Three-color spectral FRET microscopy localizes three interacting proteins in living cells

Abstract

FRET technologies are now routinely used to establish the spatial relationships between two cellular components (A and B). Adding a third target component (C) increases the complexity of the analysis between interactions AB/BC/AC. Here, we describe a novel method for analyzing a three-color (ABC) FRET system called three-color spectral FRET (3sFRET) microscopy, which is fully corrected for spectral bleedthrough. The approach quantifies FRET signals and calculates the apparent energy transfer efficiencies (Es). The method was validated by measurement of a genetic (FRET standard) construct consisting of three different fluorescent proteins (FPs), mTFP, mVenus, and tdTomato, linked sequentially to one another. In addition, three 2-FP reference constructs, tethered in the same way as the 3-FP construct, were used to characterize the energy transfer pathways. Fluorescence lifetime measurements were employed to compare the relative relationships between the FPs in cells producing the 3-FP and 2-FP fusion proteins. The 3sFRET microscopy method was then applied to study the interactions of the dimeric transcription factor C/EBPalpha (expressing mTFP or mVenus) with the heterochromatin protein 1alpha (HP1alpha, expressing tdTomato) in live-mouse pituitary cells. We show how the 3sFRET microscopy method represents a promising live-cell imaging technique to monitor the interactions between three labeled cellular components.

Figure 1

Signal components in the triple-labeled unmixed images in 3sFRET microscopy. The donor spectral bleedthrough is removed through linear unmixing; however, FRET signals still contain the acceptor spectral bleedthrough (ASBT). The λ-stack I_1S (excited by Ex1) is unmixed into the F1 (I₁₁), F2 (I₁₂), and F3 (I₁₃) channels: I₁₁ is the quenched F1 signal; I₁₂ has both FRET (FRET₁₂) and ASBT (ASBT₁₂) signals; I₁₃ includes ASBT (ASBT₁₃) and several FRET signals: FRET₁₃, FRET_23t, FRET_23d. When excited by Ex2, the specimen becomes a two-color (F2-F3) FRET system. After unmixing the λ-stack I_2S, two images are obtained in the F2 (I₂₂) and F3 (I₂₃) channels: I₂₂ is the quenched F2 signal; I₂₃ is both FRET (FRET₂₃) and ASBT (ASBT₂₃). Ex3 does not produce FRET. The image I₃₃ obtained through unmixing the λ-stack I_3S measures the F3 fluorescence intensities. Detailed explanations of each signal component are given in the section entitled: The data processing routine.

Figure 2

Image acquisition in 3sFRET microscopy. Triple- and single-labeled specimens were excited under identical imaging conditions, at optimal power levels for each excitation wavelength. (A–D) The λ-stacks of an mTFP-5aa-mVenus-10aa-tdTomato (F1-F2-F3) specimen were generated at 458-nm (Ex1) (solid line), 514-nm (Ex2) (dashed line), and 561-nm (Ex3) (dotted line) wavelengths (A), and were then unmixed using the mTFP (F1) (B), mVenus (F2) (C, solid line), and tdTomato (F3) (D, solid line) reference spectra. (E–G) Processing the unmixed images in the 3sFRET software produced the apparent FRET efficiency images for mTFP-mVenus (E₁₂ × 100) (E), mTFP-tdTomato (E₁₃ × 100) (F), and mVenus-tdTomato (E₂₃ × 100) (G). In this process, the unmixed images of the mVenus and tdTomato single-labeled specimens were also used to remove ASBT contaminations. In C, comparing the λ-stacks of an mVenus-alone specimen excited at the 458-nm (dashed line) and 514-nm (solid line) wavelengths indicated the ASBT level of mVenus. In D, the λ-stacks of a tdTomato-alone specimen excited at the 458-nm (dotted line), 514-nm (dashed line), and 561-nm (solid line) wavelengths were compared to show the ASBT levels of tdTomato. (Zeiss 510 Meta (Carl Zeiss Inc., Thornwood, NY) 63 X/1.4 NA oil).

Figure 3

FLIM-FRET measurements. The mTFP-5aa-mVenus (F1-F2), mTFP-5aa-Amber-10aa-tdTomato (F1-F3), and mTFP-5aa-mVenus-10aa-tdTomato (F1-F2-F3) constructs were analyzed by FLIM-FRET microscopy. The unquenched donor lifetime was determined from mTFP-5aa-Amber cells using a single exponential decay model. The quenched donor lifetime in the presence of an acceptor—mVenus, tdTomato, or mVenus-10aa-tdTomato—was based on a biexponential decay model. All lifetimes were determined using an estimated instrument response function (IRF) of 300 ps at the full width at half-maximum for the FLIM system (dotted line in A). (A) Representative raw data and decay fit for each construct—mTFP-5aa-Amber (dot-dashed line), mTFP-5aa-Amber-10aa-tdTomato (short-dashed line), mTFP-5aa-mVenus (long-dashed line), and mTFP-5aa-mVenus-10aa-tdTomato (solid line). (B) Corresponding fitting residuals. (C) Lifetime distributions for quenched mTFP in mTFP-5aa-mVenus-10aa-tdTomato (1.21 ns), mTFP-5aa-mVenus (1.61 ns), and mTFP-5aa-Amber-10aa-tdTomato (2.08 ns) and unquenched mTFP in mTFP-5aa-Amber (2.73 ns). These measurements clearly indicate the fastest decay of mTFP in the presence of the two acceptors (-5aa-mVenus-10aa-tdTomato) and also a faster decay of mTFP with -5aa-mVenus than with -5aa-Amber-10aa-tdTomato. (Biorad Radiance 2100 (Carl Zeiss Inc.) with Becker & Hickl SPC 150 (Becker and Hickl, Berlin, Germany) and mutliphoton excitation at 870 nm, 63 X/1.2 NA water).

Figure 4

Demonstration of the homodimerization of C/EBPα and its interaction with HP1α in live-mouse pituitary GHFT1 cells by (3sFRET microscopy. (A) The unmixed images (I₁₁, I₂₂, and I₃₃) show that C/EBPα (expressing mTFP or mVenus) and HP1α (expressing tdTomato) are co-localized in regions of centromeric heterochromatin, where the dimerization of C/EBPα (mTFP-mVenus) and its interaction with HP1α (mTFP-tdTomato and mVenus-tdTomato) are indicated by the E% (FRET efficiency percentage) images obtained in 3sFRET microscopy. (B) The mTFP-mVenus E% is negatively dependent (R = 0.69) on the donor (mTFP)/acceptor (mVenus) intensity ratios (triangles), where the mTFP intensity is determined from I₁₁ plus both FRET₁₂ and FRET₁₃, and the mVenus intensity is determined from I₂₂ plus FRET₂₃. (Inset) In a narrow donor/acceptor ratio range (0.3∼0.5), the mTFP-mVenus E% is independent (R = 0.01) of the acceptor level. These data indicate that the dimerized CEBPα proteins form clusters in regions of centromeric heterochromatin. The mTFP-tdTomato (diamonds) and the mVenus-tdTomato (crosses) E%s are not strongly negatively dependent (mTFP, R = 0.41; mVenus, R = 0.47) on the donor (mTFP, mVenus)/acceptor (tdTomato) intensity ratios, indicating a more mixed cluster/random interaction. The tdTomato intensity is determined from I₃₃. (Zeiss 510 Meta (Carl Zeiss Inc.), 63×/1.4 NA oil).