A suspicious-looking setup
In 2024, a strange scene appeared in an ETH Zurich lab: PhD student Gabriel Giger hooked a bicycle pump, via a tube, to the micro-needle of an atomic force microscope. The needle tip—thinner than a cell—was aimed at a fungal spore. He pumped until the gauge read three times car-tire pressure, squeezing a batch of bacteria through the fungal cell wall into its cytoplasm.
The contraption looked like a prank, but it was attacking a two-billion-year-old question: how does one cell get inside another cell, and how do the two survive together?
Every cell in your body contains mitochondria—the descendants of a bacterium engulfed by an archaeon that was never digested and instead stayed on as a power plant. Chloroplasts arose from a similar event with a cyanobacterium. And in April 2024, scientists confirmed the "nitroplast" in the alga *Braarudosphaera bigelowii*—a nitrogen-fixing organelle caught mid-transition from endosymbiont to organelle, a live rerun of that ancient merger's early stages.
Endosymbiosis made all complex life possible. But no one had ever watched it begin—until Giger stepped on that bike pump.
We only knew the ending
Endosymbiosis means one cell permanently living inside another—not infection, not parasitism, but a merger into an inseparable whole:
- Mitochondria: descended from free-living alpha-proteobacteria engulfed by an archaeon ~2 billion years ago.
- Chloroplasts: descended from cyanobacteria engulfed by a eukaryotic cell.
- Insect endosymbionts: aphids rely on theirs to synthesize amino acids.
- Nitroplasts: confirmed in 2024 in a marine alga, transitional "living fossils."
- Original paper: Giger, G., et al. *Nature* 635, 415–422 (2024). "Induction of endosymbiosis by bacterial injection into a fungal host."
- Quanta Magazine: Molly Herring, "Scientists Re-Create the Microbial Dance That Sparked Complex Life" (2025-01-02)
- Nitroplast discovery: Coale, T. H., et al. *Science* (2024). "Nitrogen-fixing organelle in a marine alga."
- ASM: Vilhelmiina Haavisto, "Beyond Endosymbiosis: Discovering the First Nitroplast" (2024-06-20)
The mystery: a swallowed bacterium should normally be digested—or eat its host from within. Peaceful coexistence is astronomically unlikely. Theories (a "Goldilocks" replication rate, hitchhiking into the host's reproduction cycle, host genome mutations) were all reverse-engineered from the outcome. "We know the marriage happened, but nobody saw the first date."
The bike pump and the AFM
Microbiologist Julia Vorholt decided to stop guessing and make endosymbiosis happen in the lab. Her pragmatic strategy: replay a merger nature had already pulled off—the fungus *Rhizopus microsporus* (a rice seedling blight pathogen) and its bacterium *Mycetohabitans rhizoxinica*, which makes a toxin the fungus uses to kill rice cells. She used a fungus strain naturally lacking the bacterium, offering a clean-slate rerun.
Problem one was physical: how do you push a bacterium through a fungal cell wall? Giger softened the wall with enzymes, then used a FluidFM micro-needle—normally for single-cell microsurgery—as a syringe. But cytoplasm sprayed out of the puncture: "like a dam breaking." Internal pressure kept the bacteria out.
Lacking suitable equipment, he cannibalized his own bicycle. Pumping to three times car-tire pressure finally forced the bacteria through the wall.
> "Hooking a bike pump to an atomic force microscope to stuff bacteria into a fungus sounds like a joke, but it represents a major technical breakthrough," said evolutionary biologist Thomas Richards of Oxford.
E. coli fails: too fast means infection
First attempt: *E. coli*. It replicated too fast, exhausted nutrients, tripped the fungus's immune system, and was walled off for digestion—wiped out within days. Not any two cells can cohabit; matched replication rate is a hard threshold.
Then came the true partner, *M. rhizoxinica*: it divided at a "just right" rate, triggered no immune alarm, and both cells survived. "Just seeing that both were still alive after injection was super exciting," Giger said.
Hitchhiking to the next generation
Survival is only step one. Real endosymbionts must enter the host's reproductive cycle. Under the microscope, Giger eventually watched the bacteria wriggle into fungal spores—meaning the next generation inherits them from day one. "I had trouble sleeping; the excitement lasted a long while."
He hand-picked spores and propagated ten fungal generations. Each generation, bacterial survival rates rose and spores grew healthier. Crucially, sequencing revealed the fungal genome had mutated to accommodate the bacteria—the host was changing too, not just the bacterium. "They became unable to live apart," said Vorholt.
The overlooked half: bidirectional adaptation
Classic theory treats the host as a passive container while the bacterium adjusts and sheds genes. This experiment shows the host actively changes within ten generations. "A fundamental, overlooked question—this door is now open," Richards said. The ancient merger may have been a two-way negotiation, not one-way colonization.
> "To me, this means organisms want to live together—symbiosis is the norm," said mycologist Vasilis Kokkoris (VU University, unaffiliated).
But honesty demands noting the failure rate: most cell-to-cell unions fail—mismatched replication rates, costs exceeding benefits, insufficient selection pressure. Endosymbiosis may be common in nature's history precisely because we only see the rare successes.
After the bike pump: synthetic endosymbiosis
The practical frontier is synthetic endosymbiosis: design a bacterium for a task, let it move into a host cell, and let two organisms evolve together—e.g., pollutant-metabolizing bacteria inside plant cells, or drug-synthesizing bacteria inside fungi. "Our imagination is the only limit," said Laila Partida Martínez (Cinvestav, Mexico), who originally discovered the rice blight endosymbiosis.
Can humans get chloroplasts? Giger is candid: mammalian cells struggle to stabilize them, and photosynthesis is too energy-poor for humans—"you'd get green skin, run a little solar power, but you'd still be hungry and need pizza to make up the difference."
Mergers as the unit of innovation
Major leaps in the tree of life came not from one species optimizing itself, but from two merging into something new: mitochondria, chloroplasts, nitroplasts. Technology history rhymes: the internet merged computer networks with telephone networks; the smartphone merged phones with personal computers.
What about AI? The next leap may not be "a bigger model" but a merger—an LLM fused with a Prolog reasoner, a physics simulator, or a human. Like the bacterium shedding genes, the LLM must give something up; like the mutating host genome, we must change how we work. After enough generations, neither side can leave. That is different from the "AI as a dismissible assistant" story—mitochondria cannot be fired.
Closing metaphor
Giger solved a two-billion-year-old puzzle with an everyday object. Big mergers—cellular or technological—may not need fancier equipment, just an odd pairing of two unconnected things, enough pressure, and patience.
Two billion years ago, an archaeon swallowed a bacterium and failed to digest it—an accident. In 2024, a PhD student squeezed bacteria into a fungus with a bike pump and they lived—a designed accident.
Who is stepping on the next pump?
---
References: