Until now, however, no production method could reliably control these attributes. "The size and shape of LNPs could not be reliably controlled by any previous production method. The problem may appear simple at first glance, but in reality it requires a deep understanding of lipid nanoparticle assembly," Myerson says.

A two-step fix

Conventional production mixes two fluid streams at high speed: lipid molecules suspended in ethanol, and mRNA dissolved in an acidic buffer. Crucially, the streams aren't equal — roughly three times more mRNA solution flows than lipid solution. That imbalance encourages nanoparticles loaded with mRNA to form, but it offers little control over the particles' final size or shape.

In a study published last year in ACS Nano, the MIT team showed they could gain much better control by splitting the mixing into two steps. First, mRNA and lipids meet at equal flow rates. Then, after a short delay, more buffer is added, halting particle growth. Longer delays produce larger particles.

"This gives you the ability to play around with the residence time, which is the time it takes between the first mixer and the second mixer. If you keep that residence time really long, it means your particles will grow a lot. If you keep it really short, you can keep them really small," Devos says. "It gives you a lever over lipid nanoparticle manufacturing that wasn't available before."

The researchers also found they could reshape the particles: by changing the concentration of the buffer added in the second step, they transformed spheres into elongated particles resembling avocados. Both interventions adjust size and shape without altering the LNP's chemical composition.

Adding automation

The new paper takes that method and automates it. The team incorporated a commercially available dynamic light scattering device — an instrument that measures particle sizes in real time — into the two-step process. Researchers specify a target particle size, and the system generates particles, checks their size, and, if they're off target, adjusts the delay time and other factors to steer production back on course. The system can also produce different shapes, though measuring shapes still requires analysis outside the automated setup.

"The first paper really unlocked the new methodology to make lipid nanoparticles, to truly engineer them by size and shape," Sagmeister says. "With the second study, we automate the whole process."

Three MIT undergraduates — Joy Ren, Sofiya Chubich, and Dylan Nguyen, who joined through MIT's Undergraduate Research Opportunities Program — contributed by integrating advanced software engineering with chemical engineering on the platform.

Using the automated system, the researchers mapped how changing inputs affects particle size and shape on a much faster timescale than current methods allow. They then used that experimental data to train a machine-learning model that predicts which combination of factors will produce a particular size or shape.

Next steps

The team has filed for a patent and is commercializing the technology through a new company, BIZON Labs. After initial support from the Martin Trust Center for MIT Entrepreneurship's Researcher 2 Entrepreneur program, the venture has been accepted into MIT's flagship accelerator, delta v. The research was funded by the U.S. Food and Drug Administration and by the Koch Institute Support (core) Grant from the National Cancer Institute, with work carried out in part at MIT.nano's facilities.

For developers of RNA therapeutics, the practical payoff could be simpler testing: generating particles of different sizes for a given application becomes far easier, letting researchers identify which formulation performs best before committing to costly development.