The data looked fine. Better than fine: ecapsulation rates were high, particle size was uniform, and quality control numbers passed with room to spare. If you stopped there, you’d think the mRNA was doing what it was supposed to do -- delivering genetic instructions to produce a specific protein.

But in a lab just south of Seoul, something wasn’t adding up. The team at MEPSGEN, a Korean biotechnology company specializing in microphysiological systems and nanoparticle production, had run cell-based assays expecting routine protein expression. Instead, the cells sat still: blank, unresponsive, as if nothing had been delivered at all.

“It didn’t make sense,” said Kim Yong-tae, the company’s founder and, until recently, a professor of bioengineering at the Georgia Institute of Technology. For more than a decade, his work focused on fluid dynamics and microsystems that replicate organ-level behavior, examining how cells respond to complex signaling and how nanoparticles break down under stress.

“Everything on the outside looked perfect,” he said in an interview with Korea Biomedical Review. “But biologically, the mRNA wasn’t doing anything. The cells just sat there.”

Kim had seen enough mRNA runs to know what to expect. So when the readout came back blank, he didn’t blame the material, or the lipid nanoparticles (LNPs) that were supposed to carry it into cells. He blamed the process, specifically the shear. "Everything looked perfect on paper," he said. "But biologically, the mRNA was dead."

Shear stress isn’t something most vaccine recipients, or even most regulators, spend much time thinking about. But for Kim, it became a recurring preoccupation. It’s the mechanical force created when fluids move rapidly through narrow channels -- useful for mixing, but potentially harmful to delicate materials like mRNA.

Before the pandemic, most RNA drug development took place at lab scale. Small animals like mice could be dosed with hand-mixed batches. But once studies moved into clinical trials -- especially during the pandemic -- that approach, Kim said, no longer worked. "Larger animals and humans required higher volumes, tighter consistency and faster turnaround. So companies turned to microfluidic systems to increase throughput."

That meant faster flow, which meant more shear. That helped with mixing. Still, Kim began to wonder if it came at a cost.

“We started thinking,” Kim said. “What if that very force, this invisible stress, was slicing through the mRNA?” It wasn’t the presence of mRNA that mattered, he said. It was whether it still worked.

“We realized it wasn’t just about how much mRNA got inside the nanoparticle,” he said. “The real question was whether it was still functional once it got there.” Still alive, so to speak. “Because you can encapsulate something perfectly,” he said. “And still destroy it on the way in.”

That insight, presented at the 4th LNP Formulation & Process Development Summit held April 15 to 16 in Boston, Massachusetts, pointed to a potential blind spot in RNA drug manufacturing -- one that Kim believes could help explain why the clinical performance of Covid-19 vaccines fell short of early-stage data, and why other RNA drugs have stumbled in translation from lab to clinic.

A week later, Kim expanded on the findings and the reaction they drew. Several summit attendees, he said, approached him after the session, some visibly unsettled. “We didn’t even know that was a problem,” one of them told him.

But Kim suspects many had seen it before. They just didn’t know what they were looking at.

MEPSGEN data shows high-quality RNA batches can still fail in cells

The idea that LNPs might degrade during scale-up isn’t new. Kim himself noted that researchers have long flagged pressure-based systems such as Z-flow mixers, high-speed jets, and microfluidic reactors as sources of stress.

If 90 percent or more of the mRNA ended up inside the nanoparticle -- a process known as encapsulation -- the batch was considered good. Vaccine developers checked the box. Regulators cleared the lots. But Kim’s data told a different story. "We met all the benchmarks" he said, referring to encapsulation efficiency, particle size, and uniformity. "And still, there was no protein expression." 

To test the effect directly, Kim’s team produced LNPs under varying flow conditions, 0.3 × 10⁵ per second and 3 × 10⁵ per second, the range he said was commonly used during Covid-19 vaccine manufacturing. Encapsulation remained high. “We were getting 95 percent,” Kim said. “But the protein expression dropped off a cliff.”

The mRNA was present, he said, just “too damaged to function.”

“DNA is double-stranded, so it can take a hit,” he added. “mRNA is single-stranded. It just breaks.”

Under the microscope, the particles looked pristine: symmetrical, intact, textbook. But inside, he said, the code had already started to fall apart.

He pointed to a graph showing two curves. "At 3 × 10⁵ shear rate," he said, tapping the graph, “there was no expression. Drop it to 0.3 × 10⁵, and it comes back."

The finding held across multiple mRNA lengths and cell types. The damage wasn’t total, Kim said. It was just enough to slip past detection. “It snips,” he said. “Not shreds.” 

Kim said a full study detailing the findings is slated for publication in July.

A pandemic-era shortcut may have introduced unseen risks

Kim, who founded MEPSGEN in 2019, has moved between academia and entrepreneurship for most of his career. A longtime member of Georgia Tech’s nanomedicine lab, he spent years studying microfluidic chip design before returning to Korea to launch the company.

“At first, we didn’t think this was going to be a business,” Kim said. “It was just research, and anyone could build a mixer.” Then the pandemic hit. Vaccine makers began reaching out, Kim said, some citing his own papers without realizing he was already fielding inquiries as a founder.

By 2021, he said, the pivot was complete. MEPSGEN had retooled its platform into a scalable LNP manufacturing system called NanoCalibur, designed to “scale out instead of scaling up.” Instead of pushing higher flow rates, and with them, higher shear, the company kept its microfluidic channels small and multiplied them in parallel. “It’s like adding burners to a stove,” Kim said. “The heat stays steady, but you get more done at once.”

The approach, while not revolutionary in concept, had never been rigorously applied to LNP production. But Kim believes it solves a key bottleneck that may have persisted throughout the pandemic: how to scale RNA drugs without sacrificing their integrity.

What’s startling is how long this blind spot may have persisted. “In the pandemic, speed was everything,” Kim said. “No one had time to validate shear effects. They just needed to produce billions of doses.”

Many manufacturers, Kim said, adopted high-throughput systems that, in hindsight, may have been too harsh for fragile mRNA strands. Some used Z-infusion techniques, which involve pressurized fluid blasts akin to power-washing grime off a surface.

“You get good particle formation,” Kim acknowledged. “But what’s inside may already be compromised.” And that, he says, might explain why some people had strong reactions to certain mRNA batches, while others felt nothing at all. “There was no way to tell how much damage had occurred,” Kim said. “So they just increased the dose.”

The problem, he added, isn’t that developers were careless. It’s that no one had defined the threshold before. Shear damage wasn’t binary; it was distributed. Some mRNA survived. Some didn’t. And variability followed.

And as more RNA therapies edge toward the clinic, Kim believes the industry will need to ask a harder question: what’s the point of yield, if the cargo arrives dead?