Why is it so difficult to diagnose an infectious disease on the spot?
We measure vital signs such as temperature or blood pressure routinely, yet we have no quick way to pinpoint the cause of most infection. Our inability to identify harmful bacteria and viruses costs patients dearly. During the several days or so that doctors usually take to identify bacteria and viruses, illnesses spread and may get harder to treat, and the most vulnerable patients—newborns, the elderly, anyone with a weak immune system—can die.
These delays happen amid all the advantages of high-tech medicine. Consequences are even worse in small clinics in Africa, where it can take many more days to get test results. In that time, malaria patients can be mistakenly treated for typhoid, or Ebola patients can go unquarantined.
Tests move so slowly because the molecular fingerprints of specific infections are hidden in the human body, obscured by an abundance of normal proteins and particles. In one sample of blood, there can be just 1,000 bacteria-specific molecular markers floating among trillions of irrelevant molecules. It takes a long time for expensive, complex machines, operated by highly trained scientists in specialized laboratories, to find a large enough group of target molecules to set off an alarm.
We are now on the verge of doing much better. Instead of wasting time and endangering lives transporting samples from patients to testing facilities, we can find the molecules of disease right away, identifying them in a doctor's office while the patient waits 20 minutes. We can do this with nanoscale probes, tiny sensors just a few billionths of a meter in diameter that are nestled in a small, plastic cartridge. Place a drop of blood in the cartridge and get the results. These probes react quickly to low levels of bacterial DNA, in part because they are about the same size.
Size matters. A small wave will not shake a battleship, but it will rock a little rowboat in a clearly noticeable way. It might send spray over the sides, startling the rower. Our nanoscale probes respond to their surroundings—fluid in a blood sample—in ways that would go unnoticed by a larger sensor, and we can see that happen very quickly.
My colleagues and I are excited to see that our systems will be tested in the clinic in the coming year. And ours is only one of several promising diagnostic approaches, developed by other researchers, that also use nanoscale reactions. Scientists have refined methods of shaping materials, often atom by atom, during the past decade. Laboratories across the world are using this fine-grained control to design devices that react more quickly, to highly specific triggers, than larger predecessors. We are all cautious because we have seen real-world instances where our prototype successes have fallen short. But we are hopeful, too, that such methods will eventually help us deliver care the moment it is needed.
Fishing for disease
My research group got into this area about 10 years ago. We were looking with admiration at the simple, user-friendly handheld glucose monitors used by people with diabetes. The glucose molecules essentially complete an electrical circuit in the device by giving up some of their electrons, creating a current. More current means more blood glucose. We wondered if we could use the same approach to measure DNA and RNA sequences from bacteria or viruses that are specific markers of infection.
To make this work, we needed to find a way to attract and catch some DNA molecules from these pathogens that might be present in a sample of blood taken from a patient. We were going fishing, so we needed bait. One nice feature of any piece of DNA is that it will stick, very selectively and tightly, to another sequence of DNA that we can design and synthesize ourselves. We could create a sequence to catch, say, DNA from a staph bacterial strain. That gave us our highly specific lure. We attached that lure molecule to a sensor, a millimeter-wide gold wire, designed to give off an electric current when the bacterial DNA hit. (Gold works well because it is a good current conductor.)
But because DNA, on its own, does not release enough electrons to start sucking a detectable electric current out of the gold wire, we added an amplifier. We mixed in a metal molecule, ruthenium, to our sample. This metal has a positive electrical charge, so it is attracted to the DNA, which has a negative charge. If a DNA molecule bound itself to our sensor, the metal would come along for the ride. The metal-DNA complex readily grabs electrons from the gold wire, which starts the flow of current at a level that we can detect. By using different bait molecules on the surface of our sensors, we could spot DNA from different kinds of bacteria.
The bad news was that in situations close to real life, this method did not work. It performed well enough when we dumped a great deal of bacterial DNA—trillions of molecules—into our samples. But then we tried it with levels of DNA that were closer to what typically show up in a blood sample that a doctor might draw with a needle. Usually such a sample contains 1,000 target molecules or fewer. We tilted the scales in our favor, using a million targets, but even then we could not get a detectable signal. We were nowhere near where we needed to be.
We spent a year exploring all the variables in our system and trying to understand why we could not find smaller numbers of molecules. It was frustrating—none of the tweaks we could think of seemed to make the method more sensitive. A couple of the students in the group actually gave up and asked to be transferred to different projects. I was beginning to have doubts myself and wondered if my research group was going to survive.
Thankfully, serendipity intervened. One day, in 2004, we were discussing work on an unrelated project that also involved the use of gold but on a much smaller scale. These gold nanowires were just 10 nanometers (10 billionths of a meter) across, an amount of space that could hold only five DNA molecules. So just for fun—and because nothing else was working—we swapped these nanowires for the millimeter-sized gold wires we had been using and did a few quick and dirty experiments to see if anything would happen.
Something did. One of my then postdoctoral research associates, Rahela Gasparac, came running into my office, clutching a piece of paper with the results of the first test in her hand. The nanowires gave us a millionfold enhancement in sensitivity. For a moment we thought about heading out to celebrate. Then we realized we needed to repeat the experiments and turned around and went back to the lab. We wanted to be sure what we were seeing was real. Sure enough, it was, and we knew that we had a way to get at those 1,000 molecules that could allow us to diagnose disease.
Why did nanowires allow us to sense much lower concentrations of DNA? It was because their size has a profound influence on their shape. Shaved down to the nanoscale, these wires had spiky, little hills that did not appear in their larger cousins, whose bulk gave them a flat, smooth surface. A bait molecule on one side of a hill and one on the other side had more space around them than if they were crowded together on the flat, larger wire. Fluid could move more easily through that space, carrying with it the target molecules, and the bait and targets had much more opportunity to come in contact with one another.
These probes were good, but by hand our students could make only 10 of them a day. For real clinical use, we would need thousands of them. We turned, therefore, as so many scientists and engineers have done when they wanted to manufacture a whole lot of electrical devices, to silicon.
Chips made from silicon can be adorned with electrodes and mass-produced. We wanted to take the 10-nanometer hills featured on our nanowires—the spiky hills that enhanced sensitivity so much—and reproduce them on such a chip. After about six months, we found a good way to do this with a chemical process called electroplating. We could start with larger, microscale feature in the silicon and then use the plating chemistry to lay down finer layers of gold on top. Instead of growing nanowires, we learned it was quicker and easier to create a kind of gold dome with many spikes. By fixing bait molecules to different sides of the spikes, we mimicked the separation created by the hills in the original wires. And timing was key. If we let the plating process continue for a while, then the features would grow to an unusable size. But if we cut the time short, the features would only reach nanoscale and stop.
Over the next few years we showed that we could use these detectors to analyze markers of infectious diseases caused by bacterial pathogens and that we could establish the presence or absence of the pathogen in 20 minutes. This turnaround time was important because for diagnostic testing to be successful in a doctor's office, the result would have to be returned during the course of a typical patient visit. Another feature of our approach is what we call “multiplexing”—the ability to search for multiple pathogens at one time. We were able to create multiple gold domes on the surface of our chips and attach a different type of bait molecule to each dome. This allowed us to drop a single blood sample on the chip and analyze it for many types of pathogens. Most other approaches can look for only a single type of pathogen DNA at a time. One of our most ambitious studies looked at 20 different bacteria at once, along with the DNA signals of five common types of antibiotic resistance. We were able to find them with 99 percent accuracy.
To try to get this technology out to doctors' offices, we started a company, Xagenic. The firm, for which I serve as chief technology officer, has taken our sensor chip, built a plastic cartridge around it and developed ways to include everything inside the cartridge that is needed to run a diagnostic test. The accuracy of these cartridges in finding chlamydia and gonorrhea, two sexually transmitted diseases, will be the focus of the clinical trials to start in 2016. The tests will involve physicians and their patients in 20 different medical offices. If this first trial stage is successful, we plan to submit the data to the U.S. Food and Drug Administration and ask for clearance to launch a commercial product.
We have a lot of competition from other promising nanoscale technologies. Some assays can home in on specific types of cancer with a new degree of accuracy. Chad A. Mirkin's group at Northwestern University, for instance, has developed gold nanospheres that react with cancer DNA even before the dangerous cells have formed tumors. David Walt of Tufts University has a system that counts how many disease marker molecules are present in a patient, which can be extremely useful for cancer diagnosis and monitoring. These approaches, however, are designed for use in testing labs rather than in a doctor's office.
Still other techniques do focus on on-the-spot diagnosis, and they are making their way toward mainstream medicine. Rustem Ismagilov's group at the California Institute of Technology has a wireless device, called the SlipChip, that allows DNA detection without the need for any type of cabled power. Earlier this year Samuel Sia of Columbia University and his colleagues reported in Science Translational Medicine on a tiny blood sampler that plugs into a cell phone and uses signals from antibodies to detect HIV.
I believe one or more of these technologies—or a completely different one that we do not yet know about—will eventually work well enough for everyday medical practice. At that point, reactions that take place at millionths or billionths of a meter will produce an outsize improvement in patient health.