Inhibition of glioblastoma cell proliferation, invasion, and mechanism of action of a novel hydroxamic acid hybrid molecule

Glioblastoma multiforme is one of the most aggressive brain tumors and current therapies with temozolomide or suberoylanilide hydroxamic acid (SAHA, vorinostat) show considerable limitations. SAHA is a histone deacetylase (HDAC) inhibitor that can cause undesirable side effects due to the lack of selectivity. We show here properties of a novel hybrid molecule, sahaquine, which selectively inhibits cytoplasmic HDAC6 at nanomolar concentrations without markedly suppressing class I HDACs. Inhibition of HDAC6 leads to significant α-tubulin acetylation, thereby impairing cytoskeletal organization in glioblastoma cells. The primaquine moiety of sahaquine reduced the activity of P-glycoprotein, which contributes to glioblastoma multiforme drug resistance. We propose the mechanism of action of sahaquine to implicate HDAC6 inhibition together with suppression of epidermal growth factor receptor and downstream kinase activity, which are prominent therapeutic targets in glioblastoma multiforme. Sahaquine significantly reduces the viability and invasiveness of glioblastoma tumoroids, as well as brain tumor stem cells, which are key to tumor survival and recurrence. These effects are augmented with the combination of sahaquine with temozolomide, the natural compound quercetin or buthionine sulfoximine, an inhibitor of glutathione biosynthesis. Thus, a combination of agents disrupting glioblastoma and brain tumor stem cell homeostasis provides an effective anti–cancer intervention.


Introduction
Glioblastoma multiforme (GBM) is the most common and aggressive form of brain cancer, with limited treatment options and dismal survival rates. Current treatment involves surgical resection followed by radiotherapy and chemotherapy with temozolomide (TMZ) 1 . However, more than half of GBM patients do not respond to TMZ due to the overexpression of DNA repair enzymes, notably O 6 -methylguanine transferase [2][3][4] .
Histone deacetylase (HDAC) inhibitors exert anticancer effects by inducing cell differentiation, cell cycle arrest, and apoptotic cell death through the upregulation of tumor suppressor and cell cycle-regulatory genes 5 . Suberoylanilide hydroxamic acid (SAHA, vorinostat) is a Food and Drug Administration-approved drug for the treatment of cutaneous T cell lymphoma. It is currently in clinical trials for GBM as monotherapy and combined with radiotherapy [6][7][8][9] . Despite advancements in treatments, the median survival rate for GBM remains low (14-16 months) and new therapeutic options are urgently needed 3,10 .
In this study, we combined hydroxamic acid-the active moiety of SAHA exerting biological effects in cancer cellswith primaquine to generate a new class of hybrid anticancer agents: sahaquines. Hydroxamic acid inhibits HDACs; these enzymes are overexpressed in many cancers, including GBM 11,12 . The hydroxamic acid pharmacophore of SAHA chelates metal ions, thereby inhibiting metalloenzymes such as HDACs and matrix metalloproteinases (MMPs), which promote cancer growth and invasiveness [13][14][15] . Hydroxamic acid is a weak acid, which is favorable in the acidic tumor microenvironment as weak bases become protonated, resulting in ion trapping, lysosomal accumulation, elimination by lysosomal exocytosis, and overall decreased biological activity [16][17][18] . Primaquine can directly interfere with endosomal trafficking to the plasma membrane 19 , inhibit the multidrug resistance transporter P-glycoprotein, and autophagy, thereby sensitizing cancer cells to anti-mitotic drugs 20,21 . Considering that monotherapies have limited effectiveness in GBM, we tested sahaquine in combination with TMZ, the standard of care for GBM, quercetin, and buthionine sulfoxamine. Quercetin is an abundant flavonoid found in fruits and vegetables, such as apples and onions. Its estimated daily intake ranges from 3-40 mg, but supplements up to 1000 mg per day are considered safe 22 . Although it shows no toxicity in normal cells, several studies have shown that quercetin has anticancer effects. Its mechanism of action involves the upregulation of pro-apoptotic and downregulation of anti-apoptotic factors, cell cycle arrest, and DNA intercalation, resulting in DNA damage, activation of apoptosis, and cell death 23 . In animal studies, quercetin inhibited tumor growth and improved the lifespan of tumor-bearing mice 23,24 . Furthermore, the anticancer effects of quercetin are enhanced in combination with chemotherapeutic agents or other drugs [25][26][27] .
We investigated the loss of cell viability and invasiveness in GBM as functional read-outs of the effects of sahaquine alone or in combination with TMZ and quercetin. Sahaquine was tested in both differentiated GBM cells and brain tumor stem cells (BTSCs), which are key to tumor survival and recurrence [28][29][30][31] . Our study supports the model that sahaquine-induced cell death of GBM is mediated through multiple pathways, including inhibition of HDAC6, reduction of epidermal growth factor receptor (EGFR) protein abundance, and decreased activation of downstream kinases AKT and ERK1/2. The primaquine moiety of sahaquine contributes to the inhibition of P-glycoprotein. Considering that sahaquine significantly reduced BTSC viability and markedly inhibited GBM invasion by disruption of GBM homeostasis, further systematic studies are warranted in patient-derived organoids.

Sahaquine synthesis and physicochemical properties of the selected anticancer agents
Sahaquine is a primaquine and hydroxamic acid derivative linked with glutaric acid. It is synthesized in four steps (Fig. 1). The pharmacophore, hydroxamic acid, was introduced in the last step. Yields were good to excellent (50-88%). Sahaquine was fully characterized by conventional spectroscopic and analytical methods (melting point, IR, MS, 1 H-NMR, 13 C-NMR), and the data were consistent with the proposed structure ( Supplementary  Fig. S1). The quinoline ring of sahaquine acts as the capping group and the hydroxamic acid binds zinc. Calculations of physicochemical properties showed that TMZ is a hydrophilic compound (log P = -0.28), whereas sahaquine and particularly quercetin are more lipophilic (log P = 0.92 and 2.16, respectively) ( Table 1). The isoelectric point (pI) values of these compounds vary from 2.9 (quercetin) to 9.2 (SAHA) 32 .
Sahaquine is more potent than TMZ for killing human glioblastoma and BTSCs The half maximal inhibitory concentration (IC 50 ) value of sahaquine (10 µM) was about threefold lower than that of TMZ (31 µM), whereas it was less potent than its parent compound SAHA after 72 h incubation (Fig. 2). Sahaquine precursors were also tested, but because of the relatively high IC 50 values (>50 µM), further experiments were not pursued (Supplementary Table S1). Enhanced cell killing was achieved by combining quercetin with sahaquine in a dose-dependent manner, although quercetin alone showed limited cytotoxicity (IC 50 = 140 µM after 72 h) (Fig. 2e). Combination of TMZ with sahaquine, quercetin, or SAHA at IC 50 concentrations was more effective than any of the compounds alone (Supplementary Fig. S2). Similar results were obtained by measurements of mitochondrial metabolic activity using the MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl tetrazolium bromide) assay ( Supplementary Fig. S3).
We further tested the selected compounds on GBM tumoroids, which are more drug-resistant and representative models of brain tumors in vivo. Sahaquine and TMZ reduced tumoroid sizes by 37 and 40%, respectively, while quercetin did not have a significant effect after 7 days (Supplementary Fig. S4).
Based on the results shown in Fig. 2, we investigated the cytotoxic effects of the selected compounds on BTSCs. BTSCs are a key subpopulation of GBM tumors implicated in tumor initiation, propagation and recurrence 28,30 . In vitro BTSC cultures spontaneously formed neurospheres of approximately 100 µm in diameter within 7 days. Sahaquine and quercetin were most effective at reducing the size of BTSC aggregates and abolishing the formation of neurospheres (Fig. 3).

Sahaquine inhibits GBM invasion and P-glycoprotein activity
GBM is characterized by a diffuse brain tissue distribution 33 . Tumors commonly reoccur within a few centimeters of the original lesion, making surgical resection difficult 33 . We tested the effect of sahaquine and quercetin on GBM migration using a scratch assay, and invasion using a three-dimensional (3D) collagen matrix. Sahaquine did not significantly inhibit cell migration over 24 h, while quercetin reduced cell migration by 20% (Fig. 4a). The combination of sahaquine with quercetin  was most effective, reducing migration by 42%. This effect was not observed when combining sahaquine with TMZ. In contrast, sahaquine significantly inhibited GBM invasiveness, whereas quercetin and TMZ reduced cell invasion by 35 and 45% after 4 days, respectively (Fig. 4c). GBM invasiveness is enabled by MMP degradation of the extracellular matrix and basement membranes 34,35 . We investigated the effect of the selected compounds on the abundance of secreted MMPs using gelatin zymography and showed that quercetin decreases MMP abundance in a dose-dependent manner ( Supplementary Fig. S5). Neither sahaquine nor TMZ reduced MMP concentrations, . Horizontal bars represent the mean (SD) from at least three independent experiments (***p < 0.001 compared to the untreated control; ### p < 0.001 compared to e Q 100 µM alone or f Sq 10 µM alone; Welch's ANOVA with Games-Howell post hoc test). Cell viability was measured by counting Hoechst 33342-labeled nuclei imaged using a fluorescence microscope although the hydroxamic acid moiety in sahaquine can bind zinc within the MMP structure 36 . The primaquine moiety of sahaquine contributed to the inhibition of Pglycoprotein, as assessed by intracellular retention of calcein-AM ( Supplementary Fig. S6). The primaquine concentration within sahaquine (10 µM) effectively inhibited P-glycoprotein activity, whereas 60 µM of unincorporated primaquine was required to achieve a comparable effect. A smaller extent of P-glycoprotein inhibition by SAHA was obtained with equimolar sahaquine concentrations (10 µM) ( Supplementary Fig. S6).

Sahaquine selectively inhibits HDAC6
We further examined the HDAC inhibitory activity of sahaquine compared to its parent compound SAHA. SAHA is a pan-HDAC inhibitor that caused an increase in both acetylated α-tubulin (K40) and acetylated histone H3 (K9/K14) (Fig. 5). We hypothesized sahaquine to be selective toward HDAC6, because its bulky capping group would fit better into the wide binding site of the enzyme 37,38 . Nanomolar concentrations (100 nM) of sahaquine resulted in a 1.5-fold increase in acetylated α-tubulin compared to the untreated control, but did not affect histone acetylation (Fig. 5a). Similar results were obtained with the HDAC6-selective inhibitor ACY-1215 ( Supplementary Fig. S7). TMZ and quercetin did not inhibit HDAC6. These results were supported by Western blot analyses (Fig. 5c). HDAC6 abundance was comparable following all treatments, suggesting that sahaquine inhibited the enzyme activity without affecting its protein levels (Fig. 5d).
Sahaquine reduces EGFR abundance, ERK1/2, and AKT phosphorylation EGFR overexpression and downstream hyperactivity of ERK1/2 and AKT are associated with worse prognosis in GBM 39,40 . We assessed the abundance of these markers and HDAC6 in GBM by immunohistochemistry, and showed an increase in EGFR, dually phosphorylated (active) ERK1/2, phosphorylated (active) AKT, and HDAC6 compared to control brains (Fig. 6a). To test whether sahaquine impinges on EGFR and the activation of downstream kinases, we measured EGFR abundance, dual ERK1/2 phosphorylation (Thr202/Tyr204), and AKT phosphorylation (Ser473) by Western blotting (Fig. 6b). Sahaquine reduced EGFR concentrations in GBM. Interestingly, combining sahaquine and TMZ abrogated this inhibitory effect. Sahaquine also reduced levels of phosphorylated ERK1/2 and phosphorylated AKT, alone and in combination with quercetin or TMZ. Total ERK1/2 and AKT protein levels remained unchanged (Supplementary Fig. S8).

Discussion
Results from this study show that sahaquine is more effective than TMZ in killing glioblastoma and BTSCs, as well as inhibiting glioblastoma invasiveness. The mode of action of sahaquine implicates in part excessive α-tubulin acetylation due to the selective inhibition of HDAC6, resulting in cytoskeletal reorganization ( Supplementary  Fig. S9) and reduced invasiveness. Additional modes of action involve decreased EGFR abundance and downstream activity of AKT and ERK1/2. These results are particularly striking in combination with TMZ or quercetin.
TMZ is one of few clinically approved drugs for the treatment of GBM, but a substantial portion of newly diagnosed tumors and recurrent tumors are resistant to this drug 3,41 . HDAC inhibitors are of particular interest for GBM treatment, as their effectiveness is unaltered by mechanisms of resistance upregulated in GBM, such as mismatch-repair, O 6 -methylguanine methyltransferase and base-excision repair 3,4,42 . The pan-HDAC inhibitor SAHA is currently in clinical trials for GBM, but results so far showed marginal improvement in overall survival (5.7 months compared to 4.4 months) and several serious side effects [6][7][8][9]43,44 . This and other current therapeutic interventions for GBM are ineffective 28,29,31 .
Thus, our goal was to test a new hybrid compound. The development of hybrid molecules is one of the most active areas in therapeutics. Hybrid compounds can have multiple targets, reducing the risk of resistance, lowering effective doses, and decreasing side effects 45,46 . Sahaquine is a hybrid molecule consisting of hydroxamic acid and primaquine linked by a dicarboxylic acid. Primaquine is a strong base (pI = 13.7), but addition of the hydroxamic acid group lowers its pI to 6.48, making sahaquine a weak acid. Weak acids are more advantageous than weak bases as anticancer therapeutics, because they will not be protonated in the acidic tumor environment or trigger lysosomal exocytosis [16][17][18] . Similarly to primaquine, sahaquine can also inhibit P-glycoprotein activity ( Supplementary  Fig. S6), thereby preventing multidrug resistance.
One of the great challenges in GBM treatment is heterogeneity, both within and between tumors [47][48][49][50] . Interpatient heterogeneity has been shown through genomic and transcriptomic analyses by the Cancer Genome Atlas research network 51,52 . Intratumoral heterogeneity can be attributed to the different cellular lineages and subtypes  [56][57][58] . They are resistant to radiation 28 and chemotherapy [59][60][61] , and are thought to promote tumor recurrence 30,31 . Therefore, effective GBM treatment demands a better understanding of tumor origin and heterogeneity to identify new therapeutic targets 3,62 . Sahaquine (10 µM) abolished the formation of BTSC neurospheres and significantly reduced the size of BTSC aggregates. TMZ was less effective, even at a tenfold higher concentration (100 µM). Quercetin was as effective as sahaquine in killing BTSCs, but showed limited cytotoxicity toward differentiated GBM cells. Sahaquine eliminated both BTSCs and differentiated cancer cells.
Another factor contributing considerably to GBM recurrence is tumor invasiveness. While sahaquine abolished invasiveness and contributed to the loss of tumoroid viability, it did not markedly affect the abundance of secreted MMPs. In contrast, quercetin had limited effects on tumoroid viability, but decreased GBM invasion by inhibiting MMP secretion. Quercetin inhibits nuclear factor-κB (NF-κB) nuclear translocation, which could alter MMP expression 63,64 and enhance cell death through NF-κB-dependent regulation of apoptosis.
In an effort to reduce undesirable side effects in normal cells, selective HDAC inhibitors have been developed 65,66 . Ricolinostat (ACY-1215) is a selective HDAC6 inhibitor currently in clinical trials (phase I and II) in combination with pomalidomide for multiple myeloma 67 . Ricolinostat inhibits heat shock protein 90 deacetylation, resulting in an accumulation of unfolded proteins, disruption of protein homeostasis and cell death 68 . We show that sahaquine selectively inhibits HDAC6 at nanomolar concentrations, which distinguishes it from SAHA, which is non-selective at equimolar concentrations. Interestingly, sahaquine significantly reduced the abundance of heat shock protein 70 in GBM (Supplementary Fig. S10) and altered α-tubulin organization. We have previously shown that celastrol disrupts protein homeostasis 69 and the organization of the F-actin cytoskeleton in GBM 70 . Future studies will have to evaluate how sahaquine affects proteostasis in relation to cytoskeletal dynamics.
Many drugs currently in clinical trials aim at inhibiting proteins and proliferation pathways deregulated in GBM, notably HDAC6, EGFR, AKT, and ERK1/2 [71][72][73][74] . Our in vitro studies showing enhanced ERK1/2 and AKT phosphorylation are corroborated by immunohistochemical data in tumor sections from GBM patients (Fig. 6a), also showing markedly stronger signals for HDAC6, EGFR, phosphorylated ERK1/2, and phosphorylated AKT compared to normal brain tissue (Fig. 6a). Sahaquine can decrease the abundance of EGFR, phosphorylated AKT, and phosphorylated ERK1/2 in GBM (Fig. 6b), thereby suggesting that similar hybrid molecules are viable candidates for GBM combination therapy. Interestingly, AKT deacetylation by HDAC6 promotes cancer growth and proliferation 75 , indicating that sahaquine could reduce AKT activation through HDAC6 inhibition.
Drug resistance is a major problem in glioblastoma therapy 3,41 . A recent study of HDAC inhibitors in drugresistant melanoma implicated increased levels of reactive oxygen species 76 . Combination of sahaquine with buthionine sulfoximine, which depletes endogenous glutathione levels 77 , sensitized GBM cells to reactive oxygen species and enhanced cell death ( Supplementary Fig. S11), although buthionine sulfoximine alone in the tested concentration had no effect on GBM viability. Further analysis of the effect of sahaquine on reactive oxygen species production in GBM is warranted.
Taken together, our study reveals sahaquine as a therapeutic agent affecting multiple cellular factors and processes that are critical for GBM treatment (Fig. 7). Sahaquine is superior to the clinical standard TMZ in reducing GBM and BTSC viability, invasiveness, and markers of key survival pathways. These effects are even more profound when sahaquine is combined with TMZ, buthionine sulfoximine, or quercetin. In conclusion, sahaquine is an effective cell death inducer which eliminates not only GBM cells but also BTSCs, thus suggesting that evaluation of sahaquine in combination with other drugs merit further investigations in patient-derived organoids, and eventually in humans.
Cell culture and tissue samples U251N human glioblastoma cells were originally obtained from the American Type Culture Collection. Cells were cultured in Dulbecco's modified Eagle's media (DMEM, Gibco, Thermo Fischer Scientific, Grand Island, NY, USA) supplemented with 5% (v/v) fetal bovine serum (Wisent, St. Bruno, Canada) and 1% (v/v) penicillin-streptomycin (Thermo Fisher Scientific, Eugene, OR, USA) at 37°C with 5% CO 2 and 95% relative humidity, unless otherwise indicated. Glioblastoma samples were harvested under a protocol approved by the Montreal Neurological Hospital's research ethics board (NEU-10-066). Consent was given by all patients. At least 116 brain sections from GBM patients aged 55-76 and controls were used. Tissues were from the frontal, temporal, or parietal lobes of the cerebral cortex. Human BTSCs were expanded as neurospheres in complete NeuroCult TM proliferation media (Stemcell Technologies, Vancouver, BC, Canada). NeuroCult TM basal medium contained: NeuroCult TM NS-A proliferation supplement (1/10), recombinant human epidermal growth factor EGF (20 ng/ml), recombinant human basic fibroblast growth factor (20 ng/ml), and heparin (2 μg/ml).

BTSC viability
48EF human brain tumor cells were seeded at 5,000 cells per well in 96-well plates (Sarstedt, Nümbrecht, Germany) and treated for 7 days. Cells were then imaged using light microscopy (Leica DMI4000B) and the surface areas of the neurospheres were measured in ImageJ (version 1.51s).

Scratch assay
U251N cells were seeded in 6-well plates (Sarstedt, Nümbrecht, Germany) at 1,500,000 cells per well in 1 ml media and cultured for 24 h. The scratch was performed by gently dragging a 200 μl pipette tip across the cell monolayer, after which cells were washed with phosphatebuffered saline and incubated in DMEM with or without treatment. Cytochalasin D (40 nM, Sigma-Aldrich, St. Louis, MO, USA) served as positive control. Predetermined areas of the wells were imaged using light microscopy immediately after the scratch (time = 0 h) and after 24 h. The cell-free area of the scratch was measured in ImageJ. Cell invasion assay U251N tumoroids were prepared using the hanging drop method 78 . Drops of 30,000 cells in 20 µl medium were pipetted onto the inner side of a 10 cm Petri dish (Thermo Fisher Scientific, Eugene, OR, USA) lid. The lid was quickly flipped to cover the Petri dish filled with 20 ml phosphate-buffered saline. Hanging drops were cultured at 37°C for 48 h to allow tumoroids to form. Tumoroids were then gently scooped into a medium-filled Petri dish coated with 2% agarose and cultured for 48 h. Tumoroids were implanted in collagen gel (Advanced BioMatrix, San Diego, CA, USA) mixed with DMEM (1×) and sodium hydroxide (10 mM, Sigma-Aldrich, St. Louis, MO, USA). Gels were covered with 500 µl DMEM with or without treatment. Tumoroids were imaged using light microscopy immediately after implantation (time = 0 day) and after 4 days. The area of cell invasion was measured in ImageJ.

Western blotting
Western blot analysis followed published procedures 70 . In brief, crude extracts were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and blotted onto nitrocellulose membranes. Blocked filters were probed with antibodies against acetyl-α-tubulin Wells were sampled in triplicate and the optical density was measured at 595 nm using a microplate reader (Asys UVM 340, Biochrom, Holliston, MA, USA).

Gelatin zymography
U251N cells were seeded in 60-mm tissue culture dishes (Thermo Fisher Scientific, Rochester, NY, USA) at 1,500,000 cells per dish in 3 ml media and cultured for 24 h. Cells were treated in serum-deprived DMEM for 24 h. Following treatment, culture media were collected and concentrated 15-fold using 30 kDa centrifugal filters (Millipore, Cork, Ireland) following the manufacturer's recommendations. Concentrated media were separated by SDS-PAGE using gelatin (0.1%, w/v) and acrylamide (7.5%, w/v) gels under non-reducing conditions. Gels were washed for 30 min in renaturing solution (2.5% (v/v) Triton X-100 in double-distilled water) and 30 min in developing buffer (50 mM Tris, pH 7.8; 1% (v/v) Triton X-100; 1 μM ZnCl 2 , 5 mM CaCl 2 , adjusted to pH 7.45). Gels were then incubated in fresh developing buffer at 37°C overnight. Gels were stained with 0.5% (w/v) Coomassie Blue G250 (Bio-Rad, Richmond, CA, USA) dissolved in 40% (v/v) ethanol and 10% (v/v) acetic acid for 1 h, and then destained in 40% ethanol and 10% acetic acid diluted in double-distilled water, until clear bands appeared. Quantification of MMP-9 and MMP-2 abundance (as band area) was done in ImageJ.

Tumoroid viability
U251N tumoroids were prepared using the liquid overlay system 79 . The 96-well cell culture plates were coated with 75 µl of 2% (w/v) agarose (Life Technologies, Gaithersburg, MD, USA) dissolved in serum-deprived DMEM. The agarose was cooled for 30 min, then cells were seeded at 5,000 cells per well in 200 µl media, and cultured for 4 days before treatment. Cells were treated for 7 days, and then imaged using a microscope (Leica DMI4000B). The surface area of tumoroids was analyzed in ImageJ.

Calcein-AM uptake
U251N cells were seeded in 96-well black plates at 5,000 cells per well in 0.1 ml media and cultured for 24 h before treatment. Cyclosporine A (Calbiochem, Toronto, Canada) served as a positive control for the inhibition of P-glycoprotein. Following treatment, cells were incubated in phenol-free Hanks' balanced Salt solution containing calcein-AM (0.5 µM, Thermo Fisher Scientific, Eugene, OR, USA) and propidium iodide (3 µM, Sigma-Aldrich, St. Louis, MO, USA) for 30 min at 37°C. The media were replaced with fresh Hanks' balanced salt solution and cells were imaged using a fluorescence microscope (Leica DMI4000B). Cells positively labeled with propidium iodide were excluded from the analysis.

Statistics
Experiments were performed independently at least three times. Unless otherwise indicated, data are shown as mean (SD). Normality of data distribution was assessed by the Shapiro-Wilk test. For sample sizes larger than 30, the Central Limit Theorem allows the assumption of normal distribution. Equality of variances was assessed by Levene's test. If the assumptions of normality and homogeneity of variance were met, two-tailed one-way analysis of variance (ANOVA) with Tukey-Kramer's post hoc test were performed. If homogeneity of variance was not observed, Welch's ANOVA with the Games-Howell post hoc test were used. A p value smaller than 0.05 was considered statistically significant: *p < 0.05, **p < 0.01, and ***p < 0.001.