快猫短视频

HALF a minute is all it takes. After three 10-second training sessions, Glen Rains鈥檚 crack team of sniffers is ready for anything. They could be co-opted into the hunt for a corpse. They might join the search for a stash of Semtex or a consignment of drugs. Or they could have the more tedious job of checking luggage at the airport. Whatever the assignment, their role is the same: to pick up a scent no human nose can detect and pinpoint its source. These new recruits to the fight against crime are smaller, cheaper and more versatile than a sniffer dog, and more sensitive than an electronic 鈥渘ose鈥�. They are wasps.

Insects have exquisitely sensitive olfactory systems. Their antennae are covered with microscopic sensors that can detect the faintest odour. Some are also remarkably quick learners. So it is hardly surprising they have aroused the interest of the military and security services, police and customs, all badly in need of ultra-sensitive, flexible and portable odour detectors. Insects obviously have the right stuff, but can they use it to sniff out smells they would never encounter in nature 鈥� a hint of explosives, say, or a whiff of cocaine? And if so, is it possible to make a practical device that harnesses their skills?

Enter Wasp Hound, a hand-held odour detector with a team of little black wasps as its sensor. Developed by Rains, a biological engineer at the University of Georgia, his colleague Sam Utley and Joe Lewis, an entomologist at the US Department of Agriculture鈥檚 Agricultural Research Service in Tifton, Georgia, the device is still only a prototype, but the team has high hopes for it. As well as helping with security and forensics, it can also be used to assure food quality, checking grain stores for contaminants, for example. Eventually, it may even play a role in diagnosing diseases such as cancer and TB.

Wasp Hound鈥檚 beginnings were more prosaic. For decades, Lewis has studied the wasp Microplitis croceipes as a potential biocontrol agent for two serious caterpillar pests, the corn earworm and tobacco budworm. The slender wasps, which are 10 to 12 millimetres long, are parasitoids: females lay their eggs inside live caterpillars. As Lewis investigated ways of encouraging wasps into farmers鈥� fields, he found that they are ultra-sensitive to a huge range of volatile chemicals.

The nectar-feeding wasps can pinpoint a productive flower from a few stray molecules of scent. Females seeking hosts for their eggs are attracted to the alarm odours plants release when caterpillars chew them. 鈥淭hey follow the odour trail to the plant,鈥� says Rains. 鈥淭hen they home in on the scent of the caterpillar or its faeces.鈥� The wasps start life attracted to odours they associate with their host caterpillar, but they are quick to switch to new ones if they bring better results. This could mean learning any of thousands of possible smells. 鈥淭heir survival depends on being adaptable and learning new cues as their environment changes,鈥� Rains says.

A call from the US Defense Advanced Research Projects Agency, the R&D arm of the Pentagon, prompted a rather different line of enquiry. DARPA wanted to know whether the wasps could use their remarkable olfactory skills to sniff out odours they would never normally encounter in the wild. The team set to work and found it was very easy to teach a wasp an unfamiliar smell. Expose a wasp to an odour while you give it a drop of sugar water or a caterpillar, and after just three 10-second sessions the connection is fixed in its memory (Chemical Senses, vol 28, p 545).

Bombs and bodies

鈥淪o far the wasps have learned to respond to just about any odour we鈥檝e tried,鈥� Rains says. They can detect 2,4-dinitrotoluene, a molecule given off by the explosive TNT (Journal of Forensic Sciences, vol 50, p 1187). They can pick up the scent of a corpse from molecules of cadaverine and putrescine produced as microorganisms begin the process of decomposition. Back on the agricultural beat, they are adept at identifying chemicals associated with food spoilage and toxic fungi that can contaminate stored nuts and grain. 鈥淭hey detect most odours at the parts per trillion level, which is very similar to a dog,鈥� says Rains. In a contest with a commercial electronic nose, in which both wasp and machine were 鈥渢rained鈥� to detect the fungal odour 3-octanone, the wasps were the undisputed winners. 鈥淭hey were a hundred times more sensitive than the electronic nose,鈥� says Rains.

This left the team in no doubt about the wasp鈥檚 sniffing skills, but how to harness them? You can鈥檛 put insects on a leash and walk them along the queues at the airport check-in. They must be closely confined and kept focused on the job. Then there is the problem of communication. Sniffer dogs are so costly 鈥� around $15,000 each 鈥� because it takes six months to train them and they need a dedicated handler. But at least the two can communicate: a dog follows instructions and indicates a positive result by growling, crouching or pointing with its nose. It may be much quicker to train a wasp, but how can you tell when it has picked up an odour?

The answer lies a combination of insect behaviour and technology. 鈥淲e had to work out what sort of feedback we could understand,鈥� Rains says. 鈥淲e focused on the wasp鈥檚 behaviour.鈥� When a hungry wasp scents food it begins to search for it, walking around and pressing its antennae down onto what it thinks is the source of the odour. When a female wasp detects a caterpillar, it adopts what the researchers call 鈥渃oiling鈥� behaviour, rearing up on its hind legs with a characteristic bending of the antennae and readying itself to stab in its ovipositor.

Train a wasp to link a target odour to food, and it will show search behaviour when it detects that smell. Train it to link the odour to a host, and coiling behaviour indicates a hit. In theory, you can train female wasps to detect two target odours 鈥� linking TNT, say, with food, and Semtex with a host. If it searches, you have TNT; if it coils, it鈥檚 Semtex. You can be certain it鈥檚 the target odour triggering the response: once a wasp associates a particular smell with a reward it ignores all others.

For the sake of simplicity, the team began with the foraging response. 鈥淲e had the odour coming in through a hole. Trained wasps clustered round the hole and tried to stick their heads through,鈥� says Rains. The team鈥檚 first idea was to use this as a 鈥測es鈥� signal by running an electronic beam across the hole. If the wasps stuck their heads in the hole, they broke the beam and set off an alarm. It worked, but it took up to 3 minutes to get a response and that was too slow. They decided instead to monitor the wasps鈥� search behaviour with a video camera. 鈥淭hat evolved into Wasp Hound,鈥� says Rains.

The device consists of a short piece of plastic tube closed at each end with a PVC cap. At one end there is a hole, and just above it five wasps inside a ventilated transparent cartridge. At the other end of the tube is a small fan, a miniature video camera and an LED to provide light. As the fan draws air through the hole and into the cartridge, the camera records the wasps鈥� responses, sending images to a laptop to be analysed. If the black-bodied wasps don鈥檛 recognise an incoming odour, they mill about and the image is a scattering of black pixels on a white background. If they do recognise the smell, they cluster around the inlet turning the central part of the image black. The more black pixels around the centre, the stronger the response. 鈥淲e usually get a response within 25 seconds,鈥� says Rains. Five wasps turned out to be the ideal number; any fewer and the camera can鈥檛 distinguish clearly between a yes and a no result, while more wasps brings no further improvement.

鈥淚f they stick out their tongues, that鈥檚 a yes. If they don鈥檛, it鈥檚 a no鈥�

Wasp Hound passed its initial trials with flying colours. Not only did the software easily distinguish between the intense search behaviour that indicates a positive result and aimless milling about, the wasps responded only to the target odour (Biotechnology Progress, vol 22, p 2).

Many other insects have potential as sniffers, but so far only one other group is anywhere close to developing a practical device that harnesses their skills. Inscentinel, a small British company based at Rothamsted Research in Hertfordshire, has developed a prototype detector with honeybees as the sensor. The device, as yet unnamed, has successfully detected explosives in cars and is now in trials at a freight airport somewhere in the UK.

Honeybees have the advantage of being one of the best-studied of all insects, giving researchers a head start in developing a bee-based sniffer machine. To locate nectar-producing flowers and pollen, worker bees use both visual cues and scent. When they return to the hive, they pass on information about the best places to forage, indicating direction with their famous waggle dance and providing olfactory clues by passing on some of the nectar and pollen they have collected. During their four-week life, bees learn and remember a multitude of smells.

A bee鈥檚 facility to learn new associations between odours and environmental cues means it is easy to train in the same way as Rains鈥檚 sniffer wasps, simply by offering it a drop of sugar water while exposing it to a target odour. After four 6-second exposures with a short rest between them, Inscentinel鈥檚 bees connect the target smell with food and remember it. 鈥淲e have trained them to remember a whole range of chemicals that are not present in nature, such as explosives,鈥� says bee biologist Mathilde Briens. Like Microplitis wasps, bees are at least as sensitive as dogs. 鈥淭hey can detect an odour at concentrations of a few parts per trillion. That鈥檚 equivalent to a grain of salt in a swimming pool,鈥� says the company鈥檚 managing director, Rachael Carson.

Unmistakable signal

Inscentinel exploits an even simpler behavioural response than the team in Georgia to signal when its bees have detected a target odour. When a bee expects food, it automatically sticks out its tongue, and it is this 鈥減roboscis extension reflex鈥� that Inscentinel鈥檚 device monitors. The current model, the third prototype, is the size of a shoebox with a detachable drawer that contains three bees, each in its own holder (see Photo). A pump draws air past the bees鈥� heads while a miniature camera records their response and relays the images to a laptop. 鈥淚f they stick out their tongues, the software registers a yes result. If they don鈥檛, it鈥檚 a no,鈥� Carson says. 鈥淭he proboscis reflex is very clear. Their tongues are so long you can鈥檛 mistake it.鈥�

When Inscentinel began work on the device in 2001, the bees鈥� first job was in food quality control 鈥� sniffing strawberries to check they were ripe before going onto supermarket shelves. The bee machine has also sniffed out hidden packages of tobacco and identified cheap whisky masquerading as a top brand. 鈥淪ince 9/11, the interest has mainly been in detecting explosives,鈥� says Carson. Bees can learn to recognise all known explosives at levels low enough to detect traces on someone鈥檚 hands.

With DARPA and security organisations as clients, Inscentinel isn鈥檛 giving away too much about its work on explosives. However, Carson thinks the device could be in commercial production in a year鈥檚 time. In the meantime the team is looking at ways to improve it. One plan is to extend the scope of the sniffer, with banks of bees screening the air for several odours at a time, which could save crucial minutes when hunting for a bomb.

Meanwhile, Rains is optimistic that within three years his Wasp Hound will be out there sniffing for bombs at airports and locating buried bodies. By then he hopes to have a Wasp Hound that has more of the qualities of a dog 鈥� constantly on the move but able to raise the alarm in an instant. The current 25-second response time is too slow for this, but by analysing the component parts of the wasp鈥檚 search behaviour, Rains鈥檚 team hopes to find a telltale body movement that signals when it first detects a target odour. That could reduce the response time to as little as a second. 鈥淭hat might allow you to walk around with the device and find the source of an odour,鈥� says Rains.

Bombs, drugs and bodies apart, there is another field in which an insect鈥檚 鈥渘ose鈥� could be invaluable: diagnosing diseases at an early stage. Some diseases 鈥� peptic ulcers, some cancers and TB, for instance 鈥� have signature odours that appear on the breath or in urine or saliva long before a person develops obvious symptoms (快猫短视频, 8 April, p 27). Detect these, and you can start treating the disease in its earliest stages. A group at Amersham Hospital in the UK is investigating the ability of dogs to diagnose bladder cancer from odours in urine. If dogs can do it, then there is every chance insects can too. 鈥淭his looks promising,鈥� says Carson, 鈥渁lthough no one has identified what exactly the smell is yet. It may be a complex of smells and it may change over time as the tumour ages.鈥�

Screening for TB might prove easier. In Tanzania, a Belgian research group called Apopo based at Sokoine University of Agriculture has trained giant pouched rats to sniff out TB in samples of saliva. This provides a speedier diagnosis than traditional methods but the rats take four to six months to train. Insects could offer a better service. In lab trials, bees have discriminated between different bacteria, including E. coli and BCG, a weakened strain of the TB bacterium. Inscentinel is now working with the London School of Hygiene and Tropical Medicine to see if bees can pick out the scent of TB.

Does all this mean the days of the sniffer dog are numbered? Probably not. 鈥淲e don鈥檛 see insects as a replacement for dogs,鈥� says Rains. 鈥淏ut they do have lots of advantages. They cost pennies to raise. They don鈥檛 need special handling, and because they are so quick to train you can have them on call, ready to learn a new smell whenever you need it.鈥�