New instrumentation for automated time correlated single photon counting (TCSPC) from two fluorescence proteins simultaneously enabled automation of FLIM-FRET measurements to evaluate the effects of drugs on the binding of Bcl-2 family proteins fused to mCerulean3 (mCer3) and Venus fluorescence proteins. The nomenclature used here indicates proteins fused with a fluorescence protein by a superscripted c for mCer3 and a superscripted v for Venus. The superscripted letters precede the protein names for amino-terminal, and follow them for carboxyl-terminal fusions. ^VBH3-proteins were expressed by transient transfection of MCF-7 cells stably over-expressing ^CBcl-XL or ^CBcl-2 (Figure 1). Transient expression of ^VBH3-proteins ensured that there was a large enough range of expression levels within the population of cells to collect the data needed to generate binding curves (Aranovich et al., 2012; Kale et al., 2012; Liu et al., 2012). The expressed fluorescent fusion proteins retained their expected anti- or pro-apoptotic properties (see below). To generate quantitative binding curves, TCSPC data were collected for ^CBcl-XL (FRET donor) and ^VBH3-proteins (FRET acceptor), within regions of interest (ROIs) automatically identified in cells based on mCer3 signal intensity (Figure 1A). For each ROI the TCSPC data were used to determine the average fluorescence lifetime of the mCer3 donor and the fluorescence intensities of both the donor (mCer3) and the acceptor (Venus). Data from more than 8,500 ROIs were binned according to the ratio of Venus to mCer3 intensities and together with the corresponding lifetime data used to generate binding curves (Figure 1B).

Figure 1

Download asset Open asset

Figure 1—source data 1

Source data fitted to Hill equations demonstrating that ABT-263 displaced tBid and Bad but does not displace Bim from binding to Bcl-XL.

https://doi.org/10.7554/eLife.37689.004

Download elife-37689-fig1-data1-v2.xlsx

BH3-proteins engage the hydrophobic pocket of anti-apoptotic proteins through four conserved hydrophobic residues (h1-h4) within the BH3 region (Chen et al., 2005). ABT-263 prevents BH3-protein binding by competing for the binding of residues h2 and h4 to Bcl-XL (Figure 1C) (Tse et al., 2008). To quantify the importance of these residues in live cells, we expressed wild-type and mutant ^VBH3-proteins with alanine mutations in the h2 and h4 positions (BH3-2A) and measured binding to ^CBcl-XL. As expected, BH3-2A mutations in ^VBad and ^VtBid reduced binding to ^CBcl-XL dramatically (Figure 1D–E, compare black and red). However the BH3-2A mutant of ^VBimEL bound to ^CBcl-XL similarly to wild-type ^VBimEL (Figure 1F, compare black and red), suggesting interactions with other residues contribute to binding Bim to Bcl-XL. Consistent with this result ABT-263 displaced both ^VBad and ^VtBid, but not ^VBimEL from ^CBcl-XL (Figure 1D–F, compare green and cyan).

To quantify the impact of ABT-263 on ^VBH3-proteins binding to ^CBcl-XL and ^CBcl-2, and maximize the dynamic range of the assay we estimated the FLIM-FRET efficiency from the fitted binding curves at intensity ratios of Venus to mCer3 of 0.5 and 0.25, respectively. The FRET signal from the non-binding mutant ^VBad4E (in which h1, h2, h3 and h4 in the BH3 region were mutated to glutamic acid) served as a negative control for FRET due to random collisions rather than binding, and was subtracted as background (Figure 2A). The percentage of the FRET signal remaining in the presence of ABT-263 is defined here as Resistance to displacement by ABT-263 (R_ABT-263, Figure 2A). The effects of mutations on the binding of BH3-proteins can be calculated similarly and for the BH3-2A mutation is reported as R_BH3-2A. R_ABT-263 and R_BH3-2A values for all of the mutants analyzed are provided in Table 1.

Figure 2 with 4 supplements see all

Download asset Open asset

Figure 2—source data 1

Source data for the impact of small-molecule inhibitors and BH3-sequence mutations on the interactions between anti-apoptotic proteins and BH3-only pro-apoptotic proteins.

https://doi.org/10.7554/eLife.37689.006

Download elife-37689-fig2-data1-v2.xlsx

Plotting R_ABT-263 for different interactions indicated that >80% of ^CBcl-XL:^VBimEL and ^CBcl-2:^VBimEL complexes were resistant to the addition of ABT-263 (Figure 2B). In contrast, ^CBcl-XL:^VBad and ^CBcl-XL:^VtBid complexes were less than 50% resistant to the drug (Figure 2B-left, black) while the corresponding Bcl-2 complexes were less than 40% resistant (Figure 2B-left, blue, Figure 2—figure supplement 1). Calculated similarly, R_BH3-2A values reflect the minimal effect of BH3-2A mutations on complexes of ^CBcl-XL and ^CBcl-2 with ^VBimEL (R_BH3-2A ~ 80%) compared to ^VBad (R_BH3-2A ~ 2%) and ^VtBid (R_BH3-2A ~30%) complexes (Figure 1D–F, Figure 2B, and Figure 2—figure supplement 1).

To test the generality of these results in a different cell line that is resistant to the induction of cell death by BH3 proteins, complex formation was also assessed in Baby Mouse Kidney (BMK) cells in which the genes for expression of Bax and Bak were knocked out (DKO). These cells do not undergo apoptosis in response to expression of BH3-proteins and therefore retain normal cellular morphology. As expected, the R_ABT-263 and R_BH3-2A values obtained in BMK-DKO cells for ^CBcl-XL and ^VBH3-protein complexes were similar to those observed for MCF-7 cells (^CBcl-XL (Black) in Figure 2B compared with Figure 2C, Figure 1D–F and Figure 2—figure supplement 2). Similar to the results in MCF-7 cells, the resulting R_ABT-263 and R_BH3-2A values measured for ^CBcl-XL:^VBad and ^CBcl-XL:^VtBid complexes ranged from 40% to 70% lower than those of ^CBcl-XL:^VBimEL complexes confirming that the unusual stability of ^CBcl-XL:^VBim complexes is neither cell line-specific nor a result of molecular crowding that may occur when cells undergo apoptosis (Figure 2c). As an additional control to address this latter point directly lifetime measurements were compared for manually selected images of fully adherent MCF-7 cells expressing ^CBcl-XL and ^VBim-Bad that retained relatively normal morphology (alive cells) from the bottom of the imaging well. Images of cells with a more condensed morphology (dying cells) were subsequently recorded from a higher imaging plane. The distributions of lifetimes measured for the both sets of cells were very similar confirming that the changes in dying cells do not affect the lifetime measurements (Figure 2—figure supplement 3).

Compared to ABT-263 (IC50 >10 μM), the Bcl-XL specific inhibitors A-1155463 and A-1331852 more potently disrupted ^CBcl-XL:^VBad complexes with R values of ~45% and 20% respectively, corresponding to an IC50 <1.25 μM for these compounds (Figure 2D and Figure 2—figure supplement 4). Nevertheless, their impact on ^CBcl-XL:^VBimEL complexes was limited, confirming the high stability of ^CBcl-XL:^VBimEL complexes. Results for displacement of BH3 proteins from Bcl-2 using the specific inhibitor venetoclax were uninterpretable due to interference with the FLIM measurements that resulted from the broad spectrum fluorescence of the drug. As a dual inhibitor, ABT-263 enables comparison of the effects of mutations for both ^CBcl-XL:^VBimEL and ^CBcl-2:^VBimEL complexes and was therefore used for subsequent experiments.

To confirm that the stability of the complexes measured in live cells was due to direct interactions, FRET was measured for donor-fluorophore-labeled recombinant BimL or tBid with acceptor-fluorophore-labeled Bcl-XL in a cell free liposome based system. In this assay BimL was used instead of BimEL due to the difficulty in purifying BimEL and the BH3-proteins were added at amounts predetermined to result in complete binding of BimL or tBid to Bcl-XL. ABT-263 was then titrated into the samples to compete with the BH3-proteins. As expected, a concentration of ABT-263 ~10 fold higher than that of Bcl-XL displaced tBid from Bcl-XL (IC50 ~400 nM) yet had little effect on the interaction between Bcl-XL and BimL (stable at >3 μM) (Figure 2E). Thus similar to what was seen in cells by FLIM FRET, direct binding of Bim to Bcl-XL is more resistant to ABT-263 than is binding of tBid to Bcl-XL.

To test the functional consequences of this we measured protein binding and liposome permeabilization at the same time. Under conditions that support measurement of both, the IC50 for ABT-263 for FRET between cBid and Bim with Bcl-XL was 212 ± 43 and>700 nM, respectively while for liposome permeabilization the corresponding EC50’s were 65 ± 9 nM and 150 ± 16 nM, n = 3 independent duplicates,±standard error. Thus, the change in binding affinity results in a corresponding change in membrane permeabilization.

The functional properties of the fusion proteins used in FLIM-FRET experiments were examined using multiple assays in MCF-7 cells. Transient transfection by the requisite plasmids demonstrated that expression of ^VBim and ^VtBid killed MCF-7 cells as assessed morphologically for both cell and nuclear condensation and mitochondrial transmembrane potential (Figure 3A). In contrast, expression of ^VBad had little effect on cell death in this cell line suggesting that inhibition of Bcl-2 and Bcl-XL is not sufficient to induce apoptosis in MCF-7 cells.

Figure 3

Download asset Open asset

Figure 3—source data 1

Multiparametric source data for MCF-7 cell death in response to transient expression of BimEL or Bid and the protection afforded by stably expressed Bcl-XL or Bcl-2 and the dependence of MCF-7 cells on MCL-1 for survival.

https://doi.org/10.7554/eLife.37689.016

Download elife-37689-fig3-data1-v2.xlsx

To examine the effect of expression of exogenous ^CBcl-XL (confirmed by immunoblotting in Figure 3B) apoptosis was triggered with TNFα and cycloheximide (CHX) in the same cell clones used for FLIM-FRET above. Untreated and cells treated with only CHX were used as negative controls. As expected, untransfected MCF-7 cells were efficiently killed while cells expressing either ^CBcl-XL or ^CBcl-2 were highly resistant to this well-established method for the induction of apoptosis (Figure 3C).

To determine the importance of other anti-apoptotic proteins expressed endogenously in these cells, they were treated with specific small molecule inhibitors. Consistent with the results demonstrating resistance to Bad, MCF-7 cells were largely resistant to inhibition of Bcl-2 and/or Bcl-XL (Figure 3D). Neither selective nor dual inhibition of Bcl-2 and Bcl-XL induced substantial cell death in MCF-7 cell lines (all below 20% Dead) demonstrating these cells are not highly dependent on expression of either of these anti-apoptotic proteins. Nevertheless, expression of ^CBcl-2 or ^CBcl-XL reduced cell death further (Figure 3D).

In contrast, MCF-7 cells were sensitive to inhibition of MCL-1 (Figure 3E). Moreover, overexpression of ^CBcl-XL or ^CBcl-2 protected MCF-7 cells from the MCL-1 inhibitor S-63845, likely by inhibiting pro-apoptotic proteins displaced by the BH3-mimetic from MCL-1 (Figure 3E). Taken together the data demonstrate that MCF-7 cells are primarily dependent on expression of MCL-1 for survival and that the exogenously expressed fluorescence protein fusions function as predicted in this cell line.

It is commonly accepted that the only stable binding interaction between Bim and anti-apoptotic proteins is via the BH3-region of Bim (Bim BH3) binding in the hydrophobic groove formed by the BH3 1–2 regions of the anti-apoptotic proteins (Liu et al., 2010; Sattler et al., 1997). However, resistance to ABT-263 suggests that in the context of the whole protein either the Bim BH3 region binds differently than BH3-peptides to Bcl-XL or a region not included in previous in vitro experiments, such as the N-terminal and C-terminal sequences shared by all three isoforms, contains sequences conferring ABT-263 resistant binding. To test these sequences, mutants harboring deletions in ^VBimEL and mutants in which the BH3-regions were exchanged between BimEL and Bad (Figure 4A) were assayed by FLIM-FRET for resistance to ABT-263 and to the BH3-2A mutation (Figure 4B–C and Figure 4—figure supplement 1).

Figure 4 with 1 supplement see all

Download asset Open asset

Figure 4—source data 1

Source data for the calculation of R values for resistance of Bcl-XL:BimEL and Bcl-2:BimEL complexes to ABT-263 for the various mutant BH3 proteins.

https://doi.org/10.7554/eLife.37689.018

Download elife-37689-fig4-data1-v2.xlsx

The trends observed for binding of the mutants to ^CBcl-XL and ^CBcl-2 were remarkably similar. Truncation of the N-terminus of BimEL (^VBimEL-dN) had little impact on Bim binding (Table 1), but replacing the BH3 region in Bad with that in BimEL (^VBad-BimEL) increased R_ABT-263 compared with ^VBad. Consistent with this result, binding of the reciprocal mutant (^VBimEL-Bad) to either ^CBcl-XL or ^CBcl-2 was more sensitive to the drug and the BH3-2A mutation (R_ABT-263 ~60–65%, Figure 4B R_BH3-2A ~30% Figure 4C, black and blue) respectively, compared to ^VBimEL (both R values ~ 80%, Table 1). As expected, deletion of the entire BH3 region (BimEL-d20, Figure 4A) reduced the interactions with Bcl-XL and Bcl-2 to a level similar to random collisions in the membrane (R_d20 ~25%, Figure 4—figure supplement 1F and L). Unexpectedly, deletion of the C-terminal sequence (CTS) resulted in significant decreases in R_ABT-263 and R_BH3-2A (Figure 4B–C and Table 1). Combining deletion of the Bim CTS with replacement of the Bim BH3 region by that of Bad (^VBimEL-Bad-dCTS), reduced the values of R_ABT-263 and R_BH3-2A to that of ^VBad (Figure 4B–C and Table 1). Thus, both BH3 and CTS sequences contribute to Bim binding anti-apoptotic proteins.

To examine the functional importance of the Bim CTS further, the pro-apoptotic activities of selected Bim mutants were examined in MCF-7 cells with and without exogenously-expressed Bcl-XL or Bcl-2. Expression of ^CBcl-XL or ^CBcl-2 efficiently protected cells from lower levels of expression of ^VBimEL and ^VBimEL-dCTS (<1000 units of Venus intensity) however, in neither case was the protection complete at higher levels of expression (Figure 5A).

Figure 5

Download asset Open asset

Figure 5—source data 1

Multiparametric cell death data for the mutants demonstrating that the Bim CTS contributes to Bim mediated inhibition of Bcl-XL and Bcl-2.

https://doi.org/10.7554/eLife.37689.022

Download elife-37689-fig5-data1-v2.xlsx

Unexpectedly, in MCF-7 cells there was no significant difference in induction of cell death by removing the CTS from BimEL (Figure 5A). There are two factors that are likely contributing to this result. First. Bim not only inhibits anti-apoptotic proteins it also activates the executioner proteins Bax and Bak, potentially confounding analysis. As this mechanism may be relevant in multiple cell lines it is examined in more detail elsewhere (Chi et al., 2019). Second, MCF-7 cells depend on MCL-1 for survival (Figure 3E).

To avoid the contribution to cell death by inhibition of MCL-1 we made use of the Bim mutant with the Bad BH3 region (^VBimEL-Bad) as it is not expected to inhibit Mcl-1. Consistent with expectation, ^VBimEL-Bad was less effective than ^VBimEL in killing MCF-7 cells (Figure 5B). Deletion of the CTS from ^VBimEL-Bad resulted in a further loss of pro-apoptotic activity suggesting that in ^VBimEL the effect of the CTS was masked by ^VBimEL binding to and inhibiting MCL-1 (Figures 5A and 3b). Consistent with the results for ^VBimEL, inhibition of ^VBimEL-Bad by Bcl-2 and Bcl-XL was limited. While this result is also consistent with a role for MCL-1 in binding pro-apoptotic proteins displaced from Bcl-2 and Bcl-XL by ^VBimEL-Bad there are other possible explanations that will be the subject of future investigations.

To examine potential molecular mechanisms underlying the effects of deletion of the CTS from Bim in a more controlled system we examined BimL and BimL-dCTS binding to Bcl-XL and displacement of BH3 proteins from Bcl-XL using purified proteins and the FRET based assays similar to those described above (Figure 2E). When assayed directly, in incubations containing mitochondria, deletion of the CTS from BimL reduced the apparent affinity for binding to Bcl-XL (Figure 5C). As predicted from the reduced affinity of this interaction, BimL-dCTS was also less effective at displacing tBid and activated Bax from Bcl-XL (Figure 5D and E, respectively. Control FRET measurements indicated that at the concentrations used binding of cBid by Bcl-XL was saturated and that all of the proteins other than BimL-dCTS were bound to liposome membranes. Moreover, the interactions take place on the membrane surface as both tBid and activated Bax very efficiently bind to liposomes (Kale et al., 2014).

There are five hydrophobic residues/groups named h0-h4 on the hydrophobic side of the BH3 α-helix in pro-apoptotic BH3-proteins hypothesized to mediate BH3-protein binding to anti-apoptotic proteins (Figure 6A). To quantify the binding due to each of these residues without potential interference from the additional binding site within the Bim CTS, mutations were introduced into a chimeric fusion protein composed of the Bad sequence with the Bim BH3 region (^VBad-Bim in Figure 4A). The introduced mutations substitute residues in the Bim BH3 sequence at positions h0, h1, h1 +2 (an R between h1 and h2), and h3 with the corresponding residues in Bad (Figure 6A). When measured by FLIM-FRET, the relative effects of most of the individual mutations on the R_ABT-263 and R_BH3-2A values for ^VBad-Bim binding to ^CBcl-XL (black) and ^CBcl-2 (blue) reduced binding by 10–20% (Figure 6B, Figure 6—figure supplement 1 and Table 1). Among these the h0 and h1 residues contributed to Bim binding to Bcl-2 and Bcl-XL the most. Substitution of the h0 sequence with WAA reduced R_ABT-263 for ^VBad-Bim to ^CBcl-XL from 98% (95% Confidence Interval (CI) 6%) to 79% (CI 6%). Similarly, substitution of I146 with Y decreased R_ABT-263 for ^VBad-Bim to 83% (CI 5%). The reductions in R_ABT-263 for ^VBad-Bim to Bcl-2 were similar (Table 1). However, the R_ABT-263 for ^VBad-Bim binding to both Bcl-XL and Bcl-2 was unexpectedly high. As Bad also has a CTS sequence that binds to membranes and may have other as yet uncharacterized functions we examined the effect of the I146Y mutation in BimEL-dCTS. The R_ABT-263 for Bcl-XL for this mutant (BimEL-I146Y-dCTS) was 38% (CI 8%) compared to 69% (CI 5%) for ^VBimEL-dCTS binding to ^CBcl-XL suggesting a role for I146 in BimEL binding to Bcl-XL (Figure 6C). Although a reduction was also seen for ^CBcl-2 the confidence intervals overlapped suggesting there are subtle differences in Bim binding to Bcl-XL and Bcl-2. The small differences observed for substitutions at other positions within Bim-BH3 suggest either we have reached the limit of sensitivity for the FLIM-FRET assay or they may augment but individually are not critical for ABT-263 resistant binding of Bim to Bcl-XL and Bcl-2.

Figure 6 with 1 supplement see all

Download asset Open asset

Figure 6—source data 1

Source data for the calculation of R values for mutants demonstrating that the h0 and h1 residues in the Bim BH3 contribute to the resistance of Bcl-XL:Bim complexes to ABT-263.

https://doi.org/10.7554/eLife.37689.024

Download elife-37689-fig6-data1-v2.xlsx

Deletion of the Bim CTS has been shown to abrogate binding to membranes (O'Connor et al., 1998). We therefore hypothesized that the membrane binding function of this region might affect the resistance of Bcl-XL/Bim and Bcl-2/Bim complexes to either ABT-263 treatment or BH3-2A mutations by increasing the local concentration of the proteins on mitochondria. However, instead of an uninterrupted hydrophobic sequence of more than 15 residues typical of tail-anchor proteins, the Bim CTS consists of two short hydrophobic sequences and includes multiple positively charged Arg residues, two of which are near the center of the sequence (Figure 7A, green).

Figure 7 with 1 supplement see all

Download asset Open asset

Figure 7—source data 1

Source data for the experiments demonstrating that the Bim CTS binds to Bcl-XL and Bcl-2 independent of binding to membranes.

https://doi.org/10.7554/eLife.37689.028

Download elife-37689-fig7-data1-v2.xlsx

To examine the importance of sequences within the Bim CTS for ABT-263 resistant binding of ^VBimEL to ^CBcl-XL and ^CBcl-2, mutants were generated in which hydrophobic (I181, L185, I188 and V192) and hydrophilic (R186 and R190) residues of Bim were substituted with the negatively charged amino acid Glutimate (E) or the hydrophobic residue Alanine (CTS-2A), respectively (Figure 7A). The impact of these mutations on both the subcellular localization of ^VBimEL at mitochondria (measured by the Pearson’s r with MitoTracker) and R_ABT-263 values were quantified for binding to ^CBcl-XL and ^CBcl-2. Compared to ^VBimEL (Pearson’s r ~ 0.4) mitochondrial localization of all of the mutants was similarly poor (Pearson’s r 0.1–0.3) (Figure 7B–C), and only slightly greater than the Pearson’s r measured for the cytoplasmic Venus control. However, the mutants displayed different binding properties to the anti-apoptotic proteins (Figure 7C and Figure 7—figure supplement 1). For example, even though localization at mitochondria for the ^VBim-CTS2A mutant was impaired, it bound to both anti-apoptotic proteins in an ABT-263 resistant fashion (R_ABT-263 ~89%, CI 6% and 81%, CI 11%, for ^CBcl-XL and ^CBcl-2, respectively) similar to ^VBimEL (85%, CI 5% and 83%, CI 9%; Table 1 and Figure 7C). In contrast, R_ABT-263 for ^VBimEL mutants L185E (42%, CI 10%), I188E (51%, CI 8%) and V192E (70%, CI 8%) exhibited gradually increasing binding to ^CBcl-XL (Table 1) without a corresponding increase in binding to mitochondria (Figure 7C). When all of the data were analyzed in aggregate using Spearman’s rank-order correlation to include relationships that might not be linear the result was no correlation between localization for Bcl-XL and a weak correlation for Bcl-2. We suspect the latter may be due to the fact that Bcl-2 is constitutively membrane bound while Bcl-XL is located in both the cytoplasm and on the membrane. Thus in cells, localization to mitochondria and binding to anti-apoptotic proteins are independent functions of the Bim CTS.

The effect of the Bim CTS on binding to membranes and Bcl-XL was also assessed directly using purified proteins and liposome membranes. As above, binding of BimL and BimL-mutants to Bcl-XL was measured by FRET for incubations with and without liposomes. Displacement from Bcl-XL of the BimL mutants I181E, L185E, I188E and dCTS (numbered according to the BimEL sequence to facilitate comparison) were measured for different concentrations of ABT-263 (Figure 7D). Consistent with our observations in live cells, the binding of BimL-dCTS to Bcl-XL was much more sensitive to ABT-263 (IC50 ≈ 200 nM) compared to BimL (IC50 >>4 µM). These IC50 values are difficult to compare directly with R_ABT-263 values because the protein concentrations are different. Nevertheless, in this cell free assay using purified BimL and Bcl-XL, mutation of I181, L185 and L188 in the CTS region all resulted in increased sensitivity to displacement by ABT-263 (Figure 7D). As expected (Pécot et al., 2016), BimL binding to Bcl-XL in the absence of membranes was more easily displaced by ABT-263, but still required drug concentrations greater than 1 μM (compare Figure 7D with E). In solution, the mutations further decreased resistance of Bcl-XL/BimL complexes to ABT-263 (Figure 7E) confirming they augment binding of BimL to Bcl-XL independently of binding to membranes. The simplest explanation for ABT-263 resistant binding is a direct interaction between the Bim CTS and Bcl-XL in the context of the full length Bim protein.

Mutation of the BH3 residue I146 to Y slightly exacerbated the effect of deletion of the entire CTS from Bcl-XL on R_ABT-263 (Figure 6C). However, when the I146Y mutation was combined with mutation of L185E we were unable to record a decrease in R_ABT-263 (Figure 7F) suggesting that the effects of the individual point mutants are not additive. Alternatively the effect size may be too small to measure by FLIM-FRET, a result consistent with the L185E mutation having less effect on R_ABT-263 than deletion of the CTS (Table 1).

A direct interaction between the Bim CTS and Bcl-XL is incompatible with current assumptions that the Bim CTS inserts into lipid bilayers as a transmembrane helix similar to other tail-anchor proteins (Petros et al., 2004). Therefore, we expressed a version of BimEL with Venus fused to the C-terminus (BimEL^V) to prevent the Bim CTS from adopting a transmembrane topology. Distance constraints mean that this protein can only undergo FRET with Bcl-XL if the Venus protein is located on the cytoplasmic side of the membrane.

Fusion of Venus to the C-terminus of BimEL did not abolish localization of BimEL^V at mitochondria (Figure 8A and C, red bars) and did not impair ABT-263 resistant binding to ^CBcl-XL (Figure 8B–C). Moreover, the FLIM-FRET efficiency observed was higher for BimEL^V than for ^VBimEL binding to ^CBcl-XL (Figure 8B) a result inconsistent with a transmembrane topology for the Bim CTS, as it suggests that Venus on the C-terminus of Bim is closer than Venus on the N-terminus of Bim to the mCer3 on the N-terminus of Bcl-XL. These results not only confirm the two independent functions of the Bim CTS, but also suggest that when bound to membranes the residues within the Bim CTS that bind ^CBcl-XL, for example L185, are on the cytoplasmic side of the membrane.

Figure 8

Download asset Open asset

Figure 8—source data 1

Source data fitted to a Hill equation demonstrating that BimEL-venus undergoes FRET with mCer3-Bcl-XL.

https://doi.org/10.7554/eLife.37689.032

Download elife-37689-fig8-data1-v2.xlsx

To examine the topography and Bcl-XL binding of the Bim CTS in solution and membrane bound states directly, we used a well-established procedure in which an environment sensitive fluorescent dye N,N0-dimethyl-N-(Iodoacetyl)-N0-(7-nitrobenz-2-oxa-1,3-diazol-4-yl) ethylenediamine (NBD) was attached to recombinant BimL at specific sites and the environment of the dye was measured by fluorescence spectroscopy (Kale et al., 2014). A series of mutants were generated in which individual residues across the Bim CTS were replaced with cysteine to enable NBD-labeling. As a solvent exposed control, a Bim mutant was prepared with a cysteine located at position 41. The fluorescence intensities of the dye labeled mutants were then recorded in the absence or presence of liposomes, Bcl-XL, and/or the aqueous quencher iodide.

The fluorescence of NBD increases when the dye inserts into a lipid bilayer or becomes deeply buried in the interior of a protein (Johnson, 2005). When Bim bound to Bcl-XL in solution there were no significant changes in hydrophobicity at any of the positions of the probe suggesting that during complex formation these residues do not become sufficiently buried to be protected from water (Figure 9A, diagrammed in 9C). In contrast, when BimL bound to liposomes the NBD fluorescence increased at positions 179–182 and 191–195 suggesting that these two regions anchor the CTS to the lipid bilayer (Figure 9D, diagrammed in 9F). Complex formation with Bcl-XL on the membrane did not markedly change the pattern of which amino acids increased in hydrophobicity, suggesting that complex formation does not change the way the Bim CTS interacts with membranes substantially (Figure 9G diagrammed in 9I).

Figure 9 with 2 supplements see all

Download asset Open asset

Figure 9—source data 1

Source data for Interaction of the Bim CTS with liposomes and Bcl-XL measured using purified recombinant full length proteins.

https://doi.org/10.7554/eLife.37689.034

Download elife-37689-fig9-data1-v2.xlsx

Iodide quenching was used to identify residues protected by protein-protein interactions in addition to those protected by binding to membranes. Due to the relatively large size of iodide, residues involved in protein-protein interactions are often protected from quenching by iodide but unlike residues interacting with lipids the NBD fluorescence does not increase substantially because there is no protection from water (Johnson, 2005).

When Bim was incubated with Bcl-XL in solution, comparison with the control position 41 revealed that the NBD probes in the Bim CTS were not protected from iodide (Figure 9B–C, Figure 9—figure supplement 1), suggesting a relatively low affinity interaction between the Bim CTS and Bcl-XL in solution. As expected, when bound to membranes the residues at both ends of the Bim CTS that interact with the membrane were protected from iodide (Figure 9E compare residues 179–182 and 191–195). However, residues in the middle of the sequence surrounding residue 185 exhibited variable levels of protection from iodide with the NBD located at position 185 as accessible to iodide as the soluble control residue 41 (Figure 9E, Figure 9—figure supplement 1). Thus, in membrane bound Bim, the region between the two membrane binding sites in the Bim CTS remains on the surface of the membrane where it is accessible to iodide (diagrammed in Figure 9F). In contrast, when both Bcl-XL and liposomes were added the entire Bim CTS was protected from iodide (Figure 9H, Figure 9—figure supplement 1) strongly suggesting that on membranes Bcl-XL binds to or at least covers-up the central region of the Bim CTS (diagrammed in Figure 9I).

As position L185 was shown to have the largest effect on binding of BimEL to both ^CBcl-XL and ^CBcl-2 (largest decrease in R_ABT-263, Figure 7C), and is central to the region protected from iodide when BimL bound to Bcl-XL (compare Figure 9E and H) we made a series of mutations at this location for ^VBimEL-BH3-2A. In general, exchanging L185 with other amino acids with hydrophobic or polar non-charged side-chains (G, F, M, and S) did not affect the response of ^CBcl-XL:^VBimEL-mutant complexes to BH3-2A mutations significantly, except for the L185A mutant that displayed increased binding affinity (increased R_BH3-2A) (Figure 9J and Figure 9—figure supplement 2). For Bim-L185P, R_BH3-2A decreased by about 20%, possibly due to a change in the structure or rigidity of the Bim CTS. However, replacing L185 with hydrophilic residues (H, R, D or E) decreased R_BH3-2A, by ~30–60%, comparable to truncation of the CTS at position 185 (L185stop) or deletion of the entire CTS region (^VBimEL-dCTS, Figure 4C). These results strongly suggest that L185 interacts with Bcl-XL through hydrophobic interactions.

Overall our data suggest a revised view of the Bcl-XL:Bim interaction (diagrammed in Figure 9K). Because Bim is bound to Bcl-XL by two different regions, the Bim BH3 and Bim CTS, the proteins are effectively double-bolt locked together.

Plasmids encoding the mCerulean3 (mCer3), or Venus fluorescence proteins in place of the EGFP coding region in pEGFP-C1 (Clontech) were kind gifts from Mark A. Rizzo (University of Maryland) and Ray Truant (McMaster University) respectively. To generate plasmids encoding the fusion proteins the required coding regions were amplified by PCR and inserted into the plasmids encoding the appropriate fluorescence protein. All of the plasmids included coding sequences for a GGS linker sequence between the coding regions except for ^VBad (linker sequence, SGLRSRGG) and BimEL^V (linker sequence, SRGGGPVAT). All the swapping and site-directed mutants were obtained through PCR-based mutagenesis using oligonucleotides from Integrated DNA Technologies and Phusion DNA polymerase from New England Biolabs. ABT-263 was purchased from Selleckchem, A-1155463,A-1331852 and ABT-199 were from Chemietek.

The MCF-7 human breast cancer cell line was cultured in Dulbecco's Modified Eagle Medium (DMEM, ThermoFisher) supplemented with 10% of fetal bovine serum (HyClone) and non-essential amino acids (NEAA, ThermoFisher). Transfections were performed using Fugene HD reagent according to the manufacturer’s standard protocol (Promega). MCF-7 cell clones stably transfected with vectors encoding mCer3, ^CBcl-XL or ^CBcl-2 were selected in DMEM supplemented with 10% fetal bovine serum, non-essential amino acids (NEAA, ThermoFisher) and 500 μg/ml neomycin. After 3 weeks colonies were isolated and cultured as above. The baby mouse kidney (BMK) cell line in which the genes for Bax and Bak have been deleted (BMK-DKO) was a kind gift from the originator Eileen White and was cultured as above. The two originating cell lines used in the studies reported here (MCF-7 and Baby Mouse Kidney cells with both Bax and Bak knocked out) and all stably transfected clones were shown to be free of mycoplasma using a PCR based test. The MCF-7 cells were authenticated by the hospital for Sick Children facility in Toronto. Clones that stably express mCer3 or ^CBcl-XL were selected and cultured with 5 μg/mL Blasticidin S. For microscopy, cells were seeded in 384-well imaging plates (PerkinElmer) and cultured as above for 24 hr prior to transient transfection. Cells treated with ABT-263, A-1155463 or A-1331852 were incubated at 37°C for 18–24 hr in a fresh media containing 20 μM drug before imaging unless indicated otherwise.

MCF-7 cells grown in a 384 well plate (Cell Carrier Ultra) were transfected, using the TransIT-X2(Mirus) reagent and manufacturer’s protocol, with plasmids encoding Venus protein alone, or Venus protein fused to BimEL (^VBimEL), tBid (^VtBid) or Bad (^VBad). Media was changed 3 hr later, and cells were incubated 24 hr. Prior to imaging, cells were stained with 10 nM tetramethylrhodamine (TMRE, to stain mitochondria transmembrane potential, Life Technologies) and 5 µM DRAQ 5 (to stain nucleic acids, Biostatus, UK) and incubated for at least 30 mins at 37°C, 5% CO₂. Micrographs of the cells were acquired on an OPERA Phenix (PerkinElmer), using the 20x water immersion lens (NA 1.0). Data were acquired within 2 hr after staining and during imaging the cells were maintained at 37°C, 5% CO₂. Four channels were collected simultaneously: mCer3 (Ex 425 nm, Em 435–480 nm), Venus (Ex 488 nm, Em 500–550 nm), TMRE (Ex 561, Em 570–630), DRAQ 5 (Ex 640 nm, Em 650–760 nm). Acquisition was automated and acquisition settings were identical between wells and for each independent replicate (n = 3). Acquisition settings were different for Figures 3 and 5 therefore, the X-axis is not comparable. For the data in Figure 5 the cloning vector, pSPUTK (Stratagene Santa Clara CA) was used to dilute the plasmid encoding the ^VBH3 proteins, to reduce overexpression of Venus-tagged protein while ensuring consistent transfection efficiencies.

Cell images recorded on the OPERA Phenix were imported into Harmony high-content analysis software (PerkinElmer). The seven step pipeline built to identify live and dead cells was as follows: 1) Nuclear and total cell areas for each cell were identified using DRAQ five intensity data and a Harmony segmentation algorithm. 2) The cytoplasmic area was obtained by subtracting the nuclear area from the total cell area. 3) Cell images that extended to or beyond the any edge of the micrograph were removed from the analysis as the total area is uncertain. 4) The total intensity of TMRE within each cell image was calculated. 5) Nuclear area and morphology features (e.g. roundness) were calculated. 6) Cell area and morphology features were calculated. As positive and negative controls,>100 cells treated with 500 ng/ml TNFα and cycloheximide (apoptotic cells) and >100 untreated cells, respectively were analyzed similarly for each experiment. 7) The intensity of Venus and mCerulean3 was measured for all selected cells. Data from the control cells were used to train a linear classifier in the Harmony software. Each cell death curve represents measurements for at least 1000 cells for each of three replicates.

To confirm that the linear classifier reported data similar to manual analysis a subset of the data was reanalyzed with manual thresholds to score cells as, ‘TMRE negative’ (TMRE intensity at least two standard deviations lower than the average in an untreated well) or ‘small nuclear area’ (Nuclear area at least two standard deviations lower than the average in an untreated well).

Cell death was also measured as a function of the amount of the transiently expressed protein, such as ^VBimEL, in cells. For this analysis individual cells were classified as alive or dead and then results were binned by Venus intensity. For these experiments the Venus intensity serves as an estimate of the amount of the BH3 protein produced in the cell. The range of intensities within each bin was kept consistent for each transfectant, for all three independent replicates. Mean % Dead and standard error was plotted for each intensity bin.

Steady-state fluorescence images and fluorescence lifetime images were both acquired using a custom-built Alba confocal microscope from ISS (Champagne, Illinois), which measures fluorescence lifetime by time correlated single photon counting (TCSPC). Channels of mCer3 and Venus were acquired simultaneously using 445 nm and 514 nm pulse interleaved excitation (PIE) through a 442/512/561 multiband filter and emission split by a 520nm-longpass filter was collected through 459–499 nm and 528–555 nm bandpass filters, respectively. All images were acquired as 256 pixels x 256 pixels at 25°C using a 60 × 1.3 NA PlanApo water immersion objective and laser powers set to minimize photobleaching and photon pileup during acquisition with 0.1 ms per pixel dwell time and 5-frame repeat. TCSPC-data were processed using VistaVision software (ISS) and the average lifetime for each pixel was obtained by fitting FLIM pixels binned to ensure a total photon count >1000 for each set of binned pixels analyzed. The lifetime and photon count for each pixel in both channels were exported for additional analysis using ImageJ.

For Bcl-2 family proteins the problem of measuring protein:protein interactions in live cells (Herce et al., 2013) is particularly acute as the proteins often function differently at membranes therefore, assays using cell extracts do not capture either the correct or the complete set of activities (Lovell et al., 2008). Moreover, solubilizing membranes requires detergents artificially increase and decrease Bcl-2 family protein:protein interactions (Lovell et al., 2008). The use of purified proteins and cell free assays has provided much useful information, however, due to the difficulty of purifying full-length Bcl-2 family proteins, many assays in vitro are based on truncated proteins or peptides and the environment does not well represent that of an intact cell.

We used FLIM-FRET in live cells using fluorescent protein tagged full-length proteins to quantitatively study protein:protein interactions for Bcl-2 family proteins and to assess the impact of small-molecule inhibitors on these interactions (Aranovich et al., 2012; Kale et al., 2012; Liu et al., 2012). This technique removes the disadvantages of biochemical methods, but has its own limitations. For example the over-expression of a fusion protein with a fluorescent protein may result in protein concentrations within cells that are above the Kds for the interactions, affecting the relative Kds measured. For these reasons it is difficult to compare interactions between different pairs of proteins, instead the method is best suited to comparing the effects of small molecules and mutations on interactions.

In our application of FLIM-FRET the ratio of Venus to mCer3 intensities was used as an estimate of the relative amounts of the proteins to plot binding curves. Stable expression of mCer3 and using measurements from cells with little variation in amount of mCer3 allows ratio intensities to serve as a reasonable approximation for generating binding curves, but distorts the meaning of the Hill coefficients obtained. Moreover, in some cells high expression of Venus reduces the expression of mCer3 and for the some of the proteins analyzed this effect can dominate at high ratios of Venus to mCer3. As a result, in some experiments the data at ratios of Venus to mCer3 greater than 0.5 are not well described by a standard binding curve (e.g. Figure 2—figure supplement 1, panel C, black). For this reason quantitative comparisons were restricted to those regions of the curves that are fit using a standard binding curve with a Hill slope. For ^VtBid and ^VBad it was not possible to measure binding to ^CBcl-XL at ratios of Venus to mCer3 intensities greater than ~0.7 in the presence of ABT-263 because the released ^VBH3-proteins killed the cells.

Another complication with using any kind of FRET in live cells is the impact of collisions that can generate false positive signals. Fortunately, the latter can be distinguished by measuring binding curves using FLIM-FRET. Collisions vary directly with protein concentration and therefore generate straight lines rather than curves that saturate (e.g. Figure 4—figure supplement 1E, red).

To limit the effect of compounds or mutagenesis potentially changing both the binding affinity and the maximum FRET efficiency to different extents and still generate a single descriptor that can be used to describe the interactions, we introduced the concept of Resistance (R) of a complex to a specific compound (R_ABT-263) or mutation (R_BH3-2A) (Figure 2A). Quantitative analysis based on values of R allows comparison between multiple samples each that results in a unique binding curve and that was generated from measurements of thousands of different intracellular ROIs that would be hard to compare otherwise. R values cannot replace the original binding curves for detailed determination of how the interaction is disturbed. Binding curves for ^CBcl-XL and ^VBad-Bim-I146Y with and without the BH3-2A mutation allowed quantification of the decrease in binding affinity due to the I146Y mutation (Figure 6).

To generate binding curves, ROIs containing mitochondria were selected, and the corresponding intensities in both mCer3 and Venus channels and the average lifetime for mCer3 were extracted from the original TCSPC data using a customized ImageJ script. ROIs were automatically discarded from the analysis if the ^CBclXL signal was too low and therefore noisy, or close to saturated, or the Venus signal was close to saturated. The ratio of Venus to mCer3 was calculated using the intensities of each channel. FLIM-FRET efficiency (E%) was calculated for each ROI as: E%=(1-τi/τ0)×100%, where τi is the mean lifetime for that ROI and τ0 is the average lifetime for mCer3 for all ROIs not expressing detectable Venus. FLIM-FRET efficiency was then distributed into bins of the same size according to Venus:mCer3 intensity ratio (bin size 0.1 for ratios lower than 0.5, 0.25 for ratios between 0.5 and 1.25, 0.5 for ratios higher than 1.25) and plotted (±se) against Venus:mCer3 ratio. The binding curves were fitted using GraphPad Prism version 5.0d for Macintosh (GraphPad Software, San Diego California USA, www.graphpad.com) with the function: E% = Emax × (I_Venus÷I_mCerulean3)^h/[Kd^h+(I_Venus÷I_mCerulean3)^h]. Emax is the maximum FLIM-FRET efficiency corresponding to saturation of donor binding sites by an acceptor; I_mCerulean3 and I_Venus are intensities of mCer3 and Venus, respectively; Kd is the relative equilibrium dissociation constant; h is the Hill slope.

BMK-DKO cells transiently expressing Venus-tagged constructs of interest were stained with 5 µM DRAQ 5 (BioStatus) to stain nucleic acids, and 500 nM MitoTracker-Red to stain mitochondria, (Life Technologies) in DMEM at 37°C and 5% CO₂, 30 min. One positive control well was additionally stained with 500 nM MitoTracker-Green (to stain mitochondria, Life Technologies). Cells were incubated at 37°C, 5% CO₂ during sample imaging. For each well, 60 images at 256 × 256 pixels resolution were recorded using a 40x water immersion (NA 1.1) lens on an Opera Phenix confocal microscope. DRAQ 5 (Ex 640 nm, Em 650–760 nm), Venus (Ex 488 nm, Em 500–550 nm) and MitoTracker-Red (Ex 561 nm Em 570–630 nm) channels were acquired sequentially, automatically switching between channels at each field of view.

An analysis pipeline was created in Cell Profiler 2.2.0. In this pipeline, the DRAQ five channel was used to identify nuclei, and total cell area with smoothing to ensure the entire cell is selected as one region. Cells that extend to the border of the image were rejected from, the analysis. The cytoplasm for each cell was identified as an object (Cytoplasm = ‘Total cell area’ – ‘nuclear area’). Objects suitable for analysis were identified by mean Venus intensity to identify cells expressing appropriate levels of the Venus-tagged proteins of interest. A median Venus intensity (low threshold) was used to discard any cells that passed the mean Venus filter due to improper segmentation (ie. overlap between an untransfected cell and a bright transfected cell). Objects were further selected by mean MitoTracker-Red intensity (high and low threshold), to ensure appropriate staining. A size filter was used to remove small dead cells or debris and any abnormally large (possibly improperly segmented) cells. Background was subtracted for each image and a Pearson’s correlation coefficient (Pearson’s r) between Venus and MitoTracker-Red signals was calculated for all remaining objects. In each replicate, a minimum of 20 cells were identified per protein of interest and mean Pearson’s r was calculated. In sum, at least 150 cells were analyzed per sample. The average and standard error of the mean Pearson’s r values determined for the three replicates was plotted using GraphPad Prism.

Full length and single cysteine mutants of Bcl-XL and tBid were purified as described previously (Kale et al., 2014). For Bim, the cDNA encoding full-length wild-type murine BimL was introduced into pBluescript II KS(+) vector (Stratagene, Santa Clara CA). Sequences encoding a polyhistidine tag followed by a TEV protease recognition site (MHHHHHHGGSGGTGGSENLYFQGT) were added to create an in-frame fusion to the N-terminus of BimL. BimL-dCTS was constructed by introducing a stop codon (TAA) after position P121 to delete the entire CTS. Single cysteine mutants were generated by PCR-based mutagenesis using oligonucleotides from Integrated DNA Technologies and Phusion DNA polymerase from New England Biolabs.

The recombinant proteins were expressed in Arabinose Induced (AI) Escherichia coli strain (Life Tech, Carlsbad, CA). E. coli cells were lysed by mechanical disruption with a French press. The cell lysate was separated on a Nickel-NTA column (Qiagen, Valencia CA) to bind the recombinant His-tag fused proteins and after washing a buffer containing 300 mM imidazole was applied to elute the proteins. This elution was then adjusted to 150 mM NaCl and applied to a High Performance Phenyl Sepharose (HPPS) column. BimL was eluted with a no salt buffer and dialyzed against a buffer containing 10 mM HEPES pH7.0, 20% Glycerol, and then flash-frozen and stored at −80°C.

Single cysteine mutants of Bcl-XL and tBid were labeled with the indicated maleimide-linked fluorescent dyes as described previously (Kale et al., 2014; Lovell et al., 2008). Single cysteine mutants of BimL were labeled with the same protocol as tBid with the exception that the labeling buffer also contained 4M urea.

Single cysteine mutants of BimL (41C) and tBid (126C) were purified and labeled with Alexa 568-maleimide. A single cysteine mutant of Bcl-XL (152C) was purified and labeled with Alexa 647-maleimide. Alexa568 labeled BimL or tBid was incubated with either Alexa647-labeled or as a control unlabeled Bcl-XL along with the indicated concentrations of ABT-263. The intensity of Alexa568 fluorescence with unlabeled or Alexa647-labeled Bcl-XL was measured as F_unlabeled or F_labeled respectively. FRET, indicating protein-protein interaction, was quantified using the decrease of Alexa568 fluorescence when BimL or tBid bound to Alexa647-labeled Bcl-XL compared to unlabeled Bcl-XL. FRET efficiency was calculated as: FRET efficiency (%)= (1-F_labeled/F_unlabeled)*100% as described previously for cBid-Bcl-XL. Binding was measured after incubation at 37°C for 1 hr.

To compare the inhibitory effect of ABT-263 on different BimL mutants or tBid binding to Bcl-XL, ABT-263 was titrated into a solution containing 10 nM BH3 protein (BimL or tBid) and 40 nM Bcl-XL. At these concentrations BH3 protein binding to Bcl-XL had saturated (BH3 protein bound = 100%) and represents the maximum FRET efficiency (F_max). Inhibition of BH3 protein binding to Bcl-XL by ABT-263 was measured based on the decrease of FRET (BH3 protein bound = F/F_max*100%). Data were fitted to a standard inhibitor dose response equation with a Hill slope using GraphPad Prism (V6.02). For BH3 proteins in which binding was poorly inhibited (BimL (Figure 2E and Figure 7D–E), BH3 protein bound was assumed to be 0% at an infinite concentration of ABT-263. Loss of protein on the cuvettes and other surfaces at low concentrations required use of protein at concentrations higher than the Kds of binding. For this reason the dissociation constant is not independent of protein concentration and the data in Figure 2E, Figure 7D–E report relative displacements comparable between figures rather than absolute affinity.

The single cysteine mutants of Bim were purified and labeled with NBD (N,N0-dimethyl-N-(Iodoacetyl)-N0-(7-nitrobenz-2-oxa-1,3-diazol-4-yl) ethylenediamine; Molecular Probes, Cat. #: D-2004). The NBD fluorescence assay and iodide quenching of NBD-labeled Bim mutants was performed as described previously for Bax (Kale et al., 2014), with the exception that the assays were performed in 100 ul volume in low protein binding 96-well plates and fluorescence was measured using a plate reader (Tecan M1000). In the experiments reported here, 20 nM NBD-labeled Bim alone or 20 nM Bim plus 50 nM of Bcl-XL were incubated with 0.2 mg/mL (final lipid concentration) of liposomes. For the ‘In solution’ control, the protein(s) were incubated with an equal volume of assay buffer instead of liposomes.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Jarkko Ylanko provided invaluable assistance in running the Opera and analyzing the data generated. Nehad Hirmiz developed the Cell Profiler Script used to analyze the colocalization data. Funding was provided by grant FDN 143312 from the Canadian Institutes of Health Research. DWA holds the Tier 1 Canada Research Chair in Membrane Biogenesis.

© 2019, Liu et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.