The background fluorescence signal provided by the split YFP system was evaluated by cloning the split YFP N-terminal fragment upstream to an HA tag (N-YFP–HA) and the C-terminal fragment (C-YFP) independently in the backbone of the EGFP expression vector. Human breast carcinoma (MCF-7) cells were transiently transfected with the two constructs and stained with anti-HA antibodies by immunofluorescence (IF; Figure 1a, top panel). A high YFP signal observed in HA-positive cells suggested that spontaneous complementation of the two fragments when expressed in the same cellular compartment was sufficient to induce fluorescence. However, when the YFP fragments were separated to different cellular compartments by fusing the N-YFP to the mitochondrial leader sequence cytochrome oxidase subunit-VIII (mito–N-YFP–HA), the levels of the YFP signal were significantly lowered (Figures 1a and b). When N-YFP and C-YFP were independently cloned upstream to HA–Bax, negligible YFP fluorescence was detected by microscopy and by the plate reader (Figures 1a and b). This low signal could be explained by the closed conformation of cytosolic Bax in non-apoptotic cells,^{13, 14} thus preventing YFP complementation.

If the assumption presented above is correct then an apoptotic signal resulting in the activation of Bax and its homodimerization should lead to an increase in the YFP signal at the mitochondria. To address this point, N-YFP–HA–Bax and C-YFP–HA–Bax were co-expressed in MCF-7 cells and 24 h later the cells were followed by live-cell imaging. Indeed, YFP background levels in viable cells were very low (Figure 1c, left top image, time 0). Treating the cells with etoposide resulted in a significant increase in YFP fluorescence at the mitochondria (Figure 1c, top images; Supplementary Movie 1). The mitochondria in these cells were labeled with mito-mCherry (Figure 1c, bottom images). Dynamic analysis of the temporal change in YFP fluorescence intensity showed an ∼15-fold increase in fluorescence following etoposide treatment (Figure 1c, right graph).

Bax is known to interact with the BH3-only BID protein, and thus we also analyzed the Bax–Bid split YFP pair. Expression of the Bax–Bid pair resulted in high background levels, and an apoptotic stimulus induced the translocation of the signal from the cytosol to the mitochondria, resulting in a maximal 2.1-fold increase in fluorescence (Supplementary Movie 2).

The ability of the split YFP system to monitor the conformational activation of Bax was evaluated using a number of Bax mutants (Figure 2). The P13A mutant, which was reported to show elevated apoptotic activity,²⁴ was detected in both the cytoplasm and the mitochondria of MCF-7 cells (left panels), and accordingly, YFP fluorescence was significantly higher than for wild-type Bax (right panels). The P168A mutant, which was reported to reside in the cytosol with an exposed N-terminus,²⁵ showed a high cytosolic YFP signal similar to the maximal YFP signal obtained with full-length YFP fused to the mitochondrial Smac protein. The S184V mutant, which was reported to reside at the mitochondria in healthy cells with a buried N-terminus,²⁶ showed a low mitochondrial YFP signal. Finally, it was reported previously that fusing a protein tag to the C-terminus of Bax results in high levels of cell death.²⁷ Indeed, cloning the split YFP in the C-terminus of Bax and its expression in MCF-7 cells resulted in high levels of cell death and maximal YFP fluorescence (Figure 2, right panel). Thus, the split YFP system enables to monitor the conformational changes associated with Bax activation.

In search for a split YFP protein pair that will ‘turn-on' once apoptosis has occurred at the mitochondria, we screened different protein pairs that can potentially target the YFP fragments to different cellular compartments in non-apoptotic cells and to the same compartment in apoptotic cells. Interestingly, protein pairs that are known to interact during apoptosis, such as the Cyt c–Apaf-1 and Smac–XIAP pairs, failed to show an increase in YFP fluorescence upon an apoptotic stimulus (Table 1). On the other hand, the Bax–Cyt c and Bax–Smac pairs, which are not known to interact, showed a low background signal and a significant increase in YFP fluorescence upon apoptosis stimulation (Table 1, and see below).

In response to an apoptotic stimulus, Bax translocates to the mitochondria, where it induces the release of Cyt c. In an attempt to visualize Bax translocation using the split YFP system, we generated C-YFP–Flag–Bax and Cyt c–N-YFP–HA. IF analysis confirmed that in healthy cells Bax was cytosolic and Cyt c resided in mitochondria, and that treatment with the TNFα death ligand resulted in the translocation of Bax to the mitochondria and release of Cyt c into the cytosol (Figure 3a). Co-transfection of MCF-7 cells with C-YFP–HA–Bax and Cyt c–N-YFP resulted in low background fluorescence (Figure 3b, top left panel, time 0). Treating the cells with TNFα resulted in a significant increase in the YFP fluorescence intensity at the mitochondria (Figure 3b, top left panels; Supplementary Movie 3).

To correlate the increase in YFP fluorescence with apoptosis, we co-expressed in the same cells a modified mCherry caspase sensor, which includes the effector caspase consensus cleavage site (DEVD) and the transmembrane domain of Bcl-xL (TD), which targets the fusion protein to the OMM.^{28, 29} TNFα treatment resulted in the release of mCherry into the cytoplasm shortly after the increase in mitochondrial YFP fluorescence (Figure 3b, compare bottom images to top images). The time scale of mCherry release into the cytoplasm relative to the increase in YFP signal is plotted in the left panel (graphs between the images) and right panel (top graph), respectively. The release of mCherry into the cytoplasm correlated with an increase in caspase activity (Figure 3b, bottom right graph).

Another inducer of cell death through the mitochondria is menadione, which generates reactive oxygen species (ROS) through redox cycling. MCF7 cells transiently transfected with C-YFP–HA–Bax and Cyt c–N-YFP were treated with 30 μM menadione and followed by live-cell imaging (DeltaVision; Figure 3c and Supplementary Movie 4). A significant increase in YFP was observed 120 min after treatment and peaked at 180 min.

Next we evaluated whether the expression of the Bax–Bax and the Bax–Cyt c split YFP pairs by hepatocellular somatic gene transfer (using hydrodynamic DNA injection³⁰) can be used to monitor apoptosis induced by anti-Fas antibody in liver hepatocytes.³¹ Male, 8- to 9-week-old FVB mice were hydrodynamically injected with either C-YFP–HA–Bax and Cyt c–N-YFP (Figure 4a), or N-YFP–HA–Bax and C-YFP–HA–Bax (Figure 4b), and 24 h later the mice were either left untreated or injected with a single i.p. dose of 0.55 μg/g anti-Fas antibody. Four hours after the anti-Fas antibody injection, two-photon microscopy showed the emergence of YFP-positive hepatocytes in the livers from the hydrodynamically injected mice, and the number of YFP-positive cells was elevated by the anti-Fas antibody injection (Figures 4a and b, left and right panels). Histological staining of floating sections from the livers of Fas-injected mice with an anti-cleaved/active caspase-3 antibody confirmed that a majority of the caspase-3-positive cells showed a strong YFP signal (Figure 4c, arrows). Similar results were obtained with the Bax–Cyt c split YFP pair (not shown).

MCF-7 cells were infected with lentiviruses containing N-YFP–Bax and Cyt c–N-YFP, stable clones were selected using blasticidine and the expression of the proteins was analyzed by Western blotting (Figure 5a). The levels of YFP upon etoposide treatment were determined by live-cell imaging (DeltaVision; Figure 5b). To assess whether overexpression of C-YFP–Bax and Cyt c–N-YFP results in hypersensitivity to apoptotic signals, we treated the cells with staurosporine, etoposide or TNFα for the indicated time periods and measured cell death. No significant differences in the level of cell death could be detected between naive MCF7 cells and cells that stably express the Bax–Cyt c split YFP pair (Figure 5c).

Human epithelial ovarian carcinoma cells (MLS) stably expressing the Bax–Cyt c split YFP pair showed a significant elevation of YFP fluorescence in response to staurosporine and cisplatin treatment (Figure 6a). Hypoxia-induced apoptosis was evaluated in multicellular tumor spheroids derived from MLS human epithelial ovarian carcinoma cells.^{32, 33} Spheroids with a diameter below 100 μm showed no YFP-positive cells (Figure 6b, upper panel), whereas larger spheroids (>150 μm in diameter) showed increased YFP fluorescence, mainly at the center of the spheroid (Figure 6b, middle and lower panels).

In vivo monitoring of chemotherapy-induced apoptosis was evaluated in a mixture of the stably expressing Bax–Cyt c MLS cells (2.7 × 10⁶ cells) and cells expressing the DsRed fluorescent protein (1.3 × 10⁶ cells) implanted into a dorsal skinfold window chamber in 8-week-old female nude mice.³⁴ (Figure 6c) Ten days following tumor inoculation, apoptosis was induced by a single i.p. injection of 12 mg/kg cisplatin (CP).³⁵ Intra-vital imaging of the tumors using a confocal microscope showed a significant increase in YFP-positive cells 22 and 32 h after CP injection, relative to that in saline-injected animals (Figure 6c, middle and bottom panels). Immunohistochemistry staining using an anti-cleaved/active caspase-3 antibody showed a correlation between YFP-positive and caspase-3-positive cells (Figure 6c, bottom right panel).

The N- and C-terminal halves of the YFP expression vectors were a gift from Professor Michael Frohman (Stony Brook University, NY, USA).²² The N-YFP (amino acids 1–173) and C-YFP (amino acids 155–238) were cloned upstream to numerous genes of interest. Specifically, rat Cyt c was cloned upstream to N-YFP in a pEFGP expression vector. C-YFP or N-YFP was cloned upstream to HA mouse–Bax in the pcDNA-3 expression vector. C-YFP and N-YFP–HA from the N-YFP–HA–Bax mentioned above were also cloned into an empty pEGFP expression vector. For IF staining and analysis of sub-cellular localization of C-YFP–Bax and Cyt c–N-YFP, the two genes were also cloned in-frame to a Flag tag or an HA tag in the pcDNA-3 or pEGFP expression vectors, respectively. The GFP–caspase sensor was a gift from Professor Claudio Brancolini (Universita' di Udine, Italy). The GFP was replaced by mCherry by PCR using specific primers. The mCherry was a gift from Professor Roger Tsien (University of California, San Diego, CA, USA). The Smac–YFP vector was a gift from Professor Andrew Gilmore (University of Manchester). All the plasmids that were generated were verified by sequencing and Western blot analyses.

MLS human epithelial ovarian carcinoma cells, generously provided by Professor RM Sutherland,⁴⁰ were cultured in α-Eagle's minimum essential medium (αMEM), and MCF-7 human breast cancer cells (ATCC) were grown in DMEM. Media were supplemented with 10% fetal calf serum (FCS), 2 mM -glutamine, 1% penicillin/streptomycin (Biological Industries, Beit Haemek, Israel). The MCF-7 cells were transfected using Lipofectamin-2000 (Invitrogen, Paisley, UK) in antibiotic-free medium according to the manufacturer's instructions. Twenty-four hours after transfection, the cells were treated with etoposide (100 μM; Sigma, St Louis, MO, USA), staurosporine (1 μM; Sigma), menadione (30 μM; Sigma), cisplatin (10 μM; Sigma), human TNFα (10 ng/ml; PeproTech, Rocky Hill, NJ, USA) and actinomycin D (2 μg/ml; Sigma).

Cells were plated in 24-well plates containing coverslips and transfected as described above. Twenty-four hour after transfection the cells were fixed with 3% paraformaldehyde in PBS and permeabilized with 0.2% Triton X-100/PBS. The cells were immunostained with anti-HA (3F10; Roche, Mannheim, Germany) or anti-Flag (M2; Sigma) antibodies, followed by Cy3-conjugated goat anti-IgG antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA, USA). Nuclei were stained with 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI, 10 μg/ml; Sigma). Images were collected with an Olympus IX70 microscope, equipped with a DeltaVision imaging system, using a × 40 PLAN-APO 1.42NA objective. Images were processed by constrained iterative deconvolution using the SoftWoRx software (Applied Precision, Issaquah, WA, USA).

MCF-7 cells were plated on 35-mm plates (ibidi GmbH, Germany) and transfected with the different DNA constructs. Twenty-four hours after transfection the cells were treated with a death signal (TNFα or etoposide) and visualized with an Olympus IX70 microscope, equipped with an incubator (37 °C 5% CO₂) and a DeltaVision imaging system (Applied Precision), using a × 60 objective. Images were collected every 20–45 min for 3–7 h and analyzed using the SoftWoRx software. Changes in YFP signal at a single-cell level were analyzed by converting a deconvolved time-lapse movie into a projection image. The maximal intensity of each cell was derived from a manually selected region of interest. Changes in fluorescence were also detected in MCF-7 cells cultured on a black, 96-well plate using a Synergy2 plate reader (Biotek, Winooski, VT, USA; excitation 500 nm, emission 540 nm). Twenty-four hours after transfection apoptosis was induced using etoposide, staurosporine or cisplatin in phenol red-free medium.

MCF-7 cells were cultured in a black, 96-well plate and transfected with C-YFP–HA–Bax and Cyt c–N-YFP. Twenty-four hours after transfection the cells were either left untreated or treated with 50 μM zVAD-fmk (Promega, Madison, WI, USA) to inhibit caspase activity (to reduce background activity), or treated with TNFα for various time periods. Caspase-3 activity was determined using the caspase-Glo 3/7 assay kit (Promega). A 100-μl volume of the reagent was added directly to the culture medium, shaken briefly and incubated at room temperature for 1 h. Luminescence was measured using the Wallac Victor2 1420 Multilabel plate reader (Perkin Elmer Life Sciences, Boston, MA, USA). Specific activity was calculated for each sample by subtracting the value obtained for the sample containing zVAD-fmk from the mean of the duplicate sample.

To stably express Cyt c–N-YFP and C-YFP–Bax in MCF-7 and MLS cells we used the pLenti6/V5 GateWay system from Invitrogen (Paisley, UK). Briefly, the C-YFP–Bax and Cyt c–N-YFP were independently cloned into the pCR8/GW/TOPOTA cloning system (K2500-20) and then introduced into the pLenti6/V5 vector using a GateWay LR-clonase (11791-020). Lentiviruses were generated in HEK293 FT cells using the ViraPower lentiviral expression system (K4975-00), according to the manufacturer's instructions. Stable clones were selected using blasticidin (5 μg ml⁻¹; Fluka, Sigma) and expression was analyzed by Western blot analysis. Cell viability was measured in 96-well plates using CellTiter Blue reagent from Promega.

All experiments on animals were approved by the Weizmann Institutional Animal Care and Use Committee (IACUC). For in vivo mouse experiments a total of 30 μg of mCherry caspase sensor and the split YFP fusion DNA were mixed with solution-A (2.7 mM KCl, 1.47 mM KH₂PO₄, 139 mM NaCl, 8.1 mM Na₂HPO₄ (pH 7.4)) to a total volume of up to 10% of the body weight of mouse. The entire DNA solution was injected rapidly into the tail vein of awake FVB male mice (age 8–9 weeks, n=4) in a restraining device, as reported previously.³⁰ To induce liver apoptosis, a single i.p. dose of 0.55 μg/g anti-Fas antibody (BD Pharmingen, San Diego, CA, USA) was injected.

For visualization of YFP fluorescence, the livers were excised. Half of the liver was imaged immediately by two-photon microscopy (2PM; Zeiss LSM 510 META NLO; equipped with a broadband Mai Tai-HP femtosecond single-box tunable Ti-sapphire oscillator, with automated broadband wavelength tuning 700–1020 nm from Spectraphysics, for two-photon excitation).

For histological analysis, liver samples were fixed in 2.5% paraformaldehyde (PFA), kept at room temperature for 24 h, followed by incubation in 30% sucrose for at least 48 h at 4 °C. Sequential 20-μm microtome sections were collected and stored at 4°C in PBS.

MLS human epithelial ovarian carcinoma cells were cultured in αMEM supplemented with 10% FCS and antibiotics.³² Aggregation of cells into spheroids was initiated by plating the cells from confluent cultures into agar-coated plates. After 48 h, the spheroids were transferred to a 250-ml spinner flask (Bellco, Vineland, NJ, USA) at a spinning rate of 80 r.p.m., where every 72 h for 10 days the medium was changed and a mixture of 95% air and 5% carbon dioxide was blown over the medium for 5 min.

Window chambers were implanted into a dorsal skinfold of 8-week-old CD-1 nude female mice as was described previously.³⁴ Tumors were initiated 24 h after surgery by co-injection of 4 × 10⁶ (in 30 μl of PBS) of MLS cells stably expressing either DsRed (one-third) or the Bax–Cyt c split YFP pair (two-third) to the center of the chamber. The tumors were imaged starting at 10 days after injection, under isoflurane anesthesia, by confocal microscopy (Zeiss LSM 510 META NLO; × 20 lens). Apoptosis was induced by i.p. injection of cisplatin (12 mg kg⁻¹,³⁵ dissolved in 0.9% NaCl). The percentage of YFP-coverage area was calculated using ImageJ (National Institutes of Health; http://rsb.info.nih.gov/ij/).

For immunohistochemical staining we used rabbit anti-active caspase-3 antibodies (1 : 50; Cell Signaling, Danvers, MA, USA), anti-rabbit biotin antibodies (1 : 200) and streptavidin-CY5 (1 : 100) (Jackson ImmunoResearch Laboratories). For DNA counterstaining, we used Hoechst (1 μg/ml; Molecular Probes, Invitrogen, Paisley, UK). Unstained floating sections were visualized using a Zeiss AXI0 observer fluorescent microscope using a × 20 objective. Images were obtained with an Olympus DP72 camera and analyzed using the Olympus cell^A software. Immunohistochemically stained sections were collected using an Olympus IX70 microscope, equipped with a DeltaVision imaging system, using a × 40 PLAN-APO 1.42NA objective.

Data are presented as the average±S.D. of multiple (n≥3) transfections. Student's unpaired two-tailed t-test was performed using the statistical analysis functions of Microsoft Excel. Differences were considered statistically significant at P<0.05.