A factory floor where the product is immunity
Walk into Prellis Biologics in Berkeley and the most important machine in the room is doing something no animal cage ever could: it is sculpting a human lymph node out of light. A two-photon laser traces structures finer than a capillary. Human immune cells, donated by real people, move into the printed scaffold. And then - this is the part that still sounds like fiction - they start making antibodies, the same way they would inside your neck during a cold.
Prellis calls the whole apparatus EXIS, an externalized human immune system. The pitch is blunt: stop borrowing mice to guess at human biology. Build the human part instead, and run experiments on it.
Mice are not people, and everybody knew it
For decades, the standard way to discover an antibody drug was to immunize a mouse, harvest its B cells, and hope the result translated to humans. It usually half-worked, which in drug development is an expensive way of saying it failed. Mouse-derived antibodies can trip the human immune system. Whole categories of targets - GPCRs, the slippery proteins behind a huge share of disease - resist the approach entirely.
The obvious fix would be to use human biology from the start. The inconvenient truth was that nobody could build a working human immune response in a dish. You cannot exactly ask a volunteer to grow you a fresh lymph node.
A lifeguard who decided light was printable
Melanie Matheu did not arrive in biotech by the usual gate. She grew up in a Redwood City family of welders and builders, met a Silicon Valley angel investor while working as a high school lifeguard, and spent eight years studying immune responses with two-photon imaging under Michael Cahalan. In 2015, staring at images of immune cells at UC San Francisco, she had a deceptively simple thought: if light can illuminate a blood vessel at fine resolution, light can build one.
She founded Prellis around 2016 and took the idea through SOSV's IndieBio accelerator. The original dream was enormous - print transplantable human organs. Then 2020 arrived, and a pandemic has a way of editing your roadmap.
From organs to immunity
COVID-19 redirected the company from printing organs you transplant to printing the immune tissue that makes antibodies. The same laser printing skill, pointed at a faster, more fundable problem. In 2022 the company brought in Michael Nohaile, PhD, as CEO; Matheu moved to CTO and kept her board seat.
What EXIS actually does
The trick is recreating a lymph node well enough that human B cells behave like they would in a body. Prellis prints the organoid - the LNO - with holographic two-photon photolithography, seeds it with immune cells pooled from many human donors, and lets the system do biology. Because the donor pool is genetically diverse, the antibody repertoire it produces is broad. AI models then screen the output and guide each round toward better candidates.
EXIS™ Platform
An externalized human immune system for de novo antibody discovery: human cells in, fully human, low-immunogenicity antibodies out, with broad epitope coverage and AI-guided optimization on top.
LNO™ - Lymph Node Organoid
The printed scaffold at the heart of it all. Built to recapitulate human lymph node biology in vitro so B cells activate and produce antibodies naturally, using PBMCs from multiple donors.
The claim Prellis makes for all this is concrete: 3 to 5 times faster development of highly specific, high-affinity antibodies, dozens of diverse candidates per target, at roughly a tenth of the prior cost. For tough target classes where mouse methods fail, that is not a marginal improvement.
A Berkeley biotech, by the year
Skeptics, meet the signatures
A platform that sounds like science fiction earns its credibility one of two ways: data, or partners who do their own diligence. Prellis has both. During COVID-19, the system generated roughly 300 human antibodies against SARS-CoV-2 - a real-world stress test, not a brochure demo.
Then the names. Bristol Myers Squibb signed a multi-target collaboration and license agreement in January 2022 - upfront payment, research funding, milestones, royalties. Sanofi followed in March with a collaboration and license option deal. Eli Lilly joined the partner list. These are companies with the resources to build their own platforms; they chose to rent Prellis's instead.
Funding, round by round
Investors include Celesta Capital, Avidity Partners, Khosla Ventures, SOSV, True Ventures and Lucas Venture Group. Bars scaled to the ~$65M total - the kind of money that buys a lot of lasers.
Human biology, plus machine learning
Strip away the lasers and the mission is straightforward: integrate human biology with machine learning to discover next-generation medicines, by harnessing immune diversity and AI. The longer goal is to retire animal-based antibody discovery for an approach that is human from the first cell and can reach targets the old methods never could.
It is, admittedly, an unfashionably literal way to build a drug company. Most AI-for-biology startups would rather simulate the immune system than grow one. Prellis decided the actual cells were worth the trouble.
The undruggable, reconsidered
If the platform scales the way Prellis believes it will, the targets that have stymied drug hunters for decades - the GPCRs, the proteins too human-specific for a mouse to model - move from "impossible" to "expensive but doable." That shifts which diseases get a serious shot at a therapy, which is the only metric that ultimately counts.
So return to that Berkeley room. The laser is still tracing a lymph node out of light. But the thing being printed is no longer a curiosity. It is, increasingly, where a real human antibody begins - no mouse, no guesswork, just human cells doing what they were always going to do, in a structure that did not exist until somebody decided light was printable.