Effective Targeting of the Survivin Dimerization Interface with Small-Molecule Inhibitors

Survivin is a member of the inhibitor of apoptosis (IAP) gene family containing a single baculovirus IAP repeat (BIR) domain, a zinc-finger fold, and an extended C-terminal helical coiled coil (1). It is a homodimer of a 16.5-kDa protein (2, 3). Ectopic survivin overexpression causes inhibition of cell death induced by intrinsic and extrinsic stimuli in cell lines (4–9) and in animals (10). Survivin is overexpressed essentially in all cancers, but not expressed in most adult normal tissues (1, 11). Survivin has also been shown to contribute to radiotherapy and chemotherapy resistance, and inhibition of survivin sensitizes cancer cells to these treatments (12–14). Treatments with molecular probes such as antisense oligonucleotide, ribozyme, siRNA, and dominant negative mutant all resulted in caspase-dependent cell death and increased apoptosis induced by radiation and anticancer drugs (6, 14–19). These findings clearly established survivin as an ideal target for discovery of anticancer therapeutics.

Unfortunately, survivin belongs to a group of proteins that are considered undruggable due to lack of enzymatic activities. Although small-molecule inhibitors have been identified that would interfere with the function of this type of proteins by blocking their interaction with other essential proteins, this approach has not led to drug candidates for clinical trials except in rare cases. In the case of survivin, targeting its expression has been attempted to avoid targeting survivin protein directly. For example, YM155, a small-molecule compound, has been identified to inhibit survivin expression by targeting its transcription (20) and later was shown to inhibit transcription factors such as SP1 (21). However, several phase II trials of YM155 showed only limited or modest at best efficacy on human cancers (21). An antisense oligonucleotide, LY2181308, that inhibits survivin expression has also been tested as a single agent in phase I trial for solid tumors (22) and in combination with docetaxel in phase II trial for castration-resistant prostate cancers (23). Unfortunately, neither of these two trials showed any benefit of using LY2181308. Thus, new inhibitors of survivin are clearly needed perhaps by developing a strategy to target directly at the survivin protein itself.

It is known that exposure of the hydrophobic interface of a dimeric protein often leads to conformational change (24, 25), which causes destabilization and degradation of the protein by proteasome or autophagy (26). Because survivin exists as a homo-dimer, we hypothesized that a small-molecule compound that inhibits survivin dimerization may promote survivin degradation via the proteasome and, thus, eliminates the protein and leads to spontaneous apoptosis. We recently developed a novel strategy to identify interfacial dimerization core units critical for homo-dimerization (27, 28). Using this strategy, we tested the above hypothesis by first identifying the dimerization core residues critical for survivin dimerization followed by in silico screening for inhibitors targeting the critical core residues as well as in vitro and cell-based assays. We identified a hit compound, LQZ-7, which dissociated dimeric survivin and induced proteasome-dependent survivin degradation. Further analysis helped identify several active analogues of the hit compound, which resulted in a potential lead compound (LQZ-7F) that inhibits survival of multiple cancer cell lines with low micromolar IC₅₀, suppresses xenograft tumor growth, and inhibits survivin in vivo.

Antibodies against FLAG tag (F3165), α-tubulin (T9026), and β-actin (A5316) were obtained from Sigma-Aldrich. The antibody against survivin (2808), the iScriptTM cDNA Synthesis Kit, the StepOnePlus™ Systems Kit, and the Metafectene Pro transfection reagent were purchased from Cell Signaling Technology, Bio-Rad Laboratories, Applied Biosystems, Biontex, respectively. The enhanced chemiluminescence reagents and CNBr-activated sepharose 4B were from GE Healthcare. Cell culture media and fetal bovine serum were from Media Tech or Applied Biosystems-Life Technologies. All other chemicals were purchased from Sigma or Fisher Scientific. The human cancer cell lines used in this study were all from ATCC and have been authenticated by short tandem repeat analysis on January 21, 2013.

Molecular dynamics (MD) simulations of survivin dimers and water trafficking were carried out using the AMBER9 package as previously described (27). Crystal structure of survivin dimer with PDB code 1F3H was acquired from RCSB protein databank (2). Zinc parameters were developed by Pang using the cationic dummy atom (cada) approach (29). Survivin dimer was solvated in a truncated octahedron box with edges no closer than 10 Å to any atom in the solute with appropriate number of counter ions added to neutralize each system. Particle Mesh Ewald was used to calculate the long-range electrostatic interactions and the nonbonded cutoff was set to 8.0 Å. Each system was equilibrated by a four-step protocol prior to the 20-ns production MD simulation.

Total buried dimeric interface areas and buried surface areas decomposed to each residue were calculated by areaimol of the CCP4 package. Dimerization core residues that have more than 75% solvent accessible area buried in the dimeric interface were selected. Survivin has one dimerization core unit consisting of four residues: Leu⁹⁸ and Phe¹⁰¹ from one chain and same residues from another chain. This dimerization core entity is further validated by its water-exchange rate determined via 20-ns water explicit MD simulation as previously described (27, 28). A sphere of 6 Å in radius was drawn from the center in the mass of the core. Water molecules that fell into the sphere during the simulation were monitored and their residue IDs were recorded by VMD program. Then, water molecules from every 20 frames (representing a 200 ps timespan) were pooled and compared with that from the previous 20 frames. Water molecules with residue IDs that existed in the current 20 frames but did not appear in the previous 20 frames were considered water molecules that moved in during the current 200 ps timespan. Water molecules with residue IDs found in the current 20 frames but did not show up again in the next 20 frames were considered water molecules that moved out during this 200 ps timespan.

Structure-based in silico screening was performed as previously described (30, 31). Briefly, the three-dimensional coordinates of survivin were acquired from PDB code 1F3H. Only one of the two chains was kept and the protein chain was prepared for docking. Molecular surface was calculated using Distributed Molecular Surface program. Partial charges and protons were added to the protein by UCSF Chimera Dock Prep module (32). In silico dock screening of 200,000 compounds from SPECS's library (www.specs.net) was performed using UCSF DOCK 6.0 program (33). The docking of each compound was first scored with the DOCK GRID scoring function (34). The top-scoring 1,000 compounds were analyzed again and re-scored using the AMBER scoring function of DOCK 6.0 package (35). These compounds were clustered using Molecular Operating Environment program and visually examined using the UCSF Chimera ViewDock function. Final 100 compounds were selected based on the combination of GRID and AMBER score, drug likeness (Lipinski's rule of five), and on consideration of maximizing compounds from different clusters.

One microgram of purified survivin was incubated with 50 μmol/L candidate compounds, DMSO vehicle control, or different concentrations of LQZ-7 before mixing with equal volume of 2 sample buffer (100 mmol/L Tris, pH 8.0, 20% glycerol, 0.005% bromophenol blue, 2% Triton X-100, and 100 mmol/L DTT) followed by incubation at room temperature for 30 minutes. After centrifugation at 11,000 × g for 10 minutes, the supernatants were separated by electrophoresis on 15% Tris/glycine polyacrylamide gel followed by transfer to PVDF membrane for Western blot analysis as previously described (36).

These assays were performed as previously described (37, 38). Briefly, MTT assay was performed by seeding 2,500 cells in 96-well plate and cultured 24 hours before addition of survivin inhibitors at different concentrations and continuous culture for 3 days. The cells were then subjected to MTT assay. For colony formation assay, 100 cells/well were seeded in 6-well plates and cultured for 24 hours before addition of survivin inhibitors or DMSO vehicle. The cells were continuously cultured in the presence of survivin inhibitors or DMSO for 10 to 14 days followed by staining with crystal violet and counting manually.

Fluorogenic assay was performed as previously described (39). Briefly, LQZ-7 was preincubated without or with 10 μg purified survivin for 30 minutes at 37°C followed by determination of fluorescence emission at 485 nm with excitation at 590 nm. A dose–response curve was fitted to an equation describing one-site binding model to determine the K_d of LQZ-7 binding to survivin using GraphPad Prism 4.0 software.

Annexin V apoptosis assay was performed using a kit as described by the manufacturer (Invitrogen). Briefly, treated cells were harvested after washing with PBS and then resuspended in annexin-binding buffer (10 mmol/L HEPES, pH 7.4, 120 mmol/L NaCl, 2.5 mmol/L CaCl₂) at the density of 1 × 10⁶ cells/mL. After addition of Alexa Fluor 488 annexin V and propidium iodide (PI), the cells were incubated for 15 minutes at room temperature followed by dilution with annexin-binding buffer and FACS analysis of fluorescence emission at 530 and 575 nm with 488 nm excitation.

The effect of survivin inhibitors on the half-life of survivin was determined as previously described (40). Briefly, PC3 or DU145 cells were pretreated with 2 μmol/L cycloheximide for 1 hour followed by incubation without or with 9 μmol/L LQZ-7 or 5 μmol/L LQZ-7F for different times. The cells were then harvested for Western blot analysis of survivin.

For the proteasome inhibitor rescue experiment, PC3 cells were seeded in 10-cm dishes at 8 × 10⁵ cells/dish and cultured for 48 hours followed by replacement with fresh media containing DMSO control, 7 μmol/L MG132, or 70 nmol/L bortezomib and incubation for 2 hours. LQZ-7 and LQZ-7F were then added to the culture to final concentrations of 10 and 5 μmol/L, respectively, and incubated for additional 24 hours. The cells were then harvested and washed with PBS and lysed for Western blot analysis of survivin.

For efficacy study, 3 × 10⁶ PC3 cells were injected subcutaneously in the flanks of 10 4- to 6-week-old male NSG (NOD/SCID/IL-2Rγ null) mice. When the tumor volume reached approximately 100 mm³, the mice were randomized into two different groups (5/group) with one group treated by vehicle control and the other by LQZ-7F at 25 mg/kg via i.p. injection once every 3 days for total of eight treatments. Tumor volume and body weight were measured every 2 days. On the 30th day after the initial treatment, mice were euthanized and the tumor tissues were harvested, weighed, and subjected to hematoxylin and eosin staining as well as Western blot and immunohistochemistry analysis of survivin.

To effectively target the dimerization domain of survivin, we first analyzed the dimerization interface to identify residues that are critical for survivin dimerization as targets for in silico screening. The interacting residues between the two identical subunits are comprised of residues 6 to 10, 93 to 99, and 101 to 102 (Leu⁶, Pro⁷, Pro⁸, Ala⁹, Trp¹⁰, Phe⁹³, Glu⁹⁴, Glu⁹⁵, Leu⁹⁶, Thr⁹⁷, Leu⁹⁸, Gly⁹⁹, Phe¹⁰¹, Leu¹⁰²). About 80% of these residues are hydrophobic, which is comparably high considering the average of nonpolar interaction is only approximately 50% in dimeric proteins. The dimeric survivin has a total calculated solvent accessible area of 18,039 Ų. The buried accessible area in the dimeric interface of a monomeric survivin is 550 Ų and it occupies only 6% of the total accessible area of a monomer (9,044 Ų), which is much less than the average value (∼20%) in dimeric or oligomeric proteins. The relatively small area and the high hydrophobicity of the interactive surface indicate that it may be a good target site for drug discovery.

Previously, we found that dimeric proteins may have dimerization core units that are sealed from water penetration and are critical for homodimerization (27). Further analysis of the dimeric interface using our previously described computational approach (27) shows that survivin has one dimerization core unit consisting of four residues, Leu⁹⁸ and Phe¹⁰¹ from one subunit and the same residues from another subunit (Fig. 1A). Water-exchange rate in this dimerization core unit was determined by performing 20-ns water explicit MD simulation and a computational method developed previously (27). As shown in Fig. 1B, few water molecules moved in or out of the dimerization core during the 20-ns simulation with an estimated water-exchange rate of 0.5 water/200 ps, which is comparable with that of the 14-3-3σ dimerization core (27). It is, however, significantly lower than that of the mutant 14-3-3σ molecule, which lost dimerization activity with a much higher water-exchange rate (>4 water/200 ps). These findings suggest that the dimerization core unit of survivin consisting of Leu⁹⁸ and Phe¹⁰¹ is tightly sealed and may be critical for the formation of stable survivin dimers. Disrupting this core formation may affect survivin dimerization. Indeed, mutation of Phe¹⁰¹ to Ala¹⁰¹ together with Leu¹⁰² to Ala¹⁰² mutation has been shown to disrupt survivin dimerization (41).

To identify small-molecule compounds that can potentially inhibit survivin dimerization, we performed in silico screening of approximately 200,000 compounds targeting the critical dimerization core residues Leu⁹⁸ and Phe¹⁰¹ in the dimeric interface using DOCK. Of the 100 top-scoring compounds of diversified structures, 49 chemical samples were commercially available and tested first for their cytotoxicity using two human cancer cell lines DU145 and PC3. As shown in Supplementary Fig. S1, only compounds 4, 7, 9, 12, 21, 36, and 42 at 20 μmol/L were able to inhibit ≥50% survival of both DU145 and PC3 cells. Consequently, these compounds were chosen for further investigation.

The selected compounds were tested for their ability to dissociate survivin dimers using purified survivin and nondenaturing PAGE analysis. As shown in Fig. 2A, compound #7 (named LQZ-7 with structure shown in Fig. 2B) was able to affect the mobility of purified survivin, presumably by dissociating the dimeric survivin into monomers. To confirm its activity and test its potential selectivity, we had LQZ-7 resynthesized and performed a dose–response analysis of its activity in dissociating dimeric survivin and an irrelevant control but similar dimeric protein, 14-3-3σ. Although 100 μmol/L LQZ-7 appears to be required to completely dissociate dimeric survivin (Fig. 2C), LQZ-7 at 100 μmol/L had no effect on 14-3-3σ dimerization (Fig. 2D). We also performed nondenaturing PAGE analysis of nascent proteins synthesized in a cell-free system in the presence of LQZ-7. As shown in Fig. 2E, 20 μmol/L LQZ-7 achieved a complete inhibition of survivin dimerization. Together, these findings suggest that LQZ-7 can inhibit dimerization of nascent survivin and dissociate existing dimeric survivin with selectivity over other homodimeric proteins such as 14-3-3σ. The finding that less LQZ-7 is required to completely inhibit dimerization of nascent survivin than to completely dissociate existing dimeric survivin suggests that inhibiting dimerization of newly synthesized proteins may be dynamically favorable than dissociating the existing dimeric proteins.

To verify that LQZ-7 indeed binds to survivin and inhibits survivin dimerization, we took advantage of the intrinsic fluorescent property of LQZ-7 and performed a fluorogenic titration assay in the presence or absence of survivin as previously described (39). Figure 2F shows that the intrinsic fluorescence of LQZ-7 dramatically increases in the presence of recombinant survivin, indicating that LQZ-7 likely interacts with survivin. The K_d of LQZ-7 binding to survivin is estimated to be approximately 0.24 ± 0.11 μmol/L.

Exposure of hydrophobic interface of a dimeric protein often leads to conformational change (24, 25), which causes destabilization and degradation of the protein by proteasome or autophagy (26). We, thus, hypothesized that LQZ-7 may cause proteasome-dependent degradation of survivin by dissociating survivin dimers and exposing the hydrophobic dimeric interface. To test this hypothesis, we first treated DU145 cells with LQZ-7 for different times followed by determination of endogenous survivin using Western blot analysis. Figure 3A shows that LQZ-7 treatment indeed reduces survivin protein level compared with vehicle control treatment. Similar results were also observed with PC3 cells (data not shown). However, LQZ-7 treatment had no effect on the level of survivin mRNA as determined using real time RT-PCR (Fig. 3B). LQZ-7 treatment also effectively reduced the level of ectopic Flag-tagged survivin in HEK293 cells under the control of an exogenous CMV promoter (Fig. 3C). These observations together suggest that survivin loss induced by LQZ-7 is likely at the protein but not mRNA level.

We next determined if LQZ-7 causes survivin degradation by examining its effect on survivin half-life. For this purpose, DU145 and PC3 cells were pretreated with cycloheximide to inhibit synthesis of new proteins followed by treatment with LQZ-7 for different times. Figure 3D and E show that the half-life of survivin is 1.5 to 2.5 hours in control-treated DU145 and PC3 cells, consistent with previously reported survivin half-life (42, 43). However, following LQZ-7 treatment the half-life of survivin was reduced to approximately 30 minutes. Co-treatment with a proteasome inhibitor, MG132 or bortezomib, reversed LQZ-7–induced survivin loss (Fig. 3F), consistent with a previous report that survivin degradation is mediated by the proteasome (44). We also found that LQZ-7 does not inhibit survivin protein synthesis as determined using [³⁵S]methionine pulse-labeling in combination with immunoprecipitation of survivin following LQZ-7 treatment (data not shown). Together, the above findings suggest that LQZ-7 may cause proteasome-dependent survivin degradation possibly by inhibiting dimerization and exposing the dimerization core residues of survivin.

Although LQZ-7 dissociates survivin dimer in vitro and causes survivin degradation in cells, cell-based cytotoxicity assays showed that the IC₅₀ of LQZ-7 in human PC3 and DU145 cells was approximately 25 μmol/L (Fig. 4B). The modest IC₅₀ may be due to the possibility that the carboxyl group in LQZ-7 (Fig. 4A) impedes its cellular permeability. To improve cellular effect of LQZ-7, we searched the SPECS database and identified six commercially available analogues (LQZ-7A, B, C, D, E, and F; Fig. 4A). The chemical samples of these analogues were obtained and tested first for their cytotoxicity to DU145 and PC3 cells compared with the initial hit LQZ-7. Figure 4B shows that five of the six analogues have much lower IC₅₀ than LQZ-7 for both cells. These analogues together with LQZ-7 were then used to test their effect on the expression of ectopic Flag-tagged survivin in HEK293 cells. As shown in Fig. 4C, although LQZ-7A, D, and E had no effect on the level of Flag-tagged survivin, LQZ-7B, C, and F all reduced survivin protein level. It appears that LQZ-7C and F completely eliminated survivin whereas the parent compound LQZ-7 did not, consistent with their lower IC₅₀ than the parent compound. The fact that LQZ-7E has high IC₅₀ and does not reduce survivin protein suggests that LQZ-7E may not bind to and inhibit survivin. It remains unknown why LQZ-7A and D have no effect on survivin level while maintaining low IC₅₀ (see Discussion).

Of the two analogous (LQZ-7C and F) that have the best IC₅₀ and ability to eliminate survivin, LQZ-7F is unique, smaller, and simpler in structure with a primary amine group as an advantage for further study (see below). Thus, we elected to pursue LQZ-7F further as a potential lead and tested its activity in suppressing the expression and inducing degradation of endogenous survivin. Similar as LQZ-7, LQZ-7F effectively suppressed endogenous survivin expression in both DU145 and PC3 cells (Fig. 4D) and increased degradation of endogenous survivin (Fig. 4E). Furthermore, proteasome inhibitors MG132 and bortezomib both were able to rescue LQZ-7F–induced survivin degradation (Fig. 4F). Thus, LQZ-7F, similar as its parent compound LQZ-7, also induces survivin degradation in a proteasome-dependent manner.

To investigate if LQZ-7F indeed binds to survivin, we took advantage of the primary amine group and immobilized LQZ-7F onto CNBr-activated sepharose for a pull-down assay using purified survivin. As shown in Fig. 4G, survivin was successfully pulled down by sepharose-immobilized LQZ-7F, but not by the control beads without LQZ-7F. Thus, LQZ-7F, similar to LQZ-7, may bind directly to survivin. To understand how LQZ-7F interacts with survivin, we performed docking analysis of LQZ-7F in the dimerization interface of survivin, which revealed two key interactions between LQZ-7F and survivin: (a) H-bond between the primary amine group of LQZ-7F and Glu⁹⁴ of survivin; (b) π–π stacking and hydrophobic interaction between the tetracyclic furazanopyrazine ring of LQZ-7F and the hydrophobic residues Trp¹⁰ and Phe⁹³ in addition to the dimerization core residues Leu⁹⁸ and Phe¹⁰¹ (Fig. 4H).

Survivin is ubiquitously up-regulated in human cancers and, thus, inhibiting survivin may be a general approach to help eradicate many cancers. Toward this goal, we tested the effect of LQZ-7F on survival of multiple human cancer cell lines representing acute myeloid leukemia and cancers of breast, colon, lung, pancreas, prostate, and ovary using MTT assay. As shown in Fig. 5A, LQZ-7F effectively inhibited survival of all cancer cell lines with IC₅₀ ranging between 0.4 and 4.4 μmol/L.

The activity of LQZ-7F in suppressing cancer cell survival was further evaluated using colony formation assay for PC3 and A549 cells. As shown in Fig. 5B, the colony formation efficiency of PC3 cells was reduced from approximately 62% of the control to approximately 52, 12, and 8% following treatments with LQZ-7F at 0.2, 0.5, and 1 μmol/L, respectively. A549 cells were more sensitive to LQZ-7F with lower IC₅₀ than PC3 cells and the colony formation efficiency were consistently reduced more from 45% of the control treatment to approximately 2.5% and 0% treated by 0.2 and 0.5 μmol/L of LQZ-7F, respectively.

Previously, it has been shown that dominant negative survivin causes spontaneous apoptosis of PC3, DU145, and LNCaP cells (12). We next tested if LQZ-7F also causes spontaneous apoptosis by inhibiting survivin as determined by Annexin V staining. For this experiment, we tested PC3 and HL-60 as representative cells because PC3 has high whereas HL-60 has intermediate IC₅₀ against LQZ-7F (Fig. 5A). As shown in Fig. 5C, 54% to 69% apoptosis for PC3 and 66% to 98% apoptosis for HL-60 cells were generated following treatments with 5 to 10 μmol/L LQZ-7F. These findings were further validated by determining the cleavage of PARP-1, a substrate of caspases during execution of apoptosis, in both PC3 and HL-60 cells following LQZ-7F treatments (Fig. 5D).

Finally, we tested survivin expression in all 13 cell lines (Fig. 5E) and performed a correlation analysis between survivin level and IC₅₀. As shown in Fig. 5F, the IC₅₀ values strongly associate with survivin protein level in these cells with a Pearson correlation coefficient of 0.52, indicating that LQZ-7F may suppress the survival of these cancer cells by acting on survivin.

Survivin has been shown to have dual functions in inhibiting apoptosis and in promoting cell-cycle progression (1). The later was thought to derive from survivin action in stablizing microtubules both in vitro and in vivo (45, 46). Thus, to further determine LQZ-7F effect on survivin, we analyzed microtubule structure following LQZ-7F treatment in PC3 cells. As shown in Fig. 6, the control DMSO-treated cells have orderly microtubule fibers in all cells. However, the microtubule structure was severely disrupted following LQZ-7F treatments. It also appears that many cells are arrested in mitotic phase with aberrant spindles following LQZ-7F treatments. These observations are consistent with the role of survivin in microtubule dynamics (47, 48) and in proper chromosome segregation and cytokinesis (49, 50).

We next determined if LQZ-7F is active in suppressing tumor growth in vivo using a xenograft animal model. For this purpose, NSG mice were first implanted subcutaneously with PC3 cells to establish xenograft tumors. PC3 was chosen because it has the highest IC₅₀ (Fig. 5). When the xenograft tumor reached the size of approximately 100 mm³, the mice were randomized into two groups and treated with 25 mg/kg LQZ-7F or vehicle control via i.p. injection once every 3 days for a total of eight treatments. As shown in Fig. 7A, the growth of xenograft tumors was significantly inhibited in the LQZ-7F–treated group compared with the tumors in vehicle control-treated group. However, the body weight of mice in the treatment group remained constant after multiple dosing (Fig. 7B), indicating that LQZ-7F may not cause major toxicity after multiple dosing. In fact, the body weight of the control group dropped slightly possibly due to disease burden of xenograft tumors. This symptom appears to be alleviated by LQZ-7F treatment, consistent with the smaller tumor size in the treatment group.

The final dissected tumors in the treatment group appear to be small and round with smooth surface whereas the tumors of the control group appear to be irregular and bigger (Fig. 7C), suggesting that the tumors in the treatment group are confined whereas the tumors in the control group are aggressive. The average weight of the tumors in the treatment group is significantly less than that in the control group (Fig. 7D). The LQZ-7F–treated tumors also appear to have more apoptotic cells as indicated by condensed chromatin than the control-treated tumors (Fig. 7E). Western blot and IHC staining analyses showed that survivin in xenograft tumors of the LQZ-7F–treated group was dramatically reduced compared with the tumors of the control-treated group (Fig. 7E and F), suggesting that the effect of LQZ-7F on xenograft tumor growth may be due to its binding to survivin and induction of survivin degradation in vivo.

In this study, using in silico screening targeting the critical dimerization core residues following detailed analysis of the buried surface area, we successfully identified a hit compound (LQZ-7) and a potential lead inhibitor (LQZ-7F) that can bind directly to survivin and cause proteasome-dependent survivin degradation. LQZ-7F has an IC₅₀ of 0.4 to 4.4 μmol/L against multiple cell lines of different human cancers and induces spontaneous apoptosis. It is also effective in suppressing xenograft tumor growth and reduces survivin level in xenograft tumors.

This study provides a proof of concept that the dimerization core units in the dimeric interface of undruggable homodimeric proteins can be targeted for drug discovery. Combining computational analysis of the dimeric interface to first identify these core units with in silico screening, as demonstrated here, is likely a viable approach that will help succeed in identifying small-molecule compounds inhibiting dimerization of the target protein.

The finding that LQZ-7 and its analogues induce survivin degradation is consistent with the concept that exposure of the hydrophobic interface of a dimeric protein often lead to conformational change (24, 25), which causes destabilization and degradation of the protein by the proteasome or autophagy (26). Interestingly, not all LQZ-7 analogues induced survivin degradation. LQZ-7E had no effect on survivin expression, consistent with its lack of inhibitory effect on cell survival, suggesting that LQZ-7E may not bind to and induce survivin degradation. However, compounds LQZ-7A and 7D are effective in suppressing cancer cell growth without inducing survivin degradation. This finding is intriguing and suggests that these two compounds may bind to survivin and inhibit its function but do not trigger proteasome-dependent survivin degradation. Alternatively, these compounds may have off-target effects that inhibit cell survival but lost their effect on survivin protein. Clearly, future studies are needed to test these possibilities.

It is noteworthy that mutation of the dimerization core residue Phe¹⁰¹ to Ala¹⁰¹ together with mutation of Leu¹⁰² to Ala¹⁰² disrupted survivin dimerization, which did not appear to result in survivin degradation (41). This observation apparently is different from our finding using small-molecule inhibitors to disrupt survivin dimerization. Although the cause for this difference is currently unknown, it is possible that the mutant proteins do not exist as true monomers in mammalian cells because they can form heterodimers by binding to other proteins such as CRM1 (41). On the other hand, the monomers induced by small-molecule inhibitor such as LQZ-7F may not form heterodimer with other proteins due to existence of the compound in the interface. It is also possible that mutation of the dimerization core residues reduced the hydrophobicity of the interface and, thus, the protein could escape from the cell quality control system while binding of the small-molecule compound to the interface may increase the hydrophobicity and attract the quality control system. Future studies are warranted to test these possibilities.

It is also noteworthy that a small-molecule compound, LLP3, was synthesized previously based on the Abbot8, a survivin inhibitor obtained via a NMR-based screening. LLP3 was thought to bind to the dimerization interface of survivin (39). However, LLP3 had no effect on survivin dimerization or its expression. Instead, it inhibited interaction between survivin and its partner Ran protein. Thus, it is not clear if LLP3 truly binds to the dimerization interface of survivin. Nevertheless, LLP3 inhibited proliferation of cancer cells but with a much higher IC₅₀ of 14 to 38 μmol/L.

No potential conflicts of interest were disclosed.

Conception and design: J. Qi, J.-Y. Liu, J.-T. Zhang

Development of methodology: J. Qi, Z. Dong, J.-Y. Liu, J.-T. Zhang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Z. Dong, J. Liu, J.-Y. Liu, R.C. Peery

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Qi, R.C. Peery, J.-Y. Liu, J.-T. Zhang

Writing, review, and/or revision of the manuscript: J. Qi, R.C. Peery, J.-Y. Liu, J.-T. Zhang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Liu, J.-T. Zhang

Study supervision: J.-Y. Liu, J.-T. Zhang

Other (pathology slides reviewing): S. Zhang

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.