How can simulations aid our understanding of nanoparticle-mediated siRNA delivery?

With artificial intelligence and simulation-based methods becoming more powerful, the increasing role of computational approaches in pharmaceutical research is undeniable. This includes Molecular Dynamics (MD) simulations, which have been under constant evolution since 1976 [1]. As MD has already proven valuable in the design of small-molecule drugs and protein research, it is now emerging as a promising tool for the development of siRNA delivery systems as well.

Short interfering RNA (siRNA) is used to downregulate disease-driving genes within an organism, offering broad applications in treating cancer, autoimmune diseases, and other conditions. Yet, due to fast degradation of exogenous nucleic acids in the body, successful delivery of siRNA strongly depends on the formulation strategy. While good responses can be achieved with N-acetylgalactosamine (GalNAc) siRNA conjugates in the liver, other formulation strategies focus on the delivery via Adeno Associated Viruses (AAVs) or the encapsulation of siRNA into nanoparticles. The materials used for siRNA nanoparticles cover a wide range from inorganic gold nanoparticles, via the more prominent lipid nanoparticles (LNPs) to polymeric nanoparticles, particularly polyplexes. Regardless of which material the carrier systems are based on, they all have a complexity in common which makes the complete understanding of a formulations’ behavior challenging. MD provides the potential to make molecular interactions, such as the internal organization of a nanoparticle, accessible while circumventing complex experimental methods such as x-ray scattering approaches.

All-atom (AA) MD simulations show the molecules of interest in single atom resolution. While this is a big advantage for the observation of details, it results at the same time in the major disadvantage of AA simulations: The high number of interactions the simulation must account for leads to a limitation concerning size of the simulated system and time span of the simulation. Nevertheless, AA MD studies reveal valuable details on nanoparticle formation. They can be used to investigate the direct interaction of polyplex-forming cationic polymers [2,3] with nucleic acids, unveiling binding strengths between polymers and the cargo, as well as preferred binding sites [4] (e.g., major grove vs. minor grove of the siRNA). Other applications allow insights on how modifications of polymers, such as the addition of hydrophobic units to hydrophilic polyethyleneimines (PEI), change the particle performance [5].

Similarly, AA simulations are used in the development of LNPs. First, they can be applied to predict the degree of protonation of an ionizable lipid [6]. This is crucial for LNP design, as the pKa of a lipid within an LNP (“apparent pKa”) can differ significantly from its pKa in solution. Second, insights on the geometry of lipid arrangements within an LNP [7,8] won from AA MD strongly contribute to the understanding of LNP systems. However, AA models are hardly capable of depicting polyplexes or LNPs at the full scale of experimentally observed nanoparticle sizes.

To access larger systems, coarse-grained (CG) force fields have been developed, the most popular being the Martini force field [9]. Coarse graining reduces the degrees of freedom in a simulation by reducing the resolution, as groups of atoms are summarized into beads with predefined properties such as degree of hydrophobicity or charge. With this approach, not only systems up to ~100 nm are simulated but also timescales of tenths of microseconds are accessible. To avoid wrong conclusions based on observations introduced by the inaccuracies of a CG model, the simplified models must be validated against AA simulations and/or experimental data before further use. Some well-established polymers in nucleic acid delivery, e.g., PEI, have therefore been simulated as AA models [10] and as larger CG models [11]. Here, the CG approach allows to cover the whole range of experimentally used molecular weights of PEI. CG simulations also expand the knowledge gained from MD to include more formulation parameters, such as the role of different N/P ratios (i.e., the ratio of positive charges in the encapsulation agent to negative charges of the RNA). This sheds light on different stages of siRNA encapsulation or the presence of unproductive polymer in a polyplex formulation [12]. Similarly, CG simulations shape the understanding on how different ionizable lipids form LNPs with substantially different properties [13]. Other simulations visualize the changes in LNPs upon pH changes in the environment, including dehydration and rearrangement of lipids in an LNP when dialyzed toward neutral pH [14]. Backmapping from CG resolution to AA is the chance to combine the advantages of both: assembling larger structures in CG MD and observing atomistic details in the final state after AA resolution was reapplied to the molecules. This method can, for example, be used to research the complex interplay of the different lipid types in an LNP [15].

As siRNA delivery does not end with the formation of a stable nanoparticle, efforts of MD research also expand to the interactions of nanoparticles in biological environments. A key application is the widely debated question of how carrier systems enable efficient endosomal escape of the siRNA, which is necessary for it to reach the site of action in the cytoplasm. Advanced microscopy methods provide valuable insights by tracking nanoparticles in combination with endosomal escape markers [16]. While these methods enable the correlation of particle properties with endosomal escape performance, they do not provide detailed information on underlying mechanisms. MD simulation is beginning to make the molecular process of endosomal escape observable [13,17], allowing for more comprehensive understanding of mechanistic backgrounds. In this regard, the influence of lipid composition and charge in an LNP on its ability to fuse with endosomal membranes can be visualized. Additionally, the role of different lipid types in the biological membranes can be more thoroughly discussed when made observable by a simulation [18].

With the rapidly growing number of publications in the field, the availability of tools and methods for faster preparation of simulations increases as well. For example, tools for automatic generation of membranes with desired composition [19,20] or flexible protocols for building LNPs in CG resolution [13] are publicly available. Certainly, this will further contribute to the progression of MD simulation in the field of siRNA delivery.

It can be concluded that MD simulations aid our understanding of nanoparticle-mediated siRNA delivery by making processes visible and understandable on a molecular level. However, simulation data must always be carefully validated and compared to experimental results to avoid over-interpretation and false conclusions. For now, the strength of MD lies in elucidation and explanation of experimental observations, while the time where it replaces experiments in the development of siRNA delivery systems is yet to come.

Funding Statement

This paper was not funded by any external sources.

Disclosure statement

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.