Metabolic FRET sensors in intact organs: Applying spectral unmixing to acquire reliable signals

Abstract

In multicellular organisms, metabolic coordination across multiple tissues and cell types is essential to satisfy regionalized energetic requirements and respond coherently to changing environmental conditions. However, most metabolic assays require the destruction of the biological sample, with a concomitant loss of spatial information. Fluorescent metabolic sensors and probes are among the most user-friendly techniques for collecting metabolic information with spatial resolution. In a previous work, we have adapted to an animal system, Drosophila melanogaster, genetically encoded metabolic FRET-based sensors that had been previously developed in single-cell systems. These sensors provide semi-quantitative data on the stationary concentrations of key metabolites of the bioenergetic metabolism: lactate, pyruvate, and 2-oxoglutarate. The use of these sensors in intact organs required the development of an image processing method that minimizes the contribution of spatially complex autofluorescence patterns, that would obscure the FRET signals. In this article, we show step by step how to design FRET-based sensor experiments and how to process the fluorescence signal to obtain reliable FRET values.

Keywords: Drosophila; Autofluorescence; FRET; Linear unmixing; Metabolic sensors.

Fig. 1.

Metabolic FRET-based sensors. Illustration of the functioning of the lactate sensor Laconic (San Martín et al., 2013). (A) Scheme of the molecular mechanism of the sensor. Due to the proximity of the fluorescent proteins, excitation of CFP (donor fluorophore) results in an energy transfer to YFP (acceptor fluorophore). Union to lactate causes the fluorescent proteins to distance and thus reduces the energy transfer, and the emission of the acceptor fluorophore (YFP). (B) Diagram that shows why CFP and YFP can function as a FRET pair: the emission spectrum of CFP overlaps with the absorbance spectrum of YFP (shadowed area).

Fig. 2.

Autofluorescence varies spatially. (A) Autofluorescence in a wing imaginal disc of a third instar larva. (B) Fluorescence profile across the dotted red line in A.

Fig. 3.

A spectral unmixing approach for correcting autofluorescence pixel-by-pixel. Larval brain lobe expressing Laconic, a lactate FRET-based sensor. (A) Expression map for the lactate sensor Laconic expressed in the larval central nervous system (elav-Gal4 -> UAS-Laconic). One half of the brain lobe (black arrow) displays higher expression levels than the other half (white arrow). (B) FRET map for Laconic based on its signal obtained from the image shown in A. The straightforward ratiometric method was employed. By comparing this panel with A, it is possible to see how the apparent FRET signal is biased by the expression pattern of the sensor. (C) Autofluorescence emission spectrum of the anterior lobe of a Drosophila larval brain upon excitation at 458 nm (the excitation maximum of the donor fluorophore, CFP). (D) Laconic emission spectrum of the anterior lobe of a Drosophila larval brain upon excitation at 458 nm (the excitation maximum of the donor fluorophore CFP). (E) FRET map of the sensor Laconic after correcting the signal of the image shown in A employing the linear unmixing-based method. In this case, the weighting constant K was pre-estimated as the average of the estimated values of five brain lobes from flies not expressing the sensor (UAS-Laconic line not carrying the driver elav-Gal4). Note that the difference between the two halves of the lobe displaying different expression levels of the sensor (black and white arrows) was strongly reduced. (F) Fret signal in the right and left halves of the lobe shown in A, obtained either with the straightforward method used in cell culture (map shown in panel B) or after applying the linear unmixing method (map shown in panel E). The image processing algorithm reduced to a minimum the difference between the two halves of the lobe.