The Jurkat leukemic T cell line (clone E6.1) was from American Type Culture Collection (ATCC, catalog number TIB-152). Jurkat cells were cultured at 37°C in 5% CO₂ in RPMI supplemented with 10% fetal bovine serum (FBS) (Hyclone), and DMEM, 10% fetal bovine serum, respectively). Control (CT) and Actin CytoFRET (ACT) reporter cell lines were generated as previously described (Larbret et al., 2013). Antibodies against β-actin (C4) were purchased from Santa Cruz Biotechnology, cofilin (D3F9), ERK2 (9108), profilin (C56B8), ubiquitin (P4D1), and phospho-ADF/cofilin (3312) were purchased from Cell Signaling Technology. Anti-gelsolin (ab11081) was from Abcam. Anti-thymosin β4 (AF6796) and recombinant human CXCL12/SDF-1 were from R&D Systems. WP1130, b-AP15 and PR-619 were from Merck. VLX1570 and AMD3100 were from Selleckchem. All other chemicals were from Merck. Plasmid DNA were transfected with Lipofectamine 2000 (Thermo Fischer Scientific).

Plasmid constructions, cloning, and DNA sequencing were carried out according to standard protocols. For single color (EGFP or mOrange) and EGFP or mOrange fused to actin, cloning strategies were described previously (Larbret et al., 2013).

Cofilin-1 shRNA and control shRNA lentiviral particles were purchased from Santa Cruz Biotechnology. To generate cofilin-1 knockdown cells, 1 × 10⁵ Jurkat cells were infected with 10 μl of lentiviral particles encoding either non targeting control shRNA sequence or cofilin shRNA for 12 h, washed and cultured for 48 h. After a second infection cycle, cells were cultured with puromycin to select infected cells. Transduction and shRNA efficiency was assessed by immunoblotting with antibody against cofilin.

Actin FRET signals were recorded on actin biosensor cell line (FRET-ACT) using a ratiometric method (Larbret et al., 2013). Briefly, untreated or treated FRET-CT and FRET-ACT cell lines were analyzed on a FACSCanto II flow cytometer (BD Biosciences) by using 488 nm donor excitation wavelength. Donor fluorescence was collected in the EGFP channel with a 530/30 nm filter, while mOrange acceptor signal was measured with a 610/15 nm filter. FRET ratio was determined as the ratio between acceptor (mOrange) and donor (EGFP) fluorescence intensity under the donor excitation wavelength. The FRET ratio that is determined in untreated or DMSO-treated FRET-ACT cells reflects the degree of basal actin polymerization. For each sample, a minimum of 5,000 cells that fell within the adjusted morphology gate was analyzed. For real time flow cytometry analysis, measurements were performed using an ARIA II cytometer (BD Biosciences) equipped with a heating sample injection chamber that maintains samples at 37°C during the analysis.

For treatment with chemical compounds, FRET-CT and FRET-ACT cell lines were seeded in 96 well plates at a density of 1 × 105 cell/ml the day before FACS analysis. Cells were then incubated with each compound or DMSO for 1 h at 37°C before flow cytometry analysis. A FACSCanto II flow cytometer equipped with a carousel loader was used. Each compound was screened at three different concentrations, with DMSO treatment as control. To calculate actin FRET efficiency in the presence of a compound, we used the following formula: Compound FRET efficiency (%) = [1-(DMSO FRET ratio/Compound FRET ratio)] × 100.

CM-H₂DCFDA (Thermo Fisher Scientific) was dissolved in DMSO and stored as 100 mM at –20°C. Cells were incubated with 10 μM CM-H₂DCFDA for 30 min at 37°C. Then, cells were washed twice with PBS and exposed to pharmacological treatments during 1 h at 37°C. CM-H₂DCFDA fluorescence was detected by flow cytometry with an excitation wavelength of 488 nm and emission wavelength of 535 nm.

Cell death was evaluated using an Apoptosis Detection Kit (BD Biosciences) as described previously (Didier et al., 2018). Briefly, 10⁵ cells were washed twice with PBS, stained with Annexin-V and propidium iodide in binding buffer for 15 min at room temperature. Cells were then analyzed by flow cytometry using a FACSCanto II and the DIVA8 software (BD Biosciences).

Proteostat aggresome detection kit (Enzo Life Sciences) was used for aggresome staining. Briefly, Jurkat cells were washed with Ice cold PBS, then fixed with 3.7% formaldehyde in PBS, and permeabilized with 0.5% Triton in PBS, 1% bovine serum albumin for 20 min at room temperature. Cells were washed and labeled with 200 μl of proteostat red dye and incubated for 30 min at room temperature. Analysis was performed by flow cytometry.

Cells were lysed in ice-cold RIPA lysis buffer: 1% NP-40 in 50 mM NaCl, 50 mM Tris-HCl, pH 7.4, 0.1% SDS, 0.5% sodium deoxycholate, 150 mM NaCl, 2 mM EDTA, 50 mM NaF supplemented with protease and phosphatase inhibitors (Roche Diagnostics) for 30 min on ice. Then, samples were centrifuged at 7,500 g for 10 min at 4°C. Supernatant (soluble fraction) was separate from pellet (insoluble fraction). Oxyblot detection kit (Abcam) was used for the detection of cofilin oxidation. The procedure for derivatization of protein carbonyls with DNPH and subsequent detection using the Oxyblot kit basically follow the procedure as outlined in the kit brochures. Briefly, 20 μg protein was derivatized with DNPH and separated on a 10% pre-cast gel (BioRad, United States) using standard SDS polyacrylamide electrophoresis. Following separation, the gels were transferred to a PVDF membrane and probed with the primary polyclonal anti-DNP and secondary antibodies anti-rabbit IgG provided in the kit and developed using enhanced chemiluminescence (GE Healthcare).

Jurkat T cells were electroporated with GFP-cofilin expression vector (Addgene Plasmid #50859). 48 h post transfection cells were lysed in ice cold RIPA lysis buffer supplemented with 10 mM DTT. 2 ml of clarified lysate was incubated with 50 µl of GFP-Trap^® (Chromotek) for 1 h at 4°C, and washed three times in IP buffer for 5 min each. The washed GFP-Trap^® (Chromotek) resin was incubated for 2 min in IP buffer adjusted to pH 2.0 in order to acid-elute the samples.

Jurkat cell migration was assessed on uncoated polycarbonate Transwell-96 well support plates with 5-μm pore size (Corning). Cells were serum-starved for 4 h before the experiment. 1 h before the experiment, chemical compounds were added to the medium. Then, medium alone or CXCL12 (100 ng/ml) was added in the lower chamber, and the treated cell suspension (5 × 10⁵ cells) was added in the upper chamber. Cell migration in the lower well was measured after 90 min of migration and quantified by flow cytometry as described before (Ticchioni et al., 2002).

Jurkat cells were adhered on poly-l-lysine coated slides (Thermo scientific), fixed with 3.7% paraformaldehyde. Cells were then permeabilized with 0.1% Triton X-100, washed and incubated overnigh with PBS supplemented with 1% bovine serum albumin (BSA) in the presence of anti-Cofilin antibodies. Following incubation with Alexa Fluor 594-conjugated goat anti-rabbit cross-adsorbed secondary antibodies (Thermo Fischer Scientific) for 30 min and nuclei staining with DAPI, coverslips were mounted in ProLong antifade mounting reagent (Thermo Fisher Scientific). Slides were then analyzed on a Axiovert 200M microscope (Zeiss) equipped with Orca CCD camera (Hamamatsu).

All data are presented as the mean ± SD of at least three independent determinations. p-values were determined using the Prism software (GraphPad, United States). Data sets were tested for normality, no outliers were removed and Student’s T tests were performed to determine statistical significance. p-values of 0.05 (*), 0.01 (**) and 0.001 (***) were considered statistically significant.

In order to evaluate the impact of post-translational modifications on actin dynamics, we employed Actin CytoFRET, a method that we developed for real-time detection by flow cytometry of FRET signals generated by actin filament remodelling in engineered Jurkat leukemic T cell lines (Larbret et al., 2013). Briefly, the principle of the method is based on the use of biosensor cell lines expressing actin monomers fused to EGFP or mOrange fluorescent proteins. A cell line (CT) expressing actin-EGFP and unfused mOrange serves as a negative control (Figure 1A). Within the Actin CytoFRET cell line (ACT), the interaction of actin-EGFP and actin-mOrange during actin assembly generates a FRET signal that is recorded by flow cytometry using a ratiometric method (Larbret et al., 2013). Increase of FRET signal is characterized by a decrease in GFP fluorescence intensity and an increase in mOrange intensity upon 488 nm wavelength excitation laser. The actin stabilizing agent Jasplakinolide and the actin depolymerizing drug Latrunculin B were used as controls for monitoring actin assembly and disassembly, respectively. While Jasplakinolide induced an increase of FRET signals, Latrunculin B resulted in a decrease of FRET signals (Figure 1B). Using this method, we screened a small library of compounds interfering with PTMs such as acetylation, methylation, oxidation, phosphorylation and ubiquitination for their activity on actin cytoskeleton. ACT and CT cell lines were incubated with either DMSO or increasing doses of the indicated compounds. The majority of tested molecules had no effect on actin dynamics. On the contrary, the treatment with four inhibitors of DUBs, WP1130, b-AP15, VLX1570 and PR-619 (Supplementary Figure S1) induced a strong dose dependent increase of actin FRET signals that compared to the effect of Jasplakinolide (Figure 1B). At a concentration of 30 μM, FRET signals were augmented by WP1130, b-AP15, VLX1570 and PR-619 by 20.3, 15.8, 7.6 and 8.1%, respectively (Figure 1B). Dose responses performed with these compounds showed that the EC50 for increasing actin FRET efficiency was 4.9, 5.6, 8.6 and 19.1 μM for WP1130, b-AP15, VLX1570 and PR-619, respectively (Figure 1C). In contrast, UPS inhibition by Bortezomib and MG132 or the inhibition of the E1 ubiquitin-activating enzyme by Pyr-41 did not affect actin reorganization (Figure 1B). Our screening strategy thus identifies multiple DUB inhibitors (DUBi) as novel inducers of actin assembly.

To confirm the effect of DUBi on actin polymerization, we analyzed ACT cells by confocal microscopy following treatment with WP1130 and b-AP15. DUB inhibition by WP1130 and b-AP15 resulted in a dramatic redistribution of actin-mOrange fluorescence in a cortical ring (Figure 2A). Actin remodeling is a crucial event regulating cell motility (Olson and Sahai, 2009; Pollard and Cooper, 2009). We thus examined the effect of DUB inhibition on leukemic T cell chemotactic migration. Jurkat cells were treated or not with DUBi and the CXCR4 inhibitor AMD3100 for 60 min and equal numbers of cells were allowed to migrate for 90 min towards a gradient of CXCL12 in Transwell chambers as described before (Ticchioni et al., 2002). The number of migrating cells in lower chambers was then determined by flow cytometry. The chemotaxis-driven migration of leukemic cells treated with WP1130, b-AP15, VLX1570, PR-619 and AMD3100 was dramatically impaired compared to cells treated with DMSO (more than 90% of reduction) (Figure 2B). Note that the dose of WP1130 and b-AP15 that blocked cell migration had no impact on Jurkat cell viability as shown by FACS analysis of annexin V/PI staining (Supplementary Figure S2). We next examined whether the inhibition of leukemic cell migration by DUBi originated from an impaired response to CXCL12 of its receptor CXCR4. Real time analysis of actin FRET following CXCL12 stimulation showed that the rapid reorganization of actin that was triggered by CXCL12 was completely abolished by a co-treatment with WP1130, b-AP15 and AMD3100 (Figure 2C). Importantly, DUB inhibition had no effect on the levels of CXCL12 receptor CXCR4 (Supplementary Figure S3). These observations show that DUB inhibition promotes exacerbated actin polymerization, which impairs leukemic cell chemotactic migration.

We next wish to characterize how DUB inhibition resulted in actin reorganization. The Jurkat ACT cells were pre-treated with the F-actin depolymerizing drug Latrunculin B (LB) or the Arp2/3 complex inhibitor CK-666. As shown in Figure 3A, the pre-treatment with LB abolished the elevation of actin FRET signals induced by WP1130 and b-AP15, whereas the pre-treatment with CK-666 dramatically reduced it, indicating the involvement of actin monomers nucleation and/or assembly in this process. We therefore hypothesized that DUB inhibition could affect the expression and/or the activity of actin-binding protein involved in actin nucleation or polymerization (Pollard, 2016). Western blot analysis of lysates from Jurkat cells treated with WP1130 and b-AP15 revealed that DUB inhibition did not affect the protein levels of ABPs such as profilin, gelsolin, cofilin and thymosin β4. However, a decrease of the phosphorylation of cofilin serine 3 was observed upon WP1130, b-AP15, VLX1570 and PR-619 treatment (Figure 3B) (Supplementary Figure S4). The phosphorylation of the actin-severing protein cofilin suppresses its depolymerizing and nucleation activity, thereby pointing at cofilin as a key target of DUBs in actin remodeling. Dose-response and kinetic analysis of DUBi treatment further showed a correlation between cofilin dephosphorylation and actin FRET Efficiency (Figures 3C,D). In contrast, MG132 and two other DUBi, 8TQ and HBX41108, which generated no actin FRET signals (Figure 1B), had no impact on the levels of cofilin phosphorylation (Figure 3E). Importantly, Jurkat cell treatment with Latrunculin B and CK-666 did not affect DUBi-induced cofilin dephosphorylation (Figure 3F), indicating that cofilin dephosphorylation was not secondary to actin depolymerization.

To confirm that cofilin is a major target for cytoskeletal changes induced following DUB inhibition, we stably knock-downed cofilin in the ACT cell line (Figure 4A). Surprisingly, cofilin-depleted leukemic cells exhibited a very low modification of basal actin FRET signal (+5%) compared to control cells (Figure 4B). However, the reduction of actin FRET efficiencies observed in shCofilin ACT versus shCtrl ACT reporter cells treated with WP1130 and b-AP15 revealed that actin polymerization induced by DUBi required the expression of cofilin (from 19% for shCtrl to 9% for shCofilin treated with WP1130 and from 13 to 4% for cells treated with b-AP15) (Figure 4C). Confocal microscopy analyses further showed that cofilin depletion in Jurkat cells resulted in increased polymerization of actin filaments associated with uropod-like structures and elongated cells (Figure 4D), a phenotype that is consistent with the described role of cofilin (Nishita et al., 2005). We next examined the involvement of Slingshot Protein Phosphatase (SSH) and LIM domain kinase (LIMK), two known upstream regulators of cofilin phosphorylation and activity (Huang et al., 2006; Prunier et al., 2017). DUB inhibition with WP1130 and b-AP15 has no detectable impact on LIMK phosphorylation and SSH1 expression as measured by immunoblot analysis (Figure 4E). These results indicate that actin reorganization induced by DUBi involves a non-degradative and LIMK-independent modulation of cofilin activity.

We next wished to clarify the molecular mechanism underlying the effects of DUB inhibition on actin cytoskeleton. DUB inhibition is linked to increased protein ubiquitination followed by proteasome-mediated proteolytic degradation and/or aggresome formation.(Kapuria et al., 2010; D’Arcy et al., 2011). Consistently, we observed that the treatment of Jurkat cells with WP1130 and b-AP15 induced the accumulation of polyubiquitinated proteins (Supplementary Figure S5A). However, the direct inhibition of 26S proteasome by MG-132 and bortezomid had no effect on either basal (Figure 1B) or DUBi-induced actin FRET signals (Figure 5A), indicating that DUB-mediated actin regulation is not a direct consequence of the accumulation of polyubiquitinated proteins. Furthermore, in contrast to WP1130 or MG-132, b-AP15 treatment did not result in aggresome formation (Supplementary Figure S5B), suggesting that the effect of DUB inhibition on actin remodeling did not follow protein accumulation within aggresome-like structures. In support to this, blocking the regulator of aggresome formation HDAC6 (Kawaguchi et al., 2003) with tubastatin A or trichostatin A had no impact on actin assembly that was promoted by WP1130 and b-AP15 (Figure 5A) (Supplementary Figure S5C). Recent studies have linked DUB inhibition to increased oxidative stress and ROS generation (Didier et al., 2018; Charbonneau et al., 2019; Zhang et al., 2019). Conversely, DUBs can be inactivated by ROS (Cotto-Rios et al., 2012; Lee et al., 2013). Keeping in mind that cofilin activity can be modulated by oxidative signals (Klemke et al., 2008; Cameron et al., 2015), we examined the involvement of ROS in DUBi-induced actin rearrangement. The pre-treatment of Jurkat ACT cells with the antioxidant compounds DTT, NAC, α-tocopherol and DPI abrogated actin FRET signals that were induced by WP1130 and b-AP15 (Figure 5A). Consistently, the addition of DTT or NAC prevented the generation of ROS induced by DUB inhibition (Figure 5B) and the dephosphorylation of cofilin that was induced by WP1130 and b-AP15 (Figure 5C). In addition, the pretreatment of Jurkat cells with NAC or DTT restored the ability of leukemic cells to migrate against CXCL12 when DUB activity was inhibited (Figure 5D). As a control of the effect of ROS on actin reorganization in leukemic cells, the treatment with Auranofin, a compound known for its capacity to induce ROS in leukemic cells (Fiskus et al., 2014) (Figure 5E), was found to increase actin FRET signals (Figure 5F), decreased cofilin serine 3 phosphorylation (Figure 5G), and blocked CXCL12-induced chemotactic cell migration in ROS-dependent manner (Figure 5D).

These results showing that DUBs regulate cofilin function through a ROS-dependent pathway led us to investigate the impact of DUB inhibition on cofilin oxidation, a process linked to actin cytoskeleton dynamics in T cells (Samstag et al., 2013). Jurkat cells stably expressing a GFP-cofilin construct were treated with WP1130 or b-AP15 for 30 min. Oxidated proteins were labeled with DNPH in cell lysates after pulldown with anti-GFP-Trap nanobodies and immunoblot analysis with antibodies against DNP. As shown in Figure 6A, WP1130 and b-AP15 increased cofilin oxidation by 2.6 and 2.2 fold, respectively, compared to DMSO-treated conditions. Cofilin exhibits a tendency for self-association under oxidative stress, inducing the rapid formation of dimers and oligomers (Pfannstiel et al., 2001; Goyal et al., 2013). Consistently, immunofluorescence microscopy analysis revealed that compared to control cells Jurkat cells treated with DUBi exhibit increased cofilin aggregates that partially located in the nucleus (Figure 6B). Together, these results demonstrate that DUBs inhibitors display antimigratory activity in leukemic cells through ROS-mediated cofilin oxidation and actin cytoskeleton dysfunction.