pBABE-Flag-TPPP1 [6] , pBABE-Flag-LIMK1, pBABE-FLag-LIMK1-DN [12] , myc-cofilin, myc-cofilin S3A and myc-cofilin S3D [13] plasmids were generated as previously described. pcDNA3-Flag-HDAC6 was obtained from Addgene and was originally cloned as described [14] .

U2OS (human osteosarcoma) cells were cultured and maintained in Dulbecco’s Modified Eagle Media (DMEM) supplemented with 10% Fetal Bovine Serum (FBS) at 37˚C in a humidified, 5% CO₂ atmosphere. Flag-LIMK1, Flag-LIMK1-DN, Flag-HADC6, myc-cofilin, myc-cofilin S3A, myc-cofilin S3D constructs were transiently transfected into U2OS cells using the Lipofectamine™ 2000 (Life Technologies) transfection reagent according to manufacturer’s recommendations. Cells expressing Flag-TPPP1 were generated as previously described [9] .

The following chemicals were obtained and used in this study: Y-27632 (Calbiochem), Paclitaxel (Cytoskeleton), Vincristine (Gift of Dr Maria Kavallaris, Lowy Cancer Research Centre, Sydney) and Trichostatin A (Gift of Dr Boris Sarcevic, St. Vincent’s Institute, Melbourne). Treatment of cells with Y-27632 (10 mM), Paclitaxel (500 nM), Vincristine (100 nM) and Trichostatin A (100 nM) were performed for 16 hours.

Cells were transfected with hTPPP1 ON-TARGETplus SMARTpool, hLIMK1 siRNA 5’-TGGCAAGCGTGGACTTTCA-3’ oligonucleotides or non-targeting siRNA (Dharmacon) using the Lipofectamine™ 2000 (Life Technologies) transfection reagent according to the manufacturer’s recommendations.

U2OS cell extracts (500 μg) were pre-cleared with ~2 μg of the appropriate isotype control antibodies (anti-mouse, rabbit or rat IgG) and 30 μL of Protein A (Rabbit IgG) or Protein G (Rat and Mouse IgG) sepharose beads for 2 hours at 4˚C. Cleared lysates were incubated with ~2 μg of isotype control, anti-HDAC6 (Sigma-Aldrich), anti-TPPP1 [6] or anti-LIMK1 [12] antibodies and Protein A/G sepharose overnight at 4˚C with rotation. Immunoprecipitated (IP) proteins were resolved by SDS-PAGE and subjected to immunoblot analysis.

Immunoblotting was performed as previously described [9] . Antibodies against phospho-cofilin (Ser3; #3311), phospho-MLC (Ser18, Thr19; #3674) and GAPDH-HRP (#3683) were obtained from Cell Signaling. Anti-a-tubulin (T5168), acetyl-a-tubulin (Lys40; T6793) and HDAC6 (SAB1406911) antibodies were purchased from Sigma-Aldrich. Additional antibodies used were TPPP1 [6] , LIMK1 [12] , LIMK2 [15] , Flag 9H1 clone [16] and c-myc tag clone 4A6 (Upstate). Immunoblot quantification was performed using the ImageQuant software. Data are expressed as mean ± S.E.M and sample groups were analyzed by unpaired two-tailed t-tests.

Immunofluorescence was performed as previously described [9] . Cells were incubated with anti-acetyl-a-tubulin (1:1000; Sigma-Aldrich) followed by anti-mouse IgG Alexa Fluor 488 (1:400; Life Technologies), anti-atubulin-FITC (1:400; Sigma-Aldrich, F2168) or Phalloidin Alexa Fluor 488 (1:400; Life Technologies, A12379) and Hoechst (1:10,000; Life Technologies). Images were acquired using a Zeiss AxioObserver LSM 780 inverted microscope with a 63x/1.46 oil immersion DICIII lens. Files were extracted using the Image J (v1.47) software. Image overlay was performed with the Adobe Photoshop (v11.0.2) software.

Cells were plated in triplicate on 96-well plates and 24 hours later they were incubated with the MTT reagent (CellTiter 96^® AQueous Non-Radioactive Cell Proliferation Assay kit (Promega)) for 2 hours at 37˚C in 5% CO₂. Converted MTT dye was analyzed at OD490 nm (λ 490) using a spectrophotometer plate reader (Titertek Multiskan Plus, Lab Systems). Data are expressed as mean ± S.E.M and sample groups were analyzed by unpaired two-tailed t-tests.

Previous studies reported that the LIM-kinase (LIMK) family members LIMK1 and LIMK2 interact with the microtubule regulatory protein Tubulin Polymerization Promoting Protein 1 (TPPP1) [6] [8] [17] . Contrary to the initial data indicating that TPPP1 is an in vitro LIMK substrate [6] [17] , we established that the LIMKs do not phosphorylate TPPP1 and that it is a substrate of ROCK in cells [9] . TPPP1 interacts with and inhibits Histone deacetylae 6 (HDAC6) activity [9] [11] . Given the proximal interactions between LIMK1/TPPP1 and TPPP1/HDAC6, we hypothesized that LIMK1, TPPP1 and HDAC6 form a complex. Immunoprecipitation of LIMK1, TPPP1 or HDAC6 from U2OS osteosarcoma cells demonstrated that they reciprocally co-precipitated each other, indicating that they form a trimeric complex (Figures 1(a)-(c)).

LIMK1 and TPPP1 are both substrates of ROCK [3] [9] . To investigate the impact of LIMK1 and TPPP1 phosphorylation on the stability of the LIMK1/TPPP1/HDAC6 complex, we performed immunoprecipitation studies with U2OS cell lysates that were treated with the ROCK inhibitor Y-27632 (10 mM) or vehicle. Our results demonstrate that decreased ROCK signaling (Figure 1(d)) resulting in reduced LIMK1 and TPPP1 phosphorylation increased the interaction between LIMK1/TPPP1/HDAC6 (Figures 1(e)-(g)). Further analysis of the complex dynamics revealed that stimulation of ROCK activity by treatment of cells with FBS (10%) (Figure 1(h)) reduced their interaction (Figures 1(g)-(k)). Therefore, these results demonstrate that LIMK1, TPPP1 and HDAC6 form a trimeric complex under basal conditions, and that activation of ROCK signaling, resulting in LIMK1 and TPPP1 phosphorylation, reduces LIMK1/TPPP1/HDAC6 interaction.

TPPP1/HDAC6 interaction results in the inhibition of HDAC6 activity [9] . Since HDAC6 catalyzes the deacetylation of Lys40 on β-tubulin, the TPPP1/HDAC6 interaction promotes microtubule (MT) stability via increasing its acetylation. Since LIMK1, TPPP1 and HDAC6 are associated in cells, we examined the possibility that altered LIMK1 expression modulates tubulin acetylation in cells. Analysis of cells overexpressing Flag-LIMK1 revealed a significant increase in acetyl-tubulin levels (Figures 2(a) and (b)). Conversely, the analysis of cells treated with LIMK1 siRNA revealed that reduced LIMK1 expression resulted in decreased acetyl-tubulin levels (Figures 2(c) and (d)), similar to our previously published observations with LIMK2 [8] . Therefore, our results demonstrate that LIMK1 is an integral component of the complex, which along with TPPP1 is required to inhibit

HDAC6 activity.

Our findings suggest that the dynamics of the complex is determined by the level of the components, particularly the un-phosphorylated forms of LIMK1 and TPPP1. Therefore, we hypothesized that the trimer may also modulate LIMK1 activity by protecting it from phosphorylation and therefore its activation. To test this possibility, we analyzed the level of cofilin phosphorylation, the major substrate of LIMK1, in cells expressing altered TPPP1 levels. Analysis of cells overexpressing Flag-TPPP1 revealed that increased TPPP1 expression resulted in a reduction in cofilin phosphorylation (Figure 2(e)) as well as a decrease in cellular filamentous actin (F-actin) levels (Figure 2(f)). Furthermore, analysis of cells transiently transfected with TPPP1 siRNA revealed that knockdown of TPPP1 resulted in an increase in cofilin phosphorylation (Figure 2(g)) and F-actin levels (Figure 2(h)).

Thus far, our results indicate that the LIMK1/TPPP1/HDAC6 trimer autonomously regulates actin and microtubule morphology. However, the inter-cytoskeletal arrangement of actin and microtubules in cells suggests that the downstream effects of the complex may be the result of cross talk between the two networks. To rule out this possibility, we analyzed tubulin acetylation levels in cells overexpressing wild-type myc-cofilin, myc-cofilin S3A (phospho-inhibitory), myc-cofilin S3D (phospho-mimetic) or vector. We found that changes in cofilin activity did not alter tubulin acetylation in cells (Figures 3(a) and (b)). Similarly, we investigated the possibility that differences in tubulin stability may affect cofilin phosphorylation and therefore have an impact on the actin cytoskeleton. For this, we analyzed cofilin phosphorylation levels in cells overexpressing HDAC6 or treated with the HDAC inhibitor Trichostatin A (TSA). We show that neither increased HDAC6 expression (Figures 3(c) and (d)) nor reduced HDAC activity (Figures 3(e) and (f)) altered cofilin phosphorylation or F-actin levels in cells. Overall, these results demonstrate that the LIMK1/TPPP1/HDAC6 trimer has a dual role of inhibiting HDAC6 activity and preventing LIMK1 activation to increase MT acetylation and stability as well as increasing cofilin activity and actin depolymerization.

The microtubule network undergoes dynamic polymerization and depolymerization according to the cells functional requirements. These dynamics are interrupted by chemotherapeutic agents such as vincristine and paclitaxel to enforce cell cycle checkpoints and ultimately initiate cell death programs. Our results demonstrate that the LIMK1/TPPP1/HDAC6 complex promotes MT stabilization. Therefore, we hypothesized that via increasing MT rigidity the complex could reduce the efficacy of the chemotherapeutic agents vincristine and paclitaxel, MT destabilizing and stabilizing compounds, respectively. Analysis of cell viability demonstrated that overexpression of LIMK1 or TPPP1 reduced vincristineand paclitaxel-induced cell death compared to control (Figures 4(a) and (c)), whereas siRNA-mediated knockdown of LIMK1 or TPPP1 increased drug-induced cell death compared to control (Figures 4(b) and (d)). These results suggest that reduced acetyl-tubulin levels correlate with increased sensitivity to chemotherapeutic agents. To further test this possibility we overexpressed HDAC6 and analyzed vincristineand paclitaxel-induced cell death. We show that reduced acetyl-tubulin levels sensitize cells to drug-induced cell death (Figure 4(e)).