by Jeffrey M. Perkel
Flip through any biological science journal today and chances are good you'll find at least one fluorescent micrograph. Painted in bright splashes of primary colors—red and green, yellow and blue—such cellular images both inform and amaze, their proliferation serving as a testament to the ease with which such data can be captured.
But looks can be deceiving. Yes, technically the experiments are becoming more and more routine, with premade kits and turnkey systems democratizing multicolor fluorescence microscopy. Nevertheless, developing a fluorescence assay from scratch remains a devilishly tricky proposition.
"I give multiple-hour lectures on this," says Alan Stall, director of advanced cytometry technologies R&D at BD Biosciences. "It is not trivial, especially for very complex experiments."
The number of variables can be daunting. For one thing, researchers have literally dozens of fluorophores from which to choose—Invitrogen's Alexa Fluor palette alone comprises some 20 individual colors—not to mention the growing body of fluorescent proteins and quantum dots available from the likes of Clontech (Living Colors Fruit Fluorescent Proteins), Invitrogen (QDot Nanocrystals), and eBioscience (eFluor Nanocrystals).
There are also hardware considerations. What works with flow cytometry and Western blotting may not perform well in immunofluorescence experiments. And in any event, your choices will of necessity be constrained by your equipment, or more specifically, its excitation sources, filter sets, and detectors.
And then there are the intangibles: experimental matches made meticulously on paper don't always translate so well in practice. Some dyes, for instance, behave differently as conjugates than they do as free dyes. Furthermore, it's not enough to decide which fluorophores match your equipment best; you also have to decide which antigen to pair with which dye, a decision defined by considerations such as fluorophore brightness, antigen abundance, and background.
"There are a lot of moving parts," concludes Seth Gammon, in vivo product manager at Carestream Molecular Imaging.
Four flavors of fluorochromes
A fluorescent molecule is one that can absorb radiation of one wavelength, become transiently excited, and then relax. In so doing, the molecule releases its absorbed energy (minus whatever energy it lost due to molecular vibrations) as light, albeit of a longer wavelength. The distance between a dye's absorption and emission maxima is called its Stokes shift, and for most organic dyes, that interval is relatively small. Invitrogen's Alexa Fluor 488, for instance, maximally absorbs light of 496nm and emits light of 519nm, a Stokes shift of 23nm.
Fundamentally, fluorochromes come in four flavors: (1) small, highly conjugated organic molecules; (2) fluorescent proteins; (3) fluorescent nanocrystals; and (4) tandem fluorochromes. The vast majority are small organic molecules, which tend to adopt one of a few basic configurations, such as the coumarin, fluorescein, and cyanine dye skeletons. Among this class' members are such fluorescent classics as FITC and Texas Red, as well as newer entries like eBioscience's eFluors and Thermo Scientific's DyLight dyes.
Fluorescent proteins come in two basic forms. The first is the widely popular green fluorescent protein (GFP); these dyes are generally used in genetic studies—that is, studies in which the dye molecule is expressed endogenously. The other form is comprised of the phycobiliproteins, such as phycoerythrin and allophycocyanin, which tend to be employed like more traditional dyes, being directly coupled to antibodies for immunofluorescent or flow cytometry analysis.
Fluorescent nanocrystals are nanoscale structures (typically comprising a few hundred atoms of semiconducting materials such as CdSe) with several desirable features. They are tunable, meaning that each nanocrystal of a given composition (e.g., CdSe) will emit a distinct wavelength as a function of its diameter (by comparison, each organic dye in a given family is a distinct chemical form). Invitrogen's QDot Nanocrystal family, for instance, includes nine chemically identical variants with excitation maxima ranging from 525nm to 800nm. Yet, unlike organic dye families, each nanocrystal of a given composition is also universally excitable—that is, regardless of their diameters, all nanocrystals of a particular material can be excited by the same wavelength of light, typically maximally in the UV/blue part of the spectrum. That means, in theory, that multicolor experiments can be conducted using relatively simple equipment: a single excitation source and multiple emission filters. Finally, nanocrystals, unlike organic fluorophores, don't photobleach, meaning they can be imaged repeatedly and over long time periods.
Tandem fluorochromes are chimeric molecules generally comprising a phycobilliprotein and an organic dye, such as phycoerythrin-Cy7. According to Wayne Patton, chief scientific officer at Enzo Life Sciences, these dyes are essentially energy relay systems, operating via non-radiative fluorescence resonance energy transfer, which were developed to extend the color palette available to researchers into the red and near infrared portions of the spectrum. Phycoerythrin (PE), for instance, absorbs light from a 488nm argon laser and emits it at 635nm. The coupled Cy7 then absorbs that light and reemits it at 767nm, producing a 179nm Stokes shift. According to Patton, tandem dyes are used most frequently in flow cytometry applications because of their high background in immunofluorescence applications. Also, these proteins, like quantum dots, tend to be used mostly for cell surface applications, as it can be challenging to get them into cells.
Vetting the variables
Though all fluorochromes are theoretically interchangeable—that is, any dye or fluorescent entity with given absorption and emission characteristics should theoretically be usable in a given experiment—that is not generally the case in practice.
Iain Johnson, a research fellow at Invitrogen's molecular probes division in Eugene, Oregon, has identified seven features of fluorescent dyes that users should consider when selecting a fluorophore. Three relate to a dye's absorption and emission characteristics (fluorescence excitation spectrum, absorption spectrum, and emission spectrum); two describe the fluorophore's brightness (extinction coefficient and fluorescence quantum yield), and two describe the dye's ability to withstand experimental rigors (quenching and photobleaching).
In theory, a researcher could sit down with a pencil and paper (or use online tools like Invitrogen's Spectra Viewer , or BD Biosciences' Fluorescence Spectrum Viewer), compare and contrast the different variables, and identify a set of dyes that will work best with their particular hardware. But in practice, says Johnson, the process is more complicated.
"There are many decision points, but the two biggest ones are what instruments are you going to use, and how abundant is your target," says Johnson.
For instance, there's no point selecting a dye you cannot stimulate with the excitation sources available to you, nor is there any point in using dyes you cannot detect. "There is a matching process that has to go on," Johnson explains. So, if your microscope has only a 488nm laser and compatible filters and detectors, you cannot use a coumarin dye like Alexa Fluor 350, because it won't absorb that light efficiently. Instead, you would select a dye like FITC or Alexa Fluor 488, both of which can very efficiently make use of the 488nm laser line, and whose emission can be recorded on the accompanying detection hardware.
That said, it is increasingly the case that researchers need not worry about shoehorning an experiment around only a single excitation source; most instruments contain at least two. And in many cases, at least one can excite fluorochromes outside the typical reds, greens, and blues of, say, FITC, rhodamine, and DAPI.
The LI-COR Odyssey and BD FACSAria, for instance, both have far- or near-infrared lasers capable of exciting dyes on the far right of the visible spectrum. Dyes in the visible portion of the spectrum tend to produce unsatisfactory results in thick tissue sections or live animals, because biological tissues both autofluoresce and absorb visible light. Tissue culture plastics can also autofluoresce. As a result, in vivo imaging, or imaging of thick ex vivo tissue slices, can be problematic when using conventional dyes. But, near-infrared dyes, such as LI-COR Bioscience's IRDye fluors (IRDye 680, IRDye 680LT, IRDye 700DX, and IRDye 800CW), Carestream Health's Kodak X-SIGHT Large Stokes Shift Dyes, eBioscience's APC-eFluor 780, and Biotium's CF770, don't have that problem.
"There is generally less fluorescent background in plastic at these wavelengths," explains Harry Osterman, principal scientist at LI-COR. "And at 800nm, there is a natural window of transparency for tissues and cells."
Vendors have begun outfitting instruments with violet lasers (405nm) as well, and dye manufacturers have responded with such dyes as BD Horizon V450 and BD Horizon V500 from BD Biosciences, Pacific Blue and Pacific Orange from Invitrogen, eFluor 450 from eBioscience, and AmCyan, a fluorescent protein-based dye from BD Biosciences.
Experimental method also matters, says Stephen Shiflett, technical product manager for protein detection at Thermo Scientific Pierce Protein Research, particularly when it comes to brightness and photostability. "In some cases," says Shiflett, those parameters "can be mutually exclusive." In flow cytometry, for instance, where detectors have just milliseconds to process a cell's fluorescent characteristics, brightness is key. But in fluorescence microscopy, it's photostability that counts. (Photostability refers to a dye's resistance to photobleaching, the chemical breakdown of fluorescent molecules due to excitation energy.)
Antigen abundance is another key consideration. "The natural abundance of proteins varies enormously over about almost a billion-fold range in terms of copies per cell," Johnson says. If you are imaging a highly abundant protein like tubulin, dye choice doesn't matter quite so much as if, say, you are imaging a transcription factor present in just a few copies per cell.
In general, says Tony Ward, vice president of strategic marketing at eBioscience, the best strategy is to match the least abundant targets with the brightest dyes. "You always match a bright dye with a dim antigen so you can be sure to measure the antigen you are looking for," he says.
For multicolor experiments, dye choice can become even more complicated. That's because excitation and emission spectra are just that, spectra; just because a dye has an emission maximum at, say, 650nm, doesn't mean it doesn't also emit at other nearby wavelengths as well. And when those spectra overlap, "bleedthrough" can become a problem. Suppose, for instance, that you are doing a three-color flow cytometry experiment with dyes in the green, orange, and red channels, says Nick Thomas, principal scientist for cell technologies R&D at GE Healthcare. If the green signal was extremely bright and the orange signal sufficiently weak, the green signal could bleed into the orange channel, such that the researcher would tend to overestimate that protein's abundance, Thomas explains. (The same can happen in microscopy.)
"There isn't a straightforward answer" to the question of which dye do I use, says Thomas, "because it depends on what you are going to use it for, and with, and these are biological questions, not physics questions. They are not numbers-driven."
Making your choice
In practice, most researchers are not faced with building an assay from the ground up; their potential experimental design "space" is already constrained. For instance, you may want to use a nuclear or mitochondrial stain, the spectral characteristics of which limit your options. Or, one or more of the antibodies you need may already be labeled and in use in the lab, so there is little point to buying a new one.
"There has to be a compelling reason to switch a dye," says Thomas. "If you have an existing assay with two or three dyes and it does what you want, then to switch to another dye is not an insignificant amount of work."
In many cases, Thomas adds, the choice of dye is relatively unimportant—any comparable fluorophore will do. Instead, researchers often make selections based on price, availability, and familiarity. "In most cases, if you already have it in the lab, you'll scrounge it," he says.
That said, if you absolutely need to set up a new assay, Stall advises starting with the literature. "For most [antigen] combinations, there's usually something in the literature."
Thomas recommends going first through the paper exercise of identifying the best fluorophores for your experimental setup—coordinating spectral characteristics with hardware constraints, matching dye brightness with antigen abundance, and so on—and then optimizing experimentally.
Invariably, adjustments will be required. Fluorescence intensity and spectral characteristics of free and conjugated dyes can vary, for instance, though generally not by much. Also, some dyes are more water-soluble than others, a characteristic that has real functional consequences in immuofluorescence microscopy. For one thing, lipophilic dyes can bind directly to membranes, which can increase background, or cause background to vary between cell types.
"You have a large protein decorated with a variety of chemicals, and there are a variety of ways those molecules can interact with the cells, and only one of those is the one you want," says Thomas.
Lipophilic dyes also tend to self-quench, a phenomenon that can limit the brightness of an antibody-fluorochrome conjugate. It's basic math: as the number of dye molecules on an antibody increases, the fluorescence per molecule should increase accordingly. But if the dyes are lipophilic, they will tend to aggregate, so that fluorescence per conjugated dye will peak and then decay.
The ground-truth, as Gammon says, is that establishing fluorescence assays requires a serious commitment; that's why people buy premade kits. But, if you must do so, he advises running a head-to-head comparison (most companies will provide sample aliquots of dyes, he says). And furthermore, he adds, "if you are even thinking this [antibody] might go in vivo, just start with near-infrared. That's a rule I can definitely stand by."