Solvation Science alive:


Posted on
This image illustrates how a cosolvent molecule interacts with a specific domain of a pharmaceutical antibody called Trastuzumab.

Improving drugs with computer simulations

This image illustrates how a cosolvent molecule interacts with a specific domain of a pharmaceutical antibody. Antibodies are large and complex proteins that are part of our immune system and can be very effective drugs. They are often depicted in simplified models like the cyan structure shown here, which highlights the key structural elements. The blue bubbles depict the locations where the cosolvent molecules, here arginine, interact with the antibody domain.

Antibodies – our powerful body-police

Over many years of evolution, animals (including humans and our ancestors) developed complex and effective methods for protecting against harmful bacteria and viruses. Antibodies are a crucial part of the immune system and are incredibly powerful because of their high specificity. It's like having police officers in your body who specialize exclusively in one type of criminal. Trastuzumab, the antibody to which the domain in this image belongs, acts like a top detective targeting "bad" cells in certain types of breast cancer. It can identify and eliminate these specific harmful cells with great precision (and does not make false accusations) but is useless in identifying other harmful sources. This way, our bodies have specific task forces for different threats (of course, this is simplified). Because of the specificity, antibodies are also very promising for drug design.

The role of solvation science in drug design

Antibody-based pharmaceuticals are often administered in liquid form, intravenously. This means that patients cannot just take the medicine on their own. For both patients and medical professionals, it would be preferable if the patient can self-administer the drug. Self-administration of antibodies, however, is only viable for sub-cutaneous injection, that is injection with a small needle under the skin, which most diabetes patients use for administration of insulin. The maximum amount liquid that can be injected this way is low, which means with the dosages required for antibody-based treatment, the antibody concentrations need to be high. High concentrations can cause antibodies to clump together, losing their function, and make the liquid viscous, which makes it hard to inject. Imagine putting honey into a syringe and trying to press it through the thin needle.  To reduce viscosity in antibody formulations, cosolvents like arginine are added, but the exact way how arginine accomplishes that is not understood. Thus, solution conditions of antibody formulations are often optimized with trial-and-error methods. Imagine good conditions as needles that are hidden in a haystack. The search for the needles can take a very long time. If we would understand why for example addition of arginine to an antibody solution is so effective, we maybe could design formulations that are more effective than any haystack needle ever found so far.  

When effective cosolvents are found with trial-and-error methods, we only observe that the properties of a solution improved after addition of the cosolvent. These properties are macroscopic and it's hard to gather data on specific interactions at the protein-residue level. Advances in technology now allow us to use computer simulations to study small protein systems. These simulations, called classical molecular dynamics (MD) simulations, use basic physics to calculate the natural behavior of the system. They create a trajectory, showing atomic positions over time, like a movie. The image above is one of the many ways that a trajectory can be analyzed.

Using computer simulations to study the effects of cosolvents

So, how do you get a picture like the one above? First things first, you need a protein structure. Scientists typically share these structures in what's called the protein data bank. It's like an online library where you can download these structures. These files hold details about every atom's type and location in 3D-space. To prepare the system for simulation, you start by setting up a space around the protein. You throw in molecules you're interested in, like cosolvents or ions, and fill the gaps with water molecules. Now, you've got to describe how these atoms interact with each other using what's called a "force field." It's like giving rules for a game - determining exactly how atoms play together.

Once you've got that sorted, you assign velocities to all the atoms based on temperature. Think of it like setting the speed for each atom - hotter temperatures mean faster movement, and cooler temperatures mean slower. Then, you simulate how these atoms move and interact with each other in the next moment in time.

The image we're talking about results from a simulation that involves around 200,000 atoms. For each atom, you've got to calculate its interactions with every other atom. It's like playing a massive game of tag where every atom is "it" and has to tag every other atom. This adds up to 20 billion calculations. After all that number crunching, you get a snapshot of the system's state every femtosecond (which is a millionth of a billionth of a second - super short). Keep doing this step-by-step calculation millions of times, and you've got yourself a trajectory spanning nanoseconds. This is intense, which is why scientists rely on supercomputers for these heavy-duty calculations. But even with all that horsepower, these simulations can take weeks to wrap up. It's a real testament to the complexity and scale of molecular dynamics simulations. Aside from the computers that simulate these systems, all software used from preparing the simulation to analyzing them is open source, e.g. free to use, which also means that anyone with a computer could in principle simulate molecular systems.

Analyzing and interpreting the simulations

Once the simulation is finished, the trajectory can be analyzed. With the trajectory, we now have the location of arginine molecules around the trastuzumab domain shown above (which is called Fab domain) over many nanoseconds. Averaging over these locations, you obtain the positions in space relative to the domain where arginine molecules are most of the time. This time average is shown as the blue bubbles in the image above. Most of these bubbles in the picture are on the left side, which is where the so-called complementarity determining region (CDR) is located that is responsible for recognizing “evil” (think of it as the eyes of the antibody police officer). On the right-hand side in the image the Fab domain is normally bound to the rest of the antibody (the officer’s neck). That means if the antibody is whole, this is where the protein continues and the space would already be occupied, so that no arginine could ever move there.

Coming back to the question in the beginning, we know that arginine increases the stability of antibody solutions – but how?  In the simulation we see, that the CDRs are the parts of the Fab domain where arginine is located more often. So, one possibility is that arginine shields the CDR from interacting with other antibodies or destabilizing molecules, which then increases the stability of the protein. Coming back to the image of police officers and the CDRs as their eyes: in highly crowded environments these antibody officers might lock eyes and get stuck in a staring duel. While staring at each other, their movement is hindered (high viscosity) and they might never want to stop staring (loss of function). Arginine interacting with the CDRs can be thought of as something like a newspaper, that distracts the eyes of the officers. Therefore, we conclude that in the future we maybe should focus on giving the officers stimulating tools – adding and designing cosolvents that interact favorably with the CDRs of antibodies. This information narrows down ideas for solvent design for antibodies, which brings us hopefully closer to the goal of self-administration of antibody-based drug solutions in a patients’ comfort of home.


About the author

Tobias Prass is a PhD student under the supervision of Prof. Dr. Lars Schäfer. In his bachelor and master theses, that he finished in 2018 and 2020, respectively, he already researched antibody system using molecular dynamics simulations. His PhD project focuses on investigating high-concentration antibody solutions, how they are affected by cosolvents, and antibody-centered force-field optimization.