness between all-atom and CG i observed this small difference.
Do you think that I can impose restraints on the thickness of the membrane in order to reproduce the all-atom one?
If yes, how?
Many Thanks for your help

Martina

Please Log in or Create an account to join the conversation.

More
9 years 2 months ago - 9 years 2 months ago #4553 by mnmelo
Hi,

I wouldn't attempt to try and restrain lipids to enforce a particular thickness. Thickness is a result of the combination of several factors and you'll certainly introduce unforeseen bias into your system.

As to the bending you observe, I'm not sure I understand exactly what you mean. If you explain a bit further I might be able to help better:
Is this bending actually a tilting of the helix to better match the hydrophobic length of the membrane?
Or is it an actual bending where the axis of the helix becomes curved?
How much does it bend, and is that compatible with only a 0.1nm thickness difference per leaflet?
What other information about the system do you have in order to say that this bending is unrealistic? (For instance, did you start this CG system from an atomistic one and observe it departs from the mapped structure?)
Which Martini settings are you using to restrain secondary structure? Standard Martini parameters for helices or elastic networks?
If you're using elastic networks, did you network each helix separately, or are they meshed together?

From here we can find out the best way to continue with your system.
Let me know,
Manel
Last edit: 9 years 2 months ago by mnmelo.

Please Log in or Create an account to join the conversation.

  • audagnotto
  • audagnotto's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
9 years 2 months ago #4554 by audagnotto
Replied by audagnotto on topic complex membrane simulation - THICKNESS PROBLEM
Hi,
so for what concern the questions:
1) "Is this bending actually a tilting of the helix to better match the hydrophobic length of the membrane? Or is it an actual bending where the axis of the helix becomes curved? "
Since I am simulating just the TM domain the helices try to stay inside the membrane tilting themselves.
Actually the bending is the result of the hydrophobic mismatch which causes a bending of the alpha-helix axis. So I observed both the two situations you describe in the questions.

2)"How much does it bend, and is that compatible with only a 0.1nm thickness difference per leaflet?"
I used bendix to calculate the bending and the value is around 50 degree.

I guess this bending is unrealistic because I started from an NMR structure in micelles where the two helices were straight.
In all my atomistic simulation I never observed this kind of behaviour.

3) "Which Martini settings are you using to restrain secondary structure?"
I used both the elastic network ad the standard martini. When I used elastic network, I network the 2 helices separately and I simulated for more than 300 ns but what I observed is a high rigidity of the structure. That is the two helices didn't change from the starting point. Since I want to study if the dimeric structure is stable in this membrane I suppose that the elestic network constraints too much this simple conformation and I decided to try the standard martini where I observed the bending of the dimer.

Thanks a lot

Please Log in or Create an account to join the conversation.

  • audagnotto
  • audagnotto's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
9 years 2 months ago #4555 by audagnotto
Replied by audagnotto on topic complex membrane simulation - THICKNESS PROBLEM
Sorry I forgot also to tell you that my membrane system is asymmetric and there is also cholesterol.

Please Log in or Create an account to join the conversation.

More
9 years 2 months ago #4557 by mnmelo
Hi again,

That's an interesting situation. From what I understand, elastic networks provide too rigid a structure, whereas standard Martini readily becomes bent. I'm assuming you are matching your expectations on rigidity by comparison to atomistic behavior; let me know if you're not.

Just to be clear, why do you say the elastic network version is too rigid? Does the atomistic run bend a little and this CG one not? If that's the case you can reduce the force constant of the elastic bonds to try and find a sweet spot.

On the other hand, if by rigidity it means the helices don't separate, well, you may be finding that their interaction is too strong to sample in a couple hundred nanoseconds. It might even lie outside the reach of Martini. For accurate such measurements you'll need some sort of multidimensional restricted sampling (like a multidimensional umbrella sampling). If you decide you should go that way I point you to Periole et al., Structural Determinants of the Supramolecular Organization of G Protein-Coupled Receptors in Bilayers as an example.

Now, relative to the standard Martini bending: you mention a 50º bend. That seems to me a bit sharp. Doesn't that bend cause more than the 0.2 nm thickness difference in length? There may be other reasons for it to happen. A proline in the sequence is an obvious one (do you have one in your sequence?). And does the bend occur evenly over the helix length (a curved helix) or at a single point? (a broken helix with two straight segments).

Let me know,
Manel

Please Log in or Create an account to join the conversation.

More
9 years 2 months ago #4559 by mnmelo
One more thing that might also be biasing your observations:

In your asymmetric system how do you ensure that both leaflets have matching lateral pressures?
You must make sure both individual compositions and lipid quantities equilibrate to the same area. To make matters worse cholesterol will flip across the membrane and cause imbalances in pressure, so you must take that into account when deciding how many lipids to have in which leaflet.

We have dealt with that in Ingólfsson et al., Lipid Organization of the Plasma Membrane .

Cheers,
Manel

Please Log in or Create an account to join the conversation.

  • audagnotto
  • audagnotto's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
9 years 2 months ago #4561 by audagnotto
Replied by audagnotto on topic complex membrane simulation - THICKNESS PROBLEM
Hello,

about the rigidity that I observed using the elastic network, I meant that the helices are both straigth and not separeted. Since the separation of the dimer is an event that happens during the first 30 ns in atomistic simulation I was expected to observe it immediately at CG level. So I suppose that the elestic network works perfect for big proteins but not for a small alpha helix and I decide to try to use the normal Martini. Do you think that is better to work with elestic network changing the force constant of the elastic bonds?

I don't have prolines in my protein and the bend occurs at single point of the helix. I guess that the bending of the protein is more likely a consequence of this thickness, rather than the cause. I also observe when I compute the thickness of the membrane (considering only the phosphate) that the leafleat with higher cholesterol concentration is the one that shift, decreasing the thickness of the membrane. Maybe as you told me before the flip of cholesterol could have some influence on the thickness of the membrane.
All these CG simulation were performed using the same parameters(semiisotropic, tau p,compressibility, etc..) you provided for the plasma membrane work of Ingólfsson et al.. I am wondering if the compressibility value is not the rigth one for my system. If yes, how I can choose an other one?
I also read in the mdp file you provided for plasma membrane that you apply a restraint of 2 kJ/(mol nm^2) along the z position for POPC and PIP. Why you choose just this two lipids? Shall I have to apply the same force constant to POPC for my system?

Could you also clarify what you mean by "ensure that both leaflets have matching lateral pressures"?

many thanks for all your help
Martina

Please Log in or Create an account to join the conversation.

More
9 years 1 month ago #4581 by mnmelo
I see your problem better now.

Regarding your observations with elastic networks:
Overall you shouldn't expect a precise matching of dynamic properties between Martini and atomistic models. It is the global thermodynamic behavior (namely, partitioning) that is best matched.
Furthermore you say your helices diverge after just 30ns of atomistic simulation. That points to an obviously unstable initial structure, which is a poor reference point for comparison between models.

In this particular case it might be that the elastic network Martini is stuck in a local energy minimum. If you do observe that the atomistic system undergoes some unfolding of the helices prior to dissociation, then that's a good candidate event that's blocking Martini progression. Depending on the magnitude of secondary structure change it might be more or less feasible to recreate it with Martini.

Regarding the bent helices:
I still find odd a single point break in a helix that should be rather stiff (even in the presence of large hydrophobic mismatches). Could you send me your protein .itp to This email address is being protected from spambots. You need JavaScript enabled to view it. so I can take a look? Also indicate where the helix breaks.

For your questions on the asymmetric thickness: let's suppose your patch is 20 by 20 nm. How many lipids do you decide to put on each leaflet? That's not a straightforward decision. You should have on each leaflet the amount of lipids that yield a 20x20nm^2 average area. If you had a symmetric system this would not be important as long as you had the same number of lipids on both leaflets: if you had fewer than required for 20x20nm^2 the system would just equilibrate to a smaller area.
Now imagine that in your asymmetric system you have the right amount of lipids on the top leaflet but too few on the bottom one. The system will equilibrate to a midpoint where the top leaflet is more compressed than it should, while the bottom one is too stretched out. This is probably your case right now.

How to do it properly (part I):
-Simulate two symmetric bilayers (separately) with the composition of the top and bottom leaflets;
-Measure the equilibrium area per lipid;
-Calculate how many lipids from each mixture to have in the asymmetric bilayer so that the areas match.

Having cholesterols increases the complexity of this process: cholesterols readily flip and an asymmetric bilayer presents a chemical potential difference between leaflets, which means cholesterols will likely equilibrate towards one of them. This offsets your carefully crafted matching areas from part I. What to do now? (part II):
-Simulate the asymmetric bilayer and let the cholesterols equilibrate;
-Count the new lipid proportions on each leaflet (different because cholesterol amounts changed);
-Repeat part I with these new lipid proportions;
-Repeat part II/part I until there is no more significant change in cholesterol proportions between leaflet.
This was the procedure we used for the plasma membrane paper.

Regarding your question on the soft restraints on a small number of the lipids: we used those in order to keep our very large membrane patch mostly planar, avoiding undulations that'd require too large a simulation box.
Unless your system is so large that you face the same issues you shouldn't really worry about this.

Please Log in or Create an account to join the conversation.

Time to create page: 0.109 seconds