The E-Nose: Scientists Compete with Nature's Prolific Sniffers

Illustration: Ned Shaw Volatile odorants spewing forth from every living thing reveal a hidden trove of factors like diet, health, and genetic composition. With astonishing acuteness, most animals can read these olfactory messages and model their behaviors accordingly. "Odors are their windows on the world," one researcher says of mice. Others have noted that some honeybees recognize illness in hives and avoid them.1 Dogs have alerted their owners to such maladies as melanoma2 and epileptic seizures.3 For years, the industrial and defense arenas have used electronic sensors to detect toxic gases. Now, life scientists want to know if similar instruments, dubbed 'E-noses,' could be designed to make accurate medical diagnoses: A clinician might simply wave a wand over a patient and check for olfactory signatures. The future for such technologies evokes images more familiar to Star Trek fans, but throughout history doctors have relied on their noses to assess patients.4 Physicians have associated scrofula with a beer odor, typhoid with a brown-bread aroma, and diabetes with a chewing gum (Juicy Fruit) scent on the breath. The Latin fetor hepaticus describes malodor associated with severe liver disease--even kidney failure has a reported reek. Despite the historical associations between disease and smell, most authors of medical handbooks have ignored the sniff test, because the nose is subjective but diagnostic technologies are not. Clinical results are promising so far, researchers say. The growing complexity and usability of these E-noses keeps churning up new applications both within and outside the medical arena. The devices now monitor food quality, and applications are blooming for various infections, environmental hazards, explosives, and even innocuous odors in the home. A LESSON FROM NATURE A mouse's sense of smell is so keen, says Gary Beauchamp, director of the Monell Chemical Senses Center in Philadelphia, that a male can tell whether a female is pregnant with his pups or the pups of another male. In testing the limits of murine olfactory faculties, Beauchamp and his collaborators infected sets of mice with murine mammary tumor virus (MMTV), a well-studied infection with almost imperceptible effects. The mice could tell the difference. Mice were trained to associate a reward with the scent of either an infected mouse's urine or an uninfected one. When tested in a Y-maze experiment, the mice reliably discerned between MMTV+ and MMTV- odors even when the virus was transgenically induced.5 Beauchamp says that chemical analyses will determine what the odorants are; the study suggests that they result from an immune response to the infection. "Ultimately," he says, "we'd like to see if we can develop ways of detecting these smells with artificial nose technology." But getting the technology up to snuff is something of a challenge. Timothy Pearce, lecturer in bioengineering, University of Leicester, UK, says he looks to biology for inspiration to improve the engineering performance of such systems. His group works on mimicking the mammalian olfactory system on a silicon chip. They are also designing an autonomously controlled, unmanned helicopter to conduct chemical searches; the system is based on the circuitry of a moth's brain. One problem, he says, is figuring out where to start. From receptor to neuron to olfactory bulb to brain, infinite levels exist at which olfaction could be simulated. "A single neuron in our brain could be modeled with a supercomputer. ... Clearly, that's not practical when you're trying to develop a silicon implementation of the olfactory pathway," says Pearce. John Kauer, professor of neuroscience at Tufts-New England Medical Center in Boston, says researchers started realizing in the 1980s that using just one electronic sensor in a simple lock-and-key detection system did not give many answers. A simple yes or no when it comes to odor is simply not enough. With a number of different sensors arranged in an array, more like the scent receptors in a biological nose, a system could combine the varying signals and match the patterns they produce against a database of known scents, much like the brain does. The sensors, which play the part of smell receptors, can be such things as polymers, metal oxides, or fluorescent molecules that change their properties in response to specific volatile compounds. And even these sensors can produce various results in response to different scents at varying concentrations. For E-noses, mimicking an organism's receptor sensitivity is not hard, even one as prolific as a moth that can reportedly pick up a pheromone trail from a mile away. "The brain's the thing that's hard to reproduce," says Beauchamp. The brain makes minute distinctions between thousands of specific odorants in perfectly balanced mixtures. Kauer's examinations of salamander olfactory systems reveal a complex communication system between sensory neurons, which conduct signals from scent receptors in the nose to the relay cells in the olfactory bulb. Different odorants can stimulate the same receptors, causing the same neurons to stimulate different areas of the olfactory bulb. Not only spatial differences but also temporal differences in the stimulation patterns can be triggered over a sniff's duration.6 Kauer urges that one consider a dog that will control frequency, duration, and depth of a sniff during, say, a rendezvous with a fire hydrant. He envisions the scent's signature as a topographic map, but one that changes over time. "It's a little like watching the Rockies appear and subside." Kauer and colleague Joel White developed an E-nose with the capability of discerning both spatial and temporal pattern changes in smell. They created the device to sniff out landmines through funding from the US Office of Naval Research and the Defense Advanced Research Projects Agency. Photo: Courtesy of C. William Hanson DOCTOR NOSE: A hand held E-nose can be attached to a patient's ventilator circuit. NOSING AROUND IN THE CLINIC Clinician C. William Hanson, professor of anesthesia, surgery and internal medicine, University of Pennsylvania School of Medicine, says his interest in E-noses perked up when he read a Wall Street Journal article in the mid-1990s that profiled companies using E-noses for food and wine testing. "It occurred to me that using the sense of smell has been a medical tool from the oldest recorded medical literature," says Hanson. He contacted UK-based Osmetech and later Cyrano Sciences in Pasadena, Calif., to see if such devices could detect pulmonary infections in critically ill patients using respirators. Many bacteria have distinct metabolite signatures that give away their presence. These signatures can include such vapors as unburned hydrocarbons, aldehydes, short-chain fatty acids, and other breakdown products of amino acids, says chemistry professor Joseph Stetter, Illinois Institute of Technology, whose proposed applications for E-nose technologies include tuberculosis detection. Osmetech received FDA approval for an application to detect urinary tract infections this past November; the bacteria must be cultured in the presence of special media once the sample is taken. The goal for diagnosis of pulmonary infections, Hanson says, is detection in real time from the patient's breath. Researchers are currently calibrating the detection process with Cyrano's handheld E-nose, but the variability from patient to patient makes for a difficult background. "Some patients might have pneumonia plus liver disease, or pneumonia plus kidney disease, or pneumonia plus starvation. All those things give different characteristic odors," Hanson says. Looking for broad trends in varying backgrounds can be difficult, and they may not fully validate the process. Analyses like gas chromatography/mass spectrometry (GCMS) or high-resolution mass spectrometry can further define what the E-nose must detect to diagnose disease accurately. But, E-nose sensors sometimes respond better to certain gas species, says Stetter, citing carbon monoxide, which GCMS cannot detect. A messy background also leaves the door open to false negatives, which are troublesome for diagnoses, and downright unacceptable for land mine detection. Whether an E-nose will ever live up to its biological counterpart is open to debate; the competition is stiff. A dog, says Stetter, can "walk into a room with 50 people and find its owner blindfolded." Brendan A. Maher can be contacted at bmaher@the-scientist.com. References 1. J. Fisher-Wilson, "The nose knows: How the olfactory influences conduct," The Scientist, 15[24]:22, Dec. 10, 2001. 2. R. Masterman et al., "Olfactory and behavioral response thresholds to odors of diseased blood differ between hygienic and non-hygienic honeybees," Journal of Comparative Physiology, 187:441-52, 2001. 3. J. Church, H. Williams, "Another sniffer dog for the clinic?" Lancet, 358:930, 2001. 4. W.Z. Stitt, A. Goldsmith, "Scratch and sniff. The dynamic duo," Archives of Dermatology, 131:997-9, 1995. 5. K. Yamazaki et al., "Presence of mouse mammary tumor virus specifically alters the body odor of mice," Proceedings of the National Academy of Sciences, 99:5612-5, April 16, 2002. 6. J. Kauer, "On the scents of smell in the salamander," Nature, 417:336-42, May 16, 2002.

The E-Nose: Scientists Compete with Nature's Prolific Sniffers

Interested in reading more?

Become a Member of