Welcome to my projects. The first part of this page describes the general area in which I work, and some work that has been published. The second part describes some current projects. For the projects listed, I am looking for PhD students and Postdoctoral fellows. The list is not complete. In addition, if you would be interested in working with me on something that is broadly related, I would be most interested to hear from you! For further information, please email me.
A pure bilayer is a simple model for the many different membranes of complex composition that surround cells and organelles in living creatures. The real membranes are a mixtures of a variety of lipids and proteins. Generally, these lipids have a phospholipid headgroup and two long hydrocarbon tails. The details of the headgroup and the tails determine which type of lipid we are dealing with. In model systems, the headgroup is usually phosphatidylcholine (PC) or phosphaditylethanolamine (PE), and tails generally lauroyl (L), myristoyl (M), oleoyl (O) or palmitoyl (P) chains. Combining these elements, we get lipids like DPPC (di-palmitoyl-phosphatidylcholine), DMPC (di-myristoyl-phoshatidylcholine), POPC (palmitoyl-oleoyl-phosphatidylcholine) and DLPE (di-lauroyl-phosphatidylethanolamine). Obviously, in real life things are far more complicated and a biochemistry person will be able to use many more types of lipids, but for now it is hard enough for us to deal with these four different types.
The cool picture gives an impression of what a model bilayer looks like from our point of view. The gray atoms in the middle are the carbon atoms from the tails of the lipids. The blueish spheres are the carbons from the PC headgroup, the red spheres are oxygens and the yellow spheres are phosporous atoms. The white/red speckled noise are actually watermolecules, drawn at a smaller scale so they do not cover all of the headgroups of the lipids. It should be clear now where the word 'bilayer' comes from, a membrane has two layers of lipids that face each other with their tails.
Simulating simple bilayers like the one in the picture has been a popular pastime in the 90's. Dr. Egberts in the group of Herman Berendsen was the first to simulate DPPC in full atomic detail, published in a thesis in 1988. In 1991/1992 a number of studies of DLPE, DPPC and DMPC appeared in the literature. Nowadays many studies are available, but we are still argueing about details of how to do it. There are a lot of parameters in the models we use, there are different options for many parts of the calculations and the verification of the results from simulations using experimental data is not straight forward. One of the first things I did as a PhD student was worry about all these details, and I even tried out a few different things. I hope it contributed a little to resolving some of the problems but on the grand scale (grand in the area of lipid simulations) I do not think more simulations of DPPC will increase our understanding of bilayers much.
Back to top
Lipids are funny molecules. The ones I look at have a large water-loving headgroup, and a long water-hating tail. You can argue about the accuracy of that statement from a physical point of view, but it is easy to think about the behaviour of lipids when you think about them in those terms. What happens when you put a molecule with a water-loving headgroup and a water-hating tail in water?
The answer is simple: the molecules will try to find a way to get as many of the headgroups as possible to be in contact with the water they love, while trying to avoid contacts between the tails and water. The bilayer in the picture above is an example where that worked perfectly: all the tails are in contact with others tails, but not with water. All headgroups are in contact with water. In the bilayer, this worked great because of the relative size of the headgroups and the tails. What happens when the tails are much smaller, and the headgroup the same? If you would make a bilayer there would be a lot of empty space inside. Nature does not like vacuum, as ancient scientists already said, and there are good physical reasons for it in most cases, including this one. In this case, the lipids will form more or less spherical structures, so called micelles.
A micelle is a somewhat spherical aggregate of lipids. They are not as obviously important in biological systems as bilayers are, but they are very important in other, more technical applications. Because micelles have an interior that consists of tails that do not like water, but are lined on the outside with headgroups that do like water, they easily dissolve in water. If some molecule that does not like water is thrown in water with micelles, chances are excellent that the molecule will be much happier inside the micel than in the water layer, and it will find a way to end up in the interior of the micelle. Micelles provide a way to dissolve molecules that do not like water, in water (this happens to be the way soap works too. Coincidence? No :). The use of micelles to solubilize oil is not so interesting for us, after all we are interested in understanding biological processes more than in cleaning up oil. But, there are at least two good reasons to look at the micelle in the picture.
One reason is that the lipids used look very much like the DPPC lipids from the bilayer. They have the same headgroup, but only have one tail instead of two, and that tail only has a length of 12 instead of 16. Nonetheless, from our model point of view, they are very similar and the models and methods used for the bilayer can also be applied to the micelle, to test and compare a number of properties.
A second reason is that we are interested in the structure and
dynamics of a number of proteins that interact with bilayers. As
mentioned above, a real cell membrane is a complex mixture of
lipids and proteins. To study parts of such a complex mixture it is
important to isolate what you want to study so you know that any
observations you make can only apply to that one component and
cannot be caused by some unknown ingredient. A very powerful method
to study the structure of proteins on a membrane is NMR, but
unfortunately the NMR people have not figured out yet how to
directly study proteins bound to a membrane. What they can do
however is study proteins (peptides = small proteins) that are
bound to micelles, like ours. In principle, I could study the same
peptide using my computer simulations and compare its behaviour
when it is bound to the micelle to that when it is bound to the
bilayer. If there are large differences, the NMR people will have
to start worrying about how interesting their experiments are when
they use micelles. And if there are no large differences, it seems
reasonable to use micelles as a model. In practice it will not be
so drastic; if there are differences between how a peptide acts on
a bilayer and on a micelle you can just take those differences into
account when interpreting data from experiments on micelles. But it
would be nice to know if there are significant differences.
A third reason why I like micelles is simply that they are
somehow fascinating. If you add or remove more molecules to the
aggregate the aggregate will adapt and change size and shape. In a
way, a micelle is a big molecules that consists of many small
parts. How these parts behave individually and together is
fascinating.
Back to top
The first membrane protein I simulated was a porin. Porins are large holes that sit in the outer membrane of certain classes of bacteria. They form a barrier against pathogens but allow small molecules such as ions, sugars, water and amino-acids to flow in and out of the cell. They are of interest for a number of reasons, but one of the main reasons for me to study them was that they are probably the simplest and best studied membrane protein that we know the structure of, in great detail.
This picture is based on a crystal structure, but we have also done simulations of the whole trimer in a lipid bilayer. This is quite a complicated system, with hundreds of lipids, ca. 13000 water molecules and several dozens of sodium ions.
Porin occurs as a trimer, a protein made up of three identical units. These units are colored differently in the picture. The green spheres are the headgroups of the lipids, the thin green wires their tails. Also highlighted are the aromatic amino acids in the protein, which form a clear band marking the boundaries of the protein inside the lipid bilayer. We analysed their behaviour in more detail, described in ref. 11.
General info about alamethicin, use as model ion channel, voltage dependent, antimicrobial peptide
More pictures, including Alm, Alm helix in bilayer, Alm on surface, picture from thesis explaining believed model
Unilever was interested in developing and selling tomato sauce with a longer shelf-life after opening the bottle. Failing that, a better desinfectant soap would have been good too. Naturally, Unilever turned to theoretical physicists with these problems. The goal of my previous project in Groningen was to study certain peptides that exhibit strong anti-fungal activity. By adding such peptides to tomato sauce the sauce wouldn't spoil as fast as it does now. We studied two model peptides, examples of a very large class of amphipathic alpha-helical peptides. One of them was Dermaseptin B, a well known peptide isolated from frog skin. Because frogs tend to be wet and soggy, they need a natural defense against fungi to prevent them from rotting. The other peptide was one designed by Unilever, called MB21.
Unfortunately, we did not manage to develop a new tastier tomato sauce with a longer shelflife. However, from a biophysical point of view (simulation techniques, for instance), the project did yield some interesting results. Now I just need to actually summarize them and write them up. Someday.
I'll also put the structures of the peptides, amphipathicity and some other stuff here
Although I've mainly worked with porin, alamethicin, and the peptides dermaseptin B and MB21, I have been involved in simulations of several other membrane proteins, including Aquaporin, the nicotinic receptor and Influenza A M2 channels. More information about these is available on the homepages of the rest of the Sansom group.
Back to top
This section is more technical. One of the problems in simulating membrane proteins is that we need to treat not only a protein and water, but also a lipid bilayer. This requires an accurate description of lipids, in terms of our molecular models, as well as a good description of the interaction between proteins and lipids. Ideally, we need one set of parameters that describes protein, lipids, and solvent at the same time. Currently many forcefields have different parameters for lipids and for proteins, even though chemically they contain the same type of atoms. This should not be necessary, but there are practical reasons for doing it this way. Developing a unified forcefield with as few atomtypes as possible is still desirable. There are several forcefields already, but we are working on a unified forcefield based on GROMOS96, a forcefield that leaves out some details that are expensive to simulate and that we think we can do without.
A major problem with molecular dynamics simulations in general is that they take massive amounts of computer time but only simulate a limited timescale, at the moment routinely up to about 100 ns, or 0.1 microseconds. Statistical mechanics provides the theory to calculate much slower events in some cases, and applying this theory to interface systems like our lipid bilayers is one of my interests.
Although our membrane models are getting better, including membrane proteins, lipids, ions and solvent, there are always things missing. One of the most interesting additions would be an accurate, or even a rough but useful, treatment of the transmembrane voltage difference that exists in real cells. This is of particular interest for proteins that are activated by such a transmembrane voltage. My pet peptide, alamethicin, actually is sensitive to the presence of a transmembrane voltage, and I am using this peptide to try different approaches to incorporating in a rough way a transmembrane voltage. Even though physics puritans might cringe at the idea, so far it seems to give useful results in studies of ion channels and alamethicin.
Having a model in which nearly all the atoms are present is nice and dandy, but it also means it will take our computers weeks and months to do the calculations we want to do. Because large parts of these calculations are actually uninteresting details, there must be ways to simplify them. An obvious way is to leave out details that are uninteresting and focus on details that are interesting. However, as logical and simple as this sounds, so difficult it is to simplify our complicated models in a useful and accurate way. This too is one of the things I find interesting, and it actually links to almost all my other interests. I haven't done any actual work on this, but others have. The general trend seems to be that things get very complicated very soon, but some encouraging results have been obtained, both on very simple models and on realistic molecules such as alamethicin.
Back to top
Anion channels are responsible for transport of mainly chloride ions in cells. There is currently no detailed structural knowledge about eukaryotic anion channels. However, several channels have been studied in great detail with other techniques, yielding enough information to consider modeling approaches. In particular, the Cystic Fibrosis Transmembrane Regulator and the human ClC channels are of interest. In addition, alamethicin has been engineered by Dr. Woolley and his group in Toronto to become anion selective instead of its native cation selective state. We collaborate with him and Dr. Sansom in Oxford on understanding the origin of the anion selectivity in this model channel, and attempt to improve its selectivity. Work in this area will include modeling, bioinformatics, molecular dynamics and Brownian dynamics simulations, and continuum electrostatics calculations.
How peptides aggregate inside membranes is an open question. Recently, several experimental studies have begun to address this question, with intriguing results. Molecular dynamics simulations will be used to study the aggregation of a number of model peptides in simplified membrane environments.
Antimicrobial peptides generally disrupt the membranes of a target cell, causing lysis of the cell. How this occurs, and what determines the activity and selectivity of these peptides, is currently only known approximately. Recently, several peptides with unusual folds with strong antimicrobial activity have been identified, and their solution determined by NMR. We will study these peptides using simulation methods, in collaboration with the NMR group of Dr. Vogel in Calgary.
Molecular dynamics simulations are nice, but they also have a few problems. One interesting problem is to get related properties like partition coefficients, distributions of molecules in interfaces, and binding free energies from simulations. The theoretical background for such calculations is well developed, but there have been only a few practical applications to membrane related problems. This project is divided in many smaller projects, aimed at testing and improving simulation and modeling of the thermodynamics of interactions of proteins and small molecules with lipid bilayers and simplified model membranes. It involves molecular dynamics simulations, free energy and potential of mean force calculations, development of mean field models for membranes, and force field development and testing. Part of this work will be in collaboration with the group of Prof. Mark in Groningen (and the developers of Gromacs).
If any of these projects (or something that looks vaguely related) interests you, please email me at tieleman@ucalgary.ca.
Back to top
last modified: 12 November 2000