Can you please tell us more about the NMR-based methodology for drug design INPHARMA? Are pharmaceutical companies adopting it and what is the impact to the industry?
While I was in Goettingen, we were approached by pharmaceutical companies who would have a series of drug leads and would want to know how those drug leads bind to a certain receptor.
Half of the drug leads would crystallize with the receptor, so they could easily find how those bind, but half of them would not. They asked if we had a method that could help them with that. These receptors were usually large receptors, so it was not possible to do a de novo structure determination of the complex using classical NMR methods.
We therefore had to think of alternative routes and these drug leads are usually molecules that bind quite weakly to the receptor, with dissociation constants in the micromolar range. They exchange fast between the receptor-bound and the free state during an NMR experiment.
In this case, you can use a series of experiments, which were developed in the ’80s, called transferred NOE (nuclear Overhauser effect), which actually look at the ligand only; the ligand that keeps a memory of its state when it is bound to the receptor.
If the exchange is fast enough between the receptor-bound and the receptor-free state, the ligand keeps a memory of the receptor-bound state while it is in its free state. By looking at the ligand in its free state, you can interrogate the system about its memory on the bound state.
Now, these transferred NOEs can give you the conformation of the bound ligand, so, basically can tell you which shape the bound ligand adopts when binding to the receptor. However, this cannot tell you how the ligand binds the receptor, so which orientation it is in, in the binding pocket of the receptor, so which part of the ligand interacts with which part of the receptor. For this, we thought of an alternative method.
This resulted in InPharma, which uses two ligands that bind to the receptor in alternation; either one ligand binds or the other ligand binds, so they continuously exchange their position in the binding pocket of the receptor.
Now, when the first ligand binds to the receptor, it leaves some information there; it leaves an imprint of itself before leaving the receptor. The second ligand then comes in and learns about the first ligand, by reading out the information that the first ligand has left there.
It is a bit like if you think of one person entering a room and leaving a note for you on the desk. Then you enter the room, take this note and therefore know the first person has been there. Now, what you know is that that person has been close to the desk because they had put a note there. By making the two ligands talk in this way, you know which ligand had been close to which part of the receptor by reading it out in the parts of the ligands that have been able to communicate among each other.
If person A and person B take a note that has been left on the desk, you know that person A and person B have been at the desk. If a person C and person D communicated through a note that had been left on the window, you know that person C and person D had been passing by the window. This is the same kind of information that we collect with InPharma.
We let the two ligands talk to each other through the binding pocket of the receptor. Then, we know which parts of the two ligands have seen the same part of the binding site of the receptor.
We do this with combinations of many ligands’ pairs. We use a bioinformatics program that we have written to extract this information from the NMR data. Combining this program with docking models, we can derive how these different ligands have bound to the receptors, so which parts of the receptor they have seen, and how they have communicated with the receptor.
This is done without any need to ever look at the receptor resonances. The receptor can be unlabeled; it can be just extracted from human material, from cells. We only need very small amounts of the receptor. The only things that we need to look at are the ligands.
It’s often not so straightforward getting a crystal structure of a ligand bound to a receptor, even if the receptor crystallizes easily. The ligand might not get into the crystal or it might get into the crystal in a different conformation or in a different orientation from the one that it does in solution. It might not be soluble enough or it might destroy the crystal when it is soaked in. There are different problems that can arise when trying to crystallize a receptor together with a ligand.
This is the reason people from the pharmaceutical company approached us and asked us to help with the fact that they had cases of not being able to get a crystal structure of the ligands with the receptor. By developing this methodology, we have found a way to get the binding modes of these ligands to receptors.
I think over the years we have been convincing those in management at several pharmaceutical companies that the method is good enough and is helpful. We are in the process of training some of our colleagues on the data interpretation. The data have to be interpreted through a bioinformatic effort and a program needs to be used. We have actually developed several versions of this program.
This project started in collaboration with Christian Griesinger at the MPI, so the MPI has a version of this program. We have other versions of the program. The methodology is slowly being picked up by the pharmaceutical industry and we have already seen the first posters reporting on InPharma data acquired without our help or advice, so we are very happy with it.
Would this be something that would replace the X-ray crystallography for everything, or just in cases where they can’t get into crystals?
First of all, InPharma needs a structural model of the protein in the apo state. We still need crystallography for that or any other modeling method or X-ray structures of homologous proteins.
Second, X-ray is definitely the fastest method for solving these kinds of questions, if the receptor crystallizes with your ligand. I’m not saying that we want to substitute or to take the place of X-ray crystallography completely; we are complementary, when X-ray crystallography doesn’t work.
The two methods can also be used in parallel, to check each other. It is always very good to have information from different methods. We have also seen cases where crystal structures were not reproducing what we saw in solution. This can happen, especially if you consider that when such a small ligand needs to bind to a large protein pocket, the ligand could adopt a preferred orientation that is just driven by the three-dimensional structure of the crystal, but not really by the most favorable interaction with the binding pocket.
You said that the X-ray is the faster method. What would be the time difference?
It’s difficult to say because it depends on how skilled you are in InPharma and in crystallography and how difficult it is to get a crystal of the complex. Once the crystal is there and if you are a skilled crystallographer, crystallography is much faster than InPharma, but first, you have to get a crystal.
So, you see some posters now for pharma companies who are able to
adopt and use your methodology on their own. Do you think that could
We hope so. We are pushing and promoting the product. When the methodology needs a complicated data interpretation, sometimes people have a bit of an energy barrier to overcome. We are working on allowing people to overcome this energy barrier. We hope that the method would then be used very often, or certainly on a regular basis, by the pharmaceutical industries.
How has this been used in pharma? Do you know what kind of drugs they screen and what sort of compounds they’re looking at?
Sometimes, the pharma company don’t show their compounds. They may show you spectra and the method, but not the compounds. The methodology has been used by us to look at all kinds of ligands, from small ligands, really small, organic molecules, to natural products. I’ve seen it used in the pharma companies, especially on small peptides.
How important do you think interdisciplinary approaches will be for the future of structural chemistry?
With the advances in biology, scientists have been discovering more and more molecular machines that have key functions in cells, and have characterized their composition. Of course, now, we want to understand how these machines behave; how they work; how they change their conformation during catalysis; how they are regulated; how we can interfere with their function, for example, in a disease situation; how we can boost their function; how we can correct some of the errors that might be due to genetic defects and so on.
For this reason, we need to understand how these molecular machines work and we need to understand how their structure changes over time, as they perform their function. Of course, you can take snapshots of different structures using X-ray crystallography and this has been the classical way of studying how enzymes work, but this requires being able to trap those machines in the different states. By methods in solution, you can actually watch these molecular machines working in real-time if the reaction is slow enough on the NMR time scale or if you are able to slow it down by playing with temperature or other conditions. This is really unique to NMR in solution.
Of course, if you aim to look at molecular machines of high molecular weight, meaning several hundred of kilodaltons, you are not able to solve their structure de novo by NMR only. We are now in the position of looking at proteins of very high molecular weights in solution using methyl groups as spies that tell us how the conformation of these molecular machines changes during time or what is happening in an active site of an enzyme, for example. We are not able to solve a complete structure.
To determine the structure of large complexes, we first need the high-resolution structure of pieces of these complexes. We can then put these pieces together with the NMR restraints that we can collect watching the methyl groups of proteins. NMR is very good at zooming in and at looking at the molecular details of certain parts of the enzyme; however, we need to complement NMR with other techniques in solution such as small-angle neutron scattering, which are able to report on the conformational changes of the molecular machine on a global scale. If we combine the two kinds of information, we really have a very powerful method for solving and depicting the, let’s say, conformational life of the enzyme during its function, during its activity.
The two NMR holy grails are sensitivity and resolution. Personally, I’m always more inclined towards putting resolution on my wish list. Since we work with large molecular machines and RNA, the resolution is really an issue. Large magnets do both; they serve both purposes, so if I had to put something on my wish list, I’d definitely put big fields on it.
Dynamic nuclear polarization (DNP) is a very interesting development in terms of sensitivity, especially for solid-state NMR. In my view, it is not yet mature enough to be used easily by every lab in the world. We have carried out some DNP in collaboration with other laboratories and we’ve got very promising results, especially on the RNA. We don’t feel confident about operating the DNP machine by ourselves and in developing all the experiments and adjustments that would be necessary to really optimize the technique for RNA.
However, it is definitely something that I would like to see become a broadly applicable technique that non-experts could also use. There is, of course, also the cost issue with DNP. The running costs are very high, which could probably not be afforded by every university. For me, really, the most important thing is big fields.
Can probes play a factor in this at all? Can you use probes to improve the resolution?
You can get better sensitivity with probes. If you’re only interested in sensitivity, this is a way to go. If you are also interested in working with big molecular machines that have a complex spectrum or you work with RNA with many overlapping resonances, resolution is really an issue and the quality of the spectra changes enormously by changing the magnetic field. There is no way around that.
Can even small increments show a big improvement?
Yes, even small increments in field sometimes make the difference between being able to resolve and get information from a very overlapped region and not being able to. For example, if you run an RNA spectrum at 700 or 900 MHz, you would see a huge difference for some resonances, especially in the ribose region. It doesn’t need to be an extraordinary increment; progressive changes or improvements are important for us.
Does mass spectrometry play a role?
Mass spectrometry is picking up in its role in structural biology. There are several applications of mass spec which we would also be interested in using. We try to do this mostly in collaboration, just because you can’t have the expertise in-house for everything, so you need colleagues.
One development of mass spec is the detection of cross-links that allows you to get information about protein–protein or protein–RNA distances in big complexes.
People are using these kinds of restraints to calculate structure. It’s a little bit like running a paramagnetic relaxation enhancement experiment, where you get distances between two specific positions of two proteins. NMR with paramagnetic relaxation enhancement is much more powerful because you get the distances between your paramagnetic tag and all the methyl groups, for example, of your partner proteins, so you get many more distances. The distances are much more precise and therefore more reliable.
I would not substitute the information that you can get by paramagnetic relaxation enhancement with mass spec, but I would use it as a complementary method, so you can look at distances, for example, between lysine side chains, which are not something that you can look at with the methyl group correlations that we use in NMR for large molecules.
Another interesting aspect of mass spectrometry is the so-called native mass spectrometry, where you can, for example, look at complex stoichiometry and complex equilibria in bulk. This is also something that we are starting to do in collaboration. It is nice to have, but it’s not critical to what we do.
The component that is really key to what we do is small-angle neutron scattering, which unfortunately, can only be done at neutron reactors and there are only a few of those in the world. Europe is in quite a good position for that. At least we have access to the one in Munich and the one in France, but apparently, in the US, it’s quite difficult to do small-angle neutron scattering for biological samples. I hear that this technique is not very accessible there.
Is this complementary to NMR?
It is complementary. The information that you get with small-angle scattering is very low resolution. With neutron scattering, you can use a technique which is called contrast matching, which gives you, in combination with the deuteration of certain components of your complex, many more pieces of information than the small-angle X-ray scattering. It’s more powerful in my view, but the information stays on the global shape of the complex or of sub-components of your complex. It will never give the high-resolution, atomic information that NMR gives.
It’s also not where the research is going at the moment. Research is going towards interdisciplinarity and looking at things from different angles and different aspects, so we are definitely moving away from an “one machine does it all” or “one expertise does it all” attitude. You have to know about everything.
Do you see that as the general trend – the need for an interdisciplinary approach?
Yes, definitely. Especially if you want to publish in high-impact journals; you need to have this interdisciplinary aspect.
Nowadays, we are especially interested in exploring the possibilities that solid-state NMR offers in terms of looking at the RNA part of our complexes. In solution, we can look at large proteins or protein complexes because proteins have methyl groups and methyl groups can still be seen at very high molecular weights, but RNA is different. RNA doesn’t have any methyl groups and there are actually no chemical groups that are easily seen from the RNA part of molecular machines of several hundreds of kilodalton in solution.
When we studied the structure of the RNP methylation enzyme, we could watch the proteins in action, because we had the methyl groups of the proteins. However, we had one part of the complex, namely the RNA part, which remained obscure to us and silent.
I actually always like to compare NMR spectra with the molecule talking to us. You let the reaction start in the NMR tube and the molecule talks to you by shifting the peaks and telling you: “I’m here… I’m doing this, I’m doing that.”
The RNA wouldn’t talk to us. The reason for that is because there is no real chemical moiety in the RNA molecule that is still observable in molecular machines of high molecular weight. All resonances have just become too broad to be observed. The way we somehow got information on the NMR conformation is through small-angle neutron scattering with contrast matching, and through biochemical assays, but we would really love to be able to see the RNA talking to us and reporting on its state and on what it’s doing at specific steps during the reaction.
We believe that solid-state NMR might be the only technique that is able to give us this information. That’s why we embarked on developing a methodology to watch the RNA in solid-state NMR. When we started this project, there had been only very few things done on RNA in solid-state NMR. We had to develop the methodology to assign the NMR resonances, collect structural restraints and solve the structure of the RNA by solid-state NMR, which we did successfully.
Now, we are progressing to the next step, namely, looking at the RNA in the context of molecular machines. We have obtained really beautiful spectra of the RNA component of our large RNP methylation complex with excellent line widths and we are really excited about the possibility of now watching the RNA in the complex using solid-state NMR.
About Prof. Dr. Teresa Carlomagno
Prof. Dr Teresa Carlomagno studied chemistry in Naples, Italy. She did her PhD partly in Naples and partly in Germany, in the group led by Christian Griesinger in Frankfurt, where she also stayed on for a post-doc. Then, she moved to The Scripps Research Institute in the US to work together with James Williamson.
With Christian, Prof. Dr. Carlomagno learned all the NMR and the methodology development side of NMR. With James Williamson, she wanted to learn the wet-lab part and she was particularly interested in RNA structure at the time. RNA structure was just coming up. People mostly liked to focus on proteins, but she wanted to learn how to produce RNA and how to look at RNA interactions with proteins, so she moved to the Williamson lab where she spent two years as a post-doc.
Then, she took up a group leader position at the Max Planck Institute in Goettingen. From there, she moved to the EMBL, where she also had a group leader position for about seven years. Since summer 2015, Prof. Dr Carlomagno has been in Hannover, where she holds a professorship. She’s also associated with the Helmholtz Centre for Infection Research in Braunschweig.