Identifying bacteria by sight can be quite difficult. Why not listen to them instead?
Researchers at TU Delft in the Netherlands and the university’s spinoff company SoundCell think that bacterial infections could be diagnosed with sound. They’ve crafted a nanoscale drum kit that uses some of the world’s smallest percussion instruments to turn a bacterium’s motions into song.
Previously, the Delft researchers showed that listening to a germ’s drumbeat could quickly screen it for antibiotic resistance. Now, the same researchers have discovered that different bacteria play different sounds on the drum. They’ve shown they can identify a bacterium from its song alone.
“We can really look at the level of a single cell,” says Farbod Alijani, a mechanical engineer at TU Delft and one of the authors of a new paper. “We have that sensitivity.” Alijani and colleagues reported their latest findings this March in ACS Sensors.
How to build the world’s tiniest drum
The Delft researchers call their instrument of choice a “nanodrum.” Its drumhead is fashioned from two graphene sheets, less than 1 nanometer thick, laid atop a 8 micrometer-wide cavity. This size fits most bacteria, which are about one to 10 micrometers in length.
Several years ago, the Delft researchers noticed that, if a living bacterium settled on the graphene sheet, it would beat a pattern on the drumhead. They were detecting the life-form’s subtle motions, such as the whirling of the propellor-like flagellum the bacterium uses to move about. When the drumhead moved, it left signals on a beam of laser light reflected off the surface, allowing the researchers to record the bacterium’s motion.
The drum’s tiny size is key to pinpointing individual bacteria. The Delft researchers were not the first to capture bacteria in motion, but older methods usually had to average the movements of an ensemble of many bacteria because of their micro scale. By comparison, each graphene drumhead is small enough to isolate—and record—a single bacterium.
Graphene is key to this instrument’s construction. The material is both strong enough to support a bacterium’s weight and sensitive enough to bend with each subtle bounce on the drum.
Then, by converting its drumbeat to a soundtrack, it’s possible to literally hear the motions of a living bacterium. “It’s very noisy, like a wind tunnel,” says Aleksandre Japaridze, SoundCell’s Chief Technical Officer, who is also an author of the paper.
By contrast, “if you kill it with a drug, it’s immediately very silent, and you don’t hear anything.” In previously published work, when the researchers pumped an antibiotic onto drums played by E. coli, the drums fell quiet within hours. But when they did the same to E. coli they knew to be antibiotic-resistant, the bacteria played on, seemingly unaffected.
From one song to many
Over the following years, the Delft researchers refined their technology’s ability to screen bacteria for antibiotic resistance. Let a patient’s bacteria play the drums, then administer a given antibiotic—if the music stops, that antibiotic should work.
But their work took an unexpected turn after an attendee at a conference asked Alijani if different bacteria made different sounds. Unsure of the answer, the researchers wondered how they could find out.
It was clearly possible to tell a living bacterium from a dead one by listening alone, but separating one bacterium from another species required a more sophisticated approach. The Delft researchers recorded the drumbeats of different infectious bacteria from real patients’ samples. Instead of using raw sounds, the researchers processed them into time-frequency spectrograms, which allowed the researchers to more carefully study the patterns of each bacterium’s motion.
The researchers trained two different machine learning models to examine a spectrogram and identify its drummer as one of three species: E. coli; Staphylococcus aureus (responsible for staph infections); or Klebsiella pneumoniae (one of the germs that can cause pneumonia).
Both models, each with a different underlying architecture, scored high marks in testing: One classified bacteria with 87 percent accuracy, and the other achieved 88 percent. These results suggest that each species plays different characteristic notes when it moves on the drum.
“It’s a completely different way of interpreting the different species,” Japaridze says. “Not chemically or biologically, with markers and genes, but just purely on…mechanical behavior.”
Diagnosis through song and dance
The Delft researchers think their drums are a powerful diagnostic tool. SoundCell was originally spun off to commercialize the ability to quickly and easily determine whether a bacterium is resistant to a given antibiotic, and the researchers hope hospitals in the future will also listen to the songs of a patient’s sample to identify the infection.
Antimicrobial-resistant germs may be responsible for more than one million deaths each year and may play a part in millions more. There are many reasons that antibiotic-resistant bacteria are potent threats—one is that the tests for whether a microbe is resistant are relatively slow. Today’s tools may take days to report if a microbe is resistant to a given antibiotic. By comparison, SoundCell’s technology can do this in as little as an hour.
First, SoundCell must show their nanodrums can work in the hospital. The Delft researchers’ published work was conducted on a bulky apparatus on an optical table, within the controlled confines of a laboratory. So, SoundCell has repackaged their nanodrums into a smaller device better suited for hospital use.
SoundCell has now deployed this device at two hospitals in the Netherlands, Japaridze says, to verify their performance.
The chip projected a roughly 125-micrometer image of the Mona Lisa.
