NEDD9 Regulates Actin Dynamics through Cortactin Deacetylation in an AURKA/HDAC6–Dependent Manner

NEDD9 (neural precursor cell expressed, developmentally downregulated 9) is a scaffolding protein that is upregulated in multiple cancers, including invasive breast carcinomas (1, 2), metastatic melanomas (3), aggressive glioblastomas (4), and ovarian cancers (3, 5). There is a strong correlation between NEDD9 overexpression and an increase in metastasis in patients (6–8), suggesting that NEDD9 potentially affects cancer metastasis through the regulation of cancer cell migration. NEDD9 knockdown leads to impaired tumor cell migration in vitro and in vivo (9, 10). NEDD9 overexpression promotes mesenchymal-based cell movement, which is dependent on actin polymerization and matrix proteinase activity (3, 10–13). Several lines of evidence suggest that NEDD9 functions to promote cancer cell migration and invasion through the sequential phosphorylation of NEDD9 by FAK and Src (14), as well as activation of small GTPase Rac1 (4, 15, 16).

The formation of leading edge lamellipodia through dynamic cycles of regulated actin assembly is critical for the motility of cells (17). Lamellipodia formation requires actin nucleation and subsequent polymerization to generate filamentous (F)-actin networks used to propel the cell membrane forward (18). The role of NEDD9 in lamellipodia dynamics is unknown.

Cortactin (CTTN) is a lamellipodia protein that is essential for cancer cell migration (19). CTTN localizes to lamellipodia, where it binds actin-related Arp2/3 protein complex to activate actin nucleation and to stabilize resultant F-actin branch junctures (20). CTTN is acetylated by P300/CBP–associated factor (PCAF) at multiple lysine residues within the F-actin–binding region, preventing the association of CTTN with F-actin (21). CTTN deacetylation by histone deacetylase 6 (HDAC6) restores the ability of CTTN to bind to actin filaments. Hyperacetylation or loss of CTTN expression in mesenchymal cells impairs cell migration (22) through decreased lamellipodia persistence and stability (23).

HDAC6 is involved in both tumor cell migration and invasion, and is postulated to play a role in facilitating cancer cell metastasis (21, 24, 25). We have previously shown that NEDD9 binds to and activates oncogenic serine/threonine kinase Aurora A (AURKA), which in turn phosphorylates HDAC6 to increase its deacetylase activity (26). Although the role of AURKA in cell-cycle regulation is well established (27), recent work suggests that AURKA functions to promote tumor cell motility through multiple mechanisms, including phosphoactivation of the F-actin severing protein cofilin (28–29). However, additional mechanisms of regulation of the actin cytoskeleton by AURKA within lamellipodia are unknown.

In our current work, we report a new molecular mechanism underlining NEDD9-dependant migration through the regulation of CTTN. Our findings suggest that NEDD9 depletion significantly impedes the migration of breast cancer cells due to the accumulation of hyperacetylated CTTN, destabilizing actin networks at the leading edge. Overexpression of a deacetylation mimicking CTTN point mutant (9KR) is sufficient to rescue actin dynamics at the leading edge. Depletion or inhibition of AURKA or HDAC6 recapitulates the phenotype observed in NEDD9-deficient cells. In support of these observations, inhibition of AURKA with the small-molecule inhibitor MLN8237 (alisertib) or HDAC6 with Tubastatin A decreases the metastatic capability of NEDD9-overexpressing breast cancer cells in orthotopic xenografts. Collectively, these results indicate that AURKA and HDAC6 are critical effectors of NEDD9-mediated breast cancer metastasis by increasing the pool of deacetylated CTTN required for lamellipodia stability.

Authenticated cell lines MDA-MB-231, BT549, and HEK293T were purchased from the American Type Culture Collection, MDA-231-LN (Caliper Life Sciences) and grown based on the manufacturer's recommendations. NEDD9 wild-type (WT) and knockout (KO) fibroblasts, vectors expressing human full-length NEDD9 or truncation mutants of NEDD9 were previously described (30). The shRNA/siRNAs expressing constructs against NEDD9, HDAC6, AURKA, CTTN, and the control were purchased from Thermo Fisher Scientific (sequences available upon request). Mouse WT, C- and N-terminal CTTN truncation mutant constructs were previously described (31). The acetylation-null 9KR CTTN construct was generated by site-directed mutagenesis using full-length mouse CTTN. Tissue culture medium and supplements were purchased from ATCC. Cells were transfected with plasmid DNA and/or siRNA using nucleofection (Amaxa).

Migration assays were performed using BD FluoroBlok insets (BD Biosciences) as previously described (7).

Cells expressing mCherry-LifeAct (ref. 32; Addgene; 40908) were plated on glass bottom dishes (Fisher Scientific) for 12 hours; live cell imaging was conducted using a Nikon LiveScan swept-field confocal microscope with a 37°C heated chamber in L15 medium. Images were captured every 1 minute for 120 minutes. Kymograms were produced by extracting 1 pixel-width strips from each movie frame during lamellipodia extension, and then assembled using ImageJ (v1.40). At least 10 cells per treatment were analyzed; protrusion velocity, maximum extension, and persistence were calculated as described (Supplementary Fig. S1; refs. 23, 33).

Cells were grown to confluence and lysed as described (34). Cell lysates were clarified by centrifugation at 10,000 ×g for 10 minutes. The monomeric rabbit skeletal muscle actin (cytoskeleton) was polymerized to generate F-actin according to the manufacturer's protocol. F-actin cosedimentation assays were carried out according to a previously published protocol (34). Briefly, cell lysates were incubated with 30 μg of F-actin in 50 μL of F-actin binding buffer for 1 hour at 4°C. Reaction mixtures were centrifuged at 100,000 ×g for 1 hour at 4°C. Supernatants were transferred to fresh tubes and the pellets were incubated in 50 μL of G-buffer for 1 hour. Equal volumes of supernatant and pellet fractions were analyzed by Western blotting.

Cells were disrupted by a PTY lysis buffer (35) containing the protease inhibitor cocktail (Thermo Fisher Scientific). For immunoprecipitation, primary antibodies were incubated with samples overnight at 4°C and followed by the addition of the protein A/G-sepharose (G&E Healthcare Life Sciences) for 2 hours, precipitated by centrifugation, washed, and the immune complexes resolved by SDS-PAGE (35). The antibodies against Aurora A (AurA-N), NEDD9 (2G9), myc-tag (Santa Cruz Biotechnology), pan-acetyl lysine (Cell Signaling Technology), or CTTN (4F11) were used for immunoprecipitation. Western blotting was conducted using standard procedures (7) with antibodies against NEDD9 (2G9), α-tubulin, acetylated α-tubulin, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), β-actin (Sigma), Aurora A (BD Biosciences), GFP (BD Biosciences), Myc-tag (Santa Cruz Biotechnology); CTTN (Novus Biologicals), phAurora A/T288 (Cell Signaling Technology), GFP (Abcam; ab2900), and a custom-made anti-HDAC6 (#7) antibody raised against a recombinant fragment of human HDAC6 (890-1048aa; Thermo Scientific). Secondary antibodies (Jackson ImmunoResearch) and quantification methods were previously described (7).

Immunofluorescence was performed as previously described (35) using primary antibodies against CTTN (Novus Biologicals), CTTN (4F11), phTyr166-NEDD9 (custom made against a peptide encompassing amino acids 159–172 of human NEDD9 by 21st Century Biochemicals), acetylated tubulin (Sigma), HDAC6 (#7), and AURKA (7). The secondary antibodies included donkey anti-rabbit Alexa Fluor405, 488, 555, and anti-mouse Alexa Fluor647. F-actin was visualized by Rhodamine phalloidin staining (Life Technologies). Images were captured with a confocal microscope LSM510, ×63 objective, NA1.43 at 0.2–0.3 μm step (Carl Zeiss). Quantifications were performed on z-stack projections and collected using the standard acquisition parameters within ImageJ (NIH) and LSM Image analysis software (Carl Zeiss).

CTTN and F-actin immunofluorescence images were analyzed in ImageJ by applying the Z-stack at the sum of 10 projections per cell. Signal intensities for CTTN and F-actin were calculated as outlined in Supplementary Fig. S3.

Cells were lysed in a PTY buffer and clarified lysates were used for the HDAC activity assessment using the colorimetric HDAC activity assay (Upstate) according to the manufacturer's recommendations. To differentiate the effect between class I and II and class III cytoplasmic deacetylases, lysates were treated with 4 μmol/L of the HDAC class I and II inhibitor Trichostatin A (Upstate). Assays were done in triplicate and read on a Tecan GENios plate reader.

NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ/(NSG) immunodeficient mice (The Jackson Laboratory) were housed in the West Virginia University Animal Facility under pathogen-free conditions under an approved Institutional Animal Care and Use Committee protocol.

For orthotopic injections, luciferase-expressing MDA-MB-231LN were resuspended in PBS (1 × 10⁷ cells/mL), and 0.1 mL was injected into the 4th inguinal mammary gland of 6- to 8-week-old female mice. Treatment groups consisted of at least 6 mice, and experiments were repeated three times. Mice were imaged every week for quantitative evaluation of tumor growth and dissemination for a total of 4 weeks. Following the injection of 150 mg/kg of d-Luciferin (PerkinElmer) into the peritoneum, tumors were visualized using the IVIS Lumina-II Imaging System and Living Image-4.0 software (PerkinElmer). The differences in tumor growth between the groups were determined by bioluminescence imaging (BLI) and pathology measurements. All treatments (as described below) were initiated 1 week after the tumor cell injection.

NSG males, 6- to 8-week old, were intravenously injected with 1 × 10⁵ cells resuspended in 0.05 mL of PBS and followed by BLI every week for 2 weeks. The average lung radiance was estimated at each point in both the control and the treated animals. The lungs were collected at the end of the study and were fixed and pathologically analyzed for the number of pulmonary metastases by a pathologist as described below.

Collection, fixation, and determination of pulmonary metastases were conducted as described (7) by a pathologist (R.H. Livengood). The mean number of metastases per mm² ± SEM was determined from three independent experiments (n = 6–10 animals per treatment group).

Cells were treated with 5 to 10 nmol/L of MLN8054 (Selleckchem) for 2 to 12 hours, disrupted in a PTY buffer (35), and processed for Western blot analysis or immunoprecipitation.

Compound administration began 1 week after the injection of the tumor cells into the mammary fat pad of the mice. MLN8237 was reconstituted in 10% 2-hydroxypropyl-b-cyclodextrin, 1% sodium bicarbonate in water. Of note, 20 mg/kg/dose was administered via oral gavage twice a day for 4 days per week for a total of 2 weeks. For the lung metastasis studies, MLN8237 was administered in the same manner as above, starting 24 hours after the tumor cell tail-vein injections.

Cells were treated with 20 μmol/L Tubacin (Sigma), 10 to 20 μmol/L Tubastatin A (ChemScene), and 2 μmol/L ACY1215 (ChemieTek) or 1 μg/mL Trichostatin A (Sigma) for 2 to 12 hours, and disrupted in a PTY buffer (35) and processed for Western blot analysis or immunoprecipitation.

Compound administration began 1 week after the injection of the tumor cells into the mammary fat pad of the mice. Tubastatin A was dissolved to 70 mg/mL in dimethyl sulfoxide (DMSO) and brought to 10 mg/mL with PBS. A 40 mg/kg/dose was administered via intraperitoneal injection once a day for 5 days per week for a total of 2 weeks (36). For the lung metastasis studies, Tubastatin A was administered in the same manner as above, starting 24 hours after the tumor cell tail-vein injections. ACY-1215 was dissolved to 40 mg/mL in DMSO and brought to 24 mg/mL with saline. A 75 mg/kg/dose was administered via intraperitoneal injection once a day for 5 days per week for a total of 2 weeks (37).

Statistical comparisons were made using a one-way ANOVA. When more than two groups were analyzed, the two-way ANOVA was used. P ≤ 0.05 was considered to be significant. Experimental values were reported as the mean ± SEM. All calculations were made using the GraphPad InStat software.

The role of NEDD9 in the regulation of leading edge cortical actin dynamics during cell migration is currently unknown. To determine the role of NEDD9 on lamellar actin dynamics, we performed a kymography of the leading edge of highly migratory breast cancer cells (MDA-MB-231) expressing mCherry-LifeAct along with siRNAs targeting NEDD9 (siN1 and siN2) or nontargeting control (siCon; Fig. 1A and B and Supplementary Fig. S1). Depletion of NEDD9 results in an 87% increase in the protrusions velocity and a 30% decrease in the persistence/stability of protrusions compared with the control cells. No changes in the maximum extension length were observed (Fig. 1C). The decreased lamellipodia stability in NEDD9 knockdown cells corresponds with a 3-fold reduction in overall tumor cell migration (Supplementary Fig. S2A and S2B).

The defects in lamellipodia dynamics observed in NEDD9 knockdown cells closely resemble the phenotype generated by depletion of the actin-regulating protein CTTN (23). To evaluate if NEDD9-dependent changes in the lamellipodia involved CTTN, we determined the subcellular localization of both proteins. We found that NEDD9 colocalized with CTTN at the lamellar edge (Fig. 2A). Interestingly, there is a high degree of colocalization between CTTN and the active Y166-phosphorylated form of NEDD9 (Fig. 2B). To evaluate the possibility of an interaction between NEDD9 and CTTN, proteins were immunoprecipitated from MDA-MB-231 cells (Fig. 2C) and mouse embryonic fibroblasts generated from NEDD9 knockout or WT animals (Fig. 2D). The NEDD9–CTTN complex was present in all cells except for NEDD9 knockout fibroblasts, confirming the specificity of the NEDD9–CTTN association. Using the GFP-fusion truncation mutants of NEDD9 and full-length CTTN, we identified that the serine-rich region (351–654 aa) and the C-terminal (626–834 aa) domain of NEDD9 were capable of binding to CTTN (Fig. 2E), suggesting that one or more CTTN-binding sites exist between residues 351 and 834. Reciprocal experiments with CTTN truncation mutants determined that the N-terminal CTTN domain (1–350 aa) is responsible for binding to NEDD9 (Supplementary Fig. S2C).

Next, we tested whether NEDD9 expression impacts CTTN localization at the leading edge of the migrating cells. The control and NEDD9-depleted cells were stained with anti-CTTN antibody and phalloidin to visualize the filamentous actin (Fig. 3A). The distribution of the fluorescent signal corresponding to CTTN and the actin staining along the cell length was measured as outlined in the Supplementary Fig. S3. NEDD9 knockdown led to a significant reduction in the amount of CTTN localized to the leading edge of the cell without significant changes in the distribution of actin (Fig. 3B). Thus, in the absence of NEDD9, CTTN has a limited ability to reside at the leading edge. The binding of CTTN to F-actin is crucial for its localization at the leading edge as deletion or posttranslational modification of the F-actin binding region dislocates CTTN from the cell periphery and ablates cell migration (38, 39).

To determine the acetylation status of CTTN, we immunoprecipitated CTTN from the control and shNEDD9 cells. We found that depletion of NEDD9 led to an increase in acetylated CTTN, whereas total protein levels of CTTN were not affected (Fig. 4A). A similar phenomenon was previously observed in cells with the depletion or inhibition of HDAC6, known to play a key role in the deacetylation of CTTN (21). Indeed, depletion of HDAC6 increased CTTN hyperacetylation (Fig. 4B).

To measure the HDAC6 activity in NEDD9-depleted and control cells, we used a colorimetric HDAC activity assay (Fig. 4C). Cytoplasmic extracts from shNEDD9 cells had decreased deacetylase activity compared with the controls. To define whether this decrease was attributed to the decreased HDAC6 activity, both cell lysates were pretreated with Trichostatin A. Trichostatin A inhibits HDACs of class I and II, including cytoplasmic HDAC6, but not cytoplasmic deacetylase of class III (SIRT-1/2), which potentially can also deacetylate CTTN (40). The difference in deacetylation activity between the control and shNEDD9 was abolished by treatment with Trichostatin A, suggesting the involvement of HDAC6, but not SIRT-1/2, in the deacetylation of CTTN.

We previously reported that HDAC6 is a direct substrate of AURKA, and phosphorylation of the HDAC6 increases its activity (26). Similar to NEDD9 depletion, treatment with shRNAs against AURKA increases the acetylation of CTTN (Fig. 4D). Importantly, inhibition of AURKA or HDAC6 with specific inhibitors led to an increase in the acetylation levels of CTTN (Fig. 4E), indicating a potential clinical application. Hereby, expression and activity of AURKA is required for deacetylation of CTTN, even in the presence of HDAC6. The knockdown of AURKA also led to a 2-fold decrease in migration (Fig. 4F) of the MDA-MB-231 cells. The increase in the amount of acetylated CTTN in shNEDD9, shHD6, and shAURKA-expressing cells correlated with a decreased ability of CTTN to bind F-actin (Fig. 4G and Supplementary Fig. S4A).

To determine where the AURKA–HDAC6 complex gets activated, we have mapped the localization of HDAC6, CTTN, and AURKA. We found that CTTN is colocalized with AURKA and HDAC6 at the leading edge (Fig. 5A–B). HDAC6 was shown previously to colocalize with CTTN both at the cell membrane and cytoplasm (40). Here, we show that AURKA and NEDD9 colocalize with CTTN at the leading edge (Fig. 5A, Fig. 2A–B), indicating the potential importance of this complex in lamellipodia maturation. NEDD9, AURKA, and CTTN were also found to form a complexes when coimmunoprecipitated from several different cell lines (Fig. 5C); however, this is not the case in NEDD9-deficient cells, suggesting that NEDD9 plays a critical role in scaffolding this complex at the leading edge.

To test whether the decrease in the amount of deacetylated/active CTTN was the cause of the shNEDD9-driven deficiency in lamellipodia persistence, and the increase in velocity, we overexpressed the deacetylation mimicking mutant of CTTN-9KR or control vector in the shNEDD9 cells (Fig. 6A, Supplementary Fig. S5). A kymography analyses of the MDA-MB-231 cells transfected with 9KR mutant, but not in the negative control, shows the rescue of protrusion velocity and persistence to the level observed in the control cells (Fig. 6B and C), indicating that deacetylated CTTN is a direct and essential mediator of the NEDD9–AURKA–HDAC6 pathway of lamellipodia stabilization in the breast cancer cells.

To evaluate the potential clinical application of the inhibition of NEDD9-downstream targets, such as HDAC6 or AURKA, in suppressing migration and metastasis, we treated NSG female mice injected with the human MDA-MB-231LN cells with either vehicle, HDAC6 inhibitor-Tubastatin A, AURKA inhibitor-MLN8237/alisertib, or in combination. After the injection of the cancer cells, the tumors were allowed to grow for 1 week, then followed by 2 weeks of treatment. Bioluminescent images of the primary tumors within the mammary fat pads, which were acquired at the beginning (week 0) and at the end (week 2) of the treatment, showed no significant differences in the primary tumor growth among all treatment groups (Fig. 7A–B). Nevertheless, the application of Tubastatin A as single agent led to a 2-fold decrease in the number of pulmonary metastases when compared with the control (vehicle treated) animals. Next, we combined Tubastatin A and MLN8237 to test whether this combination would increase the efficacy of Tubastatin A against metastasis. The addition of the MLN8237 improved the effect of Tubastatin A, indicating a potential additive effect (Fig. 7C–D). A complementary analysis of the pulmonary metastases by ex vivo lungs BLI supports the pathology data (Supplementary Fig. S6A–S6B). We would also like to note that the application of the MLN8237 alone was very efficacious in comparison with the vehicle or other treatment groups, indicating that the combination of the antiproliferative and antimigrative action of the MLN8237 could be sufficient to prevent and treat metastasis, whereas the inhibition of HDAC6 could potentially only decrease migration of the tumor cells.

Next, we used intravenous injection of the MDA-MB-231LN cells to determine the impact of HDAC6 inhibitors on the lungs colonization by the circulating tumor cells (CTC). To inhibit HDAC6, we have used two specific inhibitors: (i) Tubastatin A, which is widely applied in in vitro and in vivo studies, and (ii) ACY-1215, a small-molecule inhibitor and currently in clinical trials for myeloma treatment (37). On the basis of the BLI data (Fig. 7F) and the pathology reports on the dissected lungs (Fig. 7E–G), tumor cells in circulation and the lung parenchyma were extremely sensitive to both Tubastatin A and ACY-1215 and had a limited capacity to colonize the lungs when compared with the vehicle-treated cells. The efficacy of the inhibitors against HDAC6 was confirmed by immunofluorescence and Western blot analysis of the tumor cells looking for acetylation of tubulin and CTTN (Supplementary Fig. S6C and S6D). Application of Tubastatin A, which has minimal, if any, toxicity, decreased the ability of breast cancer to disseminate in xenograft models, suggesting a potentially successful clinical application of anti-HDAC6 compounds as an antimetastatic agent. We also tested the efficacy of combination Tubastatin A and MLN8237 on the lungs colonization by CTCs. The addition of the MLN8237 improved by 2-fold the effect of Tubastatin A, indicating an additive effect (Supplementary Fig. S6E–S6G).

NEDD9 is a scaffolding protein highly expressed in many cancers and is associated with progression toward metastasis (1–9). However, little is known about the exact molecular mechanisms of NEDD9-driven metastasis. The ability of NEDD9 to regulate the migration was previously documented (10, 11, 13–16); however, the impact of NEDD9 in the regulation of cortical actin dynamics, particularly of lamellipodia-based protrusions, is unknown. In agreement with earlier published data (4, 41, 42), cells depleted of NEDD9 showed up to a 3-fold decrease in migration. The effect of NEDD9 depletion on migration is analogous to depletion of the well-known promigratory regulators CTTN and HDAC6 (21). Although the role of CTTN in cell migration has been well outlined, the critical components affecting its activity are just starting to emerge (19–21, 23, 43, 44). In our current study, we define a new signaling cascade responsible for the increased tumor cell migration via the regulation of CTTN deacetylation and AURKA-dependent HDAC6 activation. We also provide evidence of the binding of CTTN to the NEDD9–AURKA complex at the leading edge, where CTTN is known to bind HDAC6, thereby bringing the key players of this novel pathway together.

Lamellipodia stability is greatly impaired upon NEDD9 depletion, resulting in a dramatic effect on the dynamics of the leading edge of the cell, with a 30% decrease in persistence of protrusions and an 87% increase in the velocity of membrane undulations. Similar effects were shown upon knockdown of the CTTN in a fibrosarcoma cell line (23).

We found that NEDD9 colocalizes and binds to CTTN within lamellipodia, where CTTN promotes actin nucleation and stabilization (45). In addition, we determined that the carboxy-terminal region of NEDD9 (351–834 aa) is responsible for the binding of NEDD9 to CTTN. CTTN was observed to be delocalized from the leading edge of the cell upon NEDD9 depletion, indicating the importance of NEDD9 expression for retaining CTTN at the cell periphery. Previously, dislocation of CTTN from the leading edge of the cell was described to be a result of CTTN being hyperacetylated (21, 46).

According to our findings, NEDD9 influences the acetylation status of CTTN. Upon NEDD9 depletion, CTTN is hyperacetylated. NEDD9, as shown previously, is required for the activation of its downstream target, AURKA (35), which is an essential effector promoting breast cancer cell invasion/migration (29). Depletion or inhibition of AURKA led to an increase in the acetylation of CTTN as well. The resulting phenotype in all cases resembled the effect of knockdown or the inhibition of HDAC6, which is responsible for deacetylating CTTN (21). The deacetylase activity of HDAC6 was previously shown by our group to be upregulated upon AURKA phosphorylation (26). The increase in CTTN acetylation upon either AURKA knockdown or inhibition suggests that not only is the presence of AURKA necessary, but, in addition, the AURKA activity is required for the deacetylation of CTTN, therefore supporting a role of AURKA phosphorylation in the regulation of HDAC6 activity toward CTTN.

The NEDD9 knockdown phenotype was rescued by a mutant variant of CTTN, which mimics the deacetylated CTTN (9KR; ref. 21). Reexpression of the 9KR mutant restored the normal actin dynamics at the leading edge of the NEDD9-depleted cells. Both the absence of CTTN (23) and, as our study indicates, hyperacetylation of CTTN triggers reduced persistence of lamellipodia protrusions. We suggest that an increase in the acetylation of CTTN, caused by NEDD9 depletion, leads to unstable protrusions formation, which is explained by the inability of the acetylated CTTN to bind F-actin to stabilize the actin branches (21). A reduction in the lamellipodial protrusion stability theoretically can be caused not only by an increased instability of the actin network, but also through a reduction of stable focal adhesions. It is known that the depletion of CTTN leads to a decrease in the rate of focal adhesion assembly (23); however, there is no indication that CTTN hyperacetylation can cause such an effect. To the contrary, there is evidence that an increase in the CTTN acetylation leads to stabilization of the focal adhesions (46).

NEDD9, a scaffold protein, forms a complex with CTTN and AURKA at the leading edge of the cell. Previous studies have shown that HDAC6 colocalizes and directly binds CTTN at the leading edge (21). Together, these findings suggest that NEDD9 brings CTTN, AURKA, and HDAC6 into a single complex at the cell's leading edge, where HDAC6 can be subjected to phosphorylation and activation by AURKA, a critical step allowing for HDAC6 to deacetylate CTTN (26). Collectively, our data suggest that NEDD9 is an upstream regulator of CTTN at the lamellar actin subdomain, which acts through the AURKA-dependent activation of HDAC6.

Finally, we tested whether targeting HDAC6 and AURKA would reduce breast cancer cell metastasis. AURKA inhibitors (MLN8237/alisertib, PHA680632/danusertib) are currently in phase II and III clinical trials (47, 48). Because the increased expression of NEDD9 correlates highly with the overexpression of AURKA and metastatic disease in the human breast cancer (7), HDAC6 inhibitors could be used in patients with NEDD9-overexpressing tumors. Selective HDAC6 inhibitors are a promising group of anticancer compounds. One of which, ACY-1215, is currently in phase I and II clinical trials for the treatment of multiple myeloma by targeting the HDAC6-dependent proteasomal degradation pathway (37, 49). In addition, it has been shown that the chemical inhibition of HDAC6 selectively reduces the amount of AURKA in the cell via degradation (50). Therefore, we hypothesized that simultaneous administration of HDAC6 and AURKA inhibitors might have a synergistic effect on metastasis. This approach was shown to be promising in the mouse orthotropic mammary fat pad xenograft experiments using the breast cancer cell line MDA-MB-231LN, and representing a triple-negative breast cancer with high NEDD9 expression (7). HDAC6 inhibition, by the specific inhibitor Tubastatin A, in an orthotopic breast cancer model was able to decrease pulmonary metastasis, but not the primary tumor growth, similar to the AURKA inhibitor MLN8237. A combination of MLN8237 and Tubastatin A produced an intermediate phenotype, suggesting potential additive inhibitory effects of MLN8237. The application of the AURKA inhibitor alone was more efficient in eradicating metastases than the HDAC6 inhibitor. Further work is required to elucidate the lack of synergism in a combinatorial setting.

Using tail-vein injection as a model for tumor colonization, comparable efficacy in reducing breast cancer cell metastasis in lungs was achieved for both HDAC6 inhibitors, ACY-1215 and Tubastatin A. Interestingly, combination of Tubastatin A and MLN8237 proved to be very efficient in eradication of metastases, improving the efficiency of Tubastatin A by 2-fold. Therefore, HDAC6 inhibitors may be used as a comparable alternative to AURKA inhibitor therapy for treatment of invasive breast cancer that may be resistant to AURKA inhibitors.

No potential conflicts of interest were disclosed.

This article contains original work only and has not been published nor submitted elsewhere. All authors have directly participated in the planning, execution, and analysis of this study and approved the submitted version of this article.

Conception and design: V.K. Kozyreva, S.A. Weed, E.N. Pugacheva

Development of methodology: V.K. Kozyreva, L.C. Kelley, E.N. Pugacheva

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V.K. Kozyreva, S.L. McLaughlin, R.H. Livengood, R.A. Calkins, L.C. Kelley, A. Rajulapati, R.J. Ice, E.N. Pugacheva

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V.K. Kozyreva, S.L. McLaughlin, A. Rajulapati, R.J. Ice, M.B. Smolkin, E.N. Pugacheva

Writing, review, and/or revision of the manuscript: V.K. Kozyreva, S.L. McLaughlin, R.H. Livengood, S.A. Weed, E.N. Pugacheva

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V.K. Kozyreva, M.B. Smolkin, E.N. Pugacheva

Study supervision: E.N. Pugacheva

The authors thank Dr. Mikael Schaller (West Virginia University, WVU) and Mrs. Lana Yoho for their outstanding administrative support, in addition to the WVU Tissue Bank and Animal Care Facility.

The Animal Models and Imaging Facility and the Microscope Imaging Facility were supported by the MBRCC and the NIH grants P20 RR016440, P30 RR032138/GM103488, and S10RR026378. This work was supported by a grant from the NIH-NCI (CA148671 to E.N. Pugacheva), a Susan G. Komen for the Cure Foundation (KG100539 to E.N. Pugacheva) and in part by a NIH/NCRR (5 P20 RR016440-09 to the Mary Babb Randolph Cancer Center) award.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.