High-selective HDAC6 inhibitor promotes HDAC6 degradation following autophagy modulation and enhanced antitumor immunity in glioblastoma

Abstract

Glioblastoma is the most fatal type of primary brain cancer, and current treatments for glioblastoma are in- sufficient. HDAC6 is overexpressed in glioblastoma, and siRNA-mediated knockdown of HDAC6 inhibits glioma cell proliferation. Herein, we report a high-selective HDAC6 inhibitor, J22352, which has PROTAC (proteolysis- targeting chimeras)-like property resulted in both p62 accumulation and proteasomal degradation, leading to proteolysis of aberrantly overexpressed HDAC6 in glioblastoma. The consequences of decreased HDAC6 ex- pression in response to J22352 decreased cell migration, increased autophagic cancer cell death and significant tumor growth inhibition. Notably, J22352 reduced the immunosuppressive activity of PD-L1, leading to the restoration of host anti-tumor activity. These results demonstrate that J22352 promotes HDAC6 degradation and induces anticancer effects by inhibiting autophagy and eliciting the antitumor immune response in glioblastoma. Therefore, this highly selective HDAC6 inhibitor can be considered a potential therapeutic for the treatment of glioblastoma and other cancers.

1. Introduction

Glioblastoma is the most common and aggressive type of malignant brain tumor. Despite improvements in surgical and medical therapies, the median survival of glioblastoma patients has remained at approXi- mately 15 months [1]. Currently, surgical resection in combination with radiotherapy and adjuvant chemotherapeutic agents, such as temozolomide (TMZ), represents the standard of care in patients with newly diagnosed glioblastoma [2–4]. However, this therapeutic ap- proach slightly prolongs overall survival by only 4.8 months, primarily due to TMZ-induced multidrug resistance [5]. Traditional che- motherapy has limited value because of non-selective toXicity, intolerable side effects and drug resistance [3]. Targeted molecular therapies have also not proven to be effective because of tumor heterogeneity, high metastatic potential and resistance mechanisms in glioblastoma. Because of these challenges, glioblastoma remains a highly lethal dis- ease [6]. Thus, the identification of effective therapeutics and targets in glioblastoma is urgently needed.

Histone deacetylases (HDACs) have been considered rational targets for the development of novel anticancer therapeutics in the past two decades [7,8]. Currently, most HDAC inhibitors used in the clinic are non-selective HDAC inhibitors. Unfortunately, undesirable side effect have been observed for several HDAC inhibitors, including vorinostat (hereafter referred to as SAHA) and panobinostat, due to their non- selective properties. SAHA and panobinostat target not only class I HDACs (HDAC1, HDAC2, HDAC3 and HDAC8) but also class II HDACs (HDAC4, HDAC6 and HDAC9) [9,10]. Therefore, a specific and selec- tive HDAC inhibitor may have therapeutic advantages with reduced toXicity over non-selective inhibitors.

In contrast to other subtypes of HDACs, HDAC6 contains two functional catalytic domains and is a predominantly cytosolic isoform that targets a diverse set of non-histone substrates, such as α-tubulin [11], HSP90 [12] and cortactin [13]. Recent studies showed that overexpression of HDAC6 promotes cell proliferation, accelerates mi-
gration, associates with poor prognosis and contributes to drug re- sistance in cancer [14,15]. Therefore, these unique features of HDAC6 make it an excellent drug target, and selective HDAC6 inhibitors might exert significant anticancer activity [16–19]. Thus, there is considerable interest in the development of specific HDAC6 inhibitors, and some of such inhibitors are in preclinical or clinical development, for instance tubastatin A (TUB) [20] and ACY-1215 [21], but no studies have fo- cused on glioblastoma treatment. Interestingly, glioblastoma patients have significantly higher HDAC6 expression levels, and siRNA-medi- ated knockdown of HDAC6 inhibited glioma cell proliferation in a preclinical study [15]. These data suggest that the specific inhibition of HDAC6 may be a promising strategy for the treatment of glioblastoma. Therefore, we hypothesized that a highly selective small-molecule HDAC6 inhibitor might have potent activity against glioblastoma with less toXicity.

In addition, several studies have indicated that HDAC6 plays an important role in the regulation of autophagy [22,23]. Autophagy, an intracellular degradation process, has drawn significant attention due to its ability to promote the survival of cancer cells under metabolic and therapeutic stress [19,24]. In particular, adaptive resistance of glio- blastoma to radiation and TMZ is mainly promoted by autophagy, which contributes to the maintenance of cell survival in the cancer microenvironment [15]. The inhibition of autophagy may therefore have a therapeutic benefit in glioblastoma treatment [25] and may also enhance the immune response against cancer [26]. Interestingly, HDAC6 is also involved in the regulation of inflammatory and immune responses, including immune synapse formation [27], T cell functional regulation [28] and macrophage responses [29].

Taken together, previous findings led to the hypothesis that a highly
selective HDAC6 inhibitor may inhibit glioblastoma cell proliferation and migration by inhibiting autophagy and simultaneously enhancing the immune response. In this study, we characterized the small mole- cule J22352 and conducted translational research on the activities de- scribed above to provide a highly selective HDAC6 inhibitor against glioblastoma.

2. Materials and methods

2.1. 2.1 Reagents and solvents for synthesis

Isatoic anhydride (Acros, New Jersey), 2-phenylethanamine (Acros, New Jersey), CDI (Acros, New Jersey), methyl-4-(bromomethyl) benzoate (Acros, New Jersey), K2CO3 (SHOWA, Japan), NH2OH (Acros, New Jersey), NaOH (SHOWA, Japan), EtOAc, CH2Cl2, Hexanes, MeOH (DUKSAN, Korea).

2.2. Chemical synthesis of the compounds J22352 and J27820

Isatoic anhydride (50 g, 306.5 mmol) in EtOAc (350 mL) was stirred at room temperature for 10 min. After 10 min, 2-phenylethanamine was added (38 mL, 306.5 mmol) and the miXture was stirred at room tem- perature for 1 h. After 1 h, the miXture was supplemented with CDI (76.85 g, 459.75 mmol) and EtOAc (300 mL) and stirred at room tem- perature for 19 h. The resulting precipitate was collected by direct fil- tration and dried in a vacuum to generate 3-phenethylquinazoline- 2,4(1H,3H)-dione (52 g, 64%). Rf = 0.57 (EtOAc/Hexane = 1:1);ESIMS (+) m/z [M + H]+ 267. 3-Phenethylquinazoline-2,4(1H,3H)-similar procedure as J22352.

2.3. Cell lines and culture conditions

All cell lines were purchased from BCRC (Bioresource Collection and Research Center, Taiwan): U87MG (BCRC 60360), HepG2 (BCRC 60025), A549 (BCRC 60074), PANC1 (BCRC 60284), A375 (BCRC 60039), LNCaP (BCRC 60088), 22Rv1 (BCRC 60545), FHs173we (BCRC 60229),Vero (BCRC 60013) and GL261 (derived from ATCC CRL-1887). All cell lines were maintained in the medium as per the manufacturer’s protocol, and supplemented with 10% fetal bovine serum (FBS) and 100 U ml−1 penicillin/streptomycin at 37 °C with 5% CO2.

2.4. Enzymatic assay

HDAC inhibition assays were performed by the Reaction Biology Corporation (Malvern, PA) with full length human recombinant HDAC1, 2, 3, 6, 8 and 11. These HDACs substrates were all expressed in baculovirus-infected Sf9 cells. Substrate I, a monoacetyl fluorogenic peptide derived from p53 (RHKKAc, residues 379–382), was used for HDAC 1, 2, 3, 6 and 11 enzymatic assays. Substrate II, a fluorogenic
diacetyl peptide derived from p53 (RHKAcKAc, residues 379–382) was used especially for HDAC8. The miXture of each reaction contained 50 mM Tris-HCl, pH 8.0, 137 mM NaCl, 2.7 mM KCl, 1 mM MgCl2, and 1 mg/mL Bovine serum albumin. The substrates in 1% (v/v) DMSO (final concentration in the reaction miXture) were added into the re- action miXture and incubated for 2 h at 30 °C. Compounds were tested in the 10-dose IC50 mode in duplicate with 3-fold serial dilutions which started at 10 μM.

2.5. Cell proliferation assay

Cell proliferation was assessed using the MTT assay. Cells were seeded in 96-well plates at a density of 5 × 103 cells per well and al- lowed to attach for 24 h before treatment. A serial dilution of the test compounds, SAHA, and tubastatin A were added to the culture medium such that the final concentration of DMSO was 0.1% in all reactions. After 72 h, 50 μL (5 mg/ml) of 3-(4,5-dimethylthiazol-2-yl)-2,5-di-
phynyl-tetrazolium bromide (MTT) reagent was added to each well.

After 4 h incubation at 37 °C, the supernatant was aspirated, and the formazan crystals were dissolved in 60 μL of DMSO at 37 °C for 10 min with gentle agitation. The absorbance was measured at 570 nm using a Molecular Device Microplate Reader. The results are presented as the mean ± standard error from at least three independent experiments.IC50 values were calculated based on relative viability values and concentrations by regression analysis.

2.6. In vitro scratch assay

450.63 mmol) and methyl-4-(bromomethyl) benzoate (35.47 g,150.21 mmol) in acetone (350 mL), and the miXture was stirred at 60 °C for 3 h. After 3 h, the solvent was evaporated, and the residue was re- suspended in water (1.2 L) and extracted with (1) EtOAc (1 L), (2) CH2Cl2 (1 L), (3) EtOAc (1 L), and (4) CH2Cl2 (1 L). The organic layer was dried over MgSO4 and evaporated to give a crude miXture. The crude miXture was suspended in MeOH (150 mL), filtered and dried in a vacuum to obtain the target ester (42 g, 67%). Rf = 0.43 (EtOAc/Hexane = 1:2); ESIMS (+) m/z [M + H]+ 415. NaOH (3.48 g, 86.86 mmol) in 2 M NH2OH (2 M in MeOH) (326 mL) was added, and the miXture was stirred in an ice bath for 10 min. After 10 min, the methyl ester inter- mediate (18 g, 43.43 mmol) and 2 M NH2OH (2 M in MeOH) (50 mL) were added, and the miXture was stirred at 30 °C for 2 h. The resulting solution was diluted with H2O (900 mL) and extracted with EtOAc (1 L × 2). The organic layer was dried over MgSO4 and evaporated to give Cells were seeded in triplicate in 6-well plates at a density of 1× 105 cells per well. Cells were grown in medium supplemented with 10% FBS until they reached confluence. Confluent monolayers of cells were carefully scratched with a 1000 μL pipette tip to create a wound, and three different fields of each wound were then imaged with a phase-contrast microscope at the time of the scratch and 24 h later. Three independent experiments were performed. The width of each wound was measured for each experimental condition. At beginning, the wound size was measured and scored as 100%. After 24 h, the width of the residual wound was measured, and the average percentage of wound closure was calculated using the free software UN-SCAN to calculate the number of piXels in the wound area.

2.8. Immunofluorescence autophagy flux assay

Autophagy fluX was determined by immunofluorescence micro- scopy (EVOS imaging system, Thermo Fisher Scientific Inc.) using a Cyto-ID Autophagy Detection Kit (Enzo Life Sciences, Farmingdale, NY, USA) following the manufacturer’s protocol. Briefly, cells were seeded on microscope slides coated with poly-D-lysine (P4158, Sigma, USA)
and incubated with growth medium. After treatment with compound (J22352, TUB or rapamycin) for 24 h at 37 °C, Cyto-ID solution con- sisting of 0.5 mL of PBS, 5% FBS, and 0.25 μL of Cyto-ID was added, and the cells were counter-stained with LysoTracker® Red DND-99 (Invitrogen) for 30 min at 37 °C in the dark. Cells were washed twice with PBS containing 5% FBS and analyzed for their autophagosomal content by measuring the Cyto-ID (fluorescein isothiocyanate) level and the Abs 577/EX 590 nm for LysoTracker® Red DND-99.

2.9. Characterization of T cell subsets

Blood samples were collected from mice after 14 days of compound treatment. Prior to staining with monoclonal antibody, red blood cells were lysed using 1X lysis buffer (BD Pharm Lyse™) for 10 min at room temperature and washed twice with staining buffer (2% FBS-PBS). Following the manufacturer’s protocol, the cell suspensions were in-
cubated on ice with fluorochrome-coupled antibodies to CD4+ and CD8+ (both from BD Biosciences) for 30 min. The characteristics of the T cells were confirmed by flow cytometry, and the data were analyzed by cell quest software (both from FACSCalibur, BD Biosciences). 1× 104 total events were counted for each sample.

2.10. Cytokine measurements

Cytokine quantification in whole blood was performed with BD cytometric bead array (CBA) mouse inflammatory cytokine kits (BD kit 560485) for IL-2, IL-6, IL-17, and IFN-γ according to the manufacturer’s protocols, as measured with a FACScan flow cytometer equipped with an argon laser, and evaluated with the CBA software FCAP Array v1.0
(all by BD Biosciences).

2.11. Animal studies

Male nude mice (BALB/cAnN.Cg-FoXnlnu/CrlNarl, 4–6 weeks old) and C57BL/6 (4–6 weeks old) mice were purchased from the National Applied Research Laboratories (Taiwan). To evaluate the therapeutic efficacy of HDAC6 inhibitors, a subcutaneous tumor was established by inoculating U87MG cells (1 × 106 per mouse, Xenografts in nude mice) or GL261 cells (1 × 106 per mouse, allografts in C57BL/6 mice) into the abdominal left hind flank. When tumor grafts had reached an average volume of 150 mm3, animals were randomized into groups of siX mice. For intra-peritoneal injections, compounds were prepared in PBS daily. An administration volume of 10 mL/kg of body weight was used for intraperitoneal administration. Concentrations are indicated in the main text and figures. Tumor growth was monitored every day. Tumor volume was measured from two directions with a digital caliber and calculated as follows: length × width2 × 1/2. The weight of the mice was determined as an indicator of tolerability on the same days. Tumor growth inhibition (TGI) was calculated according to the formula [1 − (T − T0)/(C − C0)] × 100, where T and T0 are the mean tumor vo- lumes at Day 14 and Day 1, respectively, for the experimental group, and C and C0 are mean tumor volumes for the vehicle control group. After 14 days of treatment, the animals were sacrificed by cervical dislocation. Tumor samples were harvested from animals and post-fiXed in 4% paraformaldehyde and weighed. Animal studies were carried out in accordance with the guidelines and approved by Institutional Animal Care and Use Committee, National Taiwan University College of Medicine (IACUC No. 20120452). All methods and procedures in this study were performed in accordance with the guidelines of the Animal Research regulations of the Agriculture Council (R.O.C.) and also the National Institutes of Health (U.S.A.) regarding laboratory animal welfare.

2.12. Statistical analyses

Statistical analyses were performed with GraphPad Prism 6 using the following tests: Bartlett’s test, one way ANOVA, Dunnett’s multiple comparisons test; p-values < 0.05 were considered as statistically significant.

3. Results

3.1. J22352 is an effective HDAC6 inhibitor

Quinazolin-4-one-based HDAC6 inhibitors with moderate selectivity have been developed by our group [30,31]. To further improve the selectivity profile, the core of quinazolin-4-one was replaced with quinazoline-2,4-dione by scaffold hopping, which enabled the attach- ment of the linker group at central position of the cap moiety, to make the inhibitor fit better into the catalytic site of HDAC6 than quinazolin- 4-one. Compounds, J22352 and J27820, were developed based on this rational design and were identified as highly selective HDAC6 in- hibitors (Fig. 1a). In an enzymatic assay using a panel of 7 HDAC iso- forms from three different classes of HDACs (class I, II, and IV), J22352 inhibited HDAC6 with an IC50 value of 4.7 nM and was 2000-fold more selective for HDAC6 than for the class I HDACs (HDAC1, HDAC2 and HDAC3), with little activity against HDAC8 (Table 1). J27820, an analog of J22352, was synthesized by replacing the 3-position of the phenylethyl group with a phenyl group to improve the water solubility. J27820 maintained potent HDAC6 inhibitory activity (IC50 = 5.3 nM), but its selectivity for HDAC6 over HDAC1 was 582, which was 4-fold lower than that of J22352 (Table 1). Thus, we chose J22352 for further study. Moreover, J22352 showed an improved selectivity profile for the specific inhibition of HDAC6 compared to existing HDAC6 inhibitors, including TUB and ACY-1215.

Next, the effects of J22352 on the HDAC6 substrate, acetylated α- tubulin, were evaluated. Treatment of the U87MG glioma cell line for 24 h with J22352 or TUB led to a dose-dependent increase in acetylated α-tubulin expression levels with no increase in the levels of acetylated histone H3, a marker of class I HDAC inhibition (Fig. 1b). Additionally, treatment with SAHA increased the levels of both acetylated α-tubulin and histone H3 (Fig. 1b). These results indicate that J22352 has a significant effect on the enzymatic inhibition of HDAC6, confirming that this compound is a more selective HDAC6 inhibitor.

HDAC6 is a microtubule-binding protein that regulates microtubule dynamics through deacetylation of α-tubulin, which is involved in cell migration[32,33]. Thus, the inhibitory effects of J22352 on glioblastoma cell migration were evaluated by a scratch assay. U87MG cells were treated with vehicle or 10 μM J22352, TUB or SAHA for 24 h after a scratch was made in a cell monolayer. While the migration of treated cells varied considerably in response to treatment, the most pronounced anti-migratory effect was noted for J22352 (Fig. 1c). This analysis showed that J22352 significantly decreased tumor cell migration, with an IC50 value of 0.21 μM at 24 h. Moreover, the ability of J22352 to inhibit migration was approXimately 5- and 18-fold greater than that of TUB and SAHA, respectively.

3.2. In vitro anti-proliferative activity of J22352

To investigate the cytotoXicity of this highly selective HDAC6 in- hibitor, J22352 was evaluated in anti-proliferative assays involving a panel of 9 human cancer cell lines. As shown in Table 2, J22352 was effective in nine cell lines derived from solid tumors at micro-molar concentrations and showed only slight cytotoXic effects on two non- cancerous cell lines (FHs173We and Vero cells). Notably, J22352 was most effective against U87MG glioma cells, with an IC50 value of 1.56 μM. The anti-proliferative effect of J22352 in U87MG cells was 2.2-fold greater than that of SAHA (Table 2). In addition, J22352 was more cytotoXic than TUB against glioblastoma, potentially due to its greater selectivity for HDAC6. Fig. 1d illustrates the dose-dependent effects of J22352 on U87MG cell proliferation in a series of concentrations.

3.3. Degradation of HDAC6 in cancer cell lines

Some studies have reported that HDAC6 is overexpressed in certain cancer cells and the high expression of HDAC6 promotes cell pro- liferation, migration and contributes to drug resistance in cancer [14,34]. Surprisingly, we found that J22352 decreased the aberrant overexpression of HDAC6 in glioblastoma. As shown in Fig. 2a, a dose- dependent decrease in HDAC6 protein abundance was observed upon treatment with J22352, J27820 and TUB, three highly selective HDAC6 inhibitors (over 500-fold selectivity for inhibiting HDAC6 over HDAC1); however, no remarkable decrease in HDAC6 abundance was observed after treatment with ACY-1215, which has weak selectivity for inhibiting HDAC6 (12-fold HDAC6/HDAC1 selectivity). To explore whether the decrease in HDAC6 abundance is associated with the ubiquitin proteasome system, we utilized the well-characterized pro- teasome inhibitor MG132, which blocks the proteolytic activity of the proteasome to reduce the degradation of ubiquitin-conjugated proteins in mammalian cells [22]. Cells were examined by treating with a combination of J22352 and MG132. As shown in Fig. 2b, no decrease in HDAC6 abundance was observed after co-treatment with J22352 and MG132. Compared to single treatment with J22352, the combination treatments stop the reduction of HDAC6 abundance. These results support the hypothesis that J22352 decreases HDAC6 abundance through the proteasomal degradation pathway. Similar results were also observed after TUB treatment, but no differences were observed upon treatment with ACY-1215 (weakly selective HDAC6 inhibitor) or SAHA (pan-HDAC inhibitor) (data not shown).
In addition to the ubiquitin proteasome system, the autophagosome-
lysosome pathway also plays an important role in regulating protein degradation in cells [22]. Thus, further studies were performed to determine whether the decrease in HDAC6 also involves the autopha- gosome-lysosome pathway. Cells were treated with bafilomycin A1 (baf-A1) [35], an autophagy inhibitor that blocks the autophagosome- lysosome pathway, or rapamycin [36], an autophagy enhancer that facilitates the autophagosome-lysosome fusion. HDAC6 levels were the same in the baf-A1- or rapamycin-treated group and the control group (Fig. 2c). Manipulating autophagosome-lysosome fusion did not affect HDAC6 levels, indicating that the decrease in HDAC6 abundance occurs predominately through proteasomal degradation and that the autophagosome-lysosome pathway is not involved. Interestingly, the combination of J22352 and baf-A1 significantly decreased HDAC6 abundance (P < 0.05). However, no decrease in HDAC6 abundance was observed upon treatment with the combination of J22352 and rapamycin. These results indicate that J22352 dominates the de- gradation of HDAC6 and that the degradation effect is attenuated by rapamycin-enhanced autophagosome-lysosome fusion. These results imply the presence of cross-talk between the ubiquitin proteasome and autophagy-lysosome systems, in which HDAC6 plays an important role.

3.4. J22352 Inhibits autophagosome-lysosome fusion and causes autophagic cancer cell death

The results of our proliferation assays prompted us to analyze the changes in cell morphology following treatment with J22352. When U87MG cells were treated with 5 μM J22352, there was a noticeable accumulation of autophagic vacuoles at different time intervals in the absence of apoptotic cell death (Fig. 3a). This result suggests that J22352 might be involved in autophagy. In keeping with this hypoth- esis, no caspase 3, 8 and 9 activity was observed after treatment for 72 h (data not shown), indicating that J22352 does not engage apoptotic pathways. However, both autophagic and apoptotic cell death were observed after treatment with SAHA, a pan-HDAC inhibitor.

To investigate the effect of J22352 on autophagy in U87MG cells, the fluorescence intensity of a LC3 (a marker of autophagy) probe was measured after treatment with J22352, TUB or rapamycin (which in- creases autophagic fluX). The accumulation of LC3-puncta (which refer to autophagic vacuoles) in the cytoplasm was observed only in the J22352-treated group. J22352 also had a stronger effect than TUB on the density (2.1-fold increase) and number (2.4-fold increase) of LC3- puncta (Fig. 3b). Western Blot analysis revealed that the expression levels of autophagy-related proteins (LC3 and p62) increased in a time- dependent manner after treatment with 10 μM J22352 (Fig. 3c). p62, a ubiquitin-binding factor that colocalizes with ubiquitinated protein aggregates, is often used as a marker of the inhibition of autophagic degradation [37].

As the expression levels of LC3 and p62 increased concomitantly, our results indicate that J22352 markedly induces the accumulation of autophagic vacuoles and inhibits autophagic fluX in glioblastoma.

To determine the underlying cause of the accumulation of LC3-po- sitive autophagic vacuoles in the cytoplasm after J22352 treatment, U87MG cells were incubated with J22352 either alone or in combina- tion with rapamycin or baf-A1, which blocks autophagic fluX by in- hibiting the fusion of autophagosomes and lysosomes. As shown in Fig. 3d, LC3 and p62 levels were dramatically increased in cells treated with J22352, confirming that J22352 blocks autophagic fluX. p62 levels were slightly reduced in cells treated with J22352 and rapamycin compared to those treated with J22352 alone, indicating that the in- hibitory effect of J22352 on autophagy was diminished by rapamycin. Conversely, treatment of cells with J22352 and baf-A1 led to an in- crease in the levels of both LC3 and p62 (Fig. 3d). Similar results were observed in TUB-treated cells, but J22352 had a greater effect on in- hibiting autophagic fluX than TUB had. This result revealed the sig- nificant inhibition of autophagosome-lysosome fusion by J22352 to modulate autophagy-mediated degradation.

A fluorescence imaging assay was used to further characterize the HDAC6 inhibitor-mediated blockade of autophagic fluX. When autop- hagosomes fuse with lysosomes, the acidic environment of the lyso- somes (labeled in red) quenches the green fluorescence of LC3, resulting in a shift from green to orange. This shift was observed in cells treated with rapamycin but not in those treated with J22352 (Fig. 3e). These results collectively indicate that autophagosomes in J22352-treated cells are unable to fuse with lysosomes; autophagic fluX is subsequently blocked, resulting in the accumulation of autophagic vacuoles.

3.5. In vivo anticancer activity of J22352

To investigate the in vivo anticancer activity of J22352 and J27820, we used a murine xenograft model in which U87MG cells were sub- cutaneously injected into the left flanks of nude mice. When tumors reached 150 mm3, mice were treated with test compounds or phos- phate-buffered saline (PBS) via intra-peritoneal injection at a dose of 10 mg kg−1 per day. Mice in the vehicle group showed substantial tumor growth. Tumor growth was only modestly inhibited by 14 days of treatment with TMZ. However, treatment with the HDAC6 inhibitor J22352 or its analog J27820 had a marked inhibitory effect on tumor growth (Fig. 4a). Treatment with J22352 led to a three-fold reduction in tumor weight compared to TMZ treatment (Fig. 4b). Notably, the marked anti-tumor effects of either HDAC6 inhibitor resulted in a > 80% tumor growth inhibition (TGI) rate, whereas that for TMZ was only 41%. No significant weight loss was observed, suggesting that J22352 and J27820 are well tolerated in mice (Fig. 4b).

3.6. J22352 enhances the immune response

Hematology assays were performed on mice after 14 days of treat- ment with J22352. Notably, the numbers of white blood cells, neu- trophils and lymphocytes were greatly increased in treated mice compared to control mice. However, there were no significant differ- ences among all experimental groups in other hematological para- meters, such as red blood cell count, hemoglobin, erythrocyte indices (MCV, MCH and MCHC) and platelets (Table 3). The increased neu- trophil and lymphocyte levels indicated that J22352 may be involved in immune system modulation.

To determine the impact of the HDAC6 inhibitors on the anti-tumor immune response, T cell function was assessed in immunocompetent C57BL/6 mice bearing tumors generated by injection of the murine glioma cell line GL261; these mice were subsequently treated with the specific HDAC6 inhibitors J22352 and J27820 at 10 mg kg−1. As ex- pected, untreated mice showed lower induction of T cells (Fig. 5a), especially CD8+ T cells, than animals not injected with tumor cells. Surprisingly, after 14 days of treatment with J22352 or J27820, we observed a dramatic increase in the CD8+ T cell population in treated mice compared to control mice (up to 3.4- and 4.2-fold increases, re- spectively). Notably, the CD8+ T cell population was markedly higher after J22352 or J27820 treatment than after ACY-1215 treatment (weakly selective HDAC6 inhibitor that is currently used in clinical trials [21]). There were no differences among all experimental groups in the CD4+ T cell population (Fig. 5b, c). We provide proof-of-concept of cancer immunotherapy by using highly selective HDAC6 inhibitors to activate CD8+ T cell proliferation.

In addition, cytometric bead arrays were utilized to measure the levels of inflammatory cytokines in peripheral blood. IL-2 levels were higher in mice treated with 20 μM J22352 or 50 μM J27820 than in vehicle-treated mice (1.2- and 2.3-fold higher, respectively) (Fig. 5d). Moreover, interferon (IFN)-γ levels were also markedly higher (by 2.4- and 3.5-fold, respectively). In both cases, these responses were increased in a dose-dependent manner. In contrast, IL-6 levels showed a dose-dependent decrease in the groups treated with J22352 or J27820 compared to the control group or the ACY-1215-treated group. And IL-17 levels showed no significant differences among all experimental groups. Together, these results indicate that HDAC6 inhibition by either J22352 or J27820 activates the host immune response and that the resulting effect is stronger than that of ACY-1215.

It has been reported that the proliferation of tumor-specific CD8+ T cells is negatively regulated by the PD-1/PD-L1-mediated adaptive immune response and that PD-L1 expression is controlled by signal transducer and activator of transcription 3 (STAT3) signaling [38,39]. We next used Western Blot to examine whether the immune response after J22352 treatment is mediated by STAT3/PD-L1 signaling or is further upregulated by the JAK/STAT3 pathway [40]. We observed a dose-dependent decrease in phospho-STAT3 and PD-L1 protein levels in U87MG cells, with IC50 values of 4.6 μM and 5.3 μM, respectively (Fig. 6a, 6b). Similar effects were observed after SAHA and TUB treatment, but the effects were less robust. However, no effects on JAK protein expression were observed in any of the experimental groups, suggesting that J22352 may exert immune regulatory activity through the STAT3/PD-L1 signaling pathway, not the JAK/STAT3 pathway.

4. Discussion

J22352, a novel compound designed by molecular engineering and based on a quinazoline scaffold, is herein identified as a potent and highly selective HDAC6 inhibitor (IC50 value: 4.7 nM) with 2000-fold selectivity over class I HDACs. Unlike other HDAC isozymes, HDAC6 is aberrantly overexpressed in some cancers and promotes tumorigenesis, metastasis and resistance to TMZ in glioblastoma [15,41–44]. In this
study, we found that J22352 markedly inhibited the glioma cell pro- liferation (U87MG) in a dose-dependent manner (Fig. 1d).

Notably, HDAC6 expression was significantly decreased after J22352 treatment in glioblastoma. As shown in Fig. 2a, only highly selective HDAC6 inhibitors such as J22352 and TUB caused the dose- dependent proteolysis of HDAC6; however there was no significant change in HDAC6 expression upon treatment with ACY-1215, which shows weak selectivity toward HDAC6 and causes apoptotic rather than autophagic cancer cell death [21].

HDAC6 has a unique structure containing two catalytic domains and a c-terminal zinc finger ubiquitin-binding domain (ZnF-UBP) that serves as an E3 ligase binding site to recruit the ubiquitin-conjugated protein [45]. According to data in a publication by Harding et al. [46], we hypothesized that the interaction between J22352 and the ZnF-UBP of HDAC6 may induce the conformational change that results in HDAC6 ubiquitination, thus causing the formation of ubiquitin-conjugated protein that is loaded onto dynein for proteasomal degradation. HDAC6 is thought to mediate the transport of the ubiquitin complex through α tubulin to the proteasome [45].

Based on the results described above, we presume that HDAC6 ex- pression level may be further decreased through the accumulation of p62, which would promote the delivery of ubiquitinated HDAC6 pro- tein for proteasomal degradation after J22352 treatment. As expected, there was a significant decrease in HDAC6 abundance after treatment with J22352 alone or in combination with baf-A1, which also causes p62 accumulation by blocking autophagic fluX (Fig. 2c). In addition, there was no change in HDAC6 abundance upon treatment with J22352 and rapamycin because the rapamycin-mediated increase in autophagic fluX and acceleration of p62 degradation in lysosomes attenuated the effect of HDAC6 degradation. However, p62 accumulation in response to baf-A1 treatment alone did not result in a decrease of HDAC6. There does not appear to be a direct causal correlation between p62 accu- mulation and HDAC6 protein degradation. The observed results may be attributed to baf-A1, which does not interact with the ZnF-UBP of HDAC6, whereas J22352 interacts with the active site of ZnF-UBP. In- terestingly, as shown in Fig. 2b, HDAC6 protein degradation was ob- served after a single treatment with J22352, and the combination of J22352 and MG132 showed no decrease of HDAC6 levels were ob- served. Collectively, these results support the conclusion that HDAC6 levels likely decrease as a result of J22352 interacting with the ZnF-UBP of HDAC6 and triggering HDAC6 ubiquitination, followed by protea- somal degradation instead of autophagosomal-lysosomal degradation.

These results are similar activity to those obtained with PROTACs (proteolysis-targeting chimeras). PROTACs are a combination of a therapeutic agent and a moiety recognized by an E3 ligase through a linker, which can recruit the ubiquitin proteasome system to degrade intracellular target proteins [47,48]. Unlike traditional drug develop- ment strategies that aim to inhibit aberrant or disease-causing protein function by using small molecules that block active sites of enzymes, the approach with PROTACs is to eliminate target protein expression rather than to inhibit it. Therefore, PROTACs have significantly broader therapeutic applicability with greater efficacy and fewer off-target ef- fects [49]. These results collectively illustrate that a highly selective HDAC6 inhibitor can be considered not only an enzymatic inhibitor but also an enzymatic eliminator which serves a PROTAC-like property leading to the proteolysis of HDAC6. However, it remains unclear why a small molecule such as J22352 which has similar functions as those of PROTACs in promoting the ubiquitination and degradation of ubi- quitin-conjugated proteins. The underlying mechanism by which HDAC6 inhibitors regulate the ubiquitin system need to be further clarified.

In addition to mediating the transportation of ubiquitinated proteins to proteasomes for degradation, HDAC6 is a microtubule-binding pro- tein and involves in cell migration through the deacetylation of α-tu- bulin [32]. As expected, scratch assays revealed that J22352 sig- nificantly decreased glioma cell migration; the inhibition of cell migration by J22352 was more than three-fold greater than that by SAHA and TUB (Fig. 1c). These results indicate that J22352 is a highly selective HDAC6 inhibitor in glioblastoma and functions by targeting the deacetylation of α-tubulin in the cytoplasm to inhibit cancer me- tastasis. Importantly, we observed that J22352 not only acetylates α- tubulin but also inhibits autophagy by blocking the fusion of autophagosomes with lysosomes, leading to increased metabolic stress and, ultimately, autophagic cancer cell death (Fig. 3). These findings are consistent with those in previous reports indicating that HDAC6 is in- volved in the regulation of autophagy [22,23]. We provide more evi- dence that a highly selective HDAC6 inhibitor interferes with autop- hagy by blocking the fusion of autophagosomes with lysosomes, an activity that has not been observed with pan-HDAC inhibitors.

Emerging evidence suggests that targeting autophagy may po- tentiate anticancer effects through the regulation of cell metabolism [50]. Uncontrolled cell division, accompanied by hypoXia and nutrient deficiency, is commonly observed in many types of solid tumors, especially in glioblastoma, and this characteristic of tumorigenesis leads to enhanced autophagy, which promotes cancer progression and metastasis [51]. This mechanism represents a survival strategy adopted by solid tumor cells, particularly in response to cellular stress induced by chemotherapy and radiation therapy [52]. Glioma cells are highly dependent on autophagy for adaptive responses to radiation and TMZ treatment [25], indicating that the inhibition of autophagy may have therapeutic benefits in glioblastoma patients. Indeed, J22352 demon- strated not only having cytotoXic effects against glioblastoma but also potential inhibition of autophagy; these dual functions of J22352 pro- vide valuable insight for better understanding and utilizing selective HDAC6 inhibitors in anticancer therapy.

The inhibition of autophagy could also enhance the host immune response against tumors [53]; this phenomenon has already been ob- served in several different cancer types [26,54]. For example, in the treatment of refractory colorectal cancer patients, the combination of SAHA (a pan-HDAC inhibitor) and hydroXychloroquine (an autophagy inhibitor) evoked an improved anticancer immune response without unacceptable side effects [55]. Recent investigations have indicated that cancer cells with high PD-L1 expression levels are more sensitive to autophagy inhibitors than those with low PD-L1 expression [56]. The underlying mechanisms linking autophagy and the immune response are not completely understood, but Noman et al. recently reported that blocking autophagy could suppress p-STAT3-dependent survival me- chanisms, making tumor cells more susceptible to attack by cytotoXic T cells [57]. Consistent with this hypothesis, our data imply that the J22352-mediated suppression of p-STAT3 may lead to the up-regulation of PD-L1.

Next, investigating the role of an HDAC6 inhibitor in the immune
regulation associated with autophagy in glioblastoma, we obtained findings that collectively show that J22352 stabilizes α-tubulin and blocks the fusion of autophagosomes with lysosomes, resulting in the accumulation of p62, which subsequently triggers proteasomal de- gradation of p-STAT3. These results confirm a report by Janji et al. on p62 accumulation and subsequent p-STAT3-mediated degradation using an autophagy inhibitor [58]. Moreover, we found that p62 ac- cumulation in response to J22352 treatment resulted in decreased p- STAT3 levels, consequently lowering PD-L1 protein expression and re- ducing its inhibitory regulation of the host immune response (Fig. 7). This finding can be attributed to the attenuation of negative regulation by PD-L1, further resulting in the reactivation of T cells targeting glioblastoma (Fig. 5). Our study shows that J22352 promotes CD8+ T cell activation through sequential inhibition of autophagy and a de- crease of PD-L1 expression (Figs. 5b, c and 6a).

Collectively, treatment of immunocompetent mice with J22352 increased the number of CD8+ T cells and the levels of the tumor-as- sociated inflammatory cytokines IL-2 and IFN-γ and reduced IL-6 levels (Fig. 5b, c, d). The increased IL-2 and IFN-γ levels after sustained J22352 treatment indicated ongoing T cell proliferation and immune activation. In addition, IL-6 was previously determined to be highly expressed in gliomas and required for glioma development [59,60]. Our results show that IL-6 expression levels significantly decrease after J22352 treatment and consequently result in the inhibition of the STAT3/PD-L1 signaling pathway. Moreover, no changes in the CD4+ T cell population and IL-17 levels after J22352 treatment indicates J22352 may elicit antitumor immune response mainly through the CD8+ T cell-mediated tumor killing instead of the CD4+ T cell-asso- ciated suppressive activity [61].

Taken together, our data demonstrate that J22352 plays dual roles in anticancer activity by inhibiting autophagy and inducing the im- mune response against glioblastoma. J22352 treatment resulted in significant TGI in vivo by up to 90% in glioblastoma-bearing mice with limited cytotoXic effects (IC50: 1.56 μM) (Fig. 4a). This result can be
ascribed to J22352 targeting the tumor microenvironment and inducing immune activation, leading to enhanced anti-tumor responses in vivo [62].

Currently, immunotherapies using antibodies to elicit specific immune responses have been evaluated as promising therapeutics against aggressive malignancies. However, approXimately 30% per- centage of patients currently benefit from anti-PD-1/PD-L1 therapy [63], and antibodies may cause unfavorable side effects known as im- mune-related adverse events (irAEs) that affect the dermatologic, gas- trointestinal, hepatic, and endocrine systems, among others [64,65]. In this study, we demonstrated that the treatment of glioblastoma with a selective HDAC6 inhibitor increased the levels of immune-activating cytokines and the proliferation of CD8+ T cells by decreasing negative regulation of PD-L1. To the best of our knowledge, this is the first report to show that a high-selective HDAC6 inhibitor can significantly inhibit glioblastoma by not only inhibiting autophagy associated with de- creased PD-L1 expression but also evoking PROTAC-like proteolytic degradation of HDAC6 through the proteasomal degradation pathway. These findings will provide new directions for targeting HDAC6 by developing PROTACs against cancers. We anticipate that a high-selective HDAC6 inhibitor, such as J22352, will provide better opportunities for the treatment of HDAC6-overexpressing ACBI1 cancers.