The promise and problem of gene therapy
Gene therapy has promised something remarkable for decades. Instead of repeatedly treating the symptoms of a disease, we could help the body make a missing protein itself or even correct the mutation that causes the disease.
The idea is powerful, but turning it into a medicine is much harder. One of the biggest challenges is delivery. Genetic material such as messenger RNA (mRNA) cannot simply be injected into the body and expected to reach the right cells. It needs a carrier to protect it and help it enter cells. But the carrier itself can also cause inflammation and toxicity.
Lipid nanoparticles (LNPs), tiny particles built from fatty molecules, are one of the most successful solutions so far. Still, they come with a dilemma that has been difficult to overcome. One of their most important components, the ionizable lipid, becomes positively charged in acidic environments, such as during particle assembly and inside the acidic compartments of cells called endosomes. This charge lets the lipid grab onto the negatively charged mRNA and helps release it inside cells. But that same positive charge can also cause toxicity.
In other words, the feature that helps LNPs work can also induce adverse effects.
Could we make the positive charge disappear when we do not need it?
This became our central question: could we keep the positive charge when it is needed and make it disappear when it is not?
The idea did not come from a single experiment. It took shape gradually, from clues in earlier studies.
Tracing the development of an idea
The first clue came from previous work showing that removing the ionizable lipid from an LNP formulation greatly reduced toxicity. Simply removing it was not a solution, because the particles would lose most of their ability to deliver mRNA. But it suggested that the problem might not be LNPs as a whole. Perhaps the ionizable lipid's charge behavior could be redesigned.
The second clue came from natural zwitterionic lipids, which carry both positive and negative groups. In these molecules, the negative charge offsets the positive one, and they are not toxic. But they cannot bind mRNA efficiently. We wondered whether we could build an ionizable lipid that is positively charged when needed for delivery, but carries a net negative charge under physiological conditions.
We had a hypothesis for why natural zwitterionic lipids cannot bind mRNA well. Their positive and negative groups sit very close together, so the two charges largely cancel each other out. If so, perhaps moving these two charged groups farther apart could preserve the benefit of the negative charge while still allowing the lipid to interact with negatively charged mRNA.
Turning an idea into a new lipid
Our first design was therefore simple. We increased the distance between the positive and negative groups within the molecule. We synthesized a small set of these lipids and tested whether they could deliver mRNA into cells.
When we first saw transfection, the result was modest, but it was exciting. It was our first indication that adding a negative group did not have to prevent a lipid from delivering mRNA.
That first success also revealed the next problem. The lipids did not package mRNA efficiently enough. We needed to improve their performance without abandoning the charge-switching idea. Another clue again came from previous research. Other studies had suggested that introducing hydroxyl groups (–OH) into ionizable lipids could improve nucleic acid encapsulation. We redesigned our lipid to introduce hydroxyl groups. The new lipids packaged mRNA much more efficiently, and their ability to deliver it into cells improved substantially.
At that point, what had started as a question became something we could systematically explore. We expanded the chemistry and built a larger family of charge-switching lipids, which we called S-lipids.
Rational design and a little bit of luck
Next, we evaluated the S-lipid library to see whether charge switching could be tuned for both delivery and compatibility with the immune system. We prepared charge-switching lipid nanoparticles (SNPs) with the S-lipids and evaluated them in vivo for mRNA delivery efficiency. One candidate, E20 stood out for its efficient mRNA delivery in animals.
Efficient delivery was only half the goal. We next tested whether SNPs could deliver mRNA without provoking a strong inflammatory response. We used peripheral blood mononuclear cells (PBMCs) as a testbed because they are a major source of LNP-induced inflammatory cytokines. E20-SNPs did not increase cytokine production, whereas the conventional LNPs we tested induced substantial amounts of cytokine production. To understand this difference, we examined common pathways through which LNPs trigger inflammation. E20-SNPs did not activate TLR4 or complement, consistent with our charge-switching design. In particular, the negative charge of E20 at physiological pH likely reduced its interactions with the immune-system components that activate the TLR4 and complement pathways.
However, the E20 SNPs delivered mRNA efficiently into cells and in vivo, which means they are triggering endosomal release, which has historically been associated with triggering inflammation because it causes the activation two inflammatory pathways: the galectin-8 signaling and the platelet-activating factor (PAF) signaling pathway. The activation of both of these pathways is associated with endosomal membrane damage. Delivering mRNA into the cytoplasm without damaging the endosome is very challenging because of the large size of the mRNA, and traditional LNPs release mRNA from endosomes by causing large amounts of damage to the endosomal membrane. Surprisingly, the E20-SNPs were able to deliver mRNA into the cytoplasm and trigger endosomal release without damaging the endosomal membrane. Together with our charge-switching design, this unexpected finding helped explain why E20-SNPs delivered mRNA efficiently while causing little inflammation.
From concept to in vivo proof
We then asked whether this behavior would matter in a disease model where inflammation was already present. In a mouse model of acute lung injury, we used SNPs to deliver IL-22 mRNA, which encodes a protein that helps repair damaged tissue. E20-SNPs were the only formulation we tested that reversed the existing lung injury and inflammation. Conventional LNPs carrying the same mRNA did not achieve this recovery. This difference was most likely because E20-SNPs did not worsen the existing inflammation.
Together, these findings suggest SNPs combine two properties that are often difficult to achieve simultaneously: highly efficient mRNA delivery and sharply reduced immune stimulation. This combination points to a platform with broad potential applications, especially where inflammation is a critical concern.