August 12, 1997

To: Paul Millman
From: John F. Presley

CBMB/NICHD Building 18T
National Institutes of Health
18 Library Drive
Bethesda, MD 20892

Introduction

We are interested in the dynamics and movement of various proteins in living cells tagged with GFP. Our experiments typically require time-lapse imaging over periods of time ranging from five minutes to several hours and typically involve the collection of more than 100 images from the same cell using a cooled CCD camera. The various commercial fluoroscein-like GFPs (with properties similar to S65T isolated by Roger Tsien [1] ) have proved sufficiently bright and photostable for our purposes. Unfortunately, existing combinations of GFP variants and filters have proved unsatisfactory for double labeling in similar experiments. This is primarily because the blue variants are rapidly photobleached, which makes capturing a long series of images from the same cell impossible in our hands. Another problem is that our microscope allows computer control only of excitation filters. Dichroics and emission filters must be switched manually. Thus, our ideal filter set would involve a common dichroic and emission filter but two separate excitation filters, one for each GFP variant. Additionally, both GFP variants should be highly photostable.

The variants of GFP we chose as a potential double-label pair were W7 (2) (DNA was kindly supplied by Roger Tsien) and 10C (3) which we constructed from EGFP from Clontech by introducing the four published amino acid substitutions (T203Y, S65G, V68L, S72A) using site-directed mutagenesis. W7 was chosen because we had observed that it was sufficiently photostable and because it has relatively short wavelength but visible excitation. 10C was chosen as it had the longest wavelength excitation and emission peaks of any published GFP variant (3). Comparison of the excitation spectra of W7 (2) and 10C (unpublished data) indicated such a filter set might be feasible. To excite W7, we used a filter that passed 400-430nm. 10C was excited with light from 480-495nm. A common dichroic (cut-on at 505nm) and emission filter (505 long-pass) was used. This sacrifices roughly half of the emission from W7 but allows switching between the two fluorophores changing only the excitation filter.

Results

To test the performance of the filters, we expressed 10C and W7 using the lac promotor in bacteria induced with IPTG. Bacteria expressing either 10C or W7 we allowed to settle onto coverslips and were photographed using a Zeiss 63X NA 1.4 Apochromat and Photometrics cooled CCD camera. Both sets of bacteria were extremely bright when imaged with the proper filters (400-430nm for W7; 480-495nm for 10C). If 10C was imaged using illumination through the 400-430nm bandpass, an extremely faint image could be seen through the eyepiece. Quantitation of CCD images showed that the crossover image was 3% as bright as with the 400-430nm filter; thus, the two filters discriminate well enough between the two labels for our purpose. We obtained similar results with fusions of W7 and 10C to several integral membrane and cytoplasmic proteins and expressed in live mammalian cells, indicating that the two GFP's work as a double label pair under realistic experimental conditions. GFP fluorescence with both probes was stable to mild fixation (2% formaldehyde in PBS for 20 minutes), and the two probes could be selectively excited after fixation of either bacteria of mammalian cells.

References

(1) Helm, R., Cubitt, A.B., Tsien, R.Y. Nature 373:663-664, 1995.
(2) Helm, R., and Tsien, R.Y. Current Biology 6:178-182, 1996.
(3) Ormo, M., et al, Science 273:1392-1395, 1996.