Persistent inhibition of pore-based cell migration by sub-toxic doses of miuraenamide, an actin filament stabilizer

Opposed to tubulin-binding agents, actin-binding small molecules have not yet become part of clinical tumor treatment, most likely due to the fear of general cytotoxicity. Addressing this problem, we investigated the long-term efficacy of sub-toxic doses of miuraenamide, an actin filament stabilizing natural compound, on tumor cell (SKOV3) migration. No cytotoxic effects or persistent morphological changes occurred at a concentration of miuraenamide of 20 nM. After 72 h treatment with this concentration, nuclear stiffness was increased, causing reduced migration through pores in a Boyden chamber, while cell migration and chemotaxis per se were unaltered. A concomitant time-resolved proteomic approach showed down regulation of a protein cluster after 56 h treatment. This cluster correlated best with the Wnt signaling pathway. A further analysis of the actin associated MRTF/SRF signaling showed a surprising reduction of SRF-regulated proteins. In contrast to acute effects of actin-binding compounds on actin at high concentrations, long-term low-dose treatment elicits much more subtle but still functionally relevant changes beyond simple destruction of the cytoskeleton. These range from biophysical parameters to regulation of protein expression, and may help to better understand the complex biology of actin, as well as to initiate alternative regimes for the testing of actin-targeting drugs.


SKOV3 cell migration through pores is impaired after low dose Miu treatment without affecting other features of migration. Migration of SKOV3 cells was analyzed in a Boyden chamber assay.
Pretreatment with Miu reduced the number of migrating cells nearly by half (Fig. 2a). In contrast, in the wound healing assay, which monitors undirected cell motility as such, migration was not affected by treatment (Fig. 2b).
To elucidate the effect of Miu treatment on chemotactic sensing a two dimensional (2D) chemotaxis assay was performed (Fig. 2c). The forward migration index (FMI) of SKOV3 cells and the velocity were not influenced after Miu treatment and thus, showed no significant inhibition of directed migration. Therefore, the inhibitor effect on pore based migration seemed not to be based on an inhibition on motility as such or on a reduced sense of direction. Additional migration associated parameters, like the adhesion and spreading of cells on a surface, which are the initial steps in cell migration, were also unaffected by Miu treatment, both quantitatively and morphologically ( Fig. 3a and b). Since intact vesicle trafficking is also mandatory for directional cell movement, we monitored overall cellular secretion in a cell line (HeLa) stably expressing a secreted version of horseradish peroxidase. This was likewise not reduced by treatment with Miu, while the positive control (Brefeldin A) worked as was to be expected. Since both, the 2D chemotaxis and the Boyden chamber assay are based on directional movement, the only difference, which might explain the different outcome in both, is the fact that the cells have to deform and squeeze through small pores in the Boyden chamber.
Interestingly, pretreatment with 20 nM Miu for 72 h did not change migration behavior of HUVECs -neither in the scratch assay, nor in the Boyden chamber ( Supplementary Figure 3b and c).

Nuclear stiffness of SKOV3 cells is enhanced after Miu treatment.
Consequently, in a next step, we probed cell and nuclear stiffness after treatment with Miu by atomic force microscopy as a measure of cell deformability. The integrated cell stiffness was only marginally increased after Miu treatment for 72 h (Fig. 4a). The nucleus is the largest organelle in the cell and is supposed to be the limiting factor during migration through restricted pores, therefore the stiffness of the nucleus was determined separately. The Miu treated cells displayed Figure 2. Analysis of SKOV3 cell migration. SKOV3 cells were pre-treated with 20 nM Miu over 72 h, during the migration process no compound was present. (a) Boyden chamber migration assay over 24 h migration through a 8 µm pore membrane. Paired two tailed t-test, *p < 0.05. (b) Wound healing assay, neg Ctrl: negative control, migration in FCS free medium. One representative image per treatment is shown. (c) Chemotaxis assay over 24 h along an FCS gradient. FMI, forward migration index, +/+ Ctrl: positive control, medium with FCS on both sides, no gradient. One-Way ANOVA/Bonferroni's Multiple Comparison Test, *p < 0.05. n = 3.
SCIenTIfIC RePORtS | 7: 16407 | DOI:10.1038/s41598-017-16759-7 a 1.4-fold higher stiffness of nuclei compared to control cells (Fig. 4b). In line with higher nuclear stiffness, the three dimensional (3D) migration of SKOV3 cells through a rat tail collagen I gel with restricting mesh size was reduced: fewer cells were able to squeeze through the restrictions on their migration track after Miu treatment. The number of cells which are immobile or die during the observation time slightly increased after Miu treatment by trend, but showed no alteration of sprouting cells, which are able to build protrusions, in comparison to control cells (Fig. 4c).

Miu treatment causes late downregulation of proteins enriched in the Wnt pathway.
Since actin is an integral part of transcription-and chromatin-remodeling complexes and plays an important role in mechano-induced signaling via the transcriptional co-activators MRTF and YAP, we investigated time dependent changes of the proteome upon treatment with 20 nM Miu. The detected proteins are depicted in a heat plot (Fig. 5a), where clusters downregulated in comparison to control are coded in green and upregulated proteins in red. There are two prominent time windows where larger clusters of proteins are transiently up-or downregulated: after 4 h of stimulation and after 56 h (Fig. 5b). We deemed it much more likely that the latter time point is functionally more important for the effects on nuclear stiffness and migration after 72 h, and, therefore, focused on this cluster for further analysis (Fig. 5b). The regulated proteins in the cluster indicated by a yellow frame in the heat map (upregulated after 56 h, Fig. 5b upper panel), are listed in Supplement Table 1. The proteins of the cluster framed in red (downregulated after 56 h, Fig. 5b lower panel) are listed in Supplement Table 2. These show different regulated pathways in enrichment analysis (Fig. 5c). The KEGG (Kyoto Encyclopedia of Genes and Genomes) Wnt-signaling pathway had a relatively high enrichment value and the lowest false discovery rate (FDR). 17 proteins annotated to this signaling pathway are significantly downregulated in SKOV3 cells after Miu  Table 3). The next best two pathways were "autophagy" and "epithelial cell signaling in H. pylori infection". The pathway analysis of the upregulated genes revealed 11 clusters by our criteria (false detection rate <0.2, enrichment >2, Supplementary Figure 2). However, the cluster with the highest enrichment ("sleep") consisted only of one protein and is functionally not relevant. When we ordered the clusters according to their P-values, "structural constituent of cytoskeleton"and "muscle contraction" were the most significant ones.

Miu treatment results in the inhibition of MRTF-associated protein expression. Further analysis
of the proteome data after 56 h stimulation with Miu was performed concerning MRTF-associated proteins. The selection of genes/proteins was collected by GeneCards ® Keyword Search 'MRTF' (Supplement Table 4). Most of the listed MRTF-associated proteins were down-regulated after 56 h (Fig. 6a). A few of them are related to migration regulating processes and interact with MRTF signaling. The acetyltransferase and transcription co-activator p300 was strongly down-regulated in proteome analysis after 56 h treatment with Miu. This was exemplarily confirmed by Western blot analysis, which indicated a reduction of protein level to 0.58-fold compared to control (Fig. 6b).

Discussion
Though intuitively attractive as an anti-tumor target, actin has been neglected in this regard for many years due to initial failures 6 and bad therapeutic indices 14 . Recently, we successfully used chondramide, an actin nucleating myxobacteriual depsipeptide 15 for inhibiting tumor growth 16,17 and metastasis 18 in vivo. Furthermore, by using chondramide, we elicited very specific effects on the differentiation of macrophages 19 or kinase signaling 16 at concentrations were an obvious destruction of the actin cytoskeleton and acute cytotoxic effects were still absent. This potential selectivity of actin binding compounds can maybe be explained by the concept that actin is not merely forming polymers with other actins and subsequently depolymerizes again, but that there is a continuous competition between actin binding proteins for binding sites on actin 11 . This complex network then allows for subtle control of the "actin-interactome" and related cell functions. With respect to actin binding small molecular compounds it has been found that they, in turn, can compete with specific actin binding proteins. So, for example, kabiramide C has been shown to compete with actin capping proteins like gelsolin 20 .
These findings prompted us to investigate the actions of miuraenamide A, an actin filament stabilizing myxobacterial compound 12 , which is synthetically relatively well accessible 9 , on tumor cell migration at sub-toxic concentrations. We focused on two issues: 1) biomechanical changes, and 2) changes in the proteome of treated cells over time.
Concerning the first aspect, we found that at a concentration of 20 nM, which did not cause changes in cell viability, the structure of actin was initially subtly changed, but recovered subsequently. At this later time (72 h) migration of cells and chemotaxis were unaffected in 2D-settings, while the migration through pores (or a collagen mesh with restricting size) posing a spatial obstacle was reduced. This was accompanied by a small rise of overall mechanical cell stiffness, but a significant increase of nuclear stiffness. This is in line with recent findings, which identified the nucleus as the most relevant cellular barrier while navigating through confined environments 21 . While the stiffness of the nuclear envelope itself is mainly determined by the lamins 22 , actin also plays an important role here: on the one hand, it is needed to deform the nucleus 23 , on the other hand, deformation bears the danger of nuclear rupture 24 , and actin is needed to form a mechanical shield for the nucleus 25 . These different actions are, again, regulated by different actin binding proteins: Arp2/3 23 and FMN2 25 , respectively. It lies close at hand that miuraenamide could change nuclear deformability by entering the competitive process for actin binding. In addition, it has recently been shown that the recruitment of different actin pools by different nucleating proteins (Arp2/3 vs. mDia) causes distinct migration behavior in cells 26 . Processes like these might explain, why we only see effects in a specific migration mode. This aspect is supported by the finding that migration in confined and low-adhesive environment (which is clearly the case in migration through pores of a Boyden chamber) elicits a different migration mode 27 . Interestingly, we did not observe this reduction of pore based migration in non-tumor cells (HUVECs). At the moment we can only speculate about the mechanistic reason for this difference between cell types. Different kinetics of actin turnover, or different relative importance of signaling pathways due to cell type could be responsible.
Concerning the second aspect, changes in the proteome due to long term treatment, there is a close relation to nuclear stiffness or deformation: changes of this parameter change gene expression 28 . Furthermore, mechanosensitive transcriptional co-activators like MRTF or YAP are closely regulated by changes in the actin equilibrium 29,30 . Since this is also a process of competition for actin binding sites 30 , we looked for effects of miuraenamide treatment on protein levels. We focused on changes which occurred in the time window briefly before the migration experiment, since we deemed these to be the functionally most relevant ones. After 56 h two very distinctly up-or downregulated protein clusters emerged. Looking for causes for the reduced migration we first concentrated on the down-regulated proteins (Supplement Table 1) and analyzed for pathways with significantly enriched regulated proteins according to KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways. The three signaling pathways with the combination highest enrichment /lowest false discovery rate were: 1) "Epithelial cell signaling in H. pylori infection", which is not specifically related to migration, but contains a number of motility associated genes, 2) "Autophagy", which is interesting, since actin has recently been shown to be related to autophagy 31 , and since autophagy influences migration 32 , and 3) "Wnt-signaling pathway", which is closely related to migration. The protein with the highest downregulation in the Wnt pathway turned out to be the histone acetyl transferase EP300 (or p300), which has already been linked to the mechanosensitive SRF/MRTF pathway 33 . This is surprising, since actin polymerization should cause an increased nuclear translocation of MRTF [34][35][36] , and, consequently, upregulation of MRTF target genes. However, when looking for MRTF target gene products, we surprisingly found some classical candidates downregulated after treatment with miuraenamide at 56 h (Supplement Table 4). Like with the recovery of the microscopic architecture of the actin cytoskeleton after some hours of low-dose miuraenamide, also MRTF signaling seems to be counter-regulated after some time by an as yet unknown mechanism. A closer look at the upregulated proteins indicated "structural constituent of cytoskeleton" and "muscle contraction" as the most significantly influenced pathways. This could be interpreted as a compensatory reaction of the cells upon the manipulation of actin by Miu.
All in all, the proteomics data do not explain the changes in nuclear stiffness by treatment with miuraenamide, since no proteins connected to the nuclear lamina are altered. The mode of action might depend on the direct effects on actin and/or actin binding proteins. The changes in protein expression seem more like a consequence of the nuclear effects than a cause for them and might indicate positive or negative feedback mechanisms.
In conclusion, low-dose treatment of tumor cells with miuraenamide over prolonged time does not cause cytotoxic effects, but inhibits migration through confined environments in a multifactorial way: nuclear stiffness is increased and, at the same time, the proteome is altered in several anti-migratory ways (especially decreased Wnt-and MRTF-signaling). It is tempting to speculate that by using the right treatment scheme it could be possible to apply actin binding compounds for eliciting specific and subtle changes in cellular behavior. It seems worthwhile to revisit some of the actin compounds yet neglected for in vivo studies under this aspect.

Methods
Study compound. Miuraenamide A was synthetized as described previously 9 .
Cultivation of cell lines. The human ovarian cell line SKOV3 was purchased from LGC Standards (ATCC-HTB-77, Wesel, Germany) and was cultivated in RPMI 1640 medium (PAN Biotech, Aidenbach, Germany) with 1% penicillin/streptomycin (1.5 mM, PAA Laboratories, Austria) and 10% (v/v) FCS (PAA Laboratories, Pasching, Austria). Absence of mycoplasms was regularly tested by PCR. Cells were used in passages 5-10 after thawing. Creation of the stably transfected HeLa-ssHRP cells was described previously 37 . These cells were cultivated in DMEM high glucose (PAN Biotech, Aidenbach, Germany) with 1% penicillin/streptomycin (1.5 mM) and 10% FCS. The cells were cultivated in an incubator with constant humidity at 37 °C and 5% CO 2 . Primary human umbilical vein endothelial cells (HUVECs) were purchased from Promocell (Heidelberg, Germany) and cultivated in endothelial cell growth medium (ECGM) from Promocell, according to the manu-facturer´s instructions. Cells were routinely used in passage 4 to 6.
For all experiments samples were collected simultaneously after staggered treatment.
Proliferation assay. The proliferation assay was performed in 96-well plates with an initial concentration of 5 × 10 3 cells/well. Cells were seeded overnight and afterwards treated for 72 h with the indicated concentrations of compound. Additional wells seeded with untreated cells were measured as day zero control. The cells were fixed and stained, after a washing step with PBS, with crystal violet/methanol (0.5% crystal violet (w/v), 20% methanol (v/v)). After elution of crystal violet with ethanol/Na-citrate (50% ethanol (v/v), 50% 0.1 M Na-citrat (w/v)), the absorption of the solution was determined using a microplate reader at 540 nm (Tecan Sunrise TM Microplate Absorbance Reader, Maennedorf, Austria). Adhesion assay. The adhesion assay was performed with pre-treated cells in 96-well plates. A concentration of 4 × 10 4 cells/well were seeded on collagen G coated wells and allowed to adhere for 90 min. Cells were stained with crystal violet/methanol and imaged with Axiovert 25 microscope and adhering cells were counted.

CellTiter-Blue
Chemotaxis assay. With the chemotaxis assay, migration of cells along an FCS gradient without spatial restriction was detected. Cells were pre-treated with the indicated compound concentration and seeded into ibidi TM µ-Slide Chemotaxis, collagen IV coated (ibidi, Munich, Germany) in compound free medium. After the cells had attached to the chamber slide, the channel was washed with medium without FCS and a gradient was generated by filling one reservoir with FCS negative medium and the other with 10% FCS medium. For negative control 10% FCS was filled in both reservoirs, which generates an environment of undirected migration. Images were taken every ten minutes and the indicated migration parameters were calculated using the chemotaxis and migration tool Qt Open Source Edition version 4.3.2.
Briefly, gels with a concentration of 2 mg/ml were mixed on ice with 0.25 × 10 6 cell/mixture. Rat tail collagen I was added at the end and immediately pipetted into ibidi TM µ-Slide Chemotaxis. For gelation, the gel was left on ice for 15 min before incubation for 1 h at 37 °C, 5% CO 2 . The chemotaxis reservoirs were filled with medium with or without 10% FCS to create a chemotactic gradient. Life cell images were taken using an inverted microscope (Eclipse Ti). Images were taken with a 10x phase contrast objective every ten minutes over 20 h and analyzed with the chemotaxis plugin in ImageJ.
HRP secretion assay. HeLa-ssHRP (signal sequence horseradish peroxidase) cells constitutively expressing a secretable form of horseradish peroxidase (HRP) 37 were seeded and pre-stimulated as described before.
After an indicated stimulation period, the medium was removed and the cells were washed five times with fresh medium. Finally, the medium was exchanged against 300 µl of fresh medium with or without compound. The secretion inhibitor Brefeldin A (Sigma Aldrich, Taufkirchen, Germany) was used as a positive control. Secretion was allowed in an incubator over 4 h, before harvesting the cells. The medium was collected as secretion probe, while cells were trypsinated (PAN Biotech, Aidenbach, Germany) and collected by centrifugation. The cell pellets were lysed with 0.5% Triton-X 100 (Merck, Darmstadt, Germany) in PBS on ice for 20 min. After an additional centrifugation at 4 °C, 14000x g for 20 min, supernatants and pellet fraction were transferred to a 96-well plate. The HRP substrate 2,2′-azino-bis((3-ethylbenzthiazoline-6-sulfonic acid), Sigma-Aldrich, Taufkirchen, Germany)) was added to the samples and the absorbance was read at 405 nm using a Tecan Microplate Reader (Tecan, Maennedorf, Austria). The signal of the supernatant was normalized to the signal in the pellet fraction.  Scientific) according to the manufacturer's instruction. Peptide mixture was then fractionated with a Trinity P1 column on a Dionex Ultimate 3000 HPLC system (both ThermoFisher Scientific) into 32 fractions. These fractions were then vacuum dried and reconstituted in 0.1% formic acid (FA) for LC-MS/MS measurement. The samples were measured on a Dionex Ultimate 3000 nanoLC (ThermoFisher Scientific) coupled to an Orbitrap Q Exactive HF mass spectrometer (ThermoFisher Scientific). NanoLC separation was carried out using an in-house packed capillary column (75 μm × 45 cm) filled with 3 μm Reprosil Gold C18 particles (Dr. Maisch GmbH, Ammerbuch, Germany) at 300 nL·min-1. Sample was loaded onto a trap column (75 μm × 2 cm) in 0.1% formic acid at 5 μL·min−1 for 10 min. The trap column was packed with 5 μm Reprosil ODS-3 particles. The analytical column was heated to 50 °C using a 30-cm capillary column heater (ASI, Pompton Plains, NJ). Solvent A was 0.1% FA, 5% DMSO in water; solvent B was 0.1% FA, 5% DMSO in acetonitrile 38,39 . For the analysis of regular peptides, the gradient was 0-1 min, 2-4% B; 1-52 min, 4-32% B; 52-54 min, 35-80% B; 54-56 min, 80% B; 56-58 min, 80-2% B; 58-60 min, 2% B. A top 25 data dependent acquisition method was used for MS. The survey scan was acquired at 60,000 resolution with a mass range of 360-1300 m/z and AGC target value of 3e6. The maximum injection time (max IT) was 100 ms. MS/MS spectra were acquired at 30,000 resolution with fixed first mass at 120 m/z. AGC target was 2e5 and max IT was 57 ms. Isolation window was 1.7 m/z and normalized collision (NCE) energy was 33. An underfill ratio of 1.0% was used and + 1, + 7 and higher, and unknown charge states were excluded. Further setting included "peptide match preferred" and the "exclude isotopes" option was turned on. Dynamic exclusion was set to 20 s.
The raw files were searched using MaxQuant (v1.5.3.30) against the UniProtKB Human Reference Proteome database (v22.07.13, 88,381 entries). Default MaxQuant parameters were used and the match-between-runs feature was enabled. The results were further processed in Perseus (v1.5.5.3). The detected proteins were grouped into ten clusters based on Pearson correlation. The clusters with down-and up-regulated proteins after stimulation for 56 h were depicted in plots with the statistic software R (R version 3.3.2). GOBPs (Gene Ontology and Biological Pathways) and KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways were plotted with a threshold of FDR < 0.5. Pathway with FDR < 0.2 (plotted in orange) and the highest enrichment values (enrichment > 2) were depicted in red with annotations.

Statistics.
The experiments were performed three times on three different days and done in triplicates each, if not otherwise noted. One-way ANOVA test with relevant posttest (Dunnett's Multiple Comparisons or Bonferroni's Multiple Comparison Test) or paired two tailed t-test were used to assess the significance between treatment groups or pairs. The statistical analysis was conducted with GraphPad Prism 5 and Microsoft Excel. Data availability. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.