Skip to Main Content

The holy grail of cancer drug targets is akin to a unicorn horn: a marker that only cancer cells have, clearly distinguishing them from healthy cells. In reality, nearly all cancer drug targets are also found on many healthy cells, leading to serious off-tumor toxicity that — in extreme scenarios — can be fatal.

Synthetic biologist Kobi Benenson might have a way around that. Inside an engineered virus, he and his colleagues at ETH Zurich packaged a programmable genetic circuit that uses multiple targets to build a profile of a cancer cell. Detailed in a mouse study recently published in Science Translational Medicine, it’s a nanoscopic biological computer that roams through the body, executing a program that seeks to recognize and kill cells matching that cancer profile, but spares healthy cells that don’t fit all the criteria.


“[Simple drugs] are like trying to catch a criminal by saying ‘everyone who wears baggy pants is a criminal’ or something like that,” Benenson explained. “With this broad criterion, we’ll catch like 99% innocent people. One really has to really be narrowed down by combining multiple pieces of information. So, it’s the same in the disease.”

The biological computer is a genetic circuit with engineered molecular switches that can make simple computations, similar to the way silicon transistors at the core of smartphones and laptops carry out calculations. Benenson’s circuit has two major components — an “AND” function and a “NOT” function — so that the computer looks for cells that have a profile of two molecules common in cancer cells, but not a third that’s common only in healthy cells. That makes the computer more likely to accurately distinguish cancer cells from healthy ones.

“So, we have this if A and B but not C type of decision,” Benenson said. “That ultimately translates into activation or lack thereof of a therapeutic that can kill the cancer cell.”


The “AND” function is made of two molecular switches on the computer’s genetic circuit that bind to designated cancer targets. For those targets, Benenson’s team used one protein common in liver cancer cells and another protein common in liver cells in general. If the first switch binds to its protein, it sends a molecular signal to the second switch. If the second switch also binds to its protein, then the circuit forces the cell to create a new protein called HSV-TK. This combines with another compound, which must be separately injected, to kill the cell.

But healthy cells also carry these targets, so, the team had a third molecular switch on the circuit recognize a compound known as let-7c, which is common in healthy cells but not cancerous ones. If this switch binds to let-7c, then it triggers a process that shuts down the computer’s kill command, saving the cell from execution.

Scientists have been working on biomolecular computers for years, said Wilson Wong, a biomedical engineer at Boston University who wasn’t involved with the research. He called Benenson’s new genetic circuit a tour de force.

“It’s quite well-done — the culmination of at least 10 years of work,” Wong said.

The first feat was squeezing the entire biological computer into the small engineered virus that delivers the circuit into cells. “Fitting everything into the size limit posed by the virus is not trivial,” Wong said. “You have to write a fancy program, but it only fits in 2 megabytes of space. That’s what it’s like.”

That system opens up an entirely new world of possible drug targets, Wong said. Most other cancer therapeutics only recognize targets that exist on the outside membranes of cells, but now microRNAs, proteins, and other intra-cellular molecules are available for engineering.

“Intra-cellular pathways were not druggable,” Wong said. “And now they are. That’s huge.”

The other feat was getting the “NOT” function to work, Wong said. Typically, cancer drugs only attack a cell when a target is present. That means, aside from allowing biological computers like this to build safety switches, it opens yet another entirely new way of getting drugs to recognize cancer cells.

“This ‘NOT’ logic they’re doing is a very, very unique thing,” Wong said. “It means that when something is missing in a cancer cell, then it will attack. No other drug can do that. If a cancer cell doesn’t have a gene or the target, then it’s usually just not druggable. It’s in the trash can right away, even if you know that this makes the cancer super unique.”

Once the team had designed the biomolecular computer, they tested it on mice with liver cancer. In one group of mice, they injected the full computer with both the “AND” and the “NOT” function. Another group of mice received a partial genetic circuit that only had the “AND” function. These mice experienced toxic side effects, but the mice who received the full computer both saw their tumors disappear and were spared from the toxicity.

“That was very pleasing to see,” Benenson said.

That doesn’t necessarily mean the therapeutic will be safe and effective in humans, Wong cautioned.

“Sometimes a lot of the targets aren’t expressed the same in mouse models as they are in humans,” he said. “That’s where things sort of fall apart. The question will be, ‘Do humans have the same profile as the mouse tissue?’ We don’t know.”

The next step will be to refine the biomolecular targeting computer and eventually test it in humans. Benenson and his colleagues at ETH Zurich created a biotech called Pattern BioSciences to do that with this therapeutic and develop other drugs.

“It makes sense to do that in a company,” Wong said. “If this was my work, I would make a company, too.”

An earlier version of this article incorrectly identified the journal in which the study was published.

Create a display name to comment

This name will appear with your comment

There was an error saving your display name. Please check and try again.