Fluorescence lifetime sorting reveals tunable enzyme interactions within cytoplasmic condensates

Abstract

Ribonucleoprotein (RNP) condensates partition RNA and protein into multiple liquid phases. The multiphasic feature of condensate-enriched components creates experimental challenges for distinguishing membraneless condensate functions from the surrounding dilute phase. We combined fluorescence lifetime imaging microscopy (FLIM) with phasor plot filtering and segmentation to resolve condensates from the dilute phase. Condensate-specific lifetimes were used to track protein-protein interactions by measuring FLIM-Förster resonance energy transfer (FRET). We used condensate FLIM-FRET to evaluate whether mRNA decapping complex subunits can form decapping-competent interactions within P-bodies. Condensate FLIM-FRET revealed the presence of core subunit interactions within P-bodies under basal conditions and the disruption of interactions between the decapping enzyme (Dcp2) and a critical cofactor (Dcp1A) during oxidative stress. Our results show a context-dependent plasticity of the P-body interaction network, which can be rewired within minutes in response to stimuli. Together, our FLIM-based approaches provide investigators with an automated and rigorous method to uncover and track essential protein-protein interaction dynamics within RNP condensates in live cells.

Figure 1.

FLIM of stress granules reveal fluorescence lifetime differences between condensed and dilute phases. (A) A schematic for time-domain FLIM data capture using Acousto-Optic Tunable Filters (AOTF) followed by transformation to frequency-domain for pixel distribution on phasor plots. The TCSPC profile (red curve) represents the I(t) function used in the integral formulas. Fluorescence lifetimes on phasor plots move from longer to shorter times in a clockwise direction. (B) Fluorescence intensity images from tile-scan time-lapses of U-2 OS cells stably expressing mNG-G3BP1 (stress granule marker) before and after a 45-min treatment with 0.5 mM NaAsO₂. (C and D) Intensity-threshold masks of the dilute phase (top) and stress granules (bottom) are generated by Cell Profiler (Bray and Carpenter, 2018). The pixels within the two masks were used to generate mNG fluorescence lifetime histograms for dilute phase (solid) and stress granules (dashed green). (E) Cartoon representation of the rapamycin-inducible FKBP and FRB dimerization system used to recruit VCP-FRB-mNG to mTQ2-G3BP1-FKBP–enriched stress granules. (F) U-2 OS cells stably co-expressing VCP-FRB-mNG (green) and mTQ2-G3BP1-FKBP (cyan) were treated with 0.5 mM NaAsO₂ for 45 min to induce stress granule formation prior to imaging. Tile-scan time-lapse images were captured before and 10 min after the addition of 200 nM rapamycin to recruit VCP to stress granules. Displayed are fluorescence-intensity images cropped from tile-scan mosaics. (G) Fluorescence-lifetime histograms of VCP-FRB-mNG within stress granules (dashed green) and dilute phase (solid) were obtained before and after treatment with rapamycin. Dashed lines demarcate cell boundaries. N = 3 biological replicates for experiments in B–D (n = 60 cells) and E–G (n = 15 cells). Scale bars = 10 μm.

Figure S1.

mNG emits photons with mono-exponential decay. (A) Fluorescence intensity image of U-2 OS cells stably expressing mNG and the corresponding wavelet-filtered phasor plot. (B–D) Fluorescence intensity images (top) and decay curves (bottom) of the same cells captured before (B) and after 45 min of oxidative stress (C), and a FITC image for calibration. The instrument response function (IRF) curve is highlighted in magenta. Scale bars = 10 μm.

Figure S2.

Phasor plot filtering is a universal approach to segment stress granule FLIM data, which is independent of the type of FLIM system. (A) Fluorescence intensity image of neuroblastoma (SK-N-BE) cells stably expressing G3BP1-eGFP captured using the BrightEyes time-tagging FLIM module (Perego et al., 2023; Rossetta et al., 2022). (B) Intensity-threshold mask generated with Cell Profiler. (C–E) Unfiltered, median-filtered, and complex wavelet-filtered G and S coordinates from a single FLIM image were plotted on phasor plots with a 1% threshold of the pixel with the highest photon counts, followed by segmentation using a GMM-covariance-derived ROI to produce masks. Scale bars = 10 μm.

Figure S3.

FLIM phasor plot-guided segmentation outperforms traditional intensity-threshold segmentation approaches for generating ER masks. (A and B) Fluorescence intensity image of U-2 OS cells stably expressing mCh-KDEL and the corresponding unthresholded, wavelet-filtered phasor plot. (C) Intensity-threshold mask generated with Cell Profiler. Inset demonstrates an incomplete mask coverage of ER tubules. (D) ER mask generated by phasor plot segmentation. Inset highlights the ability to segment dim ER tubules. (E) Intensity versus FLIM was compared by paired Student’s T test with mean and SE displayed. *P = 0.0331. N = 3 biological replicates with n = 61 cells. Scale bars = 10 μm.

Figure 2.

The condensed and dilute phases of RNP condensates are resolvable on wavelet-filtered phasor plots with machine-learning algorithms. (A) Unthresholded, wavelet-filtered phasor plots of all pixels accumulated from tile-scan images from Fig. 1 B expressing mNG-G3BP1 before (top) and during oxidative stress (bottom). Note the appearance of a stress granule cluster during oxidative stress (bottom). Insets highlight the absence and presence of a stress granule lifetime cluster before and during stress, respectively. (B and C) Fluorescence intensity images of mNG-DDX6–marked P-bodies and dilute phase in A549 cells before (top) and during oxidative stress (bottom), and their corresponding unthresholded, wavelet-filtered phasor plots. Insets highlight the presence of P-body lifetime clusters before and during stress. Note the increase in density of the P-body lifetime clusters during stress, and the corresponding decrease in dilute phase density corresponding to stress-induced P-body biogenesis and increased partitioning (bottom phasor plot). Scale bars = 10 μm. (D) Cartoon of phasor plots illustrating the relationships between ROI, covariance, and the strength of partitioning. The size and shape of the ROI are determined by the covariance of the condensate cluster. n = 60 and 91 cells in mNG-G3BP1 and mNG-DDX6 expressing conditions across three biological replicates, respectively.

Figure S4.

Unfiltered and filtered phasor plots of mNG localized to condensed and dilute phases. (A and B) Unfiltered and median-filtered, unthresholded phasor plots of mNG-G3BP1 captured from U-2 OS cells in Fig. 1, B–D before and after treatment with 0.5 mM NaAsO₂. Phasor plots include pixels accumulated from FLIM tile-scan images. The insets highlight GMM-generated ROIs, which were determined by the covariance of fluorescence lifetime clusters. (C) Unthresholded, wavelet-filtered phasor plots corresponding to Fig. 1, E–G of VCP-FRB-mNG before and after VCP recruitment to stress granules with rapamycin.

Figure 3.

Automated segmentation of wavelet-filtered phasor plots generates RNP condensate masks, which outperform intensity-threshold-based segmentation for accurately detecting smaller condensates. (A) A stress granule mask generated from GMM-covariance-guided segmentation of wavelet-filtered phasor plots from Fig. 2 A (bottom) corresponding to cells from Fig. 1 B. (B) Intensity-threshold stress granule masks from Fig. 1 C were overlayed on phasor masks from A to assess their degree of overlap. Phasor masks outperformed intensity-threshold masks for smaller stress granules (<2 μm²). (C and D) Intensity-threshold and phasor P-body masks from mNG-DDX6–expressing cells in Fig. 2 B. (E) Phasor P-body masks were overlayed on intensity-threshold masks, and vice versa, to assess their degree of overlap. Phasor-derived P-body masks outperformed intensity-threshold masks for assessing size and number. Dashed lines demarcate cell boundaries. Scale bars = 10 μm. N = 3 biological replicates, n = 91 cells.

Figure 4.

Automated condensate FLIM-FRET enables live tracking of protein–protein interactions within P-bodies across many cells through NaAsO₂ treatment. (A) Cartoon schematic of a two-cell line approach to evaluate mNG-DDX6 donor lifetimes in reference (Ref.) and experimental (Exp.) cells for the quantification of mean FRET efficiency within P-bodies. A clockwise rotation of the experimental donor lifetime cluster represents shorter lifetimes resulting from FRET. (B) Fluorescence intensity images of experimental A549 cells stably expressing mNG-DDX6 and mScarlet-Lsm14a. The reference cells displayed in Fig. 2 C were captured on the same day as the experimental cells. Scale bars = 10 μm. (C) Fluorescence lifetime kernel density plots of condensate-specific mNG-DDX6 pixels from reference (dashed) and experimental (solid) cells. (D) FLIM-FRET efficiencies of individual P-bodies before and during stress were calculated by using the mean fluorescence lifetimes of reference P-bodies (τ_D) and individual experimental P-bodies (τ_nD+A) in the formula, ${\mathrm{E}}_{\text{FRET}}=100•\left[1–({\tau }_{\text{nD}+\mathrm{A}}/{\tau }_{\mathrm{D}})\right]$, such that [n] = the total number of P-bodies. The magenta line represents the mean of individual P-body FLIM-FRET efficiencies. Statistical significance was determined by nested T test, ns = not significant. N = 3 biological replicates, n = 179 cells.

Figure S5.

Stress granule FLIM-FRET detects G3BP1 dimerization in live cells. (A) Reference and experimental A549 cells expressing mNG-G3BP1 without and with mScarlet-G3BP1, respectively, were captured by FLIM after a 45-min treatment with NaAsO₂. (B) Fluorescence lifetime kernel density curves mNG-G3BP1 from reference and experimental cells. Scale bars = 5 μm. (C) Mean FRET efficiencies of individual stress granules are plotted with the mean highlighted (magenta). n = 30 reference cells and n = 27 experimental cells across three biological replicates.

Figure S6.

Context-dependent interactions between subunits of the mRNA decapping complex occur within P-bodies. (A) A549 cell lines stably co-expressing mNG-Dcp2 (green) and mScarlet-Dcp1A (magenta) were captured before and after treatment with 0.5 mM NaAsO₂. (B) FLIM-phasor-guided masks were generated from mNG-Dcp2 and mScarlet-Dcp1A images to yield Mander’s colocalization coefficient before and after oxidative stress. Statistical significance was determined using a paired Student’s T test with mean and SE displayed. As a control, image tiles were randomized by shuffling prior colocalization analyses to ensure that Dcp2 and Dcp1A were not colocalizing by chance. (C) Mean FRET efficiencies between Dcp2 and Dcp1A in individual P-bodies are plotted with the mean highlighted (magenta). (A–C) N = 4 biological replicates with n = 194 cells. (D–G) Fluorescence intensity images of A549 cells expressing (D) mNG-EDC3 and mScarlet-Dcp1A or (n = 179 cells) (F) mNG-Dcp2 and mScarlet-EDC3 (n = 186 cells) were evaluated for condensate FLIM-FRET in panels E and G, respectively. N = 3 biological replicates. (C, E, and G) Statistical significance was determined by Student’s T test. ***P < 0.0001. All scale bars = 10 μm.

Figure 5.

A condensate FLIM-FRET approach utilizing cell-permeable dyes to capture reference and experimental measurements from P-bodies of similar molecular composition or from the same cells. (A) Cartoon schematic of obtaining phasor plots from DCP1A/1B DKO cells stably expressing mNG-Dcp2 and either Halo-Dcp1A or Halo-DDX6 before and after the addition of the JaneliaFluor (JF)549 acceptor dye. (B) Representative fluorescence intensity tile-scan time-lapse images before and after JF549 addition. The insets in the 0-min panels demonstrate the absence of fluorescence when the cells were excited with a 549 nm laser. Note the cells expressing (magenta/green boundaries) and not expressing (white boundaries) Halo-markers. Scale bar = 10 μm. (C) Unthresholded, wavelet-filtered phasor plots of mNG-Dcp2 captured from panel B. (D) Mean FLIM-FRET efficiencies of individual P-bodies from cells in panel B. Note that non-Halo– and Halo-DDX6–expressing cells yield ∼0% FRET efficiencies, thus demonstrating the sensitivity of condensate FLIM-FRET. Statistical significance was determined by nested T test. ****P < 0.00001. n = 133 and 109 cells in Halo-Dcp1A– and Halo-DDX6-–expressing conditions across three biological replicates, respectively.

Figure S7.

Gene ablation of Dcp1A and Dcp1B in U-2 OS cells resulted in an increase in P-body size and abundance. (A) Dcp1A and Dcp1B DKO U-2 OS cells were confirmed by western blot analyses of Dcp1A and Dcp1B antibodies. COXIV antibody was used as a loading control. (B) Immunofluorescence of DDX6 was performed in wild-type and Dcp1A/1B DKO cells stably expressing mNG-Dcp2. Nuclei were labeled with Hoecsht. Tile-scan images were captured with a 60× objective on a laser-scanning confocal microscope. (C–F) The nuclear stain was used to approximate cell number. Particle analyses of endogenous DDX6 and mNG-Dcp2 puncta were performed to obtain (C and D) P-body numbers per cell, and (E and F) area per P-body. N = 2 biological replicates. *P < 0.01, **P < 0.001. Source data are available for this figure: SourceData FS7.

Figure 6.

The return of mRNA decapping complex FLIM-FRET is delayed during stress recovery. (A) Cartoon schematic depicting mRNA decapping complex interaction dynamics before and during oxidative stress. (B and C) Unthresholded, wavelet-filtered phasor plots and mean FLIM-FRET efficiencies of mNG-Dcp2 and JF549-Halo-Dcp1A in individual P-bodies from DCP1A/1B DKO U-2 OS cells before, during, and after a 2-h washout of NaAsO₂, where n = 148, 132, 118 cells, respectively. (D) Condensate FLIM-FRET efficiencies performed in A549 cells co-expressing mNG-Dcp2 with either Halo-Dcp1A (n = 133, 119, 127 cells, respectively) or Halo-DDX6 (n = 108, 115, 89 cells, respectively) with a similar experimental layout to panel B. (C and D) Statistical significance was determined by nested one-way ANOVA. **P < 0.001, ***P < 0.0001, ****P < 0.00001. N = 3 biological replicates for all experiments.

Figure 7.

The potential to combine condensate FLIM-FRET with computational protein structure and interaction algorithms for efficient validation of predicted models and for FLIM-FRET experimental design. (A) AlphaFold3 model of mNG (green) fused to Dcp2 (black) and Halo (magenta) fused to Dcp1A (gold). The distance between the mNG chromophore and the Halo residue, which covalently binds compatible fluorophore dyes. Blue, cyan, and yellow correspond to high, medium, and low confidence atomic predictions. (B) A magnified view of a high confidence clash between Dcp2 and Dcp1A. (C) Quantifying the number of clashes with high (blue), medium (cyan), and low (yellow) confidence structural positions on both sides of the clash for the FRET pairs analyzed in this manuscript. Highlighted in red are low FRET efficiencies.