Structural and Biochemical Characterization of the Interaction of

Article pubs.acs.org/jnp

Structural and Biochemical Characterization of the Interaction of Tubulin with Potent Natural Analogues of Podophyllotoxin Mayra Antúnez-Mojica,† Javier Rodríguez-Salarichs,‡ Mariano Redondo-Horcajo,‡ Alejandra León,† Isabel Barasoain,‡ Á ngeles Canales,§ F. J. Cañada,‡ Jesús Jiménez-Barbero,⊥ Laura Alvarez,*,† and J. Fernando Díaz*,‡ †

Centro de Investigaciones Químicas-IICBA, Universidad Autónoma del Estado de Morelos, Cuernavaca, Morelos 62209, México Departamento de Química Orgánica I, Facultad Ciencias Químicas, Universidad Complutense de Madrid, Avenida Complutense s/n, 28040 Madrid, Spain ‡ Department of Chemical and Physical Biology, Centro de Investigaciones Biológicas, Consejo Superior de Investigaciones Científicas, Ramiro de Maeztu 9, 28040 Madrid, Spain ⊥ CIC bioGUNE Parque Tecnológico de Bizkaia, Edif. 801A-1°, 48160 Derio-Bizkaia, Spain, and Ikerbasque, Basque Foundation for Science, Maria Diaz de Haro 3, 48009 Bilbao, Spain §

S Supporting Information *

ABSTRACT: Four natural analogues of podophyllotoxin obtained from the Mexican medicinal plant Bursera fagaroides, namely, acetyl podophyllotoxin (2), 5′-desmethoxy-β-peltatin A methyl ether (3), 7′,8′-dehydro acetyl podophyllotoxin (4), and burseranin (5), have been characterized, and their interactions with tubulin have been investigated. Cytotoxic activity measurements, followed by immunofluorescence microscopy and flow cytometry studies, demonstrated that these compounds disrupt microtubule networks in cells and cause cell cycle arrest in the G2/M phase in the A549 cell line. A tubulin binding assay showed that compounds 1−4 were potent assembly inhibitors, displaying binding to the colchicine site with Kb values ranging from 11.75 to 185.0 × 105 M−1. In contrast, burseranin (5) was not able to inhibit tubulin assembly. From the structural perspective, the ligand-binding epitopes of compounds 1−3 have been mapped using STD-NMR, showing that B and E rings are the major points for interaction with the protein. The obtained results indicate that the inhibition of tubulin assembly of this family of compounds is more effective when there are at least two methoxyl groups at the E ring, along with a trans configuration of the lactone ring in the aryltetralin lignan core. transport.3,4 The formation of microtubules is a dynamic process that involves the polymerization and depolymerization of α- and β-tubulin heterodimers.5,6 Molecules that bind to tubulin interfere with this dynamic equilibrium and thus induce cell cycle arrest, resulting in cell death and apoptosis.7 A large number of tubulin-binding agents are under clinical development. Indeed, the finding that some tubulin-binding agents also target the vascular system of tumors (antiangiogenic activity) has piqued interest in these compounds.8 In fact, the

T

he discovery of lead compounds with active biological and biomedical properties is at the heart of pharmaceutical research. Natural molecules have long served as lead compounds for the development of molecules with optimized pharmacological properties for a variety of diseases, including cancer. Within this context, microtubules are excellent targets for cancer chemotherapeutic agents, given their role in essential processes for cancer growth and metastasis.1,2 Microtubules are highly dynamic cytoskeletal fibers that are composed of α- and β-tubulin heterodimers. They regulate cellular functions including mitosis, cell motility, maintenance of cell shape and structure, cell signaling, and organelle © XXXX American Chemical Society and American Society of Pharmacognosy

Received: May 31, 2016

A

DOI: 10.1021/acs.jnatprod.6b00428 J. Nat. Prod. XXXX, XXX, XXX−XXX

Journal of Natural Products

Article

Figure 1. Chemical structures of lignans 1−5 isolated from B. fagaroides.

podophyllotoxin (4), and burseranin (5), as well as their cytotoxic and antimitotic effects in A549 (lung) and A2780 (ovarian) human carcinoma cell lines. A multidisciplinary approach has been employed to characterize the molecular recognition process using a combination of natural products chemistry, biological assays, and NMR methods assisted by molecular modeling.

search for selective inhibitors of tubulin assembly or disassembly has led to the development of some of the most clinically useful antitumor drugs currently in use, such as the Vinca alkaloids (vincristine and vinblastine), the taxoids (paclitaxel and docetaxel), epothilone, among others.9−11 Until now five ligand-binding sites have been identified in βtubulin: taxane,12,13 laulimalide/peloruside,14,15 vinca alkaloid, 16,17 maytansine/rhizoxin, 12,18 and the colchicine sites.19,20 α-Tubulin has two additional sites: pironetin21 and hemiasterlin.22 Despite the intensive efforts focused on obtaining better derivatives or analogues of the existing microtubule-destabilizing agents, no substance has been found that can outperform the efficacy of podophyllotoxin (1) for inhibiting microtubule assembly.23−26 Podophyllotoxin (PODO) is one of the most well-known naturally occurring aryltetralin lignans exhibiting antitumor activity. This lignan exhibits high potent and selective anticancer and antiviral bioactivities. PODO binds at the colchicine site in the interface between α- and β-tubulin, which inhibits the assembly of tubulin into microtubules.1 As a result, PODO produces arrest in cell division in the G2/M phase. Although this compound has been extensively investigated as an antitumor agent, the clinical results were disappointing due to severe gastrointestinal side effects.27 However, its glycosylated derivatives, etoposide and teniposide, commonly referred to as epipodophyllotoxins, are currently used in clinics for the treatment of lung and testicular cancers, lymphoma, nonlymphocytic leukemia, and multiform glioblastoma. However, these derivatives were shown not to be inhibitors of microtubule assembly; instead, their antitumor properties were due to another mechanism: interaction with DNA and inhibition of DNA topoisomerase II.28,29 Plants containing a high content of podophyllotoxin, such as Bursera fagaroides (Kunth) Engl (Burseraceae), display multiple biological activities, including anticancer, antioxidant, antivirus, antiradiation, etc. Our research team has long focused on chemical and pharmacological investigations of B. fagaroides, in which aryltetralin lignans are the representative and main bioactive constituents.30,31 In vivo studies demonstrated that some of these compounds induce mitotic arrest and interfere with cell migration during embryo epiboly, and, more importantly, these compounds were shown to affect the microtubule cytoskeleton in a zebrafish model.31 Immunolocalization of yolk cell microtubules in zebrafish embryos demonstrated that these compounds interfere with tubulin by the same mechanism as the well-known tubulin destabilizer podophyllotoxin.31 Herein, we report on the biochemical and structural interactions with tubulin of four selected natural lignans isolated from B. fagaroides, acetyl podophyllotoxin (2), 5′desmethoxy-β-peltatin A methyl ether (3), 7′,8′-dehydro acetyl

■

RESULTS AND DISCUSSION Compounds 2−5 were obtained from the CH2Cl2 extract of the stem bark of B. fagaroides as described by Antúnez et al., 2016.31 These compounds have the skeleton of an aryltetralin lactone and share chemical structural equivalences with podophyllotoxin (1), such as the presence of the methylenedioxy, the trans lactone, and, except for compound 5, the methoxyl groups in the E ring (Figure 1). The natural lignans 2−5 were evaluated in vitro for their antiproliferative activity against lung (A549) and ovarian (A2780) human carcinoma cell lines using an MTT assay. The IC50 values after 48 h treatment are summarized in Table 1. Podophyllotoxin (1) was used as reference. As shown in Table 1. Cytotoxicity of Compounds 1−5 in A549 and A2780 Cell Lines IC50 (nM)a compound 1 2 3 4 5

A549 15 25 33 263 8670

± ± ± ± ±

1 2 7 75 3000

A2780 18 34 84 587 12 940

± ± ± ± ±

1 5 10 58 4000

Values are the mean ± standard error of three independent experiments. a

Table 1, the compounds were in general highly cytotoxic against both cell lines, with IC50 values ranging from 25 nM to 13 μM. After 48 h, the A549 cell line was the most sensitive one, with acetyl podophyllotoxin (2) being the most cytotoxic lignan, with IC50 values of 25 and 34 nM against A549 and A2780 cell lines, respectively. Fittingly, these values are similar to those reported for the parent compound podophyllotoxin (1), followed by 5′-desmethoxy-β-peltatin A methyl ether (3), which showed IC50 values of 33 and 84 nM, respectively. The less active compounds were the dihydronaphthalene lignan 4 (IC50 263 and 587 nM, respectively) and burseranin (5), with IC50 values of 8670 and 12 940 nM. Microtubule agents usually induce G2/M cell cycle arrest; hence, the effect of 2−5 on the cell cycle of A549 cells was studied. The effects of a 20 h treatment with various B



Article

Figure 2. Cell cycle histograms of A549 lung carcinoma cells incubated for 20 h with either DMSO (vehicle), 50 nM paclitaxel, 100 nM 1−3, 1 μM 4, or 25 μM 5. The histograms shown correspond to the lowest ligand concentration that induced maximal arrest in the G2/M phase.

Figure 3. Effect of compounds 1−5 on the microtubule network and mitotic spindles of A549 cells. Cells were incubated for 24 h with either DMSO (A), 100 nM 1 (B), 100 nM 2 (C), 100 nM 3 (D), 1 μM 4 (E), or 50 μM 5 (F). Insets are mitotic spindles from the same preparation. The scale bar (F) represents 10 μm. All panels and insets have the same magnification.

of DNA without microtubules.7 The most active compounds were 2 and 3. Indeed, at 100 nM they completely disrupted the cytoplasmic microtubules, as well as the spindle microtubules, similar to the reference compound (1). To confirm the colchicine binding site of tubulin as the biochemical target of these compounds, the ability of 2−5 to displace MTC [2-methoxy-5-(2,3,4-trimethoxyphenyl)-2,4,6cycloheptatrien-1-one] (a bona fide colchicine binding site probe) was checked by fluorescence (Figure 4). MTC undergoes a large fluorescence change upon binding to the colchicine site of tubulin (upper line).33 In contrast, in its free state, MTC (10 μM) did not show fluorescence under excitation at 350 nm (bottom line). It can be observed (Figure 4) that compounds 2−4 significantly reduced the fluorescence of the MTC in the solution in the presence of an equimolecular concentration of tubulin, strongly suggesting that these compounds compete for the colchicine binding site. Acetylpodophyllotoxin (2) and 1 were the most potent inhibitors, with 65% and 68% decrease in fluorescence, respectively. Compound 3 produced a 43% decrease, 4 displayed only a 17% decrease of fluorescence, while 5 showed a negligible decrease. Thus, the data indicate that compounds 1−3 bind strongly to the colchicine binding site, while the binding of 4 is weaker. As

concentrations of compounds 2−5 (ranging from 20 nM to 25 μM) and the percent of cells in each phase of the cell cycle were determined by flow cytometry. DMSO (2.55 μL) and paclitaxel (30−50 nM) were used as negative and positive controls, respectively, and podophyllotoxin (1) was included for comparison. Figure 2 displays the DNA histograms of A549 cells grown in the presence of lignans 2−5 and controls (Table S01, Supporting Information). The results indicated that the compounds accumulated cells in the G2/M phase of the cell cycle with a decrease in the number of cells in the G1 phase. All compounds act similarly to the parent compound podophyllotoxin (1), although with some differences in the required concentration to reach 70−80% of cells arrested in G2/M. Compounds 2 and 3 induced G2/M arrest in the same order of magnitude of the reference 1. As expected, untreated A549 are mainly in the G1 phase of the cell cycle. The effect of compounds 2−5 on the microtubule cytoplasmic network and on the morphology of mitotic spindles in A549 cells was studied using DM1A α-tubulin monoclonal antibodies and Hoechst 33342 to stain microtubules and DNA, respectively, as previously described.32 Figure 3 shows that all the natural lignans tested exerted a depolymerizing effect of the microtubule cytoskeleton and induced the appearance of abnormal type IV spindles as a ball C



Article

Table 2. Binding Constants by MTC Displacement compound 1 2 3 MTCa colchicinea a

binding constants Kb (×105 M−1)

Kd (nM)

± ± ± ±

54 58 850 2100 6.25

185 170 11.75 4.7 1600

0.66 0.47 1.02 0.3

Positive controls, data from literature.36

podophyllotoxin (1) (Kb = (185 ± 0.66) × 105 M−1), while the binding of lignan 3 is 1 order of magnitude weaker (Kb = (11.75 ± 1.02) × 105 M−1). To further study the molecular recognition features between the α/β-tubulin dimer and the ligands, NMR studies of the ligands were performed. Saturation transfer difference NMR (STD-NMR) experiments were used to map the binding epitopes of ligands 1−3, while TR-NOESY NMR methods were applied to establish the bound conformation of 3 to tubulin. First, NMR assignments were achieved using standard 1D and 2D NMR experiments (COSY, TOCSY, HSQC, HMBC) (Table S03, Supporting Information). Clear STD signals were observed for most of the NMR resonances of the ligands. Figure 6 shows the epitope mapping obtained for 3 along with the STD spectrum (Figures S01−03, Supporting Information). The analysis of the STD data revealed that 1 and 2 displayed the highest STD effects for H-3 and H-6 at ring B, followed by the protons on the E and A rings. In contrast, 3 showed the highest STDs in rings E and B. Compound 3 showed the highest STD effects for H-3, H-5′, and H-6′ (Table 3). SAR studies have demonstrated that a free hydroxyl in C-4′, together with a bulky group at C-7 (like etoposide and teniposide), is essential for topo II inhibition,37 while the presence of the C-4′ methoxy group turns the activity to microtubule disassembly. In this regard, it is worth mentioning that the large STD effect observed at the OMe group at C-4′ in 3 indicates its close proximity to the protein. Hence, the STD data suggest that the B and E rings present key interaction points with tubulin (Figure 7). This result is in agreement with the binding affinities obtained by fluorescence, since compound 3, which displays a substitution pattern on the B and E rings that differs from that of 1 and 2, displays less affinity for the protein. Finally, TR-NOESY experiments were employed to deduce the tubulin-bound conformation. Given the nonsymmetrical chemical structure of the E ring, compound 3 was selected for these experiments. The main point of flexibility in the molecule is the C-7′−C-1′ bond connecting the C and E rings (Figure 8). Negative NOE cross-peaks were clearly observed for 3 in the presence of the tubulin α/β heterodimer, thus confirming the presence of bound ligand under the NMR experiment conditions (Figures S4 and S5, Supporting Information). The key cross-peaks that determine the conformation around the C-7′−C-1′ bond are H8−H2′ and H8−H6′ (Figure 8). Both were detected in the NOESY spectra in the free and bound states, suggesting that both conformations are indeed recognized by the protein. This evidence is in line with the fact that compounds 1 and 2, which contain three OMe groups at the E ring, are also recognized by tubulin (Figure 8). In all cases, the rigidity of the trans-lactones ensures that the torsion angle between the two aromatic rings (C1(B)−C2(B)−C7′(C)−

Figure 4. Displacement of MTC from the colchicine binding site by lignans. Fluorescence emission spectra of 10 μM tubulin plus 10 μM MTC in the presence of 10 μM of each compound 1−5. The top line is the fluorescence emission spectrum of MTC in the presence of tubulin, and the bottom line is fluorescence of MTC alone.

expected, these results strongly correlate with the cytoplasmic microtubule disruption effect observed in A549 cells. Once the binding site was established, the effect of 1−5 on the assembly of purified tubulin was tested. The process is considered to be a nucleated condensation in which tubulin assembles into microtubules at the critical concentration of 3.3 μM. Polymerization does not occur at this concentration.34 At 30 μM, tubulin polymerizes to form microtubules, which can be pelleted under mild centrifugation conditions. Thus, the degree of tubulin polymerization can be evaluated from the mass of the microtubules pelleted in the presence of stoichiometric and substoichiometric concentrations of each natural lignan (1−5). The data reveal that 1 and 2 are potent substoichiometric assembly inhibitors. Figure 5 shows the presence of lignan 1 or

Figure 5. Inhibition of tubulin assembly by the natural lignans 1−5.

2 caused more than an 80% decrease in pelleted microtubules. In contrast, compound 3 or 4 produced only a medium−weak inhibition, even at overstoichiometric concentrations (Table S02, Supporting Information). Compound 5 did not show any significant inhibition. Therefore, it can be safely assumed that the presence of methoxy groups in the E ring is indispensable for the activity. Finally the binding constants of the most active ligands 1−3 to tubulin were measured by a competition assay with MTC.35 This assay was performed assuming a 1:1 stoichiometry. The data in Table 2 show that compound 2 was the most potent binder, with a very good binding affinity (Kb = (170 ± 0.47) × 105 M−1), within the same order of magnitude as the reference D



Article

Figure 6. (a) Ligand epitope mapping of the molecular recognition of 3 with the heterodimer of tubulin as inferred from STD experiments. In red, the signals showing the maximum STD effect; in orange, those moderately affected; in yellow, those affected to the lowest extent. (b) Off-resonance (blank) NMR spectrum acquired without protein irradiation. (c) STD experiment with 3 in the presence of dimeric tubulin.

Table 3. Relative Percentage of STD of Compounds 1−3 position

1 % relative STD

2 % relative STD

6 5′ 2′ 6′ 3 O-CH2-O 7′ 9α 9β O-CH3-6 O-CH3-4′ O-CH3-5′ O-CH3-3′ 7α 8′ 8 7β OAc

100

100

65 65 95 74

70 70 100 60

74 65 65

60 50 50

65 61

3 % relative STD 100 84 93 91 84

Figure 7. Schematic representation of protons that interact with tubulin. In red the most affected protons in the complex and the dotted blue box is the whole region that interacts with the protein.

88 74 79 67 63

These models (Figure 9) were refined as described in the Experimental Section and then submitted to CORCEMA-ST calculations.38 These calculations back-calculate the expected STD values from a given ligand/receptor geometry (Figure 10). Thus, the comparison between the experimental and theoretical STDs allows the selection of the binding modes that better fit the experimental STD-NMR data. Also, the relative intensities were predicted with reasonable agreement and good R-factors (Figure 10). In order to validate the NMR models, the calculated structure of podophyllotoxin bound to soluble tubulin was compared with that solved by X-ray diffraction (PDB code: 1SA1) (Figure 9A). The root-mean-square deviation (RMSD) and the maximum difference between the model and the X-ray diffraction structure were calculated using their heavy atoms. Both values were found lower than 1.0 Å (RMSD = 0.33 Å; maximum difference = 0.72 Å), indicating that the calculated

60

C1′(E)) is kept around 105°, providing the proper orientation for their simultaneous interactions with the protein. The STD and TR-NOESY data were employed to validate the structures deduced from a molecular modeling protocol. First, the X-ray coordinates of the reported crystallographic structure of podophyllotoxin bound to tubulin (PDB 1SA1) were used to build different starting geometries of 1−3, generated by manipulation of the ligand chemical structure to comply with the conformational evidence obtained from the TR-NOESY data. E



Article

Figure 8. (A and B) Representations of the two possible rotamers of 3 arising from rotation of the E ring with respect to the C ring.

Figure 9. Representation of binding poses of podophyllotoxin and its anologues 2 and 3. Podophyllotoxin is orange, and its anologues are gray. (A) Superposition of electron crystallographic structure PDB 1SA1 (cyan) and NMR-based structure (orange) of podophyllotoxin. (B) Binding poses of compounds 1 and 2. (C) Binding poses of 1 and 3 (in rotamer A of Figure 8). (D) Molecular surface of protein around the binding site of podophyllotoxin.

with dipolar interactions. Compounds 2 and 3 share the same binding pose with podophyllotoxin except for a slight displacement in different directions (Figure 9B,C). Compound 2 is shifted to the western part of the site, and compound 3 is moved to the north of the site. Both of these shifts cause weaker contacts between the amino acids at the binding site of tubulin and the ligands. Substitutions at R1 of compound 1 caused some changes in the position of the ligand within the tubulin binding site. In compound 3, the absence of the hydroxyl group decreases the binding strength with the protein (Table 2). On the other hand, the acetyl group of compound 2 still can act as a hydrogen bond acceptor, and consequently the change in affinity is not significant (Table 2, Figure 9D). Concerning position R2, the presence or absence of the methoxyl group modifies the orientation of the closest loop of tubulin. Indeed,

and the X-ray diffraction structures describe the same binding pose. In this pose, podophyllotoxin is buried in the intradimer interface forming two H-bonds with the protein: one of them between the alcohol at R1 and the backbone of Thr-α179 and the second between the equivalents OCH3 at C-3′ and C-5′ of the E-ring and the side chain of Cys-β241. These contacts result in favorable contributions to the binding enthalpy and entropy, the favorable entropic contribution being due to the presence of a symmetric plane that splits the E ring into two equal parts. This produces two energetically degenerate conformations that are positive for binding from the entropic point of view. Only the B conformer of compound 3 is capable of forming a hydrogen bond between the OCH3 at C-3′ and Cys-241 of tubulin (Figure 8), which may explain its decreased experimental Kb. The third relevant contact is established between the oxygens of the lactone group and His-8 and Thr-7, F



Article

Figure 10. Comparison between experimental and theoretical STD data (CORCEMA-ST) for the generated 3D models (Figure 9) of the dimer of α/β-tubulin bound to 1−3. The experimental STD intensities are given in red, and the calculated data in black. For compounds 1 and 2, the calculated values for H-2′ and H-6′ are mean values of both indistinguishable protons in the 1H NMR spectrum.

eliminating the favorable entropic factor due to the symmetry, the same as was deduced for compound 3. A multidisciplinary approach has been employed to validate a new generation of natural products from B. fagaroides. The more active compounds are cytotoxic, and they induce G2/M arrest and tubulin depolymerization at nanomolar concentrations. The assays related to the mechanism of action have demonstrated that they target the colchicine binding site in tubulin. Compounds 1−3 were similarly active against cancer cell lines and displayed relatively similar affinities for the binding site. In contrast, lignans 4 and 5, which show variations on the conformation and substitution pattern of the E ring, respectively, were much less active against the studied cell lines and did not show affinity to the active site of tubulin. In particular, those lignans that contain tri- or dimethoxy aryl moieties (1−3) were more cytotoxic than compound 5 or 4. Therefore, subtle variations in the chemical structure provoked drastic variations in cytotoxicity. Fittingly, compound 2 exhibited an equivalent activity with the parent compound, 1. The only chemical difference between compound 1 and compound 2 is the presence of an acetyl group at the R1 position of 2 instead of the hydroxyl group present at R1 in 1. Nevertheless, detailed NMR studies (STD and TR-NOESY) and molecular modeling approaches revealed a similar binding mode for all compounds with structural details related to the position of the substituents at the B and E rings. The C ring adopts a twist-boat conformation in the bound state, while the torsion connecting rings E and C rings remains flexible. The antimitotic effect of this family of compounds depends on the conformation of the E ring, and structural changes that affect the global geometry (the angle between the C and E rings) impact binding with the active site. It is interesting that the bioactive conformation of podophyllotoxin, determined in solution by experimental TR-NOESY NMR, was equivalent to

the same disposition of the loop was found for 1 and 2, but it was clearly modified in the case of the complex with 3. Additionally, the β-sheet is also slightly deformed, depending on the grade of methylation at the 3′ position. Finally, changes at R and R2 substitutions prevent the formation of the H-bond between the ligand and Thr-α179, perturbing the interaction between the lactone and the protein. The relationships between the structural changes in compounds 1−5 with their cytotoxicity in A549 and A2780 carcinoma cell lines (Table 4) were associated with the changes Table 4. Structure−Cytotoxicity Relationships of Compounds 1−5 chemical substitution R1 7′−8′ type bond extra ring R R2

cytotoxicity by chemical substitution Single Effect OH > OAc single ≫ double absent ≫ present Conjugated Effects R:H > OCH3 R2:OCH3 > H

IC50 by chemical substitution 1

Structural and Biochemical Characterization of the Interaction of

Recommend Documents