Genetic visualization of protein interactions harnessing liquid phase transitions

Abstract

Protein-protein interactions (PPIs) are essential components of cellular function. Current fluorescence-based technologies to measure PPIs have limited dynamic range and quantitative reproducibility. Here, we describe a genetically-encoded PPI visualization system that harnesses the dynamics of condensed liquid-phase transitions to analyze protein interactions in living cells. The fluorescent protein Azami-Green and p62-PB1 domain when fused to PPI partners triggered a rapid concatenation/oligomerization process that drove the condensation of liquid-phase droplets for real-time analysis of the interaction with unlimited dynamic range in the fluorescence signal. Proof-of-principle studies revealed novel insights on the live cell dynamics of XIAP-Smac and ERK2-dimer interactions. A photoconvertible variant allowed time-resolved optical highlighting for PPI kinetic analysis. Our system, called Fluoppi, demonstrates the unique signal amplification properties of liquid-phase condensation to detect PPIs. The findings introduce a general method for discovery of novel PPIs and modulators of established PPIs.

Figure 1. Cytosolic formation of fluorescent puncta composed of PB1 domain fused to AG.

(a) Homo-oligomerization of the p62 PB1 domain (top) and the green-emitting fluorescent protein AG (Azami-Green) (bottom). top, PB1 self-associates in an equilibrium in a front-to-back topology to form a high-molecular-weight homo-oligomer. The conserved acidic/hydrophobic and lysine/arginine residues of PB1 are indicated by red and blue bars, respectively. bottom, AG forms an obligate tetramer complex to become fluorescent (green). The hydrophobic interfaces between AG subunits are indicated by thick bars on two adjacent sides. (b) Fluorescence images of cultured cells expressing PB1 fusion to AG and the monomeric mutant of AG (mAG1), AG-PB1 and mAG1-PB1, respectively. The absence of a thick black bar indicates that mAG1 has no hydrophobic patches on its surface. (AG/mAG1 + Hoechst) Fluorescence images of Cos-7 cells transiently expressing AG-PB1 (upper) and mAG1-PB1 (lower) one day post-transfection. Images of AG/mAG1 (green) and Hoechst33342-stained nuclei (violet) were merged. The same distribution pattern was observed in approximately 50 other cells for each construct. (AG/mAG1 + PC) Fluorescence images of HeLa cells stably expressing AG-PB1 (upper) and mAG1-PB1 (lower). Their phase-contrast (PC) images were merged. A high-magnification image is shown to zoom in on fluorescent clusters of AG-PB1 (orange inset). The same distribution pattern was observed in 4 other HeLa cell clones for each construct. Scale bars, 5 μm. See also Supplementary Fig. 1.

Figure 2. Design and validation of Fluoppi, a PPI detection system using PB1/AG tags.

(a) Schematic representation of PPI-dependent formation of fluorescent puncta. Due to the interaction between X and Y, PB1-X and AG-Y build crosslinks, resulting in the concentration of AG fluorescence (green shading). PB1 and AG are depicted as shown in Fig. 1. (b) Visualization of rapamycin-induced association between FRB and FKBP in HeLa cells co-expressing PB1-FKBP and FRB-AG transiently (left) and stably (right). To show the size of puncta, a part of the cytosol (orange box) is magnified. See also Supplementary Videos 1 and 2. The same results were obtained from approximately 50 cells in both transient and stable experiments. (c) Visualization of nutlin-3-induced dissociation of p53/MDM2 complex in two HeLa cells (cell 1 and cell 2) that transiently co-expressed PB1-p53 and AG-MDM2. See also Supplementary Video 3. The same results were obtained from approximately 50 cells in 3 independent experiments. (d) Visualization of histamine-induced oscillatory association between CaM and M13 in a HeLa cell that co-expressed M13-PB1 and CaM-KO. The image acquisition interval was 1.5 seconds. KO forms an obligate dimeric complex. The hydrophobic interface between KO monomers is indicated by a thick bar on one side. See also Supplementary Video 5. Similar oscillatory changes were observed in 4 other cells. (b–d) Domain structures of transfected constructs are illustrated (leftmost). Each domain is depicted by a rectangle. The time point of image acquisition relative to drug administration is indicated above each image. (c,d) Total fluorescence intensity of cytosolic puncta in a cell (punctum intensity, P.I.) was plotted against time (rightmost). Scale bars, 10 μm. Scale bar in magnified box (b), 1 μm. See also Supplementary Fig. 2.

Figure 3. Application of Fluoppi to quantitative analysis of drug-induced PPI blockage.

(a,b) High-content analysis (HCA) approach to the quantification of nutlin-3-induced dissociation of p53/MDM2 complex in CHO-K1 cells stably co-expressing PB1-p53 and AG-MDM2. After treatment of various concentrations of nutlin-3 for 30 min, cells were fixed with 4% PFA for 10 min and their nuclei were stained with Hoechst33342. (a) Domain structures of transfected constructs and two representative images of cells treated with 0.3 and 40 μM nutlin-3. (b) Dose (nutlin-3 concentration)-response (normalized P.I.) curve. Both puncta and nuclei were segmented automatically. P.I. was obtained by dividing the total AG (green) fluorescence from all the puncta by the total number of nuclei in each field of view. Normalized to the P.I. value at the lowest concentration.(c,d) High-throughput screening (HTS) approach to the quantification of nutlin-3-induced dissociation of p53/MDM2 complex in CHO-K1 cells stably co-expressing PB1-p53 and AG-MDM2. After treatment of various concentrations of nutlin-3 for 30 min, cells were fixed and permeabilized with PBS(−) containing 0.75% PFA and 2% Triton X-100 for 10 min and their nuclei were stained with Hoechst33342. (c) Domain structures of transfected constructs and two representative images of cells treated with 0.3 and 40 μM nutlin-3. (d) A dose (nutlin-3 concentration)-response (normalized P.I.) curve. No segmentation was performed. P.I. was obtained by dividing the total AG (green) fluorescence by the total Hoechst33342 (blue) fluorescence for each field. Normalized to the P.I. value at the lowest concentration. (e–g) Quantification of AT-406-induced dissociation of Smac/XIAP complex in HEK293 cells stably co-expressing SmacNT-PB1 and XIAP-AG. (e) Domain structures of transfected constructs are depicted (leftmost). Fluorescence images of cells 5 min before and 5 and 10 min after the addition of 25 μM AT-406 are shown (middle). Puncta were segmented automatically, and cell number was counted manually. The time course of averaged P.I. is shown (rightmost). See also Supplementary Video 6. (f) Domain structures of transfected constructs and two representative images of cells treated with 0.05 and 100 μM AT-406. (g) Dose (AT-406 concentration)-response (normalized P.I.) curve. Both puncta and nuclei (Hoechst33342-stained) were segmented automatically. P.I. was obtained by dividing the total AG (green) fluorescence from all the puncta by the total number of nuclei in each field of view. Normalized to the P.I. value at the lowest concentration. (a–g) Each experiment was performed in triplicate. Scale bars, 100 μm. Scale bars in magnified boxes, 10 μm. See also Supplementary Fig. 3.

Figure 4. PcFluoppi using PB1 and Mmj for analysis of PPI kinetics.

Hydrophobic interfaces between Mmj subunits are indicated by thick bars on two adjacent sides. (a,b) Inter-punctum exchange of Mmj fluorescence in Cos-7 cells (outlined by white dotted lines) after local UV light irradiation. The photoconverted (red) fluorescence was measured in the UV-irradiated regions (1, encircled by blue dotted lines) and the intact regions (2, encircled by red dotted lines), and their intensities were plotted against time (top). (a) One day post-transfection of Mmj-PB1, cells were imaged for green and red fluorescence before and 1, 120, and 990 min after photoconversion. (b) One day post-cotransfection of PB1-p53 and Mmj-MDM2, cells were imaged for green and red fluorescence before and 1, 10, and 120 min after photoconversion. (c,d) Inter-punctum exchange of Mmj fluorescence in HEK293 cells (outlined by white dotted lines) that expressed PB1-BclXL:T2A:Mmj-BH3 (c) and PB1-p53:T2A:Mmj-MDM2 (d). In each experiment, a cell carrying two big puncta was chosen and one punctum was specifically UV irradiated. The intensities of red fluorescence from the UV-irradiated (dotted line) and intact (solid line) puncta were plotted against time (top). (a–d) Similar results were obtained from 3 other cells for each transfection. Scale bars, 10 μm. Scale bars in high-magnification (H.M.) boxes, 1 μm. See also Supplementary Fig. 4.

Figure 5. Liquid phase transitions for Fluoppi punctate structures.

(a–c) Liquid-like state inside Fluoppi punctate structures in CHO-K1 cells stably co-expressing PB1-p53 and AG-MDM2. (a) Cells were treated with 20 μM nutlin-3 to cause punctum disappearance. After the drug was washed away, formation and growth of puncta were observed. Upon touching (0 s), two round puncta (~2 μm in diameter) fused and relaxed their shape in 2 minutes. The change in aspect ratio was plotted over time (rightmost). Similar results were obtained using 4 other puncta with diameters of <3 μm. (b) Application of FRAP to a round punctum (5 μm in diameter). The change in fluorescence intensity in the bleached region was plotted over time (rightmost). Similar results were obtained from 4 other puncta with diameters of <10 μm. (c) Cells were time-lapse imaged. PC, phase-contrast images. AG + PC, fluorescence images merged with phase-contrast images. right, A montage of a time-lapse imaging (AG + PC) of proliferating cells. One cell-division event is marked by black arrows. See also Supplementary Video 7. Most large puncta (>10 μm) were irregular in shape. Similar images were obtained using another CHO-K1 clone. (d,e) TIRF microscopy images. 2D phase separation (spinodal decomposition) was seen on the surface of a Cos-7 cell co-expressing PB1-HRas and AG-cRaf after stimulation with EGF (C). The same pattern was observed in 5 other cells. This pattern was not seen when monomeric PB1 (mPB1) was used (See Figure S2B). See also Supplementary Video 8. (a–d) Domain structures of transfected constructs are illustrated (leftmost). Scale bars, 1 μm (a,b); 10 μm (c–e). Scale bar in magnified box (d), 1 μm.

Figure 6. HomoFluoppi using mAG1-PB1 tag for visualizing homo-dimerization.

(a) Schematic representation showing that due to the homo-dimerization of X, multiple mAG1-PB1-X molecules build crosslinks, resulting in the concentration of mAG1 fluorescence (green shading). (b) Pharmacologically regulated homo-dimerization of FKBP12(F36V) monitored in HEK293 cells that stably expressed PB1-mAG1-FKBP12(F36V). The homo-dimerization was induced by 500 nM HD and then blocked by 1 μM WL. Time points of image acquisition were counted after the observation was started. Puncta were segmented automatically, and cell number was counted manually. The time course of averaged P.I. is shown (rightmost). Scale bar, 100 μm. See also Supplementary Video 9. The experiment was performed in triplicate.(c) EGF-evoked homo-dimerization of ERK2 monitored in Cos-7 cells transiently expressing ERK2-mAG1-PB1. top, ERK2 homo-dimerization was quantified in 21 cells. Punctum formation was oscillatory in 9 cells and transient in 12 cells as represented by cell 1 (upper) and cell 2 (lower), respectively. Time points of image acquisition were counted after the addition of 50 ng/mL EGF. Puncta were segmented automatically. Despite heterogeneity of the temporal profile, initial punctum formation was observed 5–10 min after the addition of EGF. The time courses of P.I. normalized to the initial peak are shown (rightmost). Data points from cell 1 and cell 2 are indicated by black open circles. See also Supplementary Videos 10 and 11. bottom, Treatment of cells with DMSO, DEL22379, or U0126 for 30 min before the addition of 50 ng/mL EGF. Representative images are shown. Percentage of cells showing EGF-induced increase in P.I. in examined cells. Approximately 20 cells were observed in quintuplicate, and data are shown as mean ± s.e.m. Statistical significance (*p < 0.0001) was examined by Bonferoni’s multiple comparison test. Scale bars, 10 μm.