Shrink-wrapped E. coli that can only multiply inside a tumour

Two of six treated mice later rejected a second tumour outright, with no further treatment.

The Giant Problem

Live bacteria are unlike every other cancer drug in one respect that matters: they multiply. A drug molecule gets cleared. A bacterium that lands in the liver can settle in and keep going, and one study found detectable Salmonella in crab-eating monkeys more than forty days after treatment. The standard fix is to break the bacteria first. Knock out the genes they use to cause disease, then dose the weakened strain. The trade is a bad one. Attenuate enough to be safe and the bacteria usually lose the ability to colonise the tumour, which was the entire reason for using them. Clinical trials of attenuated Salmonella typhimurium have failed on exactly this point. The other route is to inject bacteria straight into the tumour and skip the bloodstream. That does help. It also only works on tumours you can reach with a needle, which rules out a large proportion of them. So the field has been stuck between a bacterium too dangerous to inject and one too feeble to bother with.

The Science

The coat goes on one cell at a time. Bacteria are suspended in a solution of small polymer building blocks; an initiator is added, and the pieces link up into a mesh directly on each bacterial surface, a net roughly 30 nanometres thick. Coated cells came out just over a micrometre across, and more than 90% of them were still alive. That mesh does several things at once. It hides the bacterium, so mice given coated cells had far lower blood levels of the inflammatory signals TNF-alpha and IL-6 than mice given bare ones. It shields the bacterium from attack: lysozyme, the enzyme that dissolves bacterial cell walls, visibly wrecked uncoated cells at 50 micrograms per millilitre and left coated ones intact. And because it is a physical cage, the cell inside cannot divide. Coated bacteria stayed in the blood with a half-life of 18.8 hours against 7.05 hours for bare ones, and they did not multiply in heart, liver, spleen, lung or kidney. A cage that never opens is a very safe placebo. So Jianhui Yang, Ao Peng and colleagues at Anhui Medical University in Hefei rebuilt the mesh's cross-links out of a short peptide, GPLGVRGK, that gets cut by MMP-2, an enzyme that tumour tissue produces heavily. Left alone in a dish, the coating shed under 5% of itself in a day. With the addition of MMP-2, it shed 88% in eight hours, and the bacteria inside started dividing again. The bacterium is the drug. It is E. coli Nissle 1917, a probiotic strain, carrying a plasmid that makes it do two things. It converts ammonia, the waste that piles up in tumours from their heavy glutamine use, into L-arginine, the amino acid T cells need to mount an attack. It also secretes a loose fragment of the PD-1 receptor, which soaks up the PD-L1 that tumour cells display to switch T cells off. The engineered strain made 16.3 times more arginine than ordinary Nissle. In treated tumours, arginine ran 6.45 times higher than in untreated mice. All of this was in mice.

How They Did It

The experiment worth describing is the last one. Mice with melanoma were treated on days 6, 8 and 10, had the primary tumour surgically cut out on day 13, and on that same day received a fresh injection of the same melanoma cells in the opposite flank. Nothing further was given. The question was whether the immune system had learned anything. In the treated group the new tumours grew slowly, and in two of six mice they did not grow at all. Those animals had more CD8+ T cells (the killers) inside the rechallenged tumour than any other group. A separate set of mice got melanoma cells intravenously instead, seeding the lungs; twelve days later the treated animals had visibly fewer tumour nodules on the lung surface.

Why You Should Care

Five of six mice with melanoma were alive at day 50 after treatment with the full system. Antibody checkpoint blockade on its own barely dented these tumours, which is the comparison that matters, since that antibody class is what oncologists actually have. What makes the platform worth watching is that the coating is not tied to this bacterium or this payload. The monomers and cross-linkers are commercially available, and the cross-linker is the part that decides when the capsule opens. Swap the peptide and you could in principle aim at a different enzyme in a different disease. The authors suggest the same wrapping might work on microalgae, bacteriophages and fungi. The distance to a patient is long. There is no human data of any kind here, and the paper itself notes that earlier bacterial cancer therapies were dropped from early-phase trials. A realistic read is years more animal work, larger species, longer follow-up, and a manufacturing process that can coat billions of cells reproducibly, before anyone writes a trial protocol.