Choose your target carefully.
Even before you optimize your imaging protocol, says Jeff Peterson, vice president for biology at ViSen Medical, make sure you choose the right target for a particular biological question. For example, to image a breast tumor in vivo, appropriate targets might include matrix metalloproteases at the tumor's leading edge, the proliferative-phase nuclear protein Ki-67, the estrogen-linked cathepsin D, or a variety of other molecular markers. Each different target will provide different imaging results.
Match your fluorophore to your experiment.
If you're using a traditional fluorescent label, this can be as simple as making sure the excitation and emission wavelengths match your instrumentation. Quantum dots give you more control over these parameters, says Brad Kairdolf, a graduate student in Shuming Nie's laboratory at Emory University, because the composition, coating, and size of QDs can be changed to give a particular absorption and emission spectrum. Giving attention to...
Quantify, then exploit, the interaction between the fluorophore and the organism.
The in vivo environment - pH, the protein environment, and the binding state of the conjugated label - can change the spectral and decay characteristics of fluorescent labels. Make sure you characterize these changes with carefully designed control experiments, says Tim Doyle, scientific director of the Small Animal Imaging Facility at Stanford University. Then, use them to glean additional biological information. For example, the way the fluorescence signal decays can shed light on the conformational state of the target molecule.
Collect emitted light at several wavelengths..
"When we see data graphed," says Richard Levenson, director of biomedical research at Cambridge Research & Instrumentation (CRI), "the first question asked is, 'what are the error bars?' But we're predisposed to accept that seeing is believing, so we don't ask about errors when we see an image." Remember that the eye can be fooled - there are many ways of producing a signal that looks green, for example. Using optical filtering to collect data in several different wavelength windows, Levenson constructs a spectral profile of the emitted light and matches it to the signature of different fluorescent sources. This increases the confidence that the signal at each pixel is classified correctly. (CRI has a proprietary implementation of this method as well.)