The Promise of Gas Vesicles

Words by
Niko McCarty

In the 1890s, a German microbiologist named Dr. Heinrich Klebahn took a train to the quaint town of Plön and met a friend at the train station there. The following day, the two men rowed a canoe through one of the nearby cerulean lakes, their oars slicing through the water’s calm surface.

As the men glided further into the lake, they arrived at a “yellowish clump” that appeared strangely bright in the sun’s afternoon rays. Klebahn, who had a broad mustache and goatee, donned bifocals and peered into the water. With the palm of an outstretched hand, he scooped up a chunk of the yellow substance and placed it carefully into a container. It was dense, the microbiologist noted, and seemed to be a large algae colony. But how did these cells, which appeared “heavier than the water,” as Klebahn later wrote, manage to float in the middle of a lake without sinking?

The algae in question were Gleotrichia echinulata, a widespread species of cyanobacteria with a stunning, yellow hue. These algae are “one of the most interesting, at least one of the most striking, of the plankton,” wrote Klebahn, who transported the cells back to his laboratory in Hamburg, one hundred kilometers to the south.1

(Left) Plöner Schloss, a castle built in the 1630s, overlooks Lake Plön. (Right) Heinrich Klebahn photographed as a young man. Credit: Fielmann AG.

Once alone with his microscope, Klebahn peered at the cells and observed spindly tendrils radiating from a central sphere (the cells look a bit like a Koosh ball). Klebahn was aware of prior studies which speculated that Gleotrichia algae refract the light—and thus appear “oddly bright” in the sun—due to gas bubbles hidden within them. Intrigued, Klebahn did what any inquisitive scientist might: He tried to pop them.

Klebahn placed the algae cells into an enclosed chamber beneath his light microscope, and then pumped in air until its internal pressure rose to twice, then three-times, normal atmospheric levels. But the gas bubbles did not pop or even change in size as he applied greater pressures. From this observation, Klebahn correctly inferred that the “refracting entities” within the cells were not bubbles at all but rather rigid structures.2 He called them gas vacuoles and published his results in a niche German journal, called Flora, in 1895.

At the time of publication, Klebahn could not have anticipated that his observations would prove most useful not during his lifetime but more than a hundred years hence. Much like Gregor Mendel’s experiments on peas, which laid the foundation for modern genetics but languished in obscurity for decades, Klebahn’s discovery of these rigid structures—now referred to as gas vesicles—is only just bearing fruit.

After decades of study, scientists figured out that gas vesicles are little more than protein shells that trap gas, rather than liquids or solids. And this singular quirk endows them with a supremely useful trait: They show up on ultrasound scans.

Gleotrichia echinulata cells have spindly tendrils radiating from a spherical core. They also possess gas vesicles, a type of protein shell used to control buoyancy. Credit: Barry Rosen.

Ultrasound machines emit high-frequency sound waves. Those sound waves cycle up and down millions of times per second, well outside the limits of human hearing. If a sound wave travels through soft tissue and then encounters bone, a great deal of energy is directed back because the two materials have a large difference in acoustic properties.3 But if a wave passes from one type of soft tissue to another with similar properties, a smaller amount of energy is reflected. By measuring the length of time that it takes for a soundwave to enter and return, an ultrasound device constructs an internal image of the body. Gas vesicles show up brightly on ultrasound scans because they reflect soundwaves in a way that tissues in the body, which are composed of solids and liquids, cannot.

Seeking new tools to study cells in situ, scientists have plucked the genes responsible for making gas vesicles in algae and inserted them into myriad organisms that don’t normally make gas vesicles, including E. coli and human cells. When these engineered cells are then placed into a mouse, one can use ultrasound to “hear” the cells as they move around. One can even explode the gas vesicles using high pressure soundwaves, akin to molecular warheads, with astounding implications for medicine.

This is the story of how a peculiar protein shell, first discovered in a German lake more than one hundred years ago, is adding a fundamentally new dimension to biology. For the first time, gas vesicles are enabling biologists to study cells not only in plates and petri dishes but inside the body itself.


“Heavier than Water”

Klebahn’s 19th century paper briefly percolated through the halls of elite European institutions but was then largely forgotten. He died in 1942, decades before his paper received the acclaim it rightly deserved.

A rediscovery of Klebahn’s work began in 1965 when two botanists at Iowa State University, C.C. Bowen and T.E. Jensen, reported a much more detailed analysis of gas vesicles in the journal Science. Bowen and Jensen made careful measurements of gas vesicles’ dimensions and also exposed them to higher pressures than Klebahn’s primitive equipment could muster. Gas vesicles measure up to one micron in length, they noted, and are “spike-like” in appearance.

By 1977, scientists had begun studying the amino acids comprising gas vesicle proteins. But the technology available at that time was poor, and initial reports were limited to an analysis of short protein fragments. A full sequence of GvpA, a protein that forms the core structure of a gas vesicle, was only decoded in 1986. A couple years later, scientists discovered that another protein, called GvpC, binds to GvpA on the exterior of gas vesicles and lends them additional strength. More than a dozen different proteins are now known to play a supporting role in folding and forming gas vesicles, which cells use to control buoyancy. As the sun dips in the sky in the late afternoon, light-metabolizing microbes make gas vacuoles and float up toward the lake’s surface to capture more of the fading light for photosynthesis.

Even after a half-century of molecular biology research, scientists didn’t know until recently how gas vesicles physically trap gas while occluding water. Last year, researchers at Delft University of Technology in the Netherlands unveiled a molecular structure of gas vesicles taken from a bacterial species called Bacillus megaterium at a resolution of 3.2 angstroms, or roughly twice the length of a single carbon-carbon bond. The structure revealed tiny pores in the protein shell that allow gas molecules to diffuse in, while a hydrophobic interior keeps water out.

For more than a century, scientists thought of gas vesicles as little more than a natural curiosity and, later, as a way for microbes to regulate buoyancy. It was only once gas vesicles were fished from the natural world and engineered in the laboratory that the next great leap in their evolution happened.

Light and Sound

Progress in biology has long been driven by light. Microscopes allow one to “see” by redirecting the light passing through a thinly sliced specimen back into the human eye. A Dutch spectacle maker, Zacharias Janssen, was likely the first person to make a compound microscope, sometime around the year 1600. Janssen’s device could magnify specimens by 20 or 30 times.

It was another Dutchman, Antonie van Leeuwenhoek, who made seminal improvements to the microscope’s design and boosted its magnification to 200 times or greater, using his homemade device to view “animalcules,” sperm, and blood. One could reasonably argue that, in the centuries since, biologists’ understanding of the cellular world has been enabled by the microscope more than any other tool. After all, observation begets understanding.

A transmission electron micrograph of a single E. coli cell that had been engineered to express gas vesicles. The cell measures 2 micrometers across, and each lighter-colored structure is an individual gas vesicle. Credit: Anupama Lakshmanan/Caltech.

But light has limits. In the visible wavelengths, it cannot penetrate tissue and so cannot be used to see inside the body. For this reason, biologists mainly observe external, or in vitro, phenomena, such as cells in petri dishes. Our understanding of the natural world has a blind spot in its center: the human body.

X-rays can be used to image structures inside the body, but only because they emit high-energy photons that blast through tissue. Fat and muscle cells are built from light elements, such as carbon and oxygen, that cannot absorb the photons; bones show up as a ghostly white because they consist of metals, such as calcium, that can.

As high-energy photons pass through the body, however, they damage cells in their path. Too much exposure can induce cancer-causing mutations or fry tissue. The only way to “safely” view individual cells from the body, historically, has been to slice up a cadaver or to take a biopsy by inserting a hollow needle through the skin and then studying the cells in culture.

Another option for peering inside the body is to forego light entirely. Magnetic resonance imaging, or MRI, machines have improved rapidly in resolution and can resolve features that span just 0.4 millimeters across. But a blood cell measures about seven microns across, nearly 60 times smaller than the maximum MRI resolution.

This is why, for the last 400 years, it has not been possible to directly and safely observe cells within the body. The physics simply does not make sense for visible light, magnetism or X-rays. But ultrasound is different.

Dr. Karl Dussik, an Austrian physician, was among the first to suggest that ultrasound could be used to diagnose diseases within the body in 1941. With help from engineers at the Physics Institute of the University of Vienna, Dussik fabricated a quartz ultrasound generator, complete with a transmitter and receiver. His inspiration stemmed from, of all places, a scientific paper from the mid-1930s that suggested ultrasonic waves could be used to detect “schools of fishes and boats at sea.”

The use of ultrasound first took off in obstetrics, where the technology was routinely used to detect abnormalities during pregnancy. A Scottish physician, Ian Donald, was the first to publish an ultrasound image of a fetus in 1958. He also invented the first obstetric ultrasound machine, called the Diasonograph, in 1963.

Prolonged exposure to ultrasound can cause nausea, headaches, and vertigo. In the late 1940s, military mechanics working on jet engines—for several hours at a time, mind you—began reporting an “ultrasonic sickness.” However, studies over the last six decades have repeatedly affirmed that ultrasound is safe, at least when used in the short durations common to medical imaging and treatment. About 72 times more power from sound waves can safely pass through the human body compared to radio waves since the latter emits a small amount of non-ionizing radiation.

The largest weakness of ultrasound is its resolution, or ability to distinguish between two closely spaced objects. Higher-frequency sounds generally produce higher-resolution images but do not penetrate tissue as deeply. Ultrasound’s resolution peaks around 30-50 microns, four times larger than the length of a red blood cell.

(Left) A micrograph of a cyanobacterium with cylindrical gas vesicles stacked in hexagonal layers. 31,500x magnification. (Right) The passage of text from a 2010 review that inspired Mikhail Shapiro to engineer gas vesicles. Credit: Adapted from A.E. Walsby, 1994 & M. Fischbach & C. Voigt, Biotechnology Journal (2010).

Mikhail Shapiro, a young biochemical engineer at Berkeley, had been thinking on how to improve ultrasound resolution throughout his Ph.D., during which he used several imaging techniques, including MRI, to study rat brains. In 2010, Shapiro picked up a review article and noticed two short paragraphs—included almost as an afterthought—about gas vesicles. His five-minute read would later awaken gas vesicles from a century-long torpor, transforming them into a sort of microscope for the 21st century; a tool to image individual cells moving within a body using ultrasound.

Moving Molecules

Shapiro has black hair and a broad smile. While listening, he has a tendency to lean forward, gaze sharply into your eyes, and steeple his fingers.

A neuroscientist and engineer by training, Shapiro completed his Ph.D. at MIT in four years. In his thesis, he wrote about a desire to invent tools that “could provide information about the concentrations or activities of specific molecules in living subjects.” During his Ph.D., Shapiro evolved a small bacterial protein in the laboratory, called BM3h, that could sense dopamine, a neurotransmitter in the brain. When dopamine levels are low, the BM3h proteins brighten an MRI signal; when dopamine is high, the signal fades away. When Shapiro injected BM3h into the brains of rats, he found he could use MRI to quantify dopamine levels in real-time and with submillimeter resolutions.

A few years later, Shapiro moved to the University of California, Berkeley to start a postdoctoral fellowship. That’s when he read about gas vesicles in the review article and resolved himself to the task of transforming ultrasound from a relatively low-resolution tool to one able to track individual cells in motion.

Spending a small amount of money from a Miller Fellowship, Shapiro ordered several microbes known to produce gas vesicles in the mail and cultured them in flasks. He carefully placed each flask on the top shelf above his lab bench, so they would be closer to the ceiling lights for photosynthesis. While sifting through the basement of an old campus building in Berkeley, Shapiro discovered an abandoned ultrasound machine from the 1990s, which he would later use to probe the microbes and see ultrasound contrast from gas vesicles for the first time. Shapiro purified vesicles from two types of microbes, Anabaena and Halobacterium, and imaged them in the laboratory at various ultrasound frequencies, from 4.8 up to 17 MHz. When he increased the ultrasound pressure beyond a threshold, the gas vesicles burst open—pop!

Mikhail Shapiro gives a lecture in Cambridge, Massachusetts as a Ph.D. student. Credit: Jean Baptiste Paris.

But Shapiro did not stop there; he wanted to know if it was actually possible to see gas vesicles directly inside of a mouse. So he ruptured some Halobacterium cells, carefully purified their gas vesicles, and injected them into mice. “Control” mice were injected with saltwater. The animals injected with the gas vesicles had clear ultrasound contrast, whereas the control animals did not. When Shapiro then ramped up the ultrasound pressure and popped the vesicles, any difference in the contrast disappeared between the animals. His 2014 paper, published in Nature Nanotechnology, was the first demonstration that gas vesicles could be imaged and popped in vivo.

An engineer’s mind, however, does not sit idle for long. An initial observation often raises more questions than it answers, leading to a deep desire to delve further and refine the system.

By 2016, Shapiro had set up a research laboratory at the California Institute of Technology in Pasadena and began to devote his attention entirely to gas vesicles. His first aim was to take the genes encoding gas vesicles in nature and shuttle them into a model organism that doesn’t normally make gas vesicles, such as E. coli. But this is difficult because gas vesicles are made by “a cluster of genes that have to work together to form the structure,” according to an article in Nature Methods. One cannot simply pluck gas vesicle genes from a cyanobacterium and plop them into E. coli or a human cell. Most gas vesicles are built from at least eight different genes, all of which work in delicate concert.

Taking gas vesicle genes from Bacillus megaterium—a bacterium called the “big beast” because a single cell can measure 100 microns in length, roughly the thickness of a human hair—and porting them into E. coli cells proved an experimental dead-end. The engineered cells made some “small, bicone-shaped gas vesicles” but were not “detectable by ultrasound, most probably because the small gas vesicles…[had] weak acoustic scattering,” Shapiro later wrote. In other words, the gas vesicles were either sickly and collapsed under normal ultrasound pressures or else dissipated before they could be detected.

A major breakthrough came in January 2018, when a paper in Nature indicated, for the first time, that gas vesicles could be expressed in E. coli cells.

The insight leading to this success was to mix-and-match genes from multiple organisms so that they would work in concert in E. coli. A postdoctoral fellow, Raymond Bourdeau, took the gvpA gene from Anabaena flos-aquaea, which makes strong gas vesicles and has high acoustic scattering, and remixed it with other genes from B. megaterium. By combining DNA from multiple organisms and inserting it into E. coli, the cells formed functional gas vesicles.

This was the first demonstration that an organism, with absolutely no evolutionary reason for making protein shells to control buoyancy, could be engineered to make gas vesicles nonetheless. Each E. coli bacterium made about 100 gas vesicles, which together occupied about 10 percent of the cells’ interior. About 50 million gas vesicles, packed into a space of one milliliter, were required to detect an ultrasound signal; however, the E. coli gas vesicles were a bit sickly and did not efficiently scatter soundwaves.

Cyanobacteria (green) make gas vesicles and float to the top of their containers. Credit: Mikhail Shapiro

Just last year Shapiro’s group succeeded in making engineered gas vesicles that produce contrast 38-fold higher than their initial versions. These enhanced gas vesicles were discovered by using computer algorithms to sift through the genomes of 288 different microbial species and then mix-and-matching those genes together to identify optimized gas vesicle structures. In 2019, a Ph.D. student in the lab, Arash Farhadi, also became the first person to successfully express gas vesicle genes in human cells, thereby setting the stage for imaging individual microbial or human cells as they move through the body.

Fished from Nature

At least a dozen laboratories are currently using gas vesicles to study living organisms. The vesicles have been so exquisitely engineered at this point, says Farhadi, that it is possible to use them to image cells at “almost single-cell sensitivity.” A 2021 paper from Daniel Sawyer, a Ph.D. student in the Shapiro laboratory, described a method called BURST (Burst Ultrasound Reconstructed with Signal Templates) “that improves the cellular detection limit by more than 1000-fold” compared to other ultrasound methods.

BURST works by blasting gas vesicles inside lab animals with acoustic pressure. This pressure causes gas vesicles to rapidly collapse and emit a signal distinct from those produced by background tissues. A signal unmixing algorithm is then used to separate gas vesicle signals from everything else in the ultrasound data. Remarkably, this enabled Sawyer to image “individual bacteria and mammalian cells” inside the livers of mice. The initial use of BURST also, remarkably, preceded a mechanistic understanding of how gas vesicles reflect sound.

But after several years of mechanical modeling and biophysical measurements, Shapiro’s team and collaborators say they’ve figured it out. It turns out that gas vesicles behave similarly to a “crushed can of Coca-Cola,” says Farhadi. Just as a soda can buckles when squeezed and makes a loud crack! sound, gas vesicles do something similar at the nanometer scale. Sound waves compress gas vesicles, causing them to buckle, crack, and re-form millions of times each second. Their faint crackles of noise can then be picked up by ultrasound.

A mechanical model of gas vesicle buckling. (Left) A cryo-electron micrograph of a gas vesicle. (Middle) The maximum volume change of gas vesicles, as a percentage, when subjected to increasing pressures from ultrasound. (Right) A computational model depicting the buckling of gas vesicles. Credit: Salahshoor H. et al. Biophysical Journal (2022).

With a mechanistic understanding of gas vesicles, it will be easier to adapt them for new medical applications. Recent studies have shown that it’s possible to physically push gas vesicles through the body, without popping them, in order to move cells toward diseased tissues, for example, or even to explode gas vesicles to destroy tumors.

The latter finding works like this: Inject mice with cancer cells. Wait a few weeks and confirm they formed tumors. Next, inject them with bacterial cells that express gas vesicles. Some species of bacteria naturally “hunt down” and colonize the tumors due to their low-oxygen environments. After confirming that the microbes colonized a tumor, send a concentrated beam of sound waves at the gas vesicles. The high-pressure waves cause vesicles to collapse and release gas. If the ultrasound is kept on, the tiny bubbles oscillate and some merge together. These gas pockets quickly grow in size, contracting and expanding, until they suddenly cavitate. The collapse happens in less than one-millionth of a second, emitting a mechanical shock that kills nearby cancer cells.

In short, it’s now possible to pinpoint the locations of engineered cells as they move through the body, control the movements of those cells, and then explode them to damage tumors or other tissues—all using sound. This could transform medicine.

Arash Farhadi (left) and Mikhail Shapiro (right) setting up an ultrasound experiment in their Caltech laboratory. Credit: Mikhail Shapiro

Just consider a pharmaceutical company that wants to test whether an engineered immune cell, such as those used for CAR-T therapy, can detect and destroy cancer cells. At least six types of CAR-T cells have already passed through clinical trials and garnered FDA approval. But during these clinical trials, only a handful of indicators are measured: the number of cells infused into the body, the size of the tumor over time, and some blood marker proxies for the immune system’s activity. Just about everything happening inside the body itself—the number of cells that make it to a tumor, the time it takes for them to reach that tumor, or the route by which they get there—is a mystery. But now, for the first time, employing a technology like BURST would make it possible to measure them.

Microscopes have existed for some four centuries, yet scientists continue using them to study cells in a dish simply because alternative methods have long been ill-equipped to study them in situ. This is a large part of the reason why attempts to translate in vitro results into real-world therapies, such as medicines, often fail. The deployment of ultrasound and gas vesicles could change this. Together, they offer biologists a sharper lens with which to study life in its holistic complexity.


Niko McCarty is a founder of Asimov Press.

Cite this essay: Niko McCarty. “The Promise of Gas Vesicles.” Asimov Press (2024). DOI:



  1. All quotes from Klebahn are translations from his 1895 publication in the journal Flora.
  2. Gas vesicles are rigid, but Klebahn’s evidence for this finding is inadmissible, according to a 1994 review from botanist Anthony E. Walsby. “A bubble of 1 micrometer in diameter in a cell would have an internal gas pressure of 0.8 MPa (0.1 MPa from atmospheric pressure, 0.3 MPa from surface tension, and 0.4 MPa from cell turgor). Doubling the external pressure to 0.2 MPa would raise the internal pressure to 90.9 MPa, which would…result in a change in diameter of only…4%; this change would be undetectable.”
  3. More specifically, this is called acoustic impedance (Z) and it’s a measure of the resistance that a material places upon the propagation of sound waves. The equation is Z = ρv, where ρ is the density of the medium and v is the speed of sound through it.
Learn More