Living cells are usually studied by interrogating large ensembles containing millions of cells, such as in a cuvette or microtiter plate. Observing entire cell populations provides only an average response, which does not yield individual cell variation and statistical distributions. To collect single-cell data, we must view individual cells under a microscope. This approach is limited by the small number of cells in a typical microscope field of view and the difficulty in tracking individual cells because they float in solution (with the exception of cells that adhere to surfaces). My laboratory has developed a method for observing thousands of individual cells over time by placing individual cells in microwells located on the end of an optical imaging fiber.
Optical fibers consist of two types of glass or plastic. The inner glass, referred to as the core, has a slightly higher refractive index than the outer ring of glass, known as the cladding. Because of the refractive-index mismatch, light is transmitted through the core over long distances via total internal reflection. Images cannot be carried over conventional optical fibers because the light signals mix within a single core so that spatial resolution is not preserved. Imaging with Arrays
Coherent fiber arrays provide a solution to this problem. Typical imaging arrays contain approximately 5000 to 50,000 individual fibers, each 3 to 7 µm in diameter, with a total array diameter of 300 to 1000 µm (see figure 1). Each fiber carries its own light signal and, depending on the nature and thickness of the cladding material between fibers, has low cross talk with adjacent fibers.
Figure 1. An array of optical fibers without cells shows no signals (left). When we add cells and excite them to induce fluorescence, we can read signals from the individual cells (right) with an array detector.
These optical imaging fibers (Schott Glass; Southbridge, MA) can form cell biosensors, which leverage the ability of living cells to respond to biologically active compounds. We fabricate the ordered microwell cell arrays on the distal end of an optical imaging fiber using a wet-chemical etching process. The microwells form as a result of the difference between the etching rates of the individual fiber core (pure silica) and clad (germania-doped silica) materials. The distal ends of the fibers comprising the array form the bottom of the etched wells. Each well is thus optically wired such that it can be individually interrogated (see figure 2).
Figure 2. Measurements of individual cells as a function of time show a distribution of responses.
We create cell arrays by randomly dispersing living cells into a fiberoptic microwell array such that each microwell accommodates only a single cell. Cells can be loaded into the wells by allowing a cell suspension to settle into wells matched to the size of the cells. The cells are encoded with fluorescent tags to identify their location within the array. In addition, a fluorescent reporter is used to correlate changes or manipulations in the local environment to responses of specific cell types. The fibers transmit the signals from the distal end containing the cells to the proximal end of the fiber, making it easy to read signals because the proximal end is not immersed in liquid. The entire array can be simultaneously interrogated and measured, yielding a rapid, repetitive, and high-density analysis method.
We randomly distribute the cells and use optical encoding schemes to identify their locations in the array. Three different methods of encoding have been used. The first two methods label the cells with conventional dyes. Live-cell encoding dyes must be rapidly incorporated into the cell, retained in the cell for a prolonged period of time, and must exhibit no toxic effects on the cell. Several lipophilic dyes have these properties for live-cell encoding. Such dyes function by incorporating a fluorophore with a hydrophobic tail into the lipid bilayer of the cytoplasmic membrane. Our first approach involved cell encoding with three lipophilic dyes: PKH26, PKH67, and DiIC18. Upon labeling, the cells remained viable in the culture, with the dyes becoming diluted in the membranes after cell divisions.
In a second encoding approach used for yeast cells, we worked with spectrally resolved concanavalin A-fluorescent dye conjugates. The strong interaction between the lectin concanavalin A (conA) and the mannoproteins on the yeast cell wall enables very efficient labeling of yeast cells. We used five different conA-dye conjugates to label five different strains of yeast.
The third approach involves genetic encoding. This approach uses genetic engineering to express different variants of green fluorescent protein (GFP), including red fluorescent protein, enhanced green fluorescent protein, and its different spectral mutants blue, cyan (ECFP), and yellow in living cells. This approach is limited, however, in terms of the number of different variants of GFP available. Studying Cells
Graduate student Yina Kuang and former postdoctoral associate Israel Biran have used this platform for a variety of biological studies. The cell arrays are mounted on an inverted epifluorescence microscope (Olympus America Inc.; Melville, NY). To address the fiber array, excitation light is focused into the proximal end of the fiber to excite the fluorescent cells on the distal end. Isotropically emitted light from the fluorophore is captured by the fiber in contact with each cell and is sent back through the fiber to a detection system that separates the excitation light from the emission light signals using optical filters.
A CCD camera at the proximal end of the fiber captures the spatially resolved fluorescence images returning through the fiber. We acquired fluorescence images (excitation: 480 nm, emission: 520 nm) every five minutes and measured the fluorescence signals using IPlab software (Scanalytics; Fairfax, VA) with a 150-ms acquisition time and 2 X 2 binning.
To demonstrate the potential of the approach, we examined yeast cells, bacterial cells, and mammalian cells (mouse fibroblasts) with the system. Yeast cells serve as a reasonable research model of mammalian cells and provide an inexpensive and simple alternative to mammalian cell assays. Yeast can be used for screening by monitoring reporter genes such as the bacterial lacZ gene, which codes for the enzyme b-galactosidase.
We used the yeast two-hybrid system to explore in vivo protein-protein interactions. Yeast cells were genetically engineered to activate transcription in response to such interactions, resulting in the expression of the lacZ reporter gene. We demonstrated the suitability of the single-cell array for two-hybrid system screening applications. Positive (interacting proteins), negative (non-interacting proteins), and wild type strains were encoded with three different dyes and randomly distributed in the array. After decoding, we added the fluorogenic b-galactosidase substrate-FDG (fluorescein d-b-D-galactopyranoside).
The signals obtained due to lacZ expression five minutes after FDG was added showed a highly stochastic response, meaning there was a tremendous amount of cell-to-cell variability between cells from the same strain. This observation has led to a number of hypotheses about why ostensibly identical cells should exhibit different response behaviors. The cell-to-cell variation would not have been observed if the assay were performed by looking at the aggregate averaged response of all cells in a microtiter well.
The arrays can also be used for biosensing. We fabricated a live cell array biosensor by immobilizing bacterial cells on the face of an optical imaging fiber containing a high-density array of microwells. Each microwell accommodated a single bacterium that was genetically engineered to respond to a specific analytein this case mercury ion, a toxin. A genetically modified E. coli strain containing the lacZ reporter gene fused to the heavy-metal-responsive gene promoter zntA, which allowed us to create a mercury-ion (Hg2+) biosensor. A plasmid carrying the gene coding for the ECFP was also introduced into this sensing strain to encode the cells to determine their locations in the array.
We measured single-cell lacZ expression when the array was exposed to mercury and could detect a response to 100 nM Hg2+ after 60 minutes of incubation. These results demonstrated that immobilized recombinant bacterial sensing cells could be used reproducibly in a biosensor device. Parallel Processing
To establish the feasibility of employing multiple cell types for screening in a single array format, we prepared two genetically constructed cell lines. One cell line responded to genotoxins and the other cell line responded to cytotoxins. In a single sequential exposure experiment, cells were first exposed to a cytotoxin and then to the genotoxin Mitomycin C to demonstrate the ability of the two cell types to differentially respond to both compounds. The GFP and ECFP fluorescence were measured to identify the respective cytotoxin and genotoxin responses from the two cell lines. Such an array may find use as a generic screening tool for analyzing the overall toxic potential of different compounds or even for monitoring water quality.
Finally, we can also use the arrays to observe cell migration. Cell migration is important for positive biological phenonema like wound healing, and disease such as cancer metastases. Graduate student Christopher DeCesare has developed an assay that enables us to rapidly monitor mammalian cell migration. Traditional assays take hours to perform whereas our fiber-based assay can be performed in five minutes. We have used the assay to observe the effects of known anti-migratory compounds and plan to extend the assay to new drug screening.
Our ongoing efforts in this area (a collaboration with professor James Collins at Boston University; Boston MA) are aimed at analyzing the intrinsic noise exhibited by complex living systems. Gene and protein expression are complex multiparameter processes involving transcription and translation with many interacting components. Complexity leads to large variability in the rates and timing of various cellular processes among individual cells from the same cell culture. By analyzing individual cells, we hope to gain an understanding of gene networks and contribute to the burgeoning field of systems biology.
Optical fiber cell arrays have demonstrated their utility for biosensing, screening, and fundamental biological studies. The ability to observe the responses from thousands of single cells over extended periods of time enables many studies that have previously been difficult or impossible to carry out. oe
David Walt is the Robinson Professor of Chemistry at Tufts University, Medford, MA.