The chaperone Hsp70 is a BH3 receptor activated by the pro - apoptotic Bim to stabilize anti - apoptotic clients

The chaperone heat shock protein (Hsp70) is crucial for avoiding protein misfolding under stress, but is also upregulated in many kinds of cancers, where its ability to buffer cellular stress prevents apoptosis. Previous research has suggested Hsp70 interacts with pro-apoptotic Bcl-2 family proteins, including Bim and Bax. However, a definitive demonstration of this interaction awaits, and insights into the structural basis and molecular mechanism remain unclear. Earlier studies have identified a Bcl-2 homology 3 (BH3) domain present in Bcl-2 family members that engages receptors to stimulate apoptosis. We now show that Hsp70 physically interacts with pro-apoptotic multidomain and BH3-only proteins via a BH3 domain, thereby serving as a novel BH3 receptor, using in vitro fluorescent polarization (FP), isothermal titration calorimetry (ITC) and cell-based co-immunoprecipitation (co-IP) further demonstrated that BimBH3 binds to a novel allosteric site in the nucleotide-binding domain (NBD) of Hsp70, Bim acts as a positive co-chaperone to promote the ATPase activity and chaperone functions. A dual role of Hsp70 ’ s anti-apoptotic function was it keeps


Introduction
Hsp70 is an ATP-dependent molecular chaperone that is abundantly expressed in most cancer cells to facilitate the maturation, activation, and stabilization of many oncogenic clients to buffer cellular stress (1)(2)(3). The chaperone function of Hsp70 is based on an allosteric mechanism, in which nucleotide-binding and conformation-specific co-chaperones regulate Hsp70-client protein interaction. As such, recruitment of Hsp70 by these co-chaperones would create a local pool of the chaperone to facilitate folding of clients (4,5).
Besides, Hsp70 inhibits caspase-independent and caspase-dependent apoptosis by directly interacting with apoptosis-inducing factors, such as AIF and APAF-1, and inhibits caspase 3 and 9 activations (6,7). However, as pivotal regulators of the mitochondrial apoptosis pathway, Bcl-2 family members have crosstalk with Hsp70 in a way that is not fully understood (8). Although a few previous reports have found Bim and Bax interactions with Hsp70 by co-IP and FRET (9-11), there is no solid evidence of the physical interactions between them. The biological consequence of these complexes is unknown.
The protein-protein interactions (PPIs) of the Bcl-2 family proteins dictate apoptosis, which is mediated by a Bcl-2 homology 3 (BH3) domain of the pro-apoptotic protein that inserts into a BH3 receptor groove on the surface of anti-apoptotic proteins (12). Recent pieces of evidence have revealed that BH3-only proteins could engage into BH3-receptor proteins besides the 16 well-known Bcl-2 family proteins (13-15), suggesting the presence of unknown BH3 receptors that have crosstalk with Bcl-2 members via a by guest on November 3, 2020 http://www.jbc.org/ Downloaded from BH3 groove.
In this study, Bim was revealed that binds Hsp70 in vitro and in cells through BH3 domain (a diagram of Bim sequence was shown in Supplementary Fig S1). Bim acts as a positive co-chaperone of Hsp70 since it increased ATPase activity and oncogenic chaperone functions of Hsp70 by stabilizing its ADP-binding conformation. In opposite to its classic role as an activator of intrinsic apoptosis, Bim was revealed that could help Hsp70 to stabilize oncogenic clients AKT and Raf-1.

Results and discussion Characterization of the binding of Hsp70 to a panel of BH3 domains.
We characterized the binding ability of Hsp70 proteins to BH3 domains of Bid, Bim, Noxa, Bax, and Bak by using in vitro biochemical assays and cell-based assays. Two Hsp70 inhibitors (VER-155008 (16) and MKT-077 (17)) were tested in parallel.
FP results illustrated that the direct interactions of Bcl-2 family proteins with Hsp70 might be mediated by the BH3 domains, where MKT-077 but not VER-155008 might also occupy.
Next, we explored whether Hsp70 could bind Bcl-2 members through BH3 domain in cells. Co-IP experiments were performed in BV173 and KCL22 cells to detect multidomain pro-apoptotic members and BH3-only pro-apoptotic members complexed with Hsp70. As shown in Fig 1A, B, the binding profiles of Hsp70 were found to be different in the two cell lines. Only Bim was observed at certain levels in complex with Hsp70 in both cell lines, while Bax and Bak were found to form heterodimers with Hsp70 only in KCL22 cells, suggesting the cell context-dependent Hsp70 complex with Bcl-2 proteins. No interactions were found between other BH3-only proteins (Bad, Puma, Noxa and Bid) with Hsp70 in either of BV173 and KCL22. We further expanded the cell lines to H23 and K562, and continuously observed Bim interactions with Hsp70, whereas interactions with other BH3-only proteins were not found. Additionally, we detected some Hsp70/Bak complex but little Hsp70/Bax complex in K562 and H23 (Supplementary Fig S3). Then, we performed competitive experiments. As shown in Fig 1A and B, BimBH3 peptide efficiently disrupted the Hsp70/Bim interaction, indicating that Bim interacts with Hsp70 via BH3 domain in cells. The control BimBH3 L62A/D67A peptide was ineffective. In addition, the Hsp70/Bim complex was potently disrupted by MKT-077, while weak to no disruption was found for VER-155008 (Fig 1A), which is consistent with the in vitro binding profile.
Of note, the level of Bim remained constant for 2-24 hr after it was released by 5 μM MKT-077, and it did not show any dose-dependent change with exposure to up to 10 μM compound (Fig 1C). Hsp70-interacting protein can be divided into two groups: clients and co-chaperones. When the ATPase activity of Hsp70 is inhibited by inhibitors, for example, MKT-077, the client protein level would decrease (18). The unchanged Bim level after MKT-077 treatment indicated that Bim is likely to be a co-chaperone of Hsp70 rather than a client. Additionally, BimBH3 peptide binds to Hsp70 at higher binding affinity (K d values at sub μM range) than the affinities of the reported client proteins (typically low to mid-micromolar) (19), which also indicated that Bim might be a co-chaperone of Hsp70 proteins rather than a client.
To explore the biological relevance of Bcl-2 family proteins with Hsp70 from a bioinformatics perspective, we examined publically available CCLE database for gene expression data and determined association of Bcl-2 family protein expression with Hsp70 expression. Associations were examined in the pan-cancer (n = 837 across 16 cancers) (Supplementary Table 1). As shown in Fig  1D and Supplementary Fig S4, there is a significant correlation between Bim and Hsp70 expression (r = 0.36, p = 10 -20 , 95% confidence interval (C.I.) of (0.31, 0.41)) (20). Compared to Bim, there were non-significant positive associations between other pro-apoptotic BH3-only and multidomain Bcl-2 family members with Hsp70.
The bioinformatics analysis highlights a related biological function between Hsp70 and Bim. Thus, we focus on Hsp70/Bim complex in the following study.

Localization of the interaction site of BimBH3 on Hsp70.
To identify the interaction site for Bim on Hsp70, we expressed the recombination NBD (residues 1-383) and SBD (residues 397-506) domains of Hsp70, and analyzed the interaction of Bim with different nucleotide-bound states of Hsp70 by ITC. As shown in Fig 2A, BimLΔC27 had a 6-fold higher binding affinity for ADP-bound full-length Hsp70 (K d = 0.89 μM) (Fig 2A, left panel) relative to ATP-bound Hsp70 (K d = 5.3 μM) (Fig 2A,  right panel). Furthermore, we measured that BimLΔC27 exhibited essentially similar binding preference and affinity for ADP-bound Hsp70 NBD domain (K d = 1.0 μM) relative to ATP-bound Hsp70 NBD (K d = 6.4 μM) (Fig 2B, left and right panel), while the interactions of the SBD domain appeared to be significantly weaker (K d = 10.2 μM) ( Fig 2C). Although BimBH3 binds both NBD and SBD, no second binding event was apparent in the ITC curve of BimB3 to full-length Hsp70, suggesting that a close contact of NBD and SBD in full-length Hsp70 might form one binding unit for Bim, as similar as previously found in Hsp70/Hsp40 complex (21). The significantly higher binding affinity of Bim to NBD than SBD showed that BimBH3 mainly interacts with Hsp70 NBD in ADP-bound state and complements SBD by weak interactions.
To test whether BimBH3 resembles that of full-length Bim, we assayed the binding affinity of BimBH3. As shown in Supplementary Fig S5, Fig S6), which is consistent with the highly similar in Hsp70 and Hsc70. Because assignments of the backbone resonances of Hsc70 NBD domain have been available in literature (22), we used spectrum of 15 N-labeled Hsc70 NBD to localize the binding site ( Fig  2D). The resonance of 6 residues (R72, V146, T177, H227, T298 and R342) showed significant decrease in peak intensity (PI1/PI2>2), and 36 residues showed significant chemical shift changes (>0.02 ppm, Fig 2E). All the residues showing significant peak intensity decrease (shown in red sticks) and more than 70% of the residues showing significant chemical changes (shown in red) were located in the subdomains IA and IIA, surrounding a cleft opposite to the nucleotide-binding pocket, and therefore this cleft is indicated as the binding site of BimBH3 (Fig 2F, red circle). To validate the binding site, we mutated H227, one of the most disturbed residues in NMR experiment, and assayed the influence on Bim binding. By FP-based binding assay, we detected that compared to wild-type Hsp70 protein, the H227A mutant resulted in a 5-fold loss of binding affinity ( Supplementary Fig S7), confirming the localization of the interaction site of Bim on Hsp70. In comparison with MKT-077 and J-domain ( Fig 2F, binding sites were labelled in green and yellow circle respectively), BimBH3 binds into a site distinct from either of them. There is a region involved in both MKT-077 and BimBH3-binding areas, indicative of mutually exclusive binding. It is consistent with the competitive effect of MKT-077 against BimBH3 in FP and co-IP assay. The NMR results indicated that BimBH3 engaged in a site which is different from the previously known sites occupied by J-domain or MKT-077. To test it, we performed FP-based binding between BimBH3 and Hsp70 in the presence or absence of J-domain, and observed that J-domain cannot influence the interaction between BimBH3 and Hsp70 (Supplementary Fig S7). The result confirmed that the binding site of Bim is different from J-domain.
Interacting proteins or small molecules that bind on Hsp70 NBD domain often induce a conformational change between ADP-like and the ATP-like state. To evaluate whether Bim has such an effect, we performed trypsin treatment of human Hsp70. Consistent with previous reports (23), the addition of either ATP or ADP decelerated the proteolysis (Fig 2G). Hsp70 saturated with ATP was primarily cleaved into three high-molecular-weight bands, including prominent bands at approximately 58 kDa (band 1) and 55 kDa (band 2). Conversely, treatment with ADP strongly favored band 2. Addition of J domain, which stimulates ATP turnover in Hsp70, converted the ATP-like pattern into an ADP-like pattern, as shown by a significantly decreased ratio of band 1/band 2 than that in ATP-like state ( ** P < 0.01, Fig  2G, bottom panel). Addition of BimBH3 to Hsp70 significantly favored band 2, which is similar to the influence of ADP and J domain. These results indicated that Bim binds preferentially to the ADP-bound state of Hsp70. The newly identified Bim's binding site is a novel allosteric site by which the ADP-bound conformation of Hsp70 could be stabilized.
So far, the solid evidence of Bim binding with Hsp70 to act as an allosteric regulator has been shown, which prompted us to evaluate the effect of BimBH3 on ATPase and chaperone activities of Hsp70. We firstly evaluated the effect of Bim on Hsp70 ATPase. As shown in Fig 3A, the combination of Hsp70 (1 μM) and BimBH3 (0.1μM), which resembled the ratio of Bim-bound Hsp70 in total Hsp70 as determined by semi-quantification assay in cells ( Supplementary Fig S8), resulted in a 3.8-fold increase of the rate of ATP consumption similarly to the addition of J domain (4.2-fold), and both of their stimulation of ATPase activity was blunted by MKT-077. The control BimBH3 L62A/D67A peptide was ineffective. We also compared BimBH3 with σ 32 , a model Hsp70's client. As shown in Supplementary Fig S9 A and B, BimBH3 could stimulate the ATPase activity of Hsp70 NBD domain by 3.5-fold, which is similar to that of full-length Hsp70 (3.8-fold). In contrast, σ 32 only stimulated full-length Hsp70 (2.2-fold) but not Hsp70 NBD. Compared to BimBH3 or J domain alone (3.6 and 4-fold respectively), σ 32 combination with BimBH3 or J domain stimulated ATP hydrolysis by 7.5 and 8.3-fold, respectively, suggesting that the combination could synergize to stimulate ATPase of Hsp70. However, the effect is abolished when using Hsp70 NBD domain ( Supplementary Fig S9B), suggesting that SBD is required for the synergistic effect.

Identification of BH3-only protein Bim as
Next, we measured the aggregation of denatured rhodanese by spectrophotometry to assess the effect of BimBH3 on Hsp70 chaperon activity. As shown in Fig 3B, the addition of either BimBH3 or J domain in Hsp70 reduced rhodanese aggregation by 40-50%, indicative of the positive effect of BimBH3 on facilitating Hsp70 to bind and stabilize denatured rhodanese. MKT-077 inhibited these positive regulations of Hsp70 ( Fig 3B). Consistent with the decreased Bim binding by H227A mutant, the effect of BimBH3 on Hsp70 ATPase and aggregation of denatured rhodanese was significantly impaired by the mutant, whereas it did not influence the effect of J domain (Fig 3A and 3B). In agreement with BimBH3 peptide, we detected similarly positive effect of BimLΔC27 on Hsp70 chaperon activity (Supplementary Fig S10). These results identified that Bim acts as a positive co-chaperone of the Hsp70 chaperone cycle.
To evaluate whether the in vitro biochemical assay represented the positive co-chaperone function of Bim in living cells, BV173 and KCL22 cells were transfected with Bim shRNA. As shown in Supplementary Fig S11, Bim shRNA transfection down-regulated Bim protein level by about 90%, while Bim silence did not affect the level and the complex state of Hsp70, Hsp40 and BAG3. Then, we immunoprecipitated Hsp70 from cell lysates of wild-type and Bim-silenced cells, respectively and performed ATPase assays. As shown by Fig 3C, Bim silence led to a downregulation of the ATPase activity of Hsp70 by 30%-40% in both BV173 and KCL22 cells, confirming the positive co-chaperone role of Bim in stimulating Hsp70 ATPase in cells.
Hsp70 is an important cancer chaperone that assists in the correct folding of oncogenic client proteins. The tumor addiction of Hsp70 may be enhanced by Bim since positive co-chaperones could upregulate the level of clients and most reported so far are pro-survival proteins (24). We then evaluated the regulation of Hsp70/Bim complex on AKT and Raf-1, which are well-established clients of Hsp70 and are two important signaling proteins that control growth and anti-apoptosis of cancer cells (25). To test if Bim is involved in the control of these clients, we by guest on November 3, 2020 http://www.jbc.org/ Downloaded from investigated whether Bim silence in BV173 affects levels of AKT and Raf-1. As shown in Fig 3D, Bim silence caused a decreased level of AKT and Raf-1 by 38% and 35% respectively, and Hsp70 silence caused a decreased level of AKT and Raf-1 by 44% and 42% respectively. However, in cells with stable Hsp70 knockdown, Bim shRNA transient transfection had little effect on the levels of AKT and Raf-1 (P>0.05, Fig 3D). Therefore, Hsp70/Bim complex is critical for stabilization of AKT and Raf-1. Similar effects of Hsp70 and Bim knockdown were detected in KCL22 and the breast cancer cell line T47D, demonstrating that these effects are widespread in multiple cancer types (Supplementary Fig S12). In line with it, when the shRNA experiments were performed in MDA-MB-435 that endogenously express a low level of Hsp70 protein, Hsp70 knockdown has no effect on the levels of AKT and Raf-1, and Bim knockdown paralleled the effect of Hsp70 ( Fig 3E). In summary, the Hsp70/Bim complex plays a critical role to facilitate the chaperone function of Hsp70 toward client proteins, which hinted that the outcome of Hsp70/Bim complex is to promote the oncogenic chaperone activity of Hsp70.
Further, a dual role of Hsp70's anti-apoptotic function was illustrated in T47D and MDA-MB-435 cells following ABT-737 treatment that Hsp70 not only antagonizes Bim to protect cells from apoptosis, but also enhances the tumor chaperone's effect on stabilizing oncogenic clients with the help of the positive co-chaperone Bim.

Role of Hsp70/Bim complex in blocking cancer cells from apoptosis.
ABT-737 is one of the most established small-molecule Bcl-2 inhibitors that induce intrinsic apoptosis by releasing pro-apoptotic Bim from Bcl-2 complexes. Its analogue ABT-199 is the first FDA approved Bcl-2 inhibitor (26,27). Although a previous report has found that mitochondria from T47D and MDA-MB-435 were similarly primed with Bim, only MDA-MB-435 cells exhibit expected sensitive to ABT-737 (28). Herein, we detected a much higher expression level of Hsp70 in T47D than that in MDA-MB-435 cells ( Fig 4A). Then, we treated the two cell lines with a gradient concentration of ABT-737 and examined the composition of Bim complexes and Hsp70 complexes. Protein levels of AKT, Raf-1 and PARP cleavage (a hallmark of apoptosis) were detected in parallel. As shown in Fig  4B, for T47D, ABT-737 treatment induced a dose-dependent release of Bim from Bcl-2, which was not rebound to Bcl-xL and Mcl-1 (Supplementary Fig  S13), whereas Hsp70/Bim complex is increased, accompanied with increased AKT and Raf-1 binding with Hsp70 and upregulation of AKT and Raf-1 protein level. Even at 0.5 μM of ABT-737 that most of Bim were released from Bcl-2, not any PARP cleavage was detected. In addition, we assayed Hsp70 interactions with Bax and Bad, two other BH3-containinting proteins that could be released from Bcl-2/Bcl-xL by ABT-737. Hsp70 showed very few interactions with Bax and Bad with little change after ABT-737 treatment (Supplementary Fig  S13), confirming the cellular activity of Hsp70/Bim complex in blocking apoptosis. In contrast, in MDA-MB-435 cells, very little Hsp70 was found in Bim complexes and no increase of Hsp70/Bim complex was detected after ABT-737 treatment ( Fig 4C). Consistently, no increase of AKT and Raf-1 complex with Hsp70 and upregulation of AKT and Raf-1 level were detected. Meanwhile, PARP cleavage was detected as by guest on November 3, 2020 http://www.jbc.org/ Downloaded from soon as Bcl-2/Bim was disrupted following 0.1 μM ABT-737 exposure, and the effects were further augmented at 0.5 μM ABT-737 ( Fig 4C). As reported by Letai A and other groups, in ABT-737 resistant cells, activated AKT could induce Bax phosphorylation and switch Bax from promoting to inhibiting apoptosis (28), while activated ERK, a downstream effector of Raf-1, could promote Mcl-1 levels to confer ABT-737 resistance (29). Herein, we revealed that it is the high expression of Hsp70 that upregulates AKT and Raf-1 with the help of Bim to antagonize ABT-737.
Tait et al previously reported that minority MOMP and DNA damage induced by non-lethal ABT-737 promotes tumorigenesis (30). Herein, we revealed that Bim released by non-lethal ABT-737 inhibits MOMP by facilitating stabilization of AKT/Raf-1. Given on the opposite role of Bim as a canonical BH3 activator to trigger MOMP by directly activating Bax/Bak, it is reasonable to suspect that the dual function of Bim might contribute to a balanced activation of Bax/Bak which led to minority MOMP and DNA damage. Thus, further study will be directed to investigate how the paradoxical role of Bim promotes tumorigenesis.
To the best of our knowledge, our work is the first one to describe that Bim binds in a new allosteric site of Hsp70 through BH3 domain and then positively regulates its chaperone activity. Bim combination with J domain has an addictive effect, and client binding can synergize with Bim or J domain to activate ATPase of Hsp70 (a model of the proposed reaction cycle was shown in Fig  4D). It reminds us a growing emergence of new partners of BH3-only protein, which endows BH3-only proteins with new roles on directly regulating the partners' activity and biological process beyond their canonical pro-apoptotic function (12,13,31). However, the new finding on Bim is unexpected. As the most predominant BH3-only protein that antagonizes all the anti-apoptosis Bcl-2-like proteins and acts as an activator of multidomain pro-apoptotic members, Bim conceals such an opposite role to help Hsp70 to stabilize oncogenic clients. There are some reports showing that cancer cells often express a significantly higher level of Bim than normal cells, which exhibits a pro-survival function (32). However, the molecular mechanism remains a mystery and people could not speculate why cancer takes the risk of overexpressing Bim. Our findings provide an unprecedented mechanism that complexed with Hsp70 at least partly contributes to the Bim balancing between surviving and apoptosis.
Hsp70 protein was demonstrated to function as a BH3 receptor to capture Bcl-2 family members via the BH3 domain for the first time. Together with two previous reports of Hsp70/Bim and Hsp70/Bax complexes by co-IP (9,10), Hsp70 can be listed into non-canonical Bcl-2-like proteins. Since the binding site of Bim is highly conserved among family members, such as Hsc70 (HSPA8), Hsp70 (HSPA1A), and GRP75 (HSPA9), as shown by sequence alignment (Supplementary Fig S14), nearly all Hsp70 family members are likely to act as BH3 receptors.
Taken together, the two-faced role of Bim in cell fate regulation and dual anti-apoptosis function of Hsp70 were revealed in this study. Moreover, the solid physical complex of Hsp70/Bim as well as the biological relevance between them as evidenced by bioinformatics analysis highlights the further exploration of the consequent biological events of Hsp70/Bim dimer in cell death and proliferation.

Materials and methods Cell Lines
Cell lines KCL22, BV-173, T47D, MDA-MB-435, K562 and H23 were purchased from American Type Culture Collection and used within 6 months from resuscitation. Cells were cultured in RPMI 1640 media (Thermo Scientific HyClone, Beijing, China) supplemented with 10% fetal bovine serum (FBS; Gibco BRL, Grand Island, NY, USA) and 100 U/mL penicillin and streptomycin at 37 °C and 5% CO 2 .

Protein Expression and Purification
Constructs of hHsp70, hHsp70 NBD (1-383) and hHsc70 NBD (1-383) cDNA were subcloned into the pHis vector with a TEV protease cleavage site for simple removal of the tag following protein production. The plasmids for full-length Hsp70, Hsp70 NBD and Hsc70 NBD were prepared externally (Takara Bio Inc., Otsu, Japan). To construct Hsp70 mutant H227A, nucleotides corresponding to residue H227 were substituted to create an alanine residue (A) with a site-directed mutagenesis kit (Clontech, Beijing, China). These proteins with an N-terminal 6×His tag were produced in E. coli strain BL21 (DE3) containing the corresponding plasmid was grown in LB medium at 37 °C to an optical density of 0.8 at 600 nm, and then induced by 0.5 mM IPTG at 37 °C for 5 hr. For NMR studies, uniformly 2 H/ 15 N labelled protein samples were produced in D 2 O minimal M9 medium, where 15 NH 4 Cl were used as sole nitrogen sources. Cells were then lysed by sonication and the lysate was cleared by centrifugation at 9000 rpm for 30 min, and Hsp70 proteins were purified from the soluble fraction using Ni-NTA resin (Qiagen), following the manufacturer's instructions. The affinity tag was captured through reverse Ni 2+ chromatography. The flow-through contained Hsp70 protein was further purified by Sephadex G-75 size exclusion chromatography (GE Healthcare).
For J domain expression, competent E. coli strain BL21 (DE3) was transformed with a plasmid which expresses bovine auxilin residues 810-910 as part of a GST fusion protein. Transformed cells were grown in 1 L of LB media to an OD of 0.4-0.7 at 37 °C, and expression of the fusion protein was induced by addition of IPTG to 1 mM. Cells were grown for 4 hr at 25 °C, and harvested by centrifugation, and suspended in 40 mL of 50 mM Tris pH 8.0, 1 mM DTT, 1 mM EDTA, 1 mM PMSF, and 5% glycerol. Cells were lysed by sonication and the lysate was cleared by centrifugation at 9000 rpm for 30 min, and the supernatant was loaded onto 5 mL of glutathione resin (GE Healthcare). The resin was washed with 50 mM Tris, pH 8.0, 1 mM DTT, 0.5 M NaCl, 1 mM EDTA, 1 mM PMSF, 0.1% Tween-20, 10 mM GSH and 5% glycerol.

Isothermal titration calorimetry (ITC)
The Hsp70 protein interactions with BimBH3 peptide or BimLΔC27 were characterized using an isothermal titration microcalorimeter, ITC200 (GE Healthcare/MicroCal, South Burlington, VT, USA) at 25 °C. The cell was loaded with 30 μM full-length Hsp70 or Hsp70 NBD domain in buffer A (5 mM MgCl 2 , 25 mM KCl, 20 mM Tris-HCl, 3 mM ADP (pH 7.5)) and buffer B (5 mM MgCl 2 , 25 mM KCl, 20 mM Tris-HCl, 3 mM ATP (pH 7.5)) respectively, by which ADP-bound full-length Hsp70, ATP-bound full-length Hsp70, ADP-bound Hsp70 NBD, or ATP-bound Hsp70 NBD domain were generated, and the injection syringe was loaded with 300 μM BH3 peptides or BimLΔC27. To assay BimBH3 peptide or BimLΔC27 binding with SBD domain, the cell was loaded with 30 μM Hsp70 SBD domain in buffer C (5 mM MgCl 2 , 25 mM KCl, 20 mM Tris-HCl (pH 7.5)). Typical titrations consisted of 12 injections of 3 μL. An additional set of injections was run in a separate experiment with buffer instead of the protein solution as a control. Before data analysis, the control values were subtracted from the main experimental data. The data were analyzed to fit a one-site model in MicroCal software.

ATPase activity
ATPase assays were carried out using Kinase-Lumi TM Plus Luminescent Kinase Assay Kit (Beyotime, China). Briefly, A master mix of full-length Hsp70 (final concentration 1 μM) was prepared in assay buffer (100 mM Tris-HCl, 20 mM KCl, and 6 mM MgCl 2 , pH 7.4]). An aliquot (50 μL) of this mixture and BimBH3 peptide (0.1 μM), J domain (0.2 μM) and σ 32 (2 μM) alone or in combination in assay buffer were added to a 96-well plate and incubated for 30 min at room temperature with or without MKT-077 (50 μM). The reaction was started by adding ATP (final concentration 40 μM). After reaction at 25 ℃, 50 μL visualization reagents in the Luminescent Kinase Assay kit were added for 10 min. The enhanced chemiluminescence signal was determined with a SpectraMax M2e (Molecular Devices). To correct for non-enzymatic hydrolysis of ATP, the absorbance of a sample formed from an identically treated ATP buffer lacking the protein was subtracted.
To detect HSP70 ATPase in BV-173, Bim shRNA-transfected BV173, KCL22 and Bim shRNA-transfected KCL22, HSP70 was co-immunoprecipitated from the above cell lysates, and then subjected to ATPase assay using Kinase-Lumi TM Plus Luminescent Kinase Assay Kit.

Rhodanese aggregation assay
The aggregation of denatured rhodanese was measured based largely on previously published procedures (34) with some modifications. Bovine liver rhodanese (30 mM; Sigma) was denatured in 6 M guanidine-HCl-30 mM morpholinepropanesulfonic acid (MOPS)-KOH (pH 7.2), 2 mM DTT at room temperature for 30 min. The denatured rhodanese was diluted to a final concentration of 1.