PAK4 methylation by the methyltransferase SETD6 attenuates cell adhesion

P21-activated kinase 4 (PAK4), a member of serine/threonine kinases family is over-expressed in numerous cancer tumors and is associated with oncogenic cell proliferation, migration and invasion. Our recent work demonstrated that the SET-domain containing protein 6 (SETD6) interacts with and methylates PAK4 at chromatin in mammalian cells, leading to activation of the Wnt/β-catenin signaling pathway. In our current work, we identified lysine 473 (K473) on PAK4 as the primary methylation site by SETD6. Methylation of PAK4 at K473 activates β-catenin transcriptional activity and inhibits cell adhesion. Specific methylation of PAK4 at K473 also attenuates paxillin localization to focal adhesions leading to overall reduction in adhesion-related features, such as filopodia and actin structures. The altered adhesion of the PAK4 wild-type cells is accompanied with a decrease in the migrative and invasive characteristics of the cells. Taken together, our results suggest that methylation of PAK4 at K473 plays a vital role in the regulation of cell adhesion and migration.


SETD6 methylates PAK4 at lysine 473 in-vitro and in cells.
To map PAK4 methylation site, we mutated several lysine residues in PAK4 protein that are highly conserved among 4 different organisms (Homo sapiens, Mus musculus, Danio rerio and Drosophila melanogaster): lysine 31, lysine 51, lysine 350, lysine 442 and lysine 473 (Supplementary Fig. S1A and Fig. 1A). In addition, we mutated lysine 78, which was identified before to be methylated in a proteomic screen at protein modification database PhosphoSitePlus (https ://www.phosp hosit e.org) 20 . The purified recombinant mutants or wild-type PAK4 proteins were subjected to in-vitro methylation assay with recombinant SETD6 (Supplementary Fig. S1B). Out of the different mutants that were tested, only the PAK4 K473R mutant showed a significant and repeatable decrease in methylation signal by SETD6 (Fig. 1B). In these methylation assays ( Fig. 1B and Supplementary Fig. S1B) SETD6 was auto-methylated, which is consistent with our previous knowledge describing the enzymatic activity of SETD6 [21][22][23] . We tested the methylation of PAK4 K473R mutant also in cells, using a pan-methyl antibody that identified methylated wild-type Flag PAK4 but not the K473R mutant (Fig. 1C). Together, these data suggest that SETD6 primarily methylates PAK4 at lysine 473 in-vitro and in cells.
Methylated PAK4 at lysine 473 upregulates β-catenin protein levels and Wnt/β-catenin target genes. Based on these data and our previous findings 13 , we hypothesized that the methylation of PAK4 at K473 mediates the activation of β-catenin. To test this hypothesis, we generated MDA-MB-231 cells stably expressing Flag PAK4 wild-type or Flag PAK4 K473R mutant that cannot be methylated by SETD6 ( Fig. 2A). Our results demonstrate that β-catenin is upregulated (total and active forms) in the presence of wild-type but not the K473R mutant in MDA-MB-231. A reduction in the β-catenin S675 phosphorylation signal was also noted upon stable over-expression of the PAK4 K473R mutant. Consistent with these findings, we performed a quantitative FACS analysis in MDA-MB-231 cells and found that active β-catenin level was increased in PAK4 wild-type, but not in PAK4 K473R stably expressing cells ( Supplementary Fig. S2A). Furthermore, isolation of the chromatin fraction revealed that the level of active β-catenin at chromatin was elevated in cells stably expressing PAK4 wild-type compare to PAK4 K473R ( Supplementary Fig. S2B), suggesting a direct regulation of gene target expression. In order to test whether these findings are specific to MDA-MB-231 cells, we examined these phenomena in the hormone dependent (estrogen and progesterone) breast adenocarcinoma cell line  55 for Homo sapiens, Mus musculus, Danio rerio and Drosophila melanogaster PAK4 protein sequences. (B) In-vitro methylation assay. Recombinant His-Sumo-PAK4 wild-type (wt) or the His-Sumo-PAK4 K473R mutant were incubated with or without His-SETD6 in the presence of 3 H-labeled SAM. Proteins were then subjected to SDS-PAGE followed by exposure to autoradiogram to detect 3 H-labeled proteins or Coomassie staining to detect all proteins. (C) Methylation assay in cells. MDA-MB-231 wild-type cells were transfected with Flag PAK4 wild-type or Flag PAK4 K473R, and both with HA SETD6 plasmids. Cell lysates were immunoprecipitated (IP) with FLAG-M2 beads, and proteins in IP and input samples were detected by Western blot with indicated antibodies. Methylation was detected with pan-methyl antibody. Uncropped gels are shown in Supplementary Fig. S9 S3A). Our previous findings indicate that depletion of SETD6 correlates with a significant reduction in the expression of some known Wnt/β-catenin target genes 13 . We therefore tested the expression levels of Wnt/β-catenin target genes by qPCR in MDA-MB-231 and MCF-7 cells. Our results demonstrate that while the expression levels of Wnt/β-catenin target genes were elevated in PAK4 wild-type cells, no change or a decrease in their expression was observed in MDA-MB-231 cells stably expressing PAK4 K473R mutant (Fig. 2B). We noted significant changes in the expression of Wnt cell-adhesion-related genes. The expression of NRCAM 24 , FN1 25 and LAMC2 25 genes was significantly upregulated in cells stably expressing wild-type PAK4 and their expression levels were reduced to a basal degree in the K473R mutant cells, in which PAK4 is not methylated by SETD6. The reduced mRNA levels of IGFBP7 26 and L1CAM 25 following stable expression of wildtype PAK4 expressing cells, were partly restored in cells stably expressing mutant PAK4 (Fig. 2B). We did not see significant changes in the expression level of most of the genes in the control MCF-7 stable cells ( Supplementary  Fig. S3B). Taken together, our data suggest that specific methylation of PAK4 at K473 orchestrates the expression of the Wnt/β-catenin target genes, among them adhesion-related genes. These findings may also indicate that PAK4 methylation plays a role in the regulation of cell adhesion process.

PAK4 methylation at K473 decreases cells adhesion.
Both PAK4 and β-catenin were linked before to the regulation of cell adhesion 8,25 . We therefore hypothesized that the methylation of PAK4 at K473 by SETD6 might be involved in the regulation of cell adhesion. A first indication for that link came from a set of experiments where we noticed that cells stably expressing PAK4 wild-type were easily detached from the cell culture plates upon serum starvation. Using crystal violet staining we observed that MDA-MB-231 cells expressing a methylateable PAK4 were easily washed from the plate surface, in contrast to cells where PAK4 is not methylated (stable PAK4 K473R or SETD6 knock-out cells), which were relatively well anchored to the surface (Supplementary Fig. S4). Consistent with these observations, cells stably expressing Flag PAK4 wild-type demonstrated a significant loss of binding to fibronectin-coated plates, in comparison to empty or Flag PAK4 K473R expressing MDA-MB-231 cells which remained adherent to the fibronectin-covered surface of the plates (Fig. 3A).

PAK4 methylation at K473 diminishes the formation of filopodia and actin structures.
Having demonstrated that PAK4 methylation at K473 negatively regulates cell adhesion, we next examined the morphology of the cells using Vybrant™ DiI Cell-Labeling Solution (staining cell membranes) and Hoechst (staining DNA and nuclei) followed by visualization by confocal microscopy (Fig. 3B, left panel). While clear filopodia structures were observed in empty and PAK4 K473R cells, the PAK4 wild-type cells displayed a smooth membrane staining suggesting reduced adhesive properties 27 . Consistent with these results, the formation of actin structures-a feature tightly linked with cell adhesion 28 and visualized using phalloidin (F-Actin) staining-was noted in cells stably expressing empty plasmid and the mutant PAK4 (Fig. 3B, right panel). Taking together, these results suggest that PAK4 methylation at K473 leads to reduction in cell adhesion and to altered adhesion-related structures.
In order to verify that the observed phenotypes are SETD6 dependent, we performed a rescue experiment using MDA-MB-231 CRISPR SETD6 knock-out cells stably expressing Flag PAK4 wild-type or Flag PAK4 K473R in the presence or absence of stably expressed Flag SETD6 complementation ( Supplementary Fig. S5A shows the protein expression levels). The fibronectin-dependent cell adhesion was lost in the PAK4 wild-type expressing cells when SETD6 is restored in these cells (Fig. 4A). As expected, the cells expressing PAK4 K473R remained www.nature.com/scientificreports/ adherent, with or without SETD6 (Fig. 4A). Consistent with these results, we could identify a significant decrease in actin and filopodia structures in the cells upon the re-addition of SETD6 ( Fig. 4B and Supplementary Fig. S5B).
In addition, rescue with SETD6 led to upregulation of Wnt/β-catenin target genes in cells stably expressing PAK4 wild-type but not in cells expressing PAK4 K473R mutant ( Supplementary Fig. S5C). Collectively, these data suggest that PAK4 methylation at K473 by SETD6 negatively regulates cell adhesion morphologies.
PAK4 K473me attenuates paxillin localization to focal adhesions. The assembly of actin structures, such as actin protrusions as well as actin stress fibers, is a cellular feature involved in the generation of focal adhesions 28 , the latter required for cell attachment, spreading and migration 29 . Paxillin, a main component of focal adhesions, was previously shown to be phosphorylated at serine 272 (S272) and activated by PAK4 to stimulate the disassembly of focal adhesions in cells 8,19 . Moreover, it was demonstrated that knock-down of PAK4 in MDA-MB-231 cells resulted in higher cell adhesion to fibronectin due to a lower turnover rate of focal adhesions (greater number and size of focal adhesions in the cells) 19 . We therefore postulated that PAK4 methylation by SETD6 at K473 will attenuate paxillin localization to focal adhesions, leading to defects in focal adhesion assembly. To test this hypothesis, we examined the localization of paxillin by immunofluorescence in MDA-MB-231 cells stably expressing empty, Flag PAK4 wild-type or Flag PAK4 K473R mutant. As shown in Fig. 5A, we found that while in empty and PAK4 K473R mutant cells, paxillin is accumulated at focal adhesions at the periphery of the cells, it is less present on the actin structures in the PAK4 wild-type cells. Consistent with these findings, cells stably expressing Flag PAK4 wild-type exhibited elevated levels of paxillin S272-ph than empty or Flag PAK4 K473R cells ( Supplementary Fig. S6A). Overall paxillin protein levels in cells stably expressing empty, Flag PAK4 wild-type or Flag PAK4 K473R, remained unaffected ( Supplementary Fig. S6B). We next hypothesized that the methylation of PAK4 by SETD6 may affect the physical interaction between PAK4 and paxillin. The interaction observed between over-expressed Flag PAK4 wild-type and endogenous paxillin in MDA-MB-231 SETD6 knock-out cells, was enhanced upon over-expression of HA SETD6 (Fig. 5B). This increased binding was observed only with Flag PAK4 wild-type and not with Flag PAK4 K473R mutant (Fig. 5B, compare lanes 2 and www.nature.com/scientificreports/ 3). Taken together, these results suggest that methylation of PAK4 by SETD6 at K473 enhances the interaction between PAK4 and paxillin and attenuates paxillin localization to focal adhesions, leading to the disassembly of focal adhesions in cells.

PAK4 K473me inhibits cell migration and invasion.
Given the close connection between focal adhesions and cell migration 30,31 , we hypothesized that the changes in cell adhesion and focal adhesion formation might also affect cell migration. To test this, we monitored the migration of cells stably expressing empty, PAK4 wild-type or PAK4 K473R mutant in a wound healing assay, where a scratch is produced on a confluent monolayer of cells and the distance the cells made towards the gap is measured. Cells stably expressing PAK4 wildtype migrated significantly slower to close the gap than the PAK4 K473R or empty plasmid expressing cells (Fig. 6A). To get a better insight into the differences in the migration patterns of these cells, we examined the migration of single cells under the same experimental setup. The acquired images were stacked into time-lapse movies using Fiji 32 , and analyzed using TrackMate (a Fiji plugin 33 ) to monitor cell movement. Calculated migration trajectories of empty and PAK4 K473R mutant cells exhibited significantly higher parameters of cell displacement, travel distance and straight-line speed of the cells, in comparison to PAK4 wild-type cells (Fig. 6B, Supplementary Fig. S7 and Movies S1-6). In contrast, cells stably expressing PAK4 wild-type travelled overall a shorter distance, moved a shorter distance from the start point to end point (displacement) and moved more slowly in a straight-line compare to empty or PAK4 K473R mutant cells. Interestingly, the morphology of the analyzed single cells visualized in Brightfield microscope, correlated well with actin structures and focal adhesion features of fixed cells shown in Figs. 3B, 4B and 5A. It seems that methylated PAK4 at K473-resulting in the impairment of focal adhesions and actin structures assembly-disrupts cell migration and therefore affects the velocity and the traveled distance of the cells (as observed in the single cell Movies S1-6). To verify that the differences are a result of cell migration and not cell proliferation, we tested the proliferation of the cells at the same conditions (Supplementary Fig. S8) and could not observe any significant difference between the different cell lines.  www.nature.com/scientificreports/ Migration is known as the movement of cells on 2D surfaces, however this characteristic is connected with cellular ability to move within 3D matrices, known as invasion. These migration findings led us to hypothesize that PAK4 K473 methylation by SETD6 affects cell invasion. Indeed, in a transwell invasion assay, where cells incubated on matrigel-coated transwell chambers were allowed to invade the membrane for 24 h, cells stably expressing empty or PAK4 K473R invaded the membrane, while PAK4 wild-type expressing cells presented impaired invasiveness (Fig. 6C). Altogether, our data support a model by which PAK4 methylation at K473 alters the generation of adhesion-related characteristics, such as filopodia and actin structures and focal adhesions. Methylated PAK4 interacts with paxillin and enhances paxillin phosphorylation at S272 which attenuates paxillin localization to the periphery of the cells and stimulates the disassembly of focal adhesions. Paxillin mislocalization leads to severe defects in cell adhesion, migration and invasion.

Discussion
The serine/threonine kinase PAK4 has many roles in the regulation of diverse oncologically-related cellular pathways and phenotypes in several types of cancers such as; gastric 34 , prostate 35 , ovarian 36 , colon 37 and breast [38][39][40] . In most cases, the kinase activity of PAK4 and the phosphorylation of its downstream substrates is required to turn these pathways on and off [34][35][36][37][38] . Here we discovered that PAK4 cellular activity is controlled by lysine methylation in breast cancer cells. We show that SETD6-mediated PAK4 methylation at K473 stimulates the activation of Wnt/β-catenin target genes and inhibits cell adhesion and migration. Formation and stabilization of cell-matrix adhesion have a vital role in cell migration 41,42 and they are important components of cell invasion during cancer progression 43 . Similarly, in our study, we show that loss of adhesion in cells stably expressing PAK4 wild-type correlates with poor ability of the cells to migrate both in a population and in single cell-based assays which result in reduced invasion.
Our data display a set of phenotypes partially overlapping those published before for PAK4. It was shown that PAK4 positively controls focal adhesions 8,27 and downregulation of PAK4 leads to elevated adhesion to fibronectin and a decreased ability to migrate in breast cancer MDA-MB-231 cells 19 . In MCF-7 breast cancer cells, PAK4 expression increased migration of the cells on vitronectin, another extracellular matrix protein, and inhibited cell spreading and adhesion to vitronectin by stimulating integrin avβ5 and specifically phosphorylating integrin  www.nature.com/scientificreports/ transformation 27 . The molecular mechanisms that govern these contradicting phenotypes, including the mRNA differential expression of β-catenin cell adhesion-related genes, remain unclear. However, one would speculate that redundant and compensating mechanisms exist to tightly regulate these delicate cellular functions. While our study focuses on PAK4 methylation in breast cancer cells, it is known that PAK4 expression is associated with promoting cell migration and invasion in other cancers as well 35,36,47 and it remains an open question if PAK4 activity in these cancers is also regulated by SETD6. We demonstrated that PAK4 methylation at K473 attenuates paxillin localization to focal adhesions as well as overall formation of filopodia structures, leading to reduced cell actin structures. Paxillin phosphorylation at S272 by PAK4 increases the turnover rate of focal adhesions 8 . Here, we show that the recruitment of paxillin to focal adhesions at the periphery of the cells depends on the methylation of PAK4 at K473. While the precise mechanism by which PAK4 methylation regulates paxillin phosphorylation and localization is still not fully clear, our data suggests that PAK4 methylation stabilizes the physical interaction with paxillin. One can speculate that this interaction may facilitate a direct or indirect recruitment of other proteins (methyl lysine binders) which stabilize the complex. The fact that PAK4 autophosphorylation occurs at the activation loop at S474 4 which is adjacent to the K473, methylated by SETD6, can also provide an elegant methyl/phospho based mechanism to explain PAK4 effect on paxillin and presumably additional adhesion-related proteins.
In recent years, the link between lysine methylation and the regulation of oncogenically-related phenotypes such as adhesion, migration and invasion has been studied in depth in the context of histones [48][49][50] . However, the role of non-histone protein methylation is now emerging to be pivotal in the regulation of these processes. It was shown before that talin1, a cytoskeletal protein vital for cell adhesion and migration, is tri-methylated by the lysine methyltransferase Ezh2. Methylation of talin1 by Ezh2 leads to protein cleavage and to disruption of talin1 interaction with F-actin, leading to disassembly of focal complexes 51 . Additionally, loss of the methyltransferase Ezh2 in dendritic cells was reported to cause formation of large focal adhesions and impaired the migration properties of the cells 51 . It still remains to be determined if SETD6 and Ezh2 are regulating cell adhesion in a similar or distinct manner, but nevertheless, these findings highlight the importance of protein lysine methyltransferases and lysine methylation signaling in the regulation of cellular processes such as cell adhesion and migration.
Plasmids. PAK4 and SETD6 sequences were amplified by PCR and were subcloned into pcDNA3.1 3×Flag and pcDNA3.1 3×HA plasmids. For viral infections, PAK4 and SETD6 were cloned into pWZL-Flag plasmid. For recombinant protein purification, PAK4 and SETD6 were cloned into pET-Sumo and pET-Duet plasmids, respectively. To generate PAK4 mutants, site-directed mutagenesis on PAK4 wild-type vector was performed using the following primers; Recombinant protein expression and purification. Escherichia coli Rosetta transformed with a plasmid expressing His-or His-Sumo tagged PAK4 wild-type, PAK4 mutant variants or SETD6 were grown in LB medium. Bacteria were harvested by centrifugation after IPTG induction and lysed by sonication on ice (25% amplitude, 1 min total, 10/5 s on/off). His-tagged proteins were purified using Ni-NTA beads (Pierce) or on a HisTrap column (GE) with the ÄKTA gel filtration system. Proteins were eluted by 0.5 M imidazole followed by dialysis to 10% glycerol in phosphate-buffered saline (PBS). Recombinant GST-SETD6 was expressed and purified as previously described 52 . In-vitro methylation assay. Methylation  Chromatin extraction. To extract chromatin fraction, biochemical fractionation was performed as previously described 53 , with the addition of another step. In brief, the chromatin pellet was resuspended in RIPA buffer including 1 mM MgCl 2 and benzonase nuclease enzyme (Sigma) and incubated on ice for 30 min. The supernatant was collected by low-speed centrifugation (5 min, 1700 × g) at 4 °C. The soluble chromatin was resolved by SDS-PAGE gel and analyzed by Western blot.
Cell migration. For cell migration (wound healing assay), cells were serum starved (0.5% FBS) over-night, then 3.5 × 10 5 cell/well were plated on fibronectin pre-coated 48-well plate. After 3.5 h, the scratch wound was produced by dragging a 200 μl pipette tip across the layer of cells. The migrated distance of the cells towards the gap was monitored by a Lionheart TM FX Automated Microscope (4×). For single cell migration assay, 5,000 cells/well were plated (in a 48-well plate), serum starved (0.5% FBS) for over-night, then the media was changed to complete media with Hoechst stain. Cell movement was monitored every 5 min for 4 h by Lionheart TM FX Automated Microscope (20×) and analyzed using TrackMate (Fiji) software 33 . Cell proliferation. For the cell proliferation assay, cells were serum starved (0.5% FBS) over-night, then 7.5 × 10 3 cell/well were plated on a fibronectin pre-coated 96-well plate. Cell proliferation was monitored by a Lionheart™ FX Automated Microscope (4×) every 2 h.
Finally, stained cells were visualized by confocal laser microscope ZEISS LSM880 Airyscan using the Plan Apochromat 63×/1.4 oil DIC M27 or Plan Apochromat 40×/1.4 Oil DIC M27 objectives 54 . Percent cells with filopodia or actin structures were manually counted. Paxillin area size was analyzed using the "YEN" threshold tool in the Fiji software 32 . For the quantification of paxillin S272-ph intensity, all images were converted to 8-bit grayscale. Following which, cell borders were traced using the wand tracing tool in Fiji and integral fluorescence density for corresponding channel was calculated within the defined area and normalized to the integral fluorescence density of DAPI staining obtained in the same area.
Invasion assay. The invasion assay was performed using ThinCerts inserts, chambers with 8 µm pore membranes (Greiner bio-one, 662638). The inserts were coated with ~ 30 µg Matrigel (Corning Matrigel Basement Membrane Matrix, Phenol Red-free, *LDEV-free, Product Number 356237) and dried for 2 h at 37 °C. Overnight serum starved cells were harvested and 5 × 10 4 cells in 0.5% FBS DMEM were added to the upper side of the inserts, and the wells were filled with 10% FBS DMEM. After 24 h, the non-invasive cells on the upper side of the inserts were removed using cotton swabs. The invaded cells were fixed and stained using the Dipp Kwik Differential Stain kit (American Mastertech Scientific). The stained cells were imaged by EVOS FL Cell Imaging System (Thermo Fisher Scientific) using 10 × objective. The images were analyzed using Fiji software 32 .

Statistical analysis.
Statistical analyses for all assays were analyzed with GraphPad Prism software, using Student's two-tailed t-test (unpaired), one-way analysis of variance (ANOVA) with a Tukey's post-test or twoway ANOVA followed by Bonferroni's post-test.