Dr. D. Mohammadyani

Johnson and Johnson

Dr. Dariush Mohammadyani is a senior scientist at Johnson and Johnson. He utilizes state-of-the-art computational approaches to analyze data and model biologics drugs. Previously, Dariush worked at Moderna Therapeutics, where for the first time, he modeled lipid nanoparticles containing RNA. During his postdoctoral fellow programs in the Department of Biophysics at Johns Hopkins University. His research has been centered on computational biophysics and computational structural biology. He received his Ph.D. from the University of Pittsburgh, Department of Bioengineering (2016). He is a co-author of more than 20 peer-reviewed journal papers, such as Nature Communications, Nature Cell Biology, Nature Chemistry, and Science Signaling.

Aug 4, 2021 Presentation at the Global Conference for Lipid Nanoparticles & Other Non-viral Nanocarriers

Computational Approach to Drug Delivery of RNA Based Therapeutics and Vaccines

Talk Transcript:


T_T Scientific_ Dariush Mohammadyani (Johnson _ Johnson)_ Computational Approach to Drug Delivery

[00:02]

Dr. Dariush Mohammadyani

Thank you, Nima. Hello everyone! And again thanks to the team of organizing for bringing together talented scientists around the globe to talk about Lipid Nano Particle. I am Dariush Mohammadyani, senior scientist at Johnson, pharmaceutical companies of Johnson and Johnson. My PhD at the university of Pittsburgh in postdoctoral study at Johns Hopkins, we are focused on computational biophysics of lipid-based systems and their interactions with proteins, excuse me. Before James and I worked at Moderna as a computational Nano scientist, very unique position to build and explore lipid nanoparticle structure in dynamics and very glad, both current and previous employers are critical players in fighting against COVID virus. Let me start first with the disclaimer contents are based on publicly available materials and do not reflect my current and previous employer's opinion.

Okay. For the outline, I thought it is good to be focused on one main message, and I tried to paint a picture that revealed the tremendous value that modeling can provide to the field of drug delivery of RNA therapeutics and vaccine and it will be true understanding, the mechanism of particle formulation, role of each component in the stability and bioactivity, and how LNP release its carbon. Well, I will touch its topic roughly here, I would be glad to speak with you during the conference or offline after the conference. Since there is another very buzzword area in artificial intelligence that use the same letters as LNP, I thought it is good to make a joke at the beginning, and if you expect to hear about NLP ‘natural language processing’, probably you are not in the right zoom meeting, I am going to talk actually about this smiley face lipid nanoparticle. Okay. I assume most of you have heard about molecular modeling and how powerful is this approach to model small systems of few atoms to peptide to proteins and even antibodies. However, you might not hear about modeling of HIV capsid or lipid nanoparticle, which are gigantic system. This is where I zoom in today.

Let me start with the state-of-the-art, atomistic modeling of HIV capsid, coupled with Cryo TM, published by Claus Sheldon, a great computational scientist who unfortunately passed away a few years ago and he is collaborator pageant john from department of structural biology at the University of Pittsburgh. This amazing revolutionary paper published, when he was in the same department as PAE June, it was very inspiring to me at the time. It is the first high resolution structure to describe the fullerene cone model of HIV capsid. Please note the dimension here, this is the main point of this slide, modeler would acknowledge, how amazing is building and running this simulation. It took a few years for Greg Ruth from university of Chicago to build a coarse grain model of the same chaps. But, they actually dive deeper to understand the function of HIV capsid and shed some light on capsid of same and even more, not only capsid lattice self-assembly but also the potential disassembly of capsid, often delivery to the cytoplasm of target cell. Is not a familiar story to people in the field of Lipid Nano Particle? Probably by now, we are convinced that we can look into the system like size of LNP using coarse grained molecular dynamic.

I will call it a novel computational microscope and it may not be wrong if I say modeling of an LNP is even a bit more complex than capsid because we are talking about the very heterogeneous system. While the static structure of an LNP is important what we can gain from modeling is dynamic of lipids, how they reorganize in various conditions, this is the critical point here. First of our coarse grain force built out there but today I would like to focus on martini core screen models, which is the real house for me. I learned to start using this model around 2010, more than a decade ago. Since then the force field has been evolved a lot been added to that protein, DNA, RNA and now we can model single and double stranded RNA with martini which is critical for mRNA therapeutics. Okay. Let us look at the building block of an LNP, already mentioned in previous talk but I quickly go through, on the top left we see the gene material as cargo with overall negative charge, this is important to remember and generally tied with ionizable lipids with positively charged when they are protonated. In the second row, you see cholesterol and phospholipid, they are body together, and is for multiple reason, and one is this is the way Mother Nature created our cell membrane, they always go together and the last part is Peg Lipid which decorates the surface. I go through each very quickly.

Okay. I think now we are ready to see some fascinating lipid nanoparticle models. This is a model built by a Peter Tillman Team from university of Calgary. This is a complete model of an LMP with all components from lipid to peg and short RNA and that. Watching this help us to move from being convinced, okay, we can do the modeling, to being confident that molecular dynamic not only revealed the structure but it can show us some essential aspect of dynamics of lipid nanoparticles component. To me watching always this jiggling and wiggling of the molecule help me to reach that kind of discovery moment, and kind of aha moment to realize something that was missing from aesthetic structure. Now, it is kind of showed up and it opened my eyes. To keep this talk non-confidential. I invite you to watch a video in YouTube where my previous manager Michelle Lina Hall presented some of our modeling, you have done together. At the time I was at Moderna and I promise you enjoy the talk where you can see the answer to this essential question, it is not just about the structure, can be modeled self-assembly of lipid nanoparticle containing RNA and the answer is a strong yes. Self-assembly meaning, how a particle forms from the beginning, from the infancy state, how do those entire lipid molecules coalesce, how RNA get encapsulates, and many other interesting questions and the next step is like how we want to utilize those that we are learning from the structure. Maybe we can ask this kind of question why do we need to know how LMP forms or what is the structure in each state. Let us say low PH, high PH, low temperature, a little bit more elevated temperature, you name it. There are many interesting questions and I think it is all for two main reasons, first from drug product development perspective to design a stable lipid Nano Particle that can tolerate various types of stresses during development.

We need to know the structure and we need to know the structure because LNP should be stable for a long term after field finish and in a reasonable condition, this is very critical, it is a huge cost saving and ease of use between minus ADA storage versus minus 5A and even room temperature. Right? Then there are structure methods and second, still LMP should be stable, stable enough from the injection side to the cell membrane to the gate that want to enter but not too stable with it enters the cell. This is the moment actually an LNP should let the RNA go. All these facts suggested that we need a precisely designed LNP structure and this is very complex optimization problem, actually. It is not easy because each conditions, each cargo, each variable needs a specific answer. This is the point here, modeling certainly can help us to guide the design of experiment and narrow the exploration of the landscape and we can work on focused areas. To put this in the context, here we can compare the component of both mRNA vaccines now from foreign tech and moderna. In terms of similarity, you see both vaccines use same lipid ratios, this is very interesting, and both use the exact same phospholipid and cluster molecules.

This is good. But then the differences, there are some main differences. One is on messenger RNA they used, probably not in terms of sequence much but differences lies in production and modification processes and another difference is ionizable lipid hearing that is very important because it is 50 percent of the total lipid, they are ionizable and they are main body of our mRNA actually and the last difference is the lipid particle of the pair, actually the polymeric part is steel peg 2000 but the lipid parts for both RNA vaccines are different. Okay, based on the composition, type of lipid, type of cargo that can be short RNA, messenger RNA, or can be more than one RNA in cases crisper cast 9 delivery, this structure can be widely different. Here I am presenting three commonly accepted structures from left to right, multi-laminal or onion-like structure. Second is nanostructure core and the last one on the right is core shell suspension. For the rest of the talk I try to focus on last two nanostructure core which is mostly accepted for short RNA systems and cortial model for mRNA-containing LNPs. Let us see how modeling provides insight into conformation and the role of each component. Starting from Peg lipid, a component with low concentration just 1.5 percent of total lipid but with a very high impact. Generally, effect 2000 is used, meaning a polymer of around 45 polyethylene glycol monomers, and in a coarse-grained language that we want to model it, we use 45 beads or kind of an atomic representation of that polymeric system connected together to build that polymeric side and on the lipidic side, it depends on the lipid structure. Peg conformation is highly dependent on the concentration. This is what the simulation reveled.

As we see in the simulation, it can be either brush like at high concentration or mushroom rock at low concentration, which is important for LNP interaction with other LNPs or other system and membrane or proteins. Second is ionizable lipid, elegantly designed lipid to function like a switch to be protonated at low PH and deprotonated at high PH. As we see in the left corner. Why do we need such a binary function actually? We generally form lipid nanoparticle at low PH, where ionizable lipid kind of covering RNA. Now we need that knowledge of electrostatic interaction, RNA is negatively short protonated ionizable lipid or positively charged, they are awesome to get together, and actually those lipids help solubilize RNA in the lipidic system. But after TFF and buffer exchange, now we can push the system to the higher PH and let LNP reorganize itself and by then probably RNAs already encapsulated and boom, ready to use. Okay. This elegant and simple simulation captures this binary function of ionizable lipid, Muscle Rama Zampur, who was a computational model in a Tillman team at the time, so they have a biliary-like system. Let us start from low PH. What they did?

They put phospholipid and ionizable lipid at low PH where we see they are highly soluble in together, and it sounds like it we have an ordered and very organized bilayer system or bilayer-like system. However, when the PH was elevated, ionizable lipid are deprotonated leading to a very nice phase separation in the center. This is among those exciting moment that actually face separation is a very good thing for biology because it helps RNA get engulfed below the surface of RNA, if it is small. Again, as I mentioned it can be Nano structure shape but if it is bigger maybe adopt different shape. Okay. Now, in this and next slide, I am going to compare LNP containing short RNA and mRNA based on available data. Let us look at what Tillman team had done and then how they peel off the structure to go to the core and understand what is happening there. But, I go in opposite direction, actually I am going to focus on the core of this nicely equilibrated mole. The core of the system contains water, shown in blue and RNA in red, we clearly see a Nano structure core, right? This is good. Suggesting RNA already brought in some water into the structure that makes sense and the reason is the backbone of RNA is hydrophilic and to satisfy those forces, RNA brings some water molecules, create some water reservoir into the structure. Now, we have those reservoirs that should be protected. Okay, first let us focus on RNA, to protect RNA have that solubility right, we need ionizable lipid in dark blue, organize themselves around all this RNA system, and we have 50 percent of those, not all of them connected to RNA, probably they are feeling some empty space as well. But in phospholipid and cholesterol, they are building monolayers around each of that water reservoir to build the shield around the water to kind of distingum from lipid core. This is what that amphiphilic lipid does, like they have hydrophilic head group and then hydrophobic tails and then hydrophilic stay close to water. This is what they do and finally Peg on the surface. Okay. Now, we saw that short RNA containing LNP. Let us look at if we have an mRNA system in the system, what is going to happen. In this PNAS paper, published by Esther Zeneca, and in previous talk also several times referred to this one.

Other use very precise method like sex and sense and propose that mRNA containing lipid nanoparticle can be described as a cortical system and it perfectly makes sense because here is not short RNA that brings few molecules, this system can be spread around all those lipids and ionizable lipids. I am not sure if anybody has built a model with actual mRNA, with the large system to capture substructure yet, but for sure this will be the next adventure from modeling community. To re-emphasize the issue, the main challenge here is not modeling the lipid or modeling short RNA, it is the modeling of long mRNA which can adopt various structures in different environments. This is an open problem for the computational people to predict first mRNA structure and then they are very interesting paper, published already but there are a lot can be done, when we know that I think we can easily build the mRNA containing LNP core screen model as well. Now that we have a good understanding of LNP structure, let us focus on next step. Using modeling to shed light on RNA released from LNP can be used really modeling to capture such an important question. Now, let us call it endosomal escape, and what we want to know is how LNP opens up. In this paper by Dan Anderson, team from MIT, focus on membrane fusion.

A mechanism that can be studied in detail using coarse grained simulation. What they showed is how lipids at the surface of lipid Nano portal plus ionizable lipid interact with the membrane of endosome. Even they studied the impact of various ionizable lipids in the process. The main idea is, in the endosome we have a lower PH and probably some enzyme facilitate this step. But, what really is important how does connection happening. Again, I want to highlight this fact that the efficacy of RNA at therapeutics is highly depend on the number of RNA that can be freed up from LNP, and reach to the cytosol because some really can easily can get degraded into endosome. In the second paper, by Paula Decusi and the teams on the same topic, this time they step even further and they clearly showed the detail of the endosomal escape mechanism. Here they showed how the contact is happening, how membranes are fusing, what caused the pore formation and how it is expanded and how the corvo getting out. As a side note here, there are some other studies around this topic that people highlighting the fact of presence of some unique lipid phase called hexagonal phase that helps facilitate opening up lipid nanoparticles and such environment. Okay, let me summarize what I have highlighted thus far based on available models. Model clearly and modeling in general, clearly has shed some light on the role of each component in lipid nanoparticle structure and dynamic. I showed how Peg covers the surface in mushroom-like conformation, phospholipid and cluster that act as a structure component of an LNP and define the boundary of LNP. We observed an impact of PH on ionizable lipid and we talked about the phase separation and organization and how different gene material can vary the structure of LNP and then finally we reviewed LNP interaction with other membrane and how lipid fusion highly tied with the drug efficacy. Here I would like to propose a few areas for models in Pharma in academia as a future direction. I will be very brief here. But I would be glad chat offline.

Like how this kind of simulation can be set up and how we can perform those very interesting and how we can tackle those kinds of questions. The first idea is, understanding something called Bleh, it is kind of a puzzle, at least it was puzzled but now with that paper published from it there, now it is a little bit more clear what is happening. Bleh is not a very interesting lipid organization governed by phase separation and the various conditions mRNA, like to be either fully lipid associated or become isolated from ionizable lipid and that caused those kind of variation that we see if you just look at this dark and dotted spot that they should mRNA, we see the phase separation here but the focus of phase separation here is not ionizable lipid, it is actually mRNA and as I mentioned, this kind of phase separation system can clearly by elegantly designing simulation can be studied. Second area to explore is modeling of the hybrid biopolymer lipid nanoparticle. Probably you have known that. There are like many studies around pure polymeric nanoparticle and again you might, here still the challenge feed polymeric system is efficacy and it is really hard to release RNA from those systems.

This is why people now are looking at the hybrid system, mixing lipid and polymer and the good news here is coarse grained modeling can be utilized here because we have a strong and good course grained model to model both polymer and lipid and this is a very sweet spot for people want to look at those hybrid SUS and the last futuristic area is actually functionalization of lipid nanoparticle to facilitate targeted delivery. One main challenge for LNP is overall to send RNA to the right tissue and to the right cell type. One hypothesis among many out there is to add some functional group to LNP surface to make it recognizable by targeted cell membrane and one idea is, okay each specific tissue and cell type, they have different composition of cell membrane and since cell signaling molecules at the surface, then by designing the right functionalized LNP, maybe we can push LNP to go to a specific tissue. This is highly feasible area to utilize modeling as people have done it and you see here, and very high impact for the future of RNA delivery. Okay. Let me get back to that one message that I said from the beginning, and I want you remember, that main message here is molecular modeling provide a high resolution insight into dynamic of lipid nanoparticles, leading to design a very stable lipid nanoparticle and highly potent medicine and I think this is my suggestion and recommendation to people in pharma in academia, we need to invest more in computational modeling to unleash really such huge potential of molecular modeling to go deeper down to understand not just static structure but dynamic of lipid nanoparticle component. Let me finish with this nice quote that the purpose of modeling is inside, not numbers and then thank you and have a good day. 

[24:44]

Nima Tamaddoni, PhD

Thanks so much. So, Alex will ask that question.

[24:52]

Alex Aust, MS, MBA

Hi, Dariush, thank you so much, upper may asked a two-parter, so the first is could you briefly talk about the role of hydrolysable peg lipid in endocytosis and endosomal escape of LNP?

[25:09]

Dr. Dariush Mohammadyani

I think what we can talk about peg is more before uptake happening. By the point we are reaching to the uptake, it is at the cell gate, probably we have lost part of that peg, and already peg is a like minor component among lipids, then maybe peg has a minimal impact under endosomal escape, and what mainly drive it is those ionizable lipid design that actually has very strong impact on the endosomal escape and there are very many things happening in the endosome, it is not just on like PH adjustment or lipid reorganization, there are some enzymes there, understanding those mechanism also very impactful. But I think, peg and peg design may have very minimal impact on that concept.

[26:07]

Nima Tamaddoni, Ph.D.

Alright, thank you also. So how do you model changes triggered by a change in PH or other of these, you know.

[26:21]

Dr. Dariush Mohammadyani

Very good question, yeah. Yes, very good question. In modeling area, overall it is hard to change fish during the simulation because like you set up the initial setting of the simulation, and you run it. But what you can do is with that binary design of ionizable lipid, you can from the beginning make your system protonated or not. You assign the charge or not to the molecule and then capture what is happening. Or you can run that like what I showed from most in amazon pools work, you can run it in low PH and then cut the simulation and change the protonation state and continue to capture the face separation.

[27:09]

Nima Tamaddoni, Ph.D.

Thank you. 

[27:11]

Dr. Dariush Mohammadyani

No problem.

[27:21]

Dr. Dariush Mohammadyani

I cannot hear you.

[27:26]

Graham Taylor, Ph.D.

So, yeah, Dariush, this is Graham from T&T Scientific, it sounds like they may be sorting out a mute echo issue, but one quick question just on the computational modeling aspect. I wanted to confirm if most of these it looks like our coarse grain simulations and I am just curious if, you know, if our goals or efforts or visions for modeling to aid in this field which is really impressive to see, you know. Are we limited at all by computational power? Is this something you need super computers for, you know? How does that play into all this?

[28:02]

Dr. Dariush Mohammadyani

Yes, absolutely! We need super computer and all the simulation these days are run by GPUs. But like having course brain and GPU coupled together, we can tackle all these problems that I just touched today very briefly. There is no issue like if we do not need to wait for next kind of revealing technology happens, if everything is there just people need to invest into that space and go deep into the structure and really peel off what is happening, you can cross-section your model, you can you can look through that, and you can look at in a dynamic fashion what is happening there, it is very beautiful. This is like what I have been doing for more than 10 years, I always enjoy looking to do simulations.

[28:51]

Nima Tamaddoni, Ph.D.

Thank you so much. Thank you. I have one question. Do you think molecular modeling can regularly be used to forecast impact of impurities on LNP formation or LMP structure?

[29:04]

Dr. Dariush Mohammadyani

Absolutely! This is what kind of drives me when Nima reached and say I think this is the right time to actually advocate this. I think we are at the very good moment. People know about all these technologies and know a little bit about the computational part and it is a very good moment to onboard this and widely invests in that area. Yes, my answer is yes, not in in gene trophy now, I am in a biologic space that likes long time back formulation and understanding question through modeling was a little bit difficult. But now we are tackling this kind of question as well. I think technology is right at the right time and we have to just leverage it and use it.

[29:52]

Nima Tamaddoni, Ph.D.

That was great! And thank you so much Dariush. That was a question actually from a scientist at bioNTech, Alexander it is a pleasure to have you here and there are more questions Dariush, you can answer with video or text, please go ahead and answer if you feel like it because we are going to go to the next stage.

[30:08]

Dr. Dariush Mohammadyani

Awesome!

[30:10]

Nima Tamaddoni, Ph.D.

Thank you so much.

[30:12]

Dr. Dariush Mohammadyani

Thank you.  

Contact Us