Migrastatin analogues with an ( E )-alkene at the ring C-3: synthesis, conformational analysis and biological evaluation

Synthesis of one acyclic and four macrocyclic analogues of migrastatin with an E -alkene at C-3 is described. The route involves ring-closing metathesis, which gave ca. 1:1 mixtures of stereoisomeric alkenes. The crystal structure for a macrolactam with an E -alkene at C-3 and Z -alkene at C-8 was obtained with a brief study of macrocycle conformation by molecular mechanics and coupling constant analysis. Amide rotamers are also observed by NMR spectroscopy. Preliminary in vitro biological study indicates that the new compounds may have limited potential as inhibitors of tumor cell migration.


Introduction
Migrastatin 1 (Scheme 1), is a polyketide [1][2][3] containing a 14-membered macrocyclic lactone, which has been reported to inhibit tumor cell migration.Structurally less complex analogues or truncated 4 analogues of migrastatin such as 2-5, have shown improved inhibition of tumour cell migration than migrastatin, in vitro 5 .7] Our group have been involved in the synthesis of migrastatin and its analogues 8 and through collaboration with Anderson and Nobis have reported that macroketone 3 inhibited E-cadherin dynamics in vivo, in a manner consistent with increased cell adhesion and reduced invasive potential. 9][12][13][14][15][16][17][18][19][20] The macrocyclic ring of migrastatin has an alkene with Z configuration at its C-3, and there are E-alkenes at C-8 and C-12.We report here the synthesis of five analogues of the core of migrastatin 6-10, including macrolactams and macrolactones with an E-alkene at C-3. Lactam analogues of the migrastatin macrolactone were found previously to be more stable in vivo and thus considered important for this reason.In addition to the synthesis we provide information on the conformation of one macrolactam, including its crystal structure.We also report preliminary data on abilities of compounds to inhibit migration of cancer cells.

Syntheses of macrocyclic compounds
Syntheses of migrastatin analogues use ring closing metathesis (RCM) 21 for macrocyclisation.Thus, dienes needed for the preparation of macrocyclic rings with E-alkenes at C-3 were first prepared from known ester 11 (Scheme 2). 22The subsequent steps we used were similar to those used by Danishefsky and co-workers in synthesis of 2-4.Thus, the reduction of 11 with DIBAL-H gave the allylic alcohol 12 5 (Scheme 2).The reaction of 12 with diphenylphosporyl azide in the presence of DBU followed by the Staudinger reaction gave an amine intermediate, which after coupling with 6-heptenoic acid in the presence of EDC gave 14, whereas direct Mitsunobu reaction of 12 with 6-heptenoic acid gave ester 13.Heating of ester 13, which has the Econfiguration at C-3, with the Grubbs-II catalyst in toluene at reflux for 45 minutes gave a mixture of cyclic alkenes (8E:8Z = 50:50, 46% isolated yield).Repeating this reaction over a longer time (90 minutes) gave rise to a modest increase amount of the E-alkene (56:43 with 42% yield).The Grubbs-II catalyst promoted reaction of amide 14 gave also a mixture of alkenes (8Z:8E = 1.2:1, 46%), with the Z-isomer preferred.Removal of the TBS protecting groups with HF-pyridine, from the intermediates obtained, gave 7-10.The stereoselectivities observed in the RCM reactions differ to those where there was a Z-alkene at C-3; preparations of 2-5 by RCM gave only E-isomers.Macrocycle 9 was subjected to saponification to give acyclic compound 6.Acyclic compounds were shown to inhibit tumour cell migratory capacity in previous work.Scheme 2. Synthesis of 7-10.Scheme 3 Synthesis of 6.

Conformation of macrolactam 10
In the case of lactam 10, crystals suitable for X-ray diffraction were obtained from EtOAc and the crystal structure was therefore determined (Figure 1).There were two crystallographically distinct molecules in the asymmetric unit and both had the same conformation.The E-configuration at C-3 as well as the Zconfiguration at C-8 and the presence of trans-amide was confirmed for 10 in the solid state.The dihedral angles as found in the crystal structure for the NH to C-10 backbone of the macrocycle are given in Table 1.Vicinal coupling constants ( 3 J values, Table 1) were also extracted from the 1 H-NMR spectrum of 10 recorded at 500 MHz (CDCl3) and these are listed also in Table 1.The MestReJ 23 programme uses a range of different equations for the computation of vicinal coupling constants (J values) in 1 H-NMR spectra.5] In Table 1, the calculated J values from Barfield-Smith [26][27] and Karplus equations as implemented by MestReJ are shown and they generally compare well with those obtained by recording the 1 H-NMR spectrum in CDCl3 for 10, with the largest discrepancy between calculated and measured 3 J values being 2.4 Hz for the coupling between the NH and one of the C-2 protons when using the Barfield-Smith equation and which was 1.8 Hz for use of the Karplus equation.Aside from the dihedral angle, the J value can be influenced by factors such as electronegative substituents and the position of those same electronegative substituents, which may contribute to differences between calculated and observed J values.The data indicates that the backbone from the amide to C-10 of the macrocycle in the crystal structure and in CDCl3 are similar.The X-ray crystal structure was also compared with the lowest energy structure generated after a conformational search, generating, minimising and evaluation energies of 10,000 conformers using Macromodel 28 (Schrodinger Inc., LLC, New York, USA) with implementation of the OPLS-AA force field. 29The two structures were superimposed (Figure 2a) and there was good correlation (RMSD = 0.25) between the lowest energy structure obtained by this conformational search with that of the X-ray crystal structure for the C-2 to C-10 backbone atoms.Five conformers were found for the lactam within 10 kJ/mol of the global minimum and these were superimposed and shown in Figure 2b.During the conformer search, the interconversion between the trans and cis amide was enabled and two of the low energy conformers were found to be the cis-rather than the trans-amide.Preliminary viewing of the 1 H-NMR spectrum for 10 gave the impression of impurities being present.1][32][33][34] The ratio was determined to be 9:1 by integration of NMR signals with the trans-amide assumed to be favoured, as supported by the trend in stability (trans amide more stable than cis amide) predicted using Macromodel.The signal at  5.39 (triplet) is assigned to the C-8 alkene proton for the lactam with the trans configured amide, whereas the triplet at  5.60 is assigned to the C-8 alkene proton of the lactam with the cisamide.There is evidence for interconversion between these amide rotamers, as provided by both 1D-NOE and 2D-NOE experiments.The basis for using 1D-NOE or 2D-NOE is that cross-peaks or correlations can be observed that are opposite in phase (negative NOEs) to the normal NOE enhancements (positive NOEs), because they arise through chemical exchange.Irradiation of the signal at  5.39 using a 1D-NOE experiment led to indirect irradiation of the signal at  5.60 through saturation transfer, indicating the nuclei interconvert.A negative NOE cross-peak, between these two signals is also apparent in the 2D-NOE spectrum of 10.Other negative crosspeak correlations are observed in the 2D spectrum, including one between the signal for H-2b at  3.54 (dd) of the major rotamer and with the signal at  3.74 ppm (dd), the latter assigned to H-2b of minor rotamer.

Evaluation of compounds as tumor cell migration inhibitors
Firstly, cell viability assays showed compounds were generally not toxic to cells (see Supp.Info.).Then four compounds 6-8 and 10 (SMGS1-4), were selected for the wound healing assay which was performed using different human cancer cell lines (pancreatic: MiaPaCa2, and breast: MDA-MB231, MDA-MB361 and MCF7) and all four compounds reduced wound closure for MiaPaCa2 cells after 24h compared to the control (Figure 3).The macrocyclic compounds SMGS2-4 reduced wound closure for MDA-MB361 cells but no significant inhibition of wound closure was observed for MDA-MB231 or MCF7 cells.None of 6-8 and 10, nor macrolactam 4 inhibited migration of these cancer cell types using the transwell migration assay (Boyden chamber assay).The Boyden chamber assay is considered a more comprehensive measure of inhibitory capacity towards cell migration.The wound repair study was also carried out for macrolactone 8 and known migration inhibitor 4 using the lung tumour A549 cell line.The wound repair assay showed cell inhibition over 24 h for macrolactam 4 with much less inhibition for 8.The transwell migration assay showed that 4 has the highest % inhibition of the A549 cell migration, but that 8 had low inhibition.

Conclusions
Herein the synthesis of analogues of migrastatin, where there is an E-alkene present at C-3, is described, providing compounds for study as potential anti-metastatic agents.Ring closure metathesis reactions showed an increased preference for the Z-or cis alkene in such analogues compared to compounds with the Z-alkene at C-3.The crystal structure of a macrolactam, and analysis of coupling constants in 1 H-NMR spectra, as well as conformational searching techniques in Macromodel provided information on the conformation of the macrocyclic ring.The presence of amide rotamers in a macrolactam derivative was supported by 1D and 2D

Experimental Section
General. 1 H-NMR spectra were recorded at 500 MHz and 13 C spectra at 126 MHz.The chemical shifts are reported in parts per million (ppm).The coupling constants are reported in Hertz (Hz) and the resonance patterns are reported from 1 H-NMR spectra with notations as the following: br (broad), s (singlet), d (double), t (triplet), q (quartet) and m (multiplet). 1 H-NMR signals were assigned with the aid of gCOSY. 13C signals were assigned with the aid of DEPT-135, gHSQCAD and gHMBCAD.IR spectra were recorded using thin film on NaCl or germanium plates.Optical rotations were determined at the sodium D line at 23 o C. TLC was performed on aluminium sheets pre-coated with Silica Gel 60 (HF254, E. Merck) and spots visualized by UV and charring with 1:20 H2SO4-EtOH or with 1:1 KMnO4 (1% w/v solution)-NaHCO3 (5% w/v solution) or with cerium molybdate stain.Flash chromatography was generally employed and was carried out using Silica Gel 60 (0.040-0.630 mm, E. Merck) and employed a stepwise solvent polarity gradient correlated with the TLC mobility.Chromatography solvents used were EtOAc, dichloromethane (Riedel-deHaen), cyclohexane, pentane and MeOH (Sigma Aldrich).Anhyd.DMF and anhydrous toluene were used as purchased from Sigma-Aldrich.Solvents THF, dichloromethane and methanol were used as obtained from a Pure-Solv TM solvent purification system.

Single crystal X-ray diffraction
An Oxford Diffraction Xcalibur system was used to collect X-ray diffraction data at ambient temperature for HPS.6] The crystal structure data has been deposited at the Cambridge Crystallographic Database (CCD).The deposition number is 1015786.

Molecular modeling
The coordinates for 10 (SMGS4) obtained from the X-ray crystal structure, provided the starting point for the conformational search The OPLSAA force field as implemented in Macromodel (www.schrodinger.com)was used.In total, for the conformational search, 10,000 conformers were generated using the MCMM method and each of these was minimised to convergence using the PRCG method.Each conformer generated was minimised in a solvent free environment and energies calculated.Conformers with energy values > 12 kJ/mol above the minimum were rejected during this analysis.

In vitro wound healing assays
The scratch assay was typically carried out to assess the influence of migrastatin analogues on cancer-cell motility.The cells were seeded at high density on FalconTM 6-well cell culture plates (Thermo Fisher Scientific, USA).After 24 h, when a 100% confluence was reached, the medium was removed and replaced with medium containing one of the migrastatin analogues or DMSO (control) at a concentration of 10μM.The monolayer was wounded by scratching the surface as uniformly and straight as possible with a pipette tip (1000 μL) at least four times.Images of cells invading the scratch were captured at indicated time points (0, 6, 12 and 24 h) by using a phase-contrast Olympus CKX41SF microscope (Olympus Corporation, Japan).The pictures were analyzed with ImageJ 1.50b (National Institutes of Health, USA).The migration rate was expressed as percentage of scratch closure and was calculated as follows: percentage of scratch closure = 100*(a -b)/a, where a is the surface area at time point 0 h, and b is the surface area that remained cell-free after 24 h of migration.The values are represented as means of three different wound fields from three independent experiments (n = 9).

Statistical analysis
Statistical analysis was carried out by using GraphPad Prism 7.0 software (GraphPad Software, USA).The invitro wound-healing assay and transwell migration assay were analyzed by using the one-way ANOVA, the Dunnett's post-hoc test and the t-test.A p-value of <0.05 was regarded as significant, whereas p-values of <0.01 and <0.001 were regarded as highly significant.The data is expressed as mean ± standard deviation.

Figure 1 .
Figure 1.X-Ray crystal structural diagram for 10.The numbering used here does not correspond with IUPAC nomenclature numbering.C2-C3 labelling here corresponds to the Z-alkene.

Figure 2 .
Figure 2. (a) Overlap of X-ray structure with lowest energy structure generated by conformational search in Macromodel is shown on the left.Overlay was generated by superimposing macrocyclic backbone atoms NH to C-10 (RMSD = 0.25) (b) Five superimposed isomers generated within 10 kJ/mol of global minimum as obtained from conformational search using Macromodel are shown on the right.Two of these five stereoisomers had a cis-amide conformation; the lowest energy conformer had the trans-amide.

Figure 3 .
Figure 3. Bar graphs show the effect of the migrastatin analogues on migration of various cell lines after 24 h of incubation.Compounds significantly retarded the process of wound healing in MiaPaCa2 cells and three compounds retarded the process for MDA-MB361 cells (M361 on graphs).No significant retardation compared to control was observed for MCF7 and MDA-MB231 cells (M231 on graphs).Results with p <0.001, and p<0.01 were considered to be highly relevant and denoted ** and ***.Results with p <0.05 were determined to be relevant and are indicated with *.