Software for lattice light-sheet imaging of FRET biosensors, illustrated with a new Rap1 biosensor

Abstract

Lattice light-sheet microscopy (LLSM) is valuable for its combination of reduced photobleaching and outstanding spatiotemporal resolution in 3D. Using LLSM to image biosensors in living cells could provide unprecedented visualization of rapid, localized changes in protein conformation or posttranslational modification. However, computational manipulations required for biosensor imaging with LLSM are challenging for many software packages. The calculations require processing large amounts of data even for simple changes such as reorientation of cell renderings or testing the effects of user-selectable settings, and lattice imaging poses unique challenges in thresholding and ratio imaging. We describe here a new software package, named ImageTank, that is specifically designed for practical imaging of biosensors using LLSM. To demonstrate its capabilities, we use a new biosensor to study the rapid 3D dynamics of the small GTPase Rap1 in vesicles and cell protrusions.

Figure 1.

Rap1 biosensor. (A) Schematic of the Rap1 biosensor showing that FRET occurs when GTP-loaded Rap1 binds the RBD, bringing mCerulean3 (blue) and YPet (yellow) together. (B) Representative Western blot showing separation of both components of the biosensor, expressed using tandem viral 2A sequences. (C) Measuring the FRET response of the biosensor by titrating the RBD component with a constant concentration of constitutively active (G12V), dominant-negative (S17N), and WT Rap1. Results are from 12 (WT and S17N) or 13 (G12V) replicates of the full titration. Error is the SEM. There is no detectable FRET for the S17N mutant. This together with high FRET efficiency yielded a sensitive biosensor. All plate assays were performed with LinXe cells. (D) The biosensor reported activation when expressed with CalDAG-GEF1 and inactivation when coexpressed with Rap1GAP1. Results from n = 5 distinct replicates are shown. Error is the SEM. (E) FRET was greatly reduced by introducing either a K48A or K48A/H49A mutations into the RalGDS affinity reagent to abrogate binding (box and whisker plot, n = 12). (F) Individual frames of Rap1 biosensor activity in HUVECs with (right panel) and without (left panel) stimulation by a small molecule activator of Epac, a Rap1 GEF (8-CPT-cAMP). Scale bars, 10 µm. (G) HUVECs expressing the biosensor were stimulated with 8-CPT-cAMP (shown in gray), and the mean cell ratio was normalized to 1 at the first time point (n = 6 cells, error is the SEM).

Figure 2.

Workflow for biosensor processing of LLSM data. The software performs initial image processing steps including flat-field correction, registration, background subtraction, photobleach correction, and bleedthrough correction. Because of great variation in the intensity values at each slice, thresholding (defining the boundary of the cell) can be done manually or with a function fit to changing values specified over a series of slices and/or time points. The surface of the cell is determined through tessellation, a process in which the edge is interpolated based on the threshold values in a rolling local region. The measure of activity (e.g., the FRET ratio or total corrected FRET [Corr. FRET]) is determined for the entire field of view, and then the surface is rendered based on edge segmentation or other surfaces of interest. The advantage of this scheme is that, without additional computation, changes in segmentation can be updated immediately and multiple surfaces such as the outer edge of the cell and regions of high activity can be shown simultaneously. After the activity measures are calculated, the images are deskewed and interpolated on a user-specified grid to generate a solid 3D-cell volume. Finally, ImageTank includes scripts to analyze biosensor activity in any region of interest or in any plane through the cell and to measure these regions or objects in physically meaningful units.

Figure 3.

Rap1 activity in different types of vesicular structures. (A) Approximately half of the Cos7 cells exhibited high Rap1 activity that extended throughout the vesicular structures. These were concentrated around the nucleus. Color scale indicates corrected FRET/CFP, excluding the highest and lowest 2% of ratio values to eliminate spurious pixels. Scale bar, 2.5 µm. (B) 3 of 12 cells showed structures partially surrounded by a shell of Rap1 activity. These were seen to both emerge from and fuse with the ventral and dorsal surfaces. Scale bars, 2.5 µm. (C) The surface of a cell, and a detail drawn with topographical lines to illustrate the position of a vesicle. The time series shows a region of the cell bisected by a plane through the vesicle and the corresponding plane in a pseudo-colored map of Rap1 activity. The vesicle is formed by a dorsal ruffle fusing with the top of the cell. Scale bars, 2.5 µm.

Figure 4.

Rap1 dynamics in ruffles. (A) The ruffles at upper left (box) are cut by the planes shown at different times. Note the changing distribution of active Rap1 and Rap1 activity distributed from the ventral surface all the way to the top of the ruffles (upper right plane). Color scale indicates corrected FRET/CFP, excluding the highest and lowest 2% of ratio values to eliminate spurious pixels. Scale bars, 5 µm. (B) A concave ruffle is cut by a series of planes parallel to the coverslip, at increasing height above the coverslip. Highest Rap1 activity is seen on the surface of the ruffle facing the direction of motion (Video 4). The shape of the ruffle is evident in the pattern of Rap1 activity on the bottom of the Cos7 cell (lowest plane). Scale bar, 5 µm. (C) A dorsal ruffle is cut by a single plane showing Rap1 activity throughout the structure and in connections to a dense region of high activity on the bottom of the Cos7 cell. Scale bar, 5 µm.