Paul L Gourley & Darryl Y Sasaki. American Scientist. Volume 89, Issue 2. Mar/Apr 2001.
The world has a wonderful transparency to a physicist who works on optics. There is little we cannot “see” by manipulating light and its interactions with matter-from the distant reaches of the universe to the submicroscopic details of materials and processes. The possibility of peering inside molecules was, in fact, what motivated the spectroscopists who conceived of one of our most powerful optical tools, the laser, in 1958. They imagined sending a light probe into the world of the very, very small. Today lasers are a way of both seeing and altering the finest details of matter.
When one of us (Gourley) began working on tiny lasers in the 1980s, lasers had undergone rapid development, producing technologies such as the compact-disk player. But many worlds remained to be conquered by laser light. It was clear, for instance, that lasers had tremendous potential in medical diagnosis and treatment.
For the past decade we and many others have been working to realize that potential. This is the story of a journey toward a new light-enabled technology, the biocavity laser, that we think will become important in the diagnosis and treatment of disease. The most recent stage of this journey has been propelled by an imaginative question. Cells, viewed by eye under the light microscope, appear quite transparent, yet they are full of complex molecules that do the important work of life. If you put a cell inside a laser cavity, the place where ordinary light is transformed into an intense, coherent beam, what might the resulting light beam tell you about the cell?
In science such a question can serve as the impetus for years of exploration and development. Designing a new device requires applying many kinds of knowledge and solving unexpected problems. Our own voyage promises shortly to produce a device with a practical use, yet of course research and development yield many other satisfactions. We have learned much along the way.
Lasers are devices engineered to emit coherent light (laser meaning light amplification by stimulated emission of radiation). Just as the atoms of gas inside a fluorescent lamp give off light after they are excited to higher energy states by electricity, so atoms in a laser can be stimulated, either by electricity or by light, to emit particles of light, or photons. Photons are emitted by excited atoms and bounced around in a controlled way until a beam of a desired quality is produced. Happily, fine-tuning the properties of the material used to build a laser, and the design of the device, allows lasers to be optimized for all kinds of uses.
To build a laser you first need to find a way to “pump” a material with electricity or light so that its atoms are excited to a higher energy level. Laser action begins when a photon strikes an excited atom, causing a second, identical photon to be emitted. If you think of the photon as a wave, what is produced in this process is a pair of waves (the original photon and its twin) with their peaks and troughs lined up exactly-coherent light.
The goal is to amplify this light by getting a chain reaction going within the material, so next you place mirrors on either side of the laser cavity to encourage resonance among the bouncing photons. They gather intensity as more and more are created and travel together; eventually, they form a concentrated wave that leaks through one of the mirrors.
The first lasers were large, table-sized platforms that gulped great quantities of optical or electrical power. The current generation of lasers produce amplified light in a very small space. Among these are solid-state lasers, a class that includes devices in which crystals of semiconductor material form both the active region and the mirrors. In many ways these are very different from the large early lasers. The most common solid-state lasers are diodes, made of stacked semiconductors with an electrical junction between them; in these lasers end facets, which serve as the mirrors, are cleaved out of the semiconductor crystal, and the laser beams are emitted from the edge of the wafer. Generally the atoms in these layers are excited by electricity supplied to the device. Such lasers are the workhorses in now-common devices such as laser printers and barcode scanners.
Lasers Turned Sideways
In 1986 Gourley and his colleagues at Sandia National Laboratories, Tim Drummond and Anthony McDonald, helped develop a new design that became a forerunner of another class of commercially important lasers. These are called VCSELs, or vertical-cavity surface-emitting lasers.
A traditional semiconductor laser cavity can be thought of as having a long rectangular shape, with mirrors on both ends and the laser beam emerging from the end. In a diode laser this cavity might be less than one-tenth of a millimeter (one ten-thousandth of a meter) long. Some new developments, however, have allowed us to create laser cavities that are on the scale of hundreds of nanometers (a nanometer is one-billionth of a meter); in fact, it is now possible to build a laser whose cavity is a fraction of the actual wavelength of the light emitted. Such cavities can have a vertical geometry; the laser device is basically a sandwich where the mirrors are the “bread,” the cavity is the filling and light is emitted from the top, not the sides.
This design is possible because of advances in microfabrication propelled by the microelectronics industry. Today, extremely thin layers of semiconducting material can be “grown”—spray-painted, really-atom by atom, using techniques that produce a well—aligned, or epitaxial, multilayered crystal. Small differences in chemical composition create a difference in their refractive index. Thus certain layers become mirrors, confining laser action to a narrow region. During experiments in the 1980s, we found that we could force the photons into just a few optical modes by carefully adjusting the distances between the mirrors. Such control of the photonic states is the key to producing coherent light without wasting energy.
In addition, we have been able to build into these lasers structures called quantum wells, which trap electrons and holes (places where atoms are missing electrons) in potential-energy levels lower than surrounding materials. This forces them to move between two particular energy states and therefore to emit light of a specific wavelength. Photons generated in quantum wells can bounce between the mirrors hundreds of times, and they have controlled energies that add to the fine-tuning of the laser output. The manufacturing technique that enables all this is called molecular-beam epitaxy. We call such extreme control of laser properties “bandgap engineering.”
As we continued the work we found that other techniques added to the control of the optical modes of our vertical-cavity laser. In 1990 Gourley, along with Joel Wendt and Allen Vawter at Sandia, began using electron-beam lithography to create structures that would allow for more perfect confinement of the light. Theorists in John D. Joannopoulos’s laboratory at the Massachusetts Institute of Technology had developed the concept of photonic crystals. These are periodic structures made of dielectric materials (electrical insulators) that forbid propagation of light in a certain frequency Wendt and Vawter used the electron beam along with reactive ion-beam etching to pattern the semiconductor into a photonic lattice, a triangular array of holes, and “write” a polymer photomask for etching the lattice structure. The lattice confines light in the plane of the cavity.
What we accomplished, then, was the “growing” and etching of a thin semiconductor structure that confined light in three dimensions. We satisfactorily reduced the amount of energy wasted, an important step in the commercial viability of the VCSEL. The distance between the mirrors in our laser could be as small as the wavelength of the light that was emitted. The microlaser we developed was photopumped-the atoms in the semiconductor were excited by pumping light in, rather than electricity-and had a cylindrical shape, so that light was emitted in a circular beam.
Looking into Cells
In order to access the active region of the laser to create the photonic lattice, Gourley, Vawter and Wendt had to separate the top mirror layer from the structure. To do this, they placed an external dielectric mirror in contact with the semiconductor surface. The open cavity prompted a fascinating new question: What would happen if you put things inside the laser cavity?
The Sandia group tried putting drops of dielectric liquid, such as water and vacuum grease, in the cavity. When the laser was pumped, the liquid participated in the “lasing”-when the spectrum of emitted light was read from a spectrometer, the liquid drops could be seen to support a well-defined optical mode. The photons were moving through the liquid drops and interacting with the atoms in the liquid in a way that produced distinctive patterns in the spectra of emitted light.
As a teenager, Gourley had peered at cells with a home microscope and noticed how transparent they looked. He knew that cells had refractive power when placed in water. Over a Thanksgiving break, Gourley went to the vacant laboratory to see what would happen if biological cells were placed in the laser cavity. The result was thrilling: Cells placed in the cavity confined the optical modes too, in ways different from liquid. After bouncing around inside the cell, the light was “reporting” the properties of the cell in the laser spectrum. As he looked inside the cavity with a microscope and camera, he saw that the lasing produced brilliant images of the cells’ optical modes.
Paul Gourley’s brother, Mark Gourley (now at Washington Hospital Center in Washington, D.C.), was working as an immunologist at the National Institutes of Health. Mark Gourley suggested placing mouse and human lymphocytes—white blood cells important in the immune system-in the laser. He traveled to Sandia with samples of normal (resting) and activated lymphocytes. Activated lymphocytes contain peptides (small proteins) that the cells make as part of an immune response; Mark and Paul wondered whether the higher concentration of protein in these cells would alter the speed of light within the cavity and therefore create a distinctive spectrum in the laser output. Their hunch proved correct; the differences in the emitted light were dramatic. When the laser detected the activated lymphocytes in a drop of blood taken from a cold-ridden lab director, funding to explore the idea of a “biocavity laser” for medical diagnosis was assured.
The most popular method for evaluating diseased tissues is fluorescent imaging and spectroscopy. Cells are stained with fluorescent dyes that interact with cell components, so that gross spectroscopic changes distinguish between normal and diseased tissue. Labeling cell components with dye is a time-consuming process and can alter the biochemistry or physiology of a cell. Therefore scientists are working on various fronts to develop far more sensitive methods that can distinguish normal from diseased tissue based on the intrinsic optical properties of the cells.
Paul and Mark Gourley realized that the information appearing on their laser spectrometer might offer such diagnostic potential. The light bouncing around inside the VCSEL had a wavelength of 850 nanometers-a wavelength where light scattering and absorption are minimal in the cell, allowing it to be transparent. The biological sample was acting as a focusing lens, helping guide the light waves to a coherent state.
Dielectric materials differ in their electrical-insulating properties, a difference known as the dielectric constant. A cell, for instance, has a higher refractive index than water, and compartments within the cell have various refractive indexes that reflect their concentrations of protein or nucleic acid. When a cell is placed in the biocavity laser, the cell compartments turn out to produce unique signatures that reflect these differences. Paul and Mark Gourley reasoned that many types of diseased cells would have distinctive signatures. just as the higher protein concentration in activated lymphocytes produced a distinctive spectrum, so might, for instance, the reduced hemoglobin in an anemic red blood cell.
From a practical standpoint, reading cell properties with a semiconductor laser clearly had great potential. Semiconductors process information with tremendous speed (witness the barcode scanner), and lasers extract information from extremely tiny samples. To extract information a laser needs only a fraction-as little as one-thousandth—of the sample needed for a traditional tissue test. With a laser, cells can be scanned repeatedly and accurately in their original biological state. If a tiny channel could be etched into the microcavity, cells could be routed in single file through the laser, creating a tiny flow cytometer-a device for continuous monitoring of cells in fluid.
Today flow cytometers have many uses. For instance, in patients infected with the human immunodeficiency virus, HIV, the severity of disease is monitored by counting the number of CD4+ cells in their blood using a cytometer. Blood is extracted from a vein and stained with special reagents for examination. If a laser could be used, a sample from a finger puncture would be sufficient, and staining would be unnecessary.
We have not attempted to distinguish the spectroscopic signature of CD4+ cells, but we have been able to demonstrate the technique with other cells. In anemia, for instance, the amount of hemoglobin in red blood cells is reduced. Standard methods for diagnosing anemia involve placing in solution a relatively large volume of whole blood, oxidizing it with potassium ferricyanide and potassium cyanide, and performing spectroscopic absorption at 540 nanometers, known to be the light-absorption peak for oxidized hemoglobin.
This method determines the average hemoglobin concentration in whole blood but cannot distinguish between cellular or plasma hemoglobin or determine the hemoglobin distribution among cells. We have placed single cells in the microlaser and found that anemic cells could be identified. Without a cell present, blood plasma produces a spectrum with a peak at 827 nanometers. In the presence of a cell, a series of spectral peaks can be seen; these peaks are shifted in the presence of hemoglobin in ways that can be predicted from the refractive index of hemoglobin solutions.
A Laser for Cancer Diagnosis
A typical mammalian cell is composed of water (70 percent), proteins (18 percent) and lipids (5 percent), along with metabolites, sugars and genetic material, which together make up the remaining 7 percent. Simpler molecules, such as water and sugar, contribute relatively little to the refractive index in the laser’s spectral region of 850 nanometers. However, more complex molecules such as proteins, RNA and DNA strongly enhance the refractive index at these longer wavelengths, and the resulting spectrum is most sensitive to the protein and genetic content of the cell.
These facts suggest that a laser method might be especially useful in detecting cancer cells because the differences between a normal cell cycle and the irregular life cycle of a cancer cell are reflected in differences in average protein concentration.
Normal human cells exist either in a dormant state called Go or in a cycle that proceeds through distinct phases of activity leading to replication. This cycle begins with a “gap” phase called G1. During the next phase, called S (for “synthesis”), the nucleus replicates its chromatin and cellular proteins, and the amount of DNA and protein doubles. G2 is a second gap phase that follows S; the cell is resting before dividing by mitosis. Most cells spend little time in G2.
Cancer cells proliferate rapidly and move more quickly through these phases. At any given time, most normal cells will be in one of the first two gap phases, G0 or G1. But cancer cells have a shortened G1 stage, and a larger percentage of cancer cells typically exist in the G2 phase, the phase in which the cell has doubled quantities of DNA and protein.
The peaks in an emission spectrum indicate optical resonance-situations where the round-trip light path is a whole number of light wavelengths. We reasoned that cells with higher DNA or protein concentrations would shift the resonance peaks to longer wavelengths. The magnitude of the shift would indicate the average biomolecular mass. In the case of cancer detection, our laser spectrometer thus ought to be able to distinguish between a normal population of cells and one with a higher mass typical of rapid proliferation.
We tested this hypothesis by flowing through the laser cavity normal human astrocytes (star-shaped cells found throughout the central nervous system) and glioblastoma cells (a cancerous form of astrocyte). As we had hoped, the cell-cycle states were easy to pick out. Cells in the G1 phase-98 percent of normal cells-shifted the resonance peak about 4.5 nanometers. A smaller peak representing a shift of about 9 nanometers showed the proportion of cells in the G2 phase-2 percent of normal cells. By contrast, a considerably larger proportion of the glioblastoma cells-about 5 percent—were in the G2 phase, and the population appeared to be much more broadly distributed between the two phases, indicating that there were many intermediate, S-phase cells. The histograms produced by the laser cytometry resemble conventional flow-cytometry histograms of DNA content in cells. Although the cell-cycle comparisons can be obtained by conventional methods, it is possible that only a few hundred cells-perhaps a sample of only one nanoliter-would be required in the biocavity laser to determine the presence of cancer cells. This is a much smaller sample than required for conventional methods. More important, the the method does not require adding fluorescent tags, so the measurement can be performed almost instantly on native cells.
Our early successes in detecting differences among cells were exciting, but many questions remained. Could we make a device that would be of real use to medicine?
We began by looking at the design of the laser cavity and how it could be optimized for medical use. We redesigned the glass chip that forms the top of the cavity, adding inlet and outlet holes, and formed flow channels by chemical etching. In all, we created 13 to 14 channels with a minimum width of 10 micrometers and a minimum length of 90 micrometers, in order to provide configurations suited to different cell sizes and velocities. We connected the channels to tubing to create a flow system through which tiny amounts of fluid could be pumped.
Here we ran into a new problem. Creating a smooth flow of biological cells through a microscopic channel in an optical device is far from a simple proposition. The problem is one encountered whenever scientists try to merge biological materials with microfabricated structures. These materials are poorly suited for each other. Proteins are large, sticky molecules that tend to adhere to glass, metal and semiconductor surfaces, forming biofilms that foul the devices and quickly compromise their performance.
Our device quickly demonstrated this problem. When we flowed the astrocytes and glioblastoma cells through our channels, they stuck on the downstream side of the channels. Within an hour or so the device was severely clogged. We tried several standard techniques, coating the semiconductor with a thin layer of silicon dioxide and then adding polyethylene glycol and silane coatings. Together these procedures increased fluid velocity and minimized cell sticking.
We do not, however, consider the problem solved. With the surface treatments the operation of our biocavity laser has greatly improved, but fouling is still observed. One of us (Sasaki) has recently conducted some experiments to determine the molecular mechanics of this problem and to understand some seemingly conflicting findings—namely, the fact that a surface that is biocompatible, supporting the adhesion and growth of cells, actually may be better at resisting biofouling when cells are flowing over it.
Most likely, Sasaki reasoned, the creation of a biofilm begins with the adsorption of proteins excreted by the cells. Proteins can rearrange their tertiary shape or unfold in reaction to their environment. Sasaki used a model protein, bovine serum albumin, to study this process. He applied two surface coatings to a semiconductor wafer-one hydrophobic, the other hydrophilic-and placed them in contact with a protein solution. Both surfaces developed protein films: The hydrophobic coating grew a matted or fused layer of denatured protein, whereas the other coating became dotted with small aggregates of folded protein.
When cells were flowed over the hydrophobic surface, the one that had developed the fused layer of denatured protein in the first experiment, they adhered to the surface as strongly as they adhered to an uncoated semiconductor surface. On the other hand, the hydrophilic coating decreased cell binding by about 60 percent. We believe that when proteins unfold as they adhere, they create a film that is “sticky” to passing cells. Albumin adsorbed on a hydrophilic surface, however, tends to maintain its structure, and the surface remains passive to cell adhesion. We concluded that the interactions of surfaces and cells can be radically changed if proteins adhere. It is unclear how important protein adhesion is in situations where cells are flowing over a surface, but these findings may explain why a biocompatible hydrophilic coating-which would be expected to support cell growth and adhesion-actually works better in a flow cell.
From Theory to Practicality
Surface-chemistry problems are one of a number of issues that face scientists working in making high-technology devices for medical use. We are optimistic that such problems can be solved and that the prospects for using laser microdevices in diagnosis are bright.
We imagine, for instance, that our biocavity laser might someday be the basis of a “smart scalpel” for cancer surgery, and we have begun collaborating with a neurosurgeon, Steve Skirboll at the University of New Mexico Cancer Center, to explore this concept. During surgery to remove a tumor, fluid aspirated from the surgical site is typically examined by a pathologist to determine whether the surgeon has removed all the tumor. This process requires a substantial investment in time (the cells must be stained and then examined) and equipment, and cannot always be completed while surgery is under way. Ideally, a surgeon would like to know immediately whether the fluid is clear of cancer cells. With a “smart scalpel” aspirated fluid would be passed immediately through a microlaser, giving a continuous statistical reading of cell protein concentration and a determination of malignancy of the sampled cells. Potentially such a device could provide real-time analysis of up to 100,000 cells per second. The information could be available in the operating room, since the laser device itself is small and portable, or perhaps at a field site such as the scene of a biological or chemical attack. We are working on the design and fabrication issues involved with such a device and estimate that it could be built for $10,000 to $50,000, far less than a conventional flow-cytometry machine.
A more everyday use for a microlaser flow cytometer would be the rapid laboratory processing of cervical cells from Pap smears or blood cells of sickle-cell anemia or HIV patients. We expect, and hope, that the biocavity laser will be one of many tiny devices that complement and enhance rapid medical diagnosis in the future. To us the cell has a wonderful transparency.