The dual mode of action of bistramide A entails severing of filamentous actin and covalent protein modification

Abstract

This study provides comprehensive characterization of the mode of action of bistramide A and identifies structural requirements of bistramide-based compounds that are responsible for severing actin filaments and inhibiting growth of cancer cells in vitro and in vivo. We rationally designed and assembled a series of structural analogs of the natural product, including a fluorescently labeled conjugate. We used TIRF microscopy to directly observe actin filament severing by this series of small molecules, which established that the combination of the spiroketal and the amide subunits was sufficient to enable rapid actin filament disassembly in vitro. In addition, we demonstrated that the enone subunit of bistramide A is responsible for covalent modification of the protein in vitro and in A549 cells, resulting in further increase in the cytotoxicity of the natural product. Our results demonstrate that bistramide A elicits its potent antiproliferative activity by a dual mechanism of action, which entails both severing of actin filaments and covalent sequestration of monomeric actin in the cell.

Small molecules that bind to monomeric or filamentous actin elicit their antiproliferative effects by impairing the ability of cells to progress through the cell cycle and undergo cytokinesis due to the defective actin cytoskeleton (1–5). Understanding the mode of action of such compounds expands our knowledge of actin biochemistry and provides opportunities for the development of new therapeutic agents (1, 2, 5). Traditionally, studies of the mechanism of action of small-molecule modulators of actin polymerization entailed the investigations of changes in fluorescence intensity during polymerization or depolymerization of pyrene- or prodan-labeled actin in solution (6–9). However, the main limitation of this method arises from the fact that the changes in the fluorescence intensity may reflect the binding of a small molecule to the filaments, especially if the binding event occurs in the close proximity to the dye, resulting in quenching of fluorescence, which could be independent of the rates of actin depolymerization (10).

Total internal reflection fluorescence (TIRF) microscopy has recently emerged as a powerful tool to investigate actin filament dynamics and its regulation by several actin-binding proteins (11–16), as well as latrunculin A (14), which is incapable of depolymerizing filamentous actin in vitro. We used TIRF microscopy to directly observe for the first time actin filament severing by a series of small molecules, which are derived from bistramide A (1)—a marine natural product that specifically and potently targets actin in the cell (17). In addition, we demonstrated that the C(1)–C(4) enone-containing subunit of this natural product plays a pivotal role in covalent modification of cellular actin, which further enhances the cytotoxicity of the corresponding bistramide-based compounds. Our study provides comprehensive characterization of the unique mode of action of bistramide A and identifies structural requirements of bistramides that are responsible for severing actin filaments and inhibiting growth of cancer cells in vitro and in vivo (18–20).

Results and Discussion

Binding to Monomeric Actin in Vitro and Cytotoxicity.

The x-ray structure of bistramide A-actin complex (23) suggested that the C(13)–C(18)-amide and the C(19)–C(38)-spiroketal subunits of the natural product played a dominant role in stabilizing the protein-small molecule interaction via a network of hydrogen-bonding contacts, as well as a series of hydrophobic interactions. To test this prediction, we assembled a series of rationally designed analogs of bistramide A using the same synthetic strategy, which was used for our synthesis of the parent natural product (24). Each compound was evaluated for its ability to bind monomeric actin by isothermal titration calorimetry (ITC) and to inhibit proliferation of A549 cells (Fig. 1A). We found that neither the spiroketal (analog 7) nor the central amide subunit (analog 8) alone displayed any observable activity. However, when both fragments were linked (analog 5), the in vitro actin binding efficiency was restored to a significant extent (K_d = 43 nM), indicating that the combination of the spiroketal and the amide subunits represented the minimal structural requirement for G-actin-binding. These observations were fully consistent with conclusions made on the basis of the x-ray crystallographic data. Furthermore, elimination of the two methyl groups at the C(23) and C(34), which participated in hydrophobic interactions with Thr-351, Phe-352, Met-355, Ile-345, Tyr-143, Leu-349 of actin, resulted in substantial loss of actin-binding efficiency (analog 6, K_d = 680 nM).

Fig. 1.

Investigation of the mechanism of action of bistramides. (A) Structures, actin-binding affinities and A549 cell growth-inhibitory constants of bistramide A (1) and analogs 2–8. (B) Visualization of severing of actin filaments by time-lapse TIRF microscopy. (C) Depolymerization of actin filaments in A549 cells. (D) Observation of covalent modification of monomeric actin by bistramide A (1) and analogs 2 and 3 by MALDI-TOF.

However, despite the potent actin-binding affinities of analogs 5 and 6, the two compounds were significantly less potent in inhibiting cell proliferation compared with the parent natural product 1. Introduction of the C(4)–C(13)-pyran subunit (analog 4) did not have significant effects on actin-binding and cell-based activity. However, incorporation of the C(1)–C(4) enone moiety in analogs 2 and 3 restored potent cell-based activity without any significant increase in G-actin binding. Interestingly, analog 2 inhibited cell growth more effectively compared with analog 5, whereas the G actin-binding affinities of the two compounds were reversed. This observation suggested for the first time that the C(1)–C(4)-enone subunit played a critical role in increasing cytotoxicity of the bistramide-based compounds—a result, which could not be rationalized using x-ray crystallographic characterization of the interaction of bistramide A with monomeric actin because three carbons and the oxygen of C(1)–C(4)-enone were disordered in the structure (23).

Severing of Filamentous Actin by Bistramide-Based Compounds in Vitro.

To gain further insight into the mode of action of bistramide A, we next investigated the interaction of the natural product and its analogs with filamentous actin. We have previously demonstrated that bistramide A rapidly decreased the fluorescence of pyrene-labeled F-actin (17). Similar observations have been reported for several other actin-binding natural products (6–9). However, this effect could provide only limited mechanistic information and may primarily reflect the binding of a small molecule to the filaments in the close proximity to the dye and quenching of pyrene fluorescence, which could be independent of the rates of actin depolymerization (10). Thus, we decided to employ TIRF microscopy to observe in real time the effects of bistramide-based compounds on actin polymerization in vitro. We polymerized G-actin containing 25% of TMR-actin and 10% of biotin-functionalized actin. The resulting actin filaments were immobilized onto the streptavidin-coated glass slides. The slides were incubated with compounds 1–8 and imaged by time-lapse TIRF microscopy. As expected, analogs 7 and 8 had no effect on actin depolymerization (Fig. 1B). In the presence of these compounds, the decrease in the number and the length of actin filaments was fully consistent with the kinetics of slow dilution-induced actin disassembly (Table 1, entries 1–3). In contrast, incubation of filamentous actin with either bistramide A (1) or simplified analogs 2–6 resulted in rapid filament depolymerization (Fig. 1B and Table 1). Analysis of the dynamics of this process revealed that each of the compounds induced multiple breaks in actin filaments and did not affect the rates of actin depolymerization at either the barbed or the pointed end (Table 1, entries 4–9).

Table 1.

Real-time observation of actin depolymerization by bistramide-based compounds using TIRF microscopy

Entry	Compound (concentration, μM)	Number of breaks* (per 100 μm of actin filaments)	Rate of actin depolimerization at barbed end, subunits/s†	Rate of actin depolimerization at pointed end, subunits/s‡
1	None	None	2.07	0.37
2	8 (200)	None	2.15	0.21
3	7 (200)	0.2	1.87	0.32
4	6 (200)	3.4	2.29	0.41
5	3 (200)	13.3	1.84	0.57
6	5 (15)	13.0	1.93	0.55
7	4 (15)	16.5	1.92	0.51
8	2 (15)	41.8	1.90	0.48
9	1 (0.75)	40.4	1.88	0.49

*Analysis of a total filament length of 300–400 μm for each compound.

^†Standard deviations are 0.4–0.6 subunits/s.

^‡Standard deviations are 0.1–0.3 subunits/s.

This study demonstrated for the first time that bistramide A was able to rapidly sever actin filaments. Furthermore, analysis of the time-lapse TIRF microscopy images enabled estimation of the relative efficiency of actin severing by bistramide-based analogs, which can be measured by calculating the number of breaks that are observed per 100 μm of actin filaments. This analysis revealed that the combination of C(19)–C(38) spiroketal subunit and the C(13)–C(18) amide segment was sufficient to enable dissolution of actin filaments (Table 1, entries 4 and 6). In addition, incorporation of the C(1)–C(4) enone appeared to increase actin-severing ability of analogs 2 and 3 (Table 1, entries 8 and 5) compared with 4, 5 and 6 (Table 1, entries 7, 6, and 4). To further examine the effect of an enone moiety on the efficiency of severing actin filaments in vitro, we measured the rates of this process over a range of concentrations of analogs 2 and 4 (1–15 μM). This study unambiguously confirmed that analog 2, which contained the enone moiety, was at least three times more potent at severing filamentous actin (Fig. 2).

Fig. 2.

Effect of analogs 2 and 4 on the rates of severing of actin filaments. (A) Actin filaments were immobilized onto glass slides and treated with different concentrations of analog 2 to induce the severing process. (B) Similar experiment was performed at variable concentration of analog 4. The rate of severing was determined using TIRF microscopy by counting the number of severing events within a 100-μm filament length per minute, which occurred before F-actin on the slide was depleted by ≈50%.

Actin Depolymerization in A549 Cells.

We also examined the ability of bistramide analogs to depolymerize actin in A549 cells. Each of the compounds was incubated with A549 cells for 2 h at a concentration that exceeded the GI₅₀ value by a factor of 5. Filamentous actin was visualized by fluorescence microscopy following Alexa Fluor 488 phalloidin staining. Bistramide A (1), as well as analogs 2 and 3 containing the C(1)–C(4)-enone, were found to be significantly more potent at inducing actin depolymerization in A549 cells compared with analogs 4, 5 and 6 (Fig. 1C). This observation was fully consistent with the enhanced cell-based activity of the compounds containing the C(1)–C(4)-enone moiety.

Covalent Modification of Actin by Enone-Containing Bistramides.

We examined the possibility that the observed increase in cytotoxicity and actin-severing ability of compounds 1, 2, and 3 could be due to the covalent modification of the protein target. We used MALDI-TOF to monitor the interaction of bistramide-based compounds with actin in vitro and found that bistramide A (1) and enone-containing analogs 2 and 3 covalently modified G-actin, resulting in the formation of higher molecular weight peaks (Fig. 1D). In contrast, no cross-linking was observed in the case of analogs 4, 5, and 6, which did not contain the enone moiety.

Due to the precedented high reactivity of Cys-374 (21, 22) we believe that the covalent interaction of bistramide A with actin is a result of the Michael addition of the thiol of Cys-374 to the C(1)–C(4)-enone portion of the natural product. This modification could not be observed by x-ray crystallography because the enone portion of bistramide, as well as the three C-terminal amino acids of actin, including Cys-374, were disordered in the crystal (23).

Having observed covalent modification of the purified actin by bistramide A and enone-containing analogs 2 and 3, we constructed BODIPY-conjugated bistramide A (9, Fig. 3) and tested whether the covalent modification of the protein could be detected in A549 cells. Initially, we treated purified actin with conjugate 9 and established cross-linking of the protein by this chemical probe using MALDI-TOF (data not shown) and fluorescence imaging of the labeled protein on polyacrylamide gel (Fig. 3, lane 1). Next, we incubated A549 cells with conjugate 9 (5 μM) for 3 h and observed efficient cross-linking of cellular actin, which was detected by fluorescence imaging of the cell lysate on polyacrylamide gel (Fig. 3, lane 2) and appeared to be identical to the purified protein (Fig. 3, lane 3). Importantly, no other covalently modified proteins were detected. A combination of MALDI-TOF and fluorescent imaging experiments using probe 9 unambiguously demonstrated that the C(1)–C(4)-enone subunit of bistramides is responsible for efficient cross-linking of actin both in vitro and in live cells.

Fig. 3.

Covalent modification of actin by BODIPY-bistramide A conjugate (9). The polyacrylamide gel was imaged at 695 nm using the excitation wavelength of 635 nm. Compound 9 was incubated with purified actin in vitro for 1 h. The resulting solution was subjected to polyacrylamide gel electrophoresis (lane 1). Compound 9 was incubated with A549 cells for 3 h. The cells were lysed and subjected to polyacrylamide gel electrophoresis (lane 2). To demonstrate identical mobility of protein bands in lanes 1 and 2, small fractions of each of the two solutions were mixed and subjected to polyacrylamide gel electrophoresis (lane 3).

Mechanism of Action of Bistramide A.

Our results demonstrate that the two main factors, which provide major contribution into the highly potent cytotoxic activity of bistramide A, are severing of filamentous actin and sequestration of monomeric actin. We believe that covalent G-actin sequestration, which is expected to further promote F-actin depolymerization, plays an important role in increasing cytotoxicity of bistramide A and the C(1)–C(4)-enone containing analogs 2 and 3. Indeed, analog 3 is significantly more cytotoxic and depolymerizes actin in cells much more effectively than analog 4, whereas the in vitro filament-severing abilities and actin-binding affinities of the two compounds are opposite. This result could only be explained by covalent sequestration of monomeric actin by enone-containing analog 3, which would lead to (i) depletion of the pool of monometic actin in the cell and subsequent actin depolymerization; and (ii) more efficient delivery of analog 3 into the cell due to the irreversible shift in chemical equilibrium after formation of the covalent actin–small molecule complex. In the absence of C(1)–C(4)-enone, analogs 4 and 5 are expected to be much less effective at sequestering monomeric actin due to reversible G-actin binding and competition with a range of other actin-binding proteins in the cell. Although additional studies of antitumor effects of bistramides are required, our results could help explaining the reported differences in the abilities of bistramides A, D, and K to inhibit tumor growth in vivo (20). The high toxicity of bistramide A in vivo could be explained by facile delivery of this compound into vital organs due to the covalent interaction with its protein target, which is abundant in healthy tissues. The reported in vivo efficacy of bistramides D and K, which are both devoid of the enone moiety, could be explained by the retained ability of these compounds to sever filamentous actin and to inhibit proliferation of rapidly dividing tumor cells.

Conclusion.

We established that bistramide A elicits potent antiproliferative activity by a unique mechanism of action, which entails both severing of actin filaments and covalent sequestration of monomeric actin. This finding was enabled by structure-based design, synthesis and detailed biochemical evaluation of a series of rationally designed synthetic analogs of the natural product. Our studies provide an explanation of the observed differences in antiproliferative activities of bistramides in vitro and in vivo and provide a foundation for a continued investigation of pharmacological properties of this class of marine natural products.

Materials and Methods

Reagents.

Actin was purified from rabbit muscle acetone powder (Pel–Freez Biologicals). TMR-actin (21) and biotin-actin (22) were prepared by reported methods. All other compounds were synthesized as described in supporting information (SI) Appendix. Alexa Fluor 488 Phalloidin was purchased from Invitrogen. Protease inhibitor mixture was purchased from Sigma.

Cell Viability Assays.

All assays were performed using at least three replicate wells for each concentration tested. Original 10–25 mM DMSO stock solution of bistramide A or its analogs was diluted to 100 μM for analogs 4–8, 10 μM for analog 3, 5 μM for analog 2 and 1 μM for bistramide A (1) with F-12K cell culture media. Two-fold serial dilutions were performed and used for cell-based assays. A549 cells were grown in 96-well white plates at the density of 1,000 cells per well in 100 μl F-12K cell culture media. Cells were allowed to attach and grow for 24 h and then treated with 30 μl of the drug solution and incubated further for 48 h. After incubation, cell viability was determined using luminescence-based commercial kit (CellTiter-Glo, Promega) and luminescence measured using a Wallac Victor 3 plate reader. To measure cell growth inhibition, viability assays were performed twice: at time point T₀ and after 48 h incubation (T₄₈). Change in the number of viable cells was measured, and GI₅₀ was calculated from sigmoidal plots.

Isothermal Titration Caloriemetry.

A solution of rabbit skeletal muscle actin (5–10 μM) was titrated with bistramide A or its analog (50–100 μM) at 25°C in buffer containing 2 mM Tris·HCl, pH 8.0, 0.2 mM CaCl₂, 0.01% NaN₃, 0.2 mM ATP, 0.2 mM 2-mercaptoethanol. Dissociation constants were calculated from the binding curve using Origin analytical software (OriginLab) with the binding model involving a single set of identical sites.

Fluorescent Visualization of F-Actin in Cells.

A549 cells grown on coverslips were incubated with analogs 4–8 (100 μM) analog 3 (2 μM) or analog 2 (500 nM) or bistramide A (150 nM) in F-12K cell culture medium for 2 h, fixed with 3% formaldehyde in PBS at 25°C for 5 min and permeabilized with 0.1% Triton X-100 at 25°C for 10 min. After washing with PBS, coverslips were stained with Alexa Fluor 488 Phalloidin and visualized by fluorescence microscopy.

Total Internal Reflection Fluorescence Microscopy.

Actin filament severing was studied using a custom-built total internal reflection fluorescence microscope. Images were collected with a 100×, 1.45 NA objective (Olympus) and an EMCCD camera (iXon, Andor Technologies). Glass flow cell (6 μl volume) was washed with 0.25 mg/ml neutravidin in F-buffer (50 mM KCl, 1 mM MgCl²⁺, 1 mM EGTA, 10 mM imidiazole, 2 mM ATP, 0.2 mM 2-mercaptoethanol, 2 min) and blocked with 1 mg/ml BSA in F-buffer (2 min). Ten microliters of 200 nM Mg²⁺-F-actin (10% biotinylated, 25% labeled with maleimide-6-tetramethylrhodamine) in TIRF-buffer (50 mM KCl, 1 mM MgCl²⁺, 1 mM EGTA, 10 mM imidiazole, 2 mM ATP, 4.5 mg/ml glucose, 0.5% 2-mercaptoethanol, 4.3 mg/ml glucose oxidase, 0.7 mg/ml catalase) was added to the flowcell and incubated for 2 min. The chamber was rinsed with TIRF-buffer and imaged. Compounds were freshly diluted from their DMSO stock solutions in TIRF-buffer and were added to the chamber immediately after the start of each experiment. Images were collected at 0.2-s exposures every 3 s for 200 frames. For kinetics studies, movies of suitable length were obtained for each of the analog concentrations (see SI Movies 1–8). Analysis was performed using ImageJ.

Interaction of Bistramide-Based Compounds with Actin by MALDI-TOF.

A 1:1 actin–drug complex was incubated for 30 min in 2 mM Tris·HCl, pH 8.0, 0.2 mM CaCl₂, 0.01% NaN₃, 0.2 mM ATP, 1 mM 2-mercaptoethanol, diluted with deionized water to 26 μM and mixed with an equal volume of a solution of 10 mg/ml sinapinic acid in 50% CH₃CN- 0.1% TFA. MALDI spectra was obtained on a Voyager–DE PRO biospectrometry workstation.

Fluorescence Visualization of Covalent Modification of Actin in Cells.

A549 cells were incubated with 5 μM BODIPY-bistramide conjugate (9) in cell culture media (F-12K, 20 ml) for 3 h. Cells were trypsinised, counted (1.6 × 10⁷ cells), centrifuged, washed with PBS (2 × 2 ml) and treated with 200 μl of lysis buffer (50 mM Pipes, pH 6.9, 50 mM NaCl, 5 mM MgCl₂, 5 mM EGTA, 5% glycerol, 10% protease inhibitor mixture, 0.1% Nonidet P-40, 0.1% Triton X-100, 0.1% Tween-20, 0.1% 2-mercaptoethanol, 0.001% Antifoam C). The cell suspension was sonicated for 30 s and incubated over ice for 1 h. Cellular debris was removed by centrifugation. Gel electrophoresis of the supernatant followed by imaging on a Bio-Rad Molecular Imager FX-PRO PLUS identified the fluorescently labeled protein bands (excitation and emission wavelengths were 635 and 695 nm, respectively).