HDAC6 is a tubulin deacetylase involved in many cellular functions related to cytoskeleton dynamics, including cell migration and autophagy. In addition, HDAC6 affects antigen-dependent CD4+ T cell activation. In this study, we show that HDAC6 contributes to the cytotoxic function of CD8+ T cells. Immunization studies revealed defective cytotoxic activity in vivo in the absence of HDAC6. Adoptive transfer of wild-type or Hdac6−/− CD8+ T cells to Rag1−/− mice demonstrated specific impairment in CD8+ T cell responses against vaccinia infection. Mechanistically, HDAC6-deficient cytotoxic T lymphocytes (CTLs) showed defective in vitro cytolytic activity related to altered dynamics of lytic granules, inhibited kinesin-1–dynactin-mediated terminal transport of lytic granules to the immune synapse and deficient exocytosis, but not to target cell recognition, T cell receptor (TCR) activation or interferon (IFN)γ production. Our results establish HDAC6 as an effector of the immune cytotoxic response that acts by affecting the dynamics, transport and secretion of lytic granules by CTLs.
Cytotoxic T lymphocytes (CTLs) are a specialized population of CD8+ T cells that provides defense against virus-infected cells and tumors. Naïve CD8+ T cells differentiate into CTLs upon antigen recognition, a process involving the synthesis and storage of cytotoxic mediators into lysosomal-derived lytic granules (Williams and Bevan, 2007). CTLs eliminate target cells by different mechanisms, including secretion of pro-inflammatory cytokines [e.g. tumor necrosis factor (TNFα) or interferon (IFN)γ; de Saint Basile et al., 2010] and FAS-L (FAS ligand; officially known as FASLG) ligation to its receptor, as well as granule-mediated apoptosis upon cell–cell contact and immune synapse formation (Ritter et al., 2013). Lytic granules fuse with the plasma membrane and release granzymes, cathepsins and perforins (Prf proteins) (Lopez et al., 2013; Pardo et al., 2009). The immune synapse acts as a focal point for exocytosis of lytic granules. Lytic granule polarization towards the target cell depends on T cell receptor (TCR) engagement, driven by the relocation of the centrosome to immune synapse. The lytic granules degranulate at a secretory domain adjacent to the TCR-enriched region within the immune synapse (de Saint Basile et al., 2010; Ritter et al., 2013).
Histone deacetylase 6 (HDAC6) is a ubiquitous cytosolic protein from the class II HDACs family with X-linked inheritance, that binds to and deacetylates α-tubulin at Lys40 (Hubbert et al., 2002; Valenzuela-Fernandez et al., 2008). HDAC6 also modulates other substrates, for example, cortactin and Hsp90. HDAC6 controls cell migration (Zhang et al., 2007), T-regulatory functions (de Zoeten et al., 2011) and CD4+ T cell activation (Serrador et al., 2004). Consistent with this, HDACs inhibitors impair some immune functions (Mosley et al., 2006; Tsuji et al., 2015). However, the precise contribution in vivo (by using Hdac6−/− mice) has not been assessed, and the mechanisms involved remain unsolved. HDAC6 also functions as a scaffold protein in T cell migration (Cabrero et al., 2006) and the transport of misfolded proteins (Kawaguchi et al., 2003). In this report, we describe the impaired killing capacity of Hdac6−/− CTLs. The molecular mechanism underlying this defect involves a scaffold role that positions HDAC6 as a protein that oversees the proper movement of lytic granules, their transport to the immune synapse and secretion towards the target cell.
RESULTS AND DISCUSSION
HDAC6 deficiency reduces the cytolytic capacity of CD8+ T lymphocytes
We examined the ability of cytotoxic T cells from Hdac6−/− mice to kill target cells in vitro. CD8+ T cells from wild-type (WT) and Hdac6−/− mice expressing the transgenic ovalbumin (OVA)-specific TCR (OT-I) were activated in vitro and cultured to generate CTLs. Cell cytotoxicity was subsequently analyzed by survival of dye-labeled EL4 target cells pulsed or not with OVA257–264 peptide (SIINFEKL). Hdac6−/− CTLs showed decreased killing activity (Fig. 1A), consistent with reduced expression of CD107a (also known as LAMP1) in Hdac6−/− CTLs upon degranulation (Fig. 1B). Likewise, CTLs from OT-I mice treated with tubastatin A, a potent HDAC6 inhibitor (Butler et al., 2010), displayed a reduced killing ability (Fig. S1A). We also detected decreased Prf1 secretion from activated (i.e. induced by anti-CD3 and anti-CD28 monoclonal antibodies) Hdac6−/− CTLs (Fig. 1C, left). Next, we assessed the secretion promoted by phorbol-12-myristate-13-acetate (PMA), to bypass TCR stimulation. Both cathepsin D and Prf1 decreased in supernatants from activated Hdac6−/− CTLs (Fig. 1C, right). Taken together, our data demonstrate that Hdac6−/− CTLs show reduced cytotoxic activity and suggest that HDAC6 controls exocytosis.
We next tested IFN-γ production; the frequency of CTLs producing IFNγ and its secretion was unaffected in activated Hdac6−/− CTLs (Fig. 1D,E), in contrast to what has been described for ACY-1215, a recently described inhibitor that is tenfold more selective for HDAC6 than for HDAC1, HDAC2 and HDAC3 (class I HDACs) that shows slight activity against HDAC8 (Tsuji et al., 2015). Likewise, treatment of CTLs from OT-I mice with tubastatin A had no significant effect (Fig. S1B). Importantly, T cell signaling induced by anti-CD3 and anti-CD28 monoclonal antibodies in Hdac6−/− cells was comparable to control, as determined by assessing PLCγ1 and Erk1 and Erk2 (Erk1/2; also known as MAPK3 and MAPK1, respectively) phosphorylation (Fig. 1F). Likewise, the increase in intracellular Ca2+ remained unchanged upon activation (Fig. 1G). As expected, tubulin acetylation at Lys40 was increased in Hdac6−/− CTLs (Fig. 1F). These results suggest that the killing defect observed does not result from a general impairment of CTLs function.
Defective in vivo and ex vivo killing in HDAC6-knockout mice
The effector activity of Hdac6−/− CD8+ T cells was tested in vivo following immunization of mice by inoculation with SIINFEKL-pulsed dendritic cells. The in vivo killing activity against the injected splenocytes as target cells (pulsed or not with SIINFEKL) was analyzed upon recovery by peritoneal lavage. Notably, Hdac6−/− mice showed reduced specific killing of target cells (Fig. 2A, left panel). However, the proportion of SIINFEKL-specific CD8+ T cells in the endogenous repertoire was not affected in Hdac6−/− mice (Fig. 2A, right panel), suggesting that the cytotoxic function rather than the number of antigen-specific CTLs could underlie the defect. Next, we examined whether the decreased cytotoxic function of the CTLs resulted in an impaired ability to prevent morbidity and/or mortality during a viral infection. To restrict HDAC6 deficiency to CD8+ T cells, we adoptively transferred Rag1−/− mice with WT or Hdac6−/− naïve CD8+ T cells and subsequently challenged with a fully replicative vaccinia virus (VACV) WR strain. This infection model mimics the immunological and clinical features of smallpox vaccination in humans (Mota et al., 2011). CD8+ T cell proliferation was comparable, or even increased (division 4) in Hdac6−/− mice (Fig. 2B). Rag1−/− mice adoptively transferred with Hdac6−/− CD8+ T cells showed increased morbidity at 9 and 11 days post-infection (dpi) (Fig. 2C). Virus titration from the lesion tissue demonstrated that Hdac6−/− immune cells exerted a less-efficient virus clearance (Fig. 2D). Consistent with our findings on the lack of effect in endogenous antigen-specific CD8+ T cell numbers, CTL expansion tracked at 13 dpi was not affected in Hdac6−/− mice (Fig. 2E, left). The proportion of activated CD8+ T cells (CD44high, with high expression of CD44) at early (5 dpi) and late stages (13–30 dpi) of the disease were similar for WT and Hdac6−/− mice (tested in peripheral blood and spleen, respectively; Fig. 2E, right). These in vivo results emphasize the role of HDAC6 in the CD8+ T-cell-dependent protection against VACV infection without affecting effector CD8+ CTL differentiation.
HDAC6 drives the terminal transport of lytic granules to the target cell
CTL killing is limited to target cells (and not neighbor cells) by the confinement of secretion to the immune synapse established between the CTL and the target cell (de Saint Basile et al., 2010). Interestingly, the intracellular colocalization between cathepsin D and Lamp1 (CD107a) was affected in Hdac6−/− CTLs conjugated with target cells, pointing to the mislocalization of lytic mediators in these cells (Fig. 3A; images and middle graphs). The decreased secretion of lytic proteins from Hdac6−/− CTLs suggests that HDAC6 regulates exocytosis of lytic granules (Fig. 1C). Indeed, the translocation of the centrosome to the contact area with the target cell was more pronounced in Hdac6−/− CTLs than in WT cells (Fig. 3A, right graph), in accordance with the effect described for the HDAC inhibitor Trichostatin A on the centrosomal polarization in CD4+ T cells (Serrador et al., 2004). This suggests that the defective exocytosis might rely on the movement of lytic granules themselves.
We thus monitored the dynamics of lytic granules at the subcortical immune synapse cytoskeleton and their release by total internal reflection fluorescence microscopy (TIRFm). CTLs were loaded with a pH-dependent lysosomal tracker, which allows the visualization of the lytic granules, and settled on to a stimulating surface to form an immune-synapse-like structure (Fig. 3B). These experiments revealed significant changes in the distribution of the lytic granules and their dynamics, with a marked decrease in the number of lytic granules detected at the immune-synapse-like structure in Hdac6−/− cells, suggesting alterations to the active transport of the granules from the centrosomal region to the plasma membrane. Indeed, the mean fluorescence intensity detected for Hdac6−/− granules was lower, which suggests a higher pH and, therefore, a different degree of maturation, although the lytic granules displayed similar sizes in WT and Hdac6−/− cells (Fig. 3C). The most remarkable difference pertained to the x–y distribution of the lytic granules, which was wider (diffusion surface) in the Hdac6−/− CTLs, with a higher diffusion coefficient, although there were similar duration times and path lengths (Fig. 3D). These data suggest that the lytic granules from Hdac6−/− CTLs are not properly targeted and/or that they dock inefficiently at the immune synapse.
Tubulin motors control the delivery of lytic granules to the plasma membrane. Whereas dynein controls lytic granule targeting to the centrosome (Burkhardt et al., 1993; Mentlik et al., 2010), the kinesin-1–Slp3–Rab27a (Slp3 is also known as SYTL3) complex directs terminal transport to the plasma membrane for exocytosis (Kurowska et al., 2012). Dynactin might also be part of this complex, linking the cargo to kinesin-1 motor (Haghnia et al., 2007; Hendricks et al., 2010). We then hypothesized that HDAC6 regulates the movement and delivery of the lytic granules at the immune synapse through kinesin-1. Using a biochemical approach, we observed that HDAC6 formed a complex with kinesin-1 light chain (KLC1) upon triggering with anti-CD3 and anti-CD28 monoclonal antibodies (Fig. 3E). Moreover, interaction of the kinesin-activator complex dynactin subunits p150-glued (also known as DCTN1) and p50-dynamitin (also known as DCTN2) was impaired in Hdac6−/− (Fig. 3F).
In summary, our data support a specific role for HDAC6 in the intracellular localization of lytic mediators and, particularly, in their exocytosis. Therefore, the catalytic and scaffold activities of HDAC6 might act at multiple levels in the control of cytotoxic-related pathways, making HDAC6 a potential candidate that could be targeted to modulate CTLs in specific diseases.
MATERIALS AND METHODS
Hdac6−/− mice were generated through targeting of exons 10 to 13 (Gao et al., 2007). They were intercrossed in a C57BL/6 background to generate wild-type and knockout littermates. Mice presenting transgenic inserts for mouse Tcra-Variable2 and Tcrb-Variable5 genes, which TCR recognizes ovalbumin257-264 peptide in the context of H2Kb MHC-I (OT-I) were crossed to female Hdac6+/−mice to generate WT and KO littermates; males were used for in vivo experimentation given that Hdac6 is an X-linked gene. Rag1−/− mice were used for adoptive transfer experiments. These studies were approved by the local Ethics Committee for Basic Research at the CNIC and the Comunidad Autónoma de Madrid.
Cytotoxic cells were produced by culturing cells upon stimulation with SIINFEKL peptide (0.5 µM, 24 h) or concanavalin A (2.5 µg/ml, 36 h) and cultured in presence of IL-2 (50–100 IU/ml) for at least 7 days. All other cells were cultured and treated as described previously (Cascio et al., 2015; Martín-Cófreces et al., 2006; Sancho et al., 2008).
Immunoprecipitation, CTL signaling and immunoblotting
Experiments were performed as described previously (Martín-Cófreces et al., 2012, 2008, 2006); anti-KLC1 antibody was from Merck Millipore (Darmstadt, Germany; KLC; at 5 μg/ml for immunoprecipitation and 1 μg/ml for western blotting), anti-HDAC6 (at 0.1 μg/ml for western blotting) from Assay Biotech (Sunnyvale, CA), and anti-p50 and -p150 from BD Pharmingen (Franklin Lakes, New Jersey, US; anti-p150 was used at 1 μg/ml for immunoprecipitation and 0.25 μg/ml for western blotting. Anti-p50 at 0.25 μg/ml for western blotting).
Measurement of intracellular variations in Ca2+ levels by using flow cytometry
The method used for intracellular Ca2+ influx has been described previously (June and Moore, 2004). In particular, 5×106 purified CD8+ CTLs generated in vitro were loaded with 2 µg ml−1 INDO-1 AM (Invitrogen Corporation) and stimulated with anti-CD3 and anti-CD28 monoclonal antibodies (BD Biosciences; Franklin Lakes, NJ) plus goat anti-Armenian-hamster IgG antibodies (Jackson Immunoresearch Laboratories; West Grove, PA; 6, 3 and 6 µg/ml, respectively).
In vitro degranulation assay
CD107a expression was monitored with Alexa-Fluor-647-conjugated anti-CD107a antibody (BD Biosciences) in monensin-pretreated CD8+ OT-I cells (5 mM) stimulated with SIINFEKL-pulsed (1 µM; 3 h, 37°C) EL4 cells. Cells were stained with phycoerythrin-conjugated anti-CD8 and FITC-conjugated anti-CD44 antibody, analyzed by fluorescence-activated cell sorting (FACS) and data were processed with FlowJo 7.6.5 (TreeStar Inc, Ashland, OR).
Confocal and total internal reflection fluorescence microscopy analysis
Cell conjugates between CTLs and EL4 cells were allowed to form (15 min) and processed as described previously (Cascio et al., 2015; Martín-Cófreces et al., 2006) under a Leica SP5 confocal microscope (Leica Microsystems; Mannheim, Germany) mounted on an inverted DMI6000 microscope fitted with a HCX PL APO 63×1.40-0.6 NA oil objective. Images were processed using Imaris software (Bitplane; Zurich, Switzerland) and Image J software (http://rsbweb.nih.gov/ij/), and assembled with Photoshop 6 software. The 3D distance from the centrosomal center of mass to the target cell-edge was measured by generating image masks from fluorescence with Imaris Software. TIRFm imaging was performed with a Leica AM-TIRF-MC-M system mounted on a Leica DMI-6000B microscope coupled to an Andor-DU8285_VP-4094 camera (Andor; Belfast, UK) fitted with a HCX-PL-APO 100.0×1.46 NA oil objective as described previously (Baixauli et al., 2011; Martín-Cófreces et al., 2012). The laser penetrance used was 90 nm (561 nm laser). The mechanical properties of the lytic granules were determined with a user-customized routine developed in Python. The software can be freely downloaded from: https://dl.dropboxusercontent.com/u/4050954/VesiclesAnalyser.zip. For more information, see the tutorial included.
Vaccinia virus infection and virus titration
Tails were scarified with vaccinia virus [VACV; 2×106 plaque-forming units (PFU)/mouse] by gently scratching (×25) with a 28 1/2 G needle. For virus titration, tails were mechanical disaggregated (1 ml of PBS), subjected to freeze-thaw cycles and sonication. Serial dilutions of the homogenates were added to monolayers of CV-1 cells seeded in 24-well plates. Cells were stained with Cristal Violet 24 h later. We observed a detection limit of 5 PFU/tail, the number of plaques was multiplied by the reciprocal of sample dilution and converted to PFU/g of tissue.
In vitro and in vivo cytotoxicity assay
For in vitro experiments, EL4 target cells were incubated with 1 μM Cell Violet and pulsed with 1 μM SIINFEKL, or with 0.1 μM Cell Violet and no SIINFEKL, washed extensively, mixed (1:1), pooled with different dilutions of effectors, plated in a 96-well U-bottomed plate for 5 h (37°C) in triplicate and analyzed by FACS. Dead cells were excluded on the basis of propidium iodide staining. The mean percentage of survival in antigen-loaded targets was calculated relative to antigen-negative internal controls in each sample. Specific lysis was calculated by using the following equation: percentage specific lysis=100×[1−(percentage of cells staining for Cell Violet at 1 µM/percentage of cells staining for Cell Violet at 0.1 µM)]. All data were normalized to the basal specific lysis in absence of effector cells. For in vivo assays, WT and Hdac6−/− mice were inoculated by intraperitoneal injection of bone marrow dendritic cells pulsed with 1 µM of SIINFEKL and lipopolysaccharide (LPS; 1 µg ml−1) for 1 h. After 7 days, CD45.1 splenocytes were prepared as targets as described above and injected intraperitoneally into recipients. Cells were recovered 24 h later by peritoneal lavage and in vivo killing levels measured (Hermans et al., 2004; Iborra et al., 2012; Sancho et al., 2008; Schulz et al., 2005).
Data were analyzed with GraphPad Prism software (La Jolla, CA) for normality (D'Agostino-Pearson or the Kolmogorov–Smirnov test for small samples). Student's t-tests or Mann–Whitney tests were used for normal or non-normal data, respectively, and two-tailed ANOVA for grouped data (Bonferroni post-test).
We thank Manuel Gomez and Miguel Vicente for critical reading of the manuscript. Experimentation was performed at Cellomics and Microscopy Units (CNIC) and Flow Cytometry Core Unit (CNIO).
N.N.-A., N.B.M.-C. and F.S.-M. designed experiments, made the figures and wrote the manuscript; N.N.A., S.I., N.B.M.-C., O.M.-G., J.V., D.S., G.M. and T.-P.Y. and E.C. collected and/or analyzed data; A.T. developed the Quant Application.
This work was supported by the Ministerio de Economía y competitividad (MINECO) [grant number SAF2014-55579-R]; Comunidad Autónoma de Madrid (CAM) [grant number INDISNET01592006]; Instituto de Salud Carlos III y Fondo Europeo de Desarrollo Regional (FEDER) [grant numbers BIOMID-PIE13/041 and RD12/0042/0056]; European Research Council (ERC) [grant number ERC-2011-AdG 294340-GENTRIS]. The Centro Nacional de Investigaciones Cardiovasculares (CNIC) is supported by the MINECO and Pro-CNIC Foundation.
The authors declare no competing or financial interests.