Structural Basis of Colchicine-Site targeting Acylhydrazones active against Multidrug-Resistant Acute Lymphoblastic Leukemia.

Tubulin is one of the best validated anti-cancer targets, but most anti-tubulin agents have unfavorable therapeutic indexes. Here, we characterized the tubulin-binding activity, the mechanism of action, and the in vivo anti-leukemia efficacy of three 3,4,5-trimethoxy-N-acylhydrazones. We show that all compounds target the colchicine-binding site of tubulin and that none is a substrate of ABC transporters. The crystal structure of the tubulin-bound N-(1'-naphthyl)-3,4,5-trimethoxybenzohydrazide (12) revealed steric hindrance on the T7 loop movement of β-tubulin, thereby rendering tubulin assembly incompetent. Using dose escalation and short-term repeated dose studies, we further report that this compound class is well tolerated to >100 mg/kg in mice. We finally observed that intraperitoneally administered compound 12 significantly prolonged the overall survival of mice transplanted with both sensitive and multidrug-resistant acute lymphoblastic leukemia (ALL) cells. Taken together, this work describes promising colchicine-site-targeting tubulin inhibitors featuring favorable therapeutic effects against ALL and multidrug-resistant cells.

Introduction

Despite the success already achieved in the treatment of children with acute lymphoblastic leukemia (ALL), relapse still occurs in around 20% of patients (Steinherz et al., 1996, Stock et al., 2013). The overall survival rate for relapsed childhood ALL is approximately 40%–60% despite intensive chemotherapy and allogeneic stem cell transplantation (Oskarsson et al., 2015, Roy et al., 2005), highlighting the need for new drugs or new drug formulations. Recent clinical trials in adults (reviewed in Soosay Raj et al., 2013) and children (Shah et al., 2016) with relapsed or refractory ALL have shown promising results with the use of liposomal vincristine. Vincristine has been in clinical use for decades and is one of the main components of every ALL protocol. Liposomal vincristine allowed the use of higher and more frequent doses of this anti-tubulin agent by circumventing its neurotoxic effects, which are likely caused by perturbations of microtubule functions that are essential for axonal transport in neurons (Shah et al., 2016, Soosay Raj et al., 2013).

Tubulin is one of the best validated targets for cancer therapy. There are dozens of new anti-tubulin agents in clinical or late preclinical development, pursuing better therapeutic indexes, i.e., a better trade-off between efficacy and toxicity (Field et al., 2015, Liu et al., 2014, Wood et al., 2001). In this context, our group has designed and tested a series of 3,4,5-trimethoxy-N-acylhydrazones some of which exhibited strong anti-microtubule and anti-leukemia activities, in vitro and in vivo, while showing modest toxicity toward normal proliferating T cells (Salum et al., 2015). In this study, we highlight the biochemical and biological properties of three 3,4,5-trimethoxy-N-acylhydrazones: compounds 7, 12, and 21, which showed the best in vitro anti-leukemic effects in our previous study (Salum et al., 2015).

Results

Compounds 7, 12, and 21 Bind to the Colchicine-Binding Site of Tubulin and Inhibit Microtubule Assembly

Our previously reported computational docking efforts propose that N-acylhydrazones could adopt an appropriate stereochemistry in the colchicine-binding site of tubulin (Salum et al., 2015). To test the interaction of compounds 7, 12, and 21 (Figure 1A) with tubulin, we assessed the UV absorbance spectrum of these compounds (Figure S1) followed by their ability to displace 2-methoxy-5-(2′,3′,4′-trimethoxy)-2,4,6-cycloheptatrien-1-one (MTC), a reversible tubulin ligand targeting the colchicine-binding site (Fitzgerald, 1976).

Figure 1

Compounds 7, 12, and 21 Are Colchicine-Binding Site Tubulin Inhibitors

(A) Derivatives from N-acylhydrazones with the best in vitro anti-leukemic effects in Salum et al. (2015).

(B) Displacement of MTC from the colchicine-binding site by N-acylhydrazones. Fluorescence intensity of MTC bound to tubulin after addition of 10 μM compounds 7, 12, and 21. Each curve represents the mean of three independent experiments. See also Figure S2.

(C) Inhibition of tubulin assembly by compounds 7, 12, 21, and podophyllotoxin after 85 min of incubation. Absorbance spectra of 25 μM tubulin in the presence of different concentrations (0.5–20 μM) of the three compounds studied or DMSO (vehicle) was monitored over time by turbidity.

See also Figure S3.

Compounds 7, 12, and 21 were capable of displacing the MTC probe, confirming their binding to the colchicine pocket of tubulin (Figure 1B). Calculation of binding constants (Figure S2) revealed that compound 21 binds tubulin with a 10 times higher affinity (binding constant of 5.0 × 107 M−1) than compounds 7 (3.3 × 106 M−1) and 12 (1.26 × 106 M−1). For comparison, podophyllotoxin's binding constant for tubulin, in the same assay, was reported to be 1.85 × 107 M−1(Antúnez-Mojica et al., 2016).

In the tubulin polymerization assay, which included podophyllotoxin as a positive control, all three compounds were effective in inhibiting tubulin polymerization (Figure 1C). Inhibition of tubulin assembly by these compounds over time is shown in Figure S3. We further investigated the effects of these compounds in cellular tubulin depolymerization assays using the A-549 cell line. Microtubule network depolymerization occurred at 200 nM for compounds 12 and 21, at 300 nM for compound 7, and at 100 nM for colchicine, which was included as positive control (Figure 2A). Compound 21 was slightly more potent than the other two compounds, because cells could be seen arrested in prometaphase and DNA was arranged in a ball of condensed chromosomes with no microtubules, which is a type IV spindle. On the other hand, in mitotic spindles seen in cells treated with compounds 7 and 12 the chromosomes were arranged in a ball enclosing several star-shaped aggregates of microtubules characteristic of type III spindles (Jordan et al., 1992). In all cases, mitotic arrest was accompanied by net microtubule depolymerization (Figure 2A).

Figure 2

Effects of Compounds 7, 12, and 21 on Microtubule Cytoskeleton and Their Cellular Uptake

(A) Effects of compounds 7, 12, and 21 on the cellular microtubule network of A-549 cells. A-549 cells were incubated for 20 h with DMSO (control, Ctr), 300 nM of 7, 200 nM of 12, 200 nM of 21, or 100 nM of colchicine. Anti-α-tubulin antibodies and Hoechst 33342 were used to stain microtubules (green) and DNA (blue), respectively. Mitotic spindles from the same preparation are shown in the right corner of each picture.

(B) Amount of compounds 7, 12, and 21 in total cellular lysates of CEM leukemia cells after 0.5, 6, or 12 h of treatment with 100 nM of the corresponding compound. The amount of each compound was normalized by the amount of protein from the same sample. Each bar represents the mean ± SD of three biological replicates. Data were analyzed by the two-way ANOVA followed by Bonferroni's post-test for mean comparisons (*p < 0.05; **p < 0.01; ***p < 0.001).

Although compounds 7 and 12 had similar effects in the MTC displacement and tubulin depolymerization assays, compound 12 was slightly more potent than compound 7 in promoting cellular microtubule depolymerization (Figure 2A), suggesting that compound 12 may have a better cellular bioavailability. Thus, the next step was to investigate the cellular uptake of compounds 7, 12, and 21, after 30 min, 6 h, and 12 h of incubation with each compound at a concentration (100 nM) that is close to their IC50 in cytotoxic assays. As expected, mass spectrometric analyses of total cellular extracts showed a significant lower cellular uptake of compound 7. Compound 12 had higher uptake after 30 min of treatment. However, its level decreased after 6 and 12 h, whereas the amount of compound 21 increased with time (Figure 2B). Whether compound 12 is subjected to cellular metabolization remains to be investigated.

Compounds 7, 12, and 21 Are Not Substrates of ATP-Binding Cassette Transporters

Overexpression of ATP-binding cassette (ABC) transporters is involved in the cellular resistance to vinca alkaloids and anti-cancer drugs in general, constituting one of the main causes of failure in cancer therapy (El-Awady et al., 2017). To evaluate if compounds 7,12, and 21 were substrates of ABC transporter proteins, cell viability assays were performed using four pairs of cell lines, each pair including a parental sensitive cell line and its resistant derivative, known to overexpress ABC transporters. All resistant cells displayed a typical multidrug resistance (MDR) phenotype, as evaluated by the calcein exclusion assay (Figure S4). Compounds 7,12, and 21 showed similar IC50 values (Table 1) and cell-cycle arrest potency (Figure S5) for both sensitive and resistant cell lines, indicating that these compounds are not good substrates of ABC transporter proteins, which was not the case for colchicine, vinblastine, and vincristine. However, a modest 4- to 9-fold increased resistance to the acylhydrazone compounds was observed for the colchicine-resistant mouse 3T3 embryonic fibroblastoid cell line. Considering the three pair of neoplastic cell lines, compound 12 showed the lowest IC50 for all six cell lines, followed by compounds 21 and 7, respectively (Table 1).

Table 1

Cytotoxicity Data (IC50 ± SD, nM) for Compounds 7, 12, 21, Colchicine, Vinblastine, and Vincristine against Human Neoplastic Cell Lines

Compounds versus Cell Lines	7	12	21	Colchicine	Vinblastine	Vincristine
A2780	144 ± 6	48 ± 1	82 ± 14	23 ± 5	0.5 ± 0.1	3 ± 0.4
A2780/AD	101 ± 11.1	20.6 ± 2.0	55.9 ± 1.4	270 ± 49	29 ± 5	457 ± 131
Ratioa	0.7	0.4	0.7	11.7	57.3	154.6
KB	103 ± 1.3	19.7 ± 5.6	43.2 ± 11	18.6 ± 0.7	0.23	0.7
KB/VB	107 ± 12.1	20.3 ± 2.4	46.9 ± 10	724 ± 228	85.9 ± 3.3	1130 ± 100
Ratio	1.04	1.03	1.09	38.9	374.3	1608
CEM	116 ± 27.6	31.4 ± 8	54 ± 0.7	14.5 ± 0.3	0.6	0.7 ± 0.2
CEM/VCR	122 ± 25	32 ± 11	50 ± 6	281 ± 5	128 ± 3	1700 ± 131
Ratio	1.05	1.02	0.93	19.42	213.65	2322.52
NIH3T3	65.7 ± 4.4	13.2 ± 4.3	109 ± 15	30 ± 3.5	1.02 ± 0.02	2.8 ± 0.7
NIH-MDR-G185	268 ± 50	95.8 ± 22	984 ± 302	783 ± 64.7	33.2 ± 0.95	898 ± 111
Ratio	4.08	7.25	9.02	26.1	32.5	320.7

Cytotoxicity results are expressed as IC50 values, the compound concentrations producing 50% cell growth inhibition, and represent the mean ± SD of three to five independent experiments.

Determined by dividing the resistant cell line IC50 mean by the sensitive cell line IC50 mean.

Cell-Cycle Arrest, DNA Damage, and Apoptosis Caused by Acylhydrazone Compounds

We choose compound 12, which presented the best in vitro cytotoxic results, as our leading compound for further mechanistic investigations. Increased in vitro proliferation of primary ALL cells was found to correlate with increased sensitivity to some chemotherapeutic drugs, including vincristine (Kaaijk et al., 2003). As shown in Figure 3, we found no correlation (p = 0.1821) between the doubling time and in vitro resistance (IC50) to compound 12 on a series of different precursor B cell ALL and T cell ALL cell lines (Table S1). To investigate how compound 12 leads to cell death, the pre-B ALL leukemia cell line RS4;11 was treated with compound 12 for 18 h and then labeled with bromodeoxyuridine (BrdU) and stained with antibodies against H2AX and PARP. Treatment with compound 12 resulted in a population of cells with DNA content in between G1 and G2, suggesting the occurrence of unequal division (Figure 4). DNA damage (H2AX) and apoptosis (PARP) occurred both at the G1 and G2 phases of the cell cycle. These results suggest that cells treated with compound 12 face cell death both as a consequence of mitotic arrest (in G2/M) and after unequal division; however, we cannot exclude the possibility of mitotic slippage followed by post-slippage cell death. As unequal division could lead to the continuous cycling of some genomically unstable cells, and the risk of secondary tumors, we investigated the generation of micronuclei. As shown in Table 2 micronuclei induction by compound 12 was comparable to that by colchicine and significantly lower than that by vincristine.

Figure 3

Cell Proliferation and Sensitivity to Compound 12 Do Not Correlate

Eleven ALL cell lines of precursor B cell ALL (Reh, RS4;11, 697, NALM-16, NALM-30) and T cell ALL (Jurkat, ALL-SIL, HPB-ALL, TALL-1, P12-ICHIKAWA, MOLT-4) were analyzed regarding their doubling time and in vitro resistance (IC50 value; see Table S1) to compound 12 at 48 h. Pearson's r correlation test resulted in a no significant correlation (p = 0.1821 and R2 = 0.1885).

Figure 4

Multiparametric Flow Cytometry Analysis of Cell Cycle, Apoptosis, and DNA Damage in RS4;11 Cells Treated with Compound 12

(A and B) Cells were treated with (A) DMSO (vehicle) 45 nM or (B) compound 12 (IC50 dose) for 18 h followed by labeling with 10 μM BrdU for 45 min. The cells were then harvested and analyzed by immunofluorescent staining and multicolor flow cytometric analysis using the BD FACSVerse Flow Cytometer. BrdU-positive cells are color-gated green, whereas BrdU-negative cells at G1 phase, between G1 and G2 phase, and G2 phase of the cell cycle are colored red, light blue, and dark blue, respectively.

Table 2

Micronuclei Formation Induced by Colchicine, Vincristine, and Compound 12

	Normal Morphology	Micronuclei	p Valuea
			Compared with Colchicine	Compared with Vincristine
Vehicle (DMSO)	199	1	<0.0001	<0.0001
Colchicine IC50	174	26	–	<0.0001
Vincristine IC50	99	101	<0.0001	–
12 IC50	163	37	0.1695	<0.0001
12 IC90	131	69	<0.0001	0.0017

Fisher's exact test. Colchicine IC50 = 13.5nM; vincristine IC50 = 1.1nM; compound 12 IC50 = 81.4nM; IC90 = 126.2nM.

Crystal Structure of the Tubulin Compound 12 Complex

To elucidate the tubulin-binding mode of compound 12, we solved its structure in complex with tubulin by X-ray crystallography. To obtain crystals of tubulin, we complexed two α/β-tubulin heterodimers from bovine brain with rat stathmin-like protein RB3 and chicken tubulin tyrosine ligase (the complex is denoted T2R-TTL) (Prota et al., 2013a, Prota et al., 2013b). Subsequently, the compound was soaked into the T2R-TTL crystals, and we were able to obtain X-ray diffraction data to 2.0 Å resolution (Table S2). We found that only the colchicine site of the α1β1-tubulin was occupied with compound 12 (Figures 5A and S6). Superimposing the T2R-TTL-compound 12 complex structure with the unliganded T2R-TTL structure (PDB: 4IHJ) showed no major conformational changes upon ligand binding (root-mean-square deviation of 0.33 Å over 1,864 Cα-atoms).

Figure 5

X-Ray Analysis of the T2R-TTL-Compound 12 Complex

(A) Chemical structure of compound 12 and overall view of the α/β-tubulin heterodimer-compound 12 complex structure. Compound 12 (green spheres) binds at the interface between α-tubulin (dark gray ribbon) and β-tubulin (light gray ribbon) in close proximity to the GTP (blue spheres)-binding site. See also Figure S6 and Table S2.

(B) Close-up view of compound 12 (green sticks) and the surrounding binding pocket formed by α- and β-tubulin using the same color code as in (A). Carbon atoms are depicted in gray for tubulin and green for compound 12; nitrogen and oxygen atoms are colored in blue and red, respectively. Interacting residues are shown as sticks and are labeled; residues βTyr202, βAla316, βIle318, and βIle378 are omitted for clarity. Secondary structural elements of tubulin are labeled in light blue.

(C) Same view and color code of the tubulin-compound 12 complex as in (B) with the apo T2R-TTL structure (purple; PDB: 4IHJ) superimposed.

(D) Same view and color code of the tubulin-compound 12 complex as in (B) with the structure of tubulin-colchicine (orange; PDB: 4O2B) superimposed.

The trimethoxyphenyl moiety of compound 12 fits in a predominantly hydrophobic pocket shaped by the side chains of βTyr202, βCys241, βLeu242, βLeu248, βAla250, βLeu252, βLeu255, βAla316, βIle318, βAla354, and βIle378 (Figure 5B). A second predominantly hydrophobic pocket is shaped by βAsn258, βMet259, βThr314, βAla316, βLys352, αSer178, and αVal181 into which the naphthalene moiety of compound 12 is inserted. The structure also reveals a water-mediated hydrogen bond between the nitrogen of the acylhydrazone linker of compound 12 and the backbone of αThr179. A previous docking study (Salum et al., 2015) proposed three additional hydrogen bonds. The first one is between the oxygen of the 3′-methoxy of compound 12 and the side chain of βCys241; our tubulin compound 12 crystal structure shows a distance between the two atoms of 3.9 Å, which seems too long to allow for the permanent formation of a stable hydrogen bond. The other two hydrogen bonds were predicted to be formed between the carbonyl group of compound 12 and the main chains of βAsp251 and βLeu255. Our tubulin compound 12 crystal structure reveals distances of 3.6 and 4.3 Å, respectively, which are on the higher side to establish stable hydrogen bonds. In conclusion, our analysis reveals that compound 12 establishes a single hydrogen bond with tubulin.

To elucidate the mode of action of compound 12 on tubulin, we superimposed several T2R-TTL crystal structures. First, we compared the tubulin-compound 12 structure with the apo structure of tubulin (PDB: 4IHJ). The major difference observed concerns the βT7 loop, which is flipped outward upon binding of the ligand (Figure 5C). Second, we superimposed the binding mode of compound 12 onto the tubulin-colchicine complex structure (PDB: 4O2B; Figure 5D). This analysis shows a large overlap of the two ligands with only a minor shift of the αT5 loop due to the differing sizes of the two ligands. The βT7 loop is flipped outward to nearly the same position for both compound 12 and colchicine. These observations indicate that compound 12, like colchicine, causes steric hindrance on the βT7 loop movement, thus locking tubulin in an assembly-incompetent conformation.

To determine the structural basis for the different tubulin-binding affinities by compounds 7, 12, and 21, we modeled compounds 7 and 21 into the colchicine-binding site based on the crystal structure of the T2R-TTL-compound 12 complex. These three acylhydrazones only differ in the size of their respective aromatic moiety, which was nicely reflected in the minor rearrangements observed after energy minimization at the end of the βH10-S9 loop and the start of βS9 strand. No perturbations of the βT7 loop or βH7 helix were observed in the models. To accommodate the bromine or methyl moieties of compounds 7 and 21, respectively, the βH10-S9 loop and the start of βS9 strand shifted slightly away from the colchicine site (Figure S7). Moreover, the 3,4,5-trimethoxy-acylhydrazone moieties minimally moved deeper into the binding pocket by maintaining all binding interactions found in the structure of compound 12. Finally, the individual changes from a naphthalene to a bromobenzene or a toluene moiety in compounds 7 and 21 only caused a minimal conformational change in the αT5 loop. These observations suggest that both compounds lacking the naphthalene moiety form more favorable interactions with the site, which agrees with their higher binding affinities.

Weight Loss Caused by Acylhydrazones 7, 12, and 21

We have previously shown that compound 12 is at least three orders of magnitude more toxic to leukemia cells than to normal proliferating lymphocytes. In a single-dose acute toxicity study, compound 12 was safe even at the maximum orally administered dose tested (1,000 mg/kg) (Salum et al., 2015). This time, we evaluated all three compounds in repeated dose and dose-escalating toxicity tests, which included also vincristine and colchicine for comparisons. Mice receiving intraperitoneal (i.p.) injections of N-acylhydrazones at a dose of 1 or 10 mg/kg/day, for 2 weeks, showed no weight loss (Figure 6A). Similar results were seen in the treatment with 0.15 mg/kg of vincristine or colchicine. However, mice treated with 1.5 mg/kg of vincristine or colchicine died after 3 to 5 days (Figure 6A), confirming the narrow therapeutic indexes of these drugs.

Figure 6

Weight Loss of C57BL/6 Mice Treated with Compounds 7, 12, 21, Vincristine, and Colchicine

(A) Weight loss of C57BL/6 mice treated with escalating daily doses of N-acylhydrazones, vincristine, and colchicine. Groups of two animals received daily i.p. injections of drugs at increasing doses of vincristine, colchicine, compound 7, compound 12, or compound 21. The doses of vincristine and colchicine were 0.15, 0.3, 0.6, 1.2, 2.4, 4.8, 9.6, 19.2, and 38.4 mg/kg. The doses of 7, 12, and 21 were 1, 2, 4, 8, 16, 32, 64, 128, and 256 mg/kg.

(B) Weight loss of C57BL/6 mice treated with repeated daily doses of N-acylhydrazones, vincristine, and colchicine. Groups of two animals received daily i.p. injections of drugs at two different dosages, during 13 consecutive days. In control (Ctr), animals were injected with vehicle (2% DMSO, 10% Tween 20, 10% glycerol in PBS); d indicates that mouse was dead on the next day. Dots are representative of a single animal. Data were analyzed by two-way ANOVA followed by Bonferroni's post-test for comparison of follow-up weights with the first weight for the corresponding animal (*p < 0.05; **p < 0.01; ***p < 0.001).

In the dose-escalating experiment mice receiving daily escalating doses of vincristine and colchicine showed abrupt weight loss followed death at a dose of 38.4 mg/kg. Increasing doses of N-acylhydrazones, on the other hand, were better tolerated. Compound 21 was the most toxic of the three N-acylhydrazones tested. It caused weight loss and death at a dose of 256 mg/kg. Mice receiving compound 7 or 12 did not show significant weight loss or any other clinical sign or behavioral alteration suggestive of intoxication, even at 256 mg/kg (Figure 6B).

In Vivo Anti-leukemia Effects of Compound 12

We have previously shown that compound 12 is able to inhibit the progression of patient-derived B cell precursor ALL cells in immunocompromised mice at a weekly i.p. dose of 1 mg/kg (Salum et al., 2015). Here we preliminarily evaluated different compound 12 treatment schemes on the survival of mice transplanted with the RS4;11 ALL cell line. Animals were treated for 4 weeks with DMSO (control); compound 12 at 1 mg/kg, once a week, i.p.; compound 12 at 0.5 mg/kg, thrice a week, every other day, i.p.; or compound 12 at 50 mg/kg, twice a week, orally. As shown in Figure 8, the dose of 1 mg/kg i.p. once a week was not sufficient to prevent leukemia progression or improve survival of mice engrafted with the RS4;11 leukemia cell line. On the other hand, compound 12 at a lower dose of 0.5 mg/kg, i.p., but given thrice a week, had a profound impact on slowing the leukemia progression (Figure 7A) and as a consequence on increasing animal survival (Figure 7B). Apparently, exposure of compound 12 to leukemia cells for a longer time may be advantageous. Oral administration of compound 12 was the second best treatment, however, at the expense of a much higher cumulative dose (100 mg/kg/week). These results suggest that compound 12 has low oral bioavailability.

Figure 8

Compound 12 Prolonged Survival of Mice Transplanted with Multidrug-Resistant ALL

(A and B) Kaplan-Meyer survival curves of mice transplanted with (A) NALM6 or the (B) multidrug-resistant N6/ADR ALL cell lines, following treatment with compound 12 (10 mg/kg, i.p., every day), vincristine (0.15 mg/kg, i.p., weekly), or vehicle. Treatment started on the fifth day after transplantation (dashed vertical line). Curves were compared by the log rank test. The indicated p values are for comparison with the curve of the control group. See also Figure S8.

Figure 7

Anti-leukemia Effect of Compound 12 at Different Dose and Administration Routes

NOD/SCID mice were transplanted with RS4;11 ALL cells. After engraftment (>0.5% leukemia cells in peripheral blood mononuclear cells), animals were randomly distributed into groups (n = 3) and treated for 4 weeks with vehicle or the indicated schemes of compound 12.

(A) Leukemia progression as estimated by the percentage of leukemia cells in peripheral blood after engraftment. Note that leukemia burden was measured every 7 days. Thus, the last measurement of leukemia in the peripheral blood does not correspond to leukemia burden at death.

(B) Kaplan-Meyer survival curves of mice following the indicated schemes of compound 12 administration. Curves were compared by the log rank test. For curves' legend see (A). The indicated p values are for the comparison with the curve of the control group.

To validate the use of compound 12 against MDR cells in vivo, we use the pre-B cell ALL NALM6 or its MDR counterpart, N6/ADR, which is characterized by P-glycoprotein overexpression (Treichel and Olken, 1992). As expected, compound 12 showed similar IC50 values against both NALM6 and N6/ADR, which was not the case for vincristine (Figure S8). Building on the better therapeutic profile of compound 12, mice were treated with 10 mg/kg, i.p., every day, whereas vincristine was administrated i.p. once a week at a concentration of 0.15 mg/kg (Szymanska et al., 2012). Both vincristine and compound 12 prolonged the overall survival of mice transplanted with NALM6, with vincristine being slightly more efficient than compound 12 (Figure 8A). However, only compound 12 was able to prolong the survival of mice transplanted with N6/ADR (Figure 8B), corroborating with our hypothesis that compound 12 is more efficient than vincristine against MDR leukemia.

Discussion

Vincristine is one of the backbone drugs of ALL treatment. It is a microtubule inhibitor whose major disadvantages are related to its limited bioavailability and high toxicity. Here, we report on three acylhydrazone derivatives with microtubule-depolymerizing activities, which bind to the colchicine-binding site of tubulin, distinctive from the vinca- and taxane-binding sites of the vinca alkaloids and taxanes currently used in clinics.

As most other colchicine-binding site agents, and in contrast to vincristine and vinblastine, compounds 7,12, and 21 were demonstrated to be active against multidrug resistance (MDR) cells. Only the colchicine-resistant mouse 3T3 embryonic fibroblastoid cell line NIH-MDR-G185 showed a modest resistance to the acylhydrazone compounds; however, it was lower than the resistance observed to colchicine, vinblastine, and vincristine. Of note, the NIH-MDR-G185 cell line was obtained through ectopic expression of the MRD1 gene (Cardarelli et al., 1995), so the levels of P-glycoprotein expression may be much higher than the levels observed in the patient-derived MDR cells. The ability to kill MDR cells may be important in the context of relapsed ALL. Although overexpression of membrane-associated ABC transporter was not associated to vincristine resistance at diagnosis (Holleman et al., 2004), relapsed ALL was shown to express increased levels of multidrug resistance genes, including P-glycoprotein, lung resistance-related protein, and multidrug resistance-associated protein when compared with diagnostic samples (Dhooge et al., 2002, Plasschaert et al., 2005, Terci Valera et al., 2004, van den Heuvel-Eibrink et al., 2000).

Among the three compounds tested, compound 12 is the one with the most favorable cytotoxic activity (Table 1; Salum et al., 2015). Intriguingly, here we show that the tubulin-binding affinity of 12 is not superior but slightly lower than that of compound 7 and 10 times lower than that of compound 21 in the MTC displacement assay. Compound 7 showed a significantly lower cellular uptake than 12, therefore explaining its lower cellular activity and suggesting that the bromobenzene moiety present in compound 7 has the lowest ability of crossing the cellular membrane compared with the toluene and naphthalene moieties present in compounds 21 and 12, respectively. On the other hand, the cellular uptake of 21 was even better than that of 12. As compounds 12 and 21 share the same mode of action on tubulin (see below), how can the better cytotoxic activity of 12 be explained, considering that 21 showed higher tubulin-binding affinity and cellular uptake? α-tubulin and β-tubulin are encoded by multigene families, whose members may hold differences in the amino acid composition of their colchicine-binding pocket. Compounds 12 and 21 may have different affinities with each type of tubulin isotype (Kumbhar et al., 2016, Santoshi and Naik, 2014). Leukemia cells express a mixture of different α- and β-tubulin isotypes (data not shown), which is probably not matched by the calf brain tubulin composition used in the MTC displacement assay. Thus the in vitro tubulin binding was not a perfect surrogate for the leukemia tubulin binding. Besides, the possibility exists that the higher cellular uptake of 21 is due to off-target or unspecific cellular binding.

Concomitant analysis of cell cycle, DNA damage, and apoptosis revealed that compound 12 promoted DNA damage and apoptosis not only in G2/M but also in G1. Death in G1 probably occurred after unequal division of cells previously arrested in G2/M. Although unequal division could lead to secondary tumors, we found micronuclei formation by compound 12 to be lower than that by vincristine, even at the dose of IC90. Cells in S-phase were not sensitive to compound 12, which agrees with the fact that no correlation was found between the proliferating index and compound 12's IC50 values on a series of ALL cell lines.

The crystal structure of the tubulin-compound 12 complex confirmed the binding of compound 12 to the colchicine-binding site of tubulin, which is located between the α- and β-tubulin subunits and shaped by residues of loop T5 of α-tubulin and strands S8, S9, and S10, loop T7, and helices H7 and H8 of β-tubulin (Ravelli et al., 2004). Free tubulin assumes a “curved” conformation. For tubulin to polymerize into microtubules this curved conformation has to undergo a conformational change to reach the “straight” structural form of tubulin. During this conformational change, the βT7 loop moves into the colchicine-binding site; however, in the presence of a ligand, the βT7 loop is sterically hindered to assume the straight conformation. As a consequence, tubulin is locked in its curved conformation and becomes assembly incompetent (Prota et al., 2014, Ravelli et al., 2004). Our structural analyses suggest that such a mechanism is indeed also valid for compound 12. One peculiarity of compound 12 is that it forms only a single hydrogen bond with tubulin, involving the nitrogen of the acylhydrazone linker of compound 12 and the backbone of α-tubulin (residue αThr179). Other colchicine-binding site inhibitors so far characterized were shown to make at least two hydrogen bonds, one of which is always with β-tubulin (Bueno et al., 2018, Gaspari et al., 2017, McNamara et al., 2015, Prota et al., 2014, Wang et al., 2016, Zhou et al., 2016).

Energy-minimized models of both compounds 7 and 21 bound to the colchicine site revealed the same mechanism of action, with no perturbations of the βT7 loop or βH7 helix. Three very small differences were seen: (1) the 3,4,5-trimethoxy-acylhydrazone moiety of 7 and 21 move minimally deeper into the binding pocket; (2) the βH10-S9 loop and the start of βS9 strand shift slightly away from the colchicine-binding site to accommodate the bromine or methyl moieties of compounds 7 and 21, respectively; and (3) a minimal conformational change in the αT5 loop was recorded. No additional interactions of compounds 7 and 21 with tubulin were observed. These observations suggested that the higher affinities of both compounds 7 and 21 when compared with 12 do not derive from the formation of additional interactions, but rather from the formation of more favorable interactions in the absence of the large naphthalene moiety, which occupies a partially hydrophobic pocket in the crystal structure of compound 12. The modeling data further suggest that the difference in affinity between compounds 7 and 21 likely derives from the differing van der Waals radii of their bromine and methyl moieties, respectively.

An attractive feature of these molecules is that they can be administered in higher doses than vincristine or colchicine. As previously shown (Salum et al., 2015), we here confirm one of these compounds' (compound 12) strong anti-leukemia activity as a single agent. Our exploratory in vivo toxicity test, measuring mice weight loss, revealed that significantly higher doses of N-acylhydrazones can be administrated daily to mice when compared with vincristine and colchicine, with no weight loss or clinical sign of toxicity. Compound 21, however, was lethal at the highest dose tested in the escalating dose study. Whether this increased toxicity is related to its higher neuronal (calf brain tubulin) tubulin-binding affinity or higher cellular uptake remains to be investigated.

Remarkably, compound 12 prolonged the overall survival of mice transplanted with the RS4;11 ALL cell line when given at a dose of only 0.5 mg/kg thrice a week; this dose given thrice a week was shown to be better than 1 mg/kg given once a week. In addition, despite low oral bioavailability, compound 12 was shown to be effective against leukemia when administered by gavage. Pharmacokinetic studies are warranted to improve the treatment schedule. Importantly, we validated the use of compound 12 for the treatment of animals transplanted with a multidrug-resistant ALL cell line, although the increased survival in this case was shorter than in mice transplanted with the RS4;11 cells, even when higher doses of compound 12 were used (10 mg/kg every day). This probably reflects the higher aggressiveness of NALM6/N6/ADR, and in any case, vincristine showed no efficacy against MDR cells compared with compound 12.

In conclusion, we describe new promising colchicine-binding site tubulin inhibitors for the treatment of ALL, featuring a favorable therapeutic index and activity against multidrug-resistant cells. In addition, we provide a structural basis to understand how N-acylhydrazones destabilize microtubules, leading to cell death.

Limitations of the Study

Although this study provided information regarding the cellular uptake of compounds 7, 12, and 21, subsequent studies are necessary to investigate the differential metabolization of these compounds by the cell, in particular compound 12, whose cellular concentration appeared to decrease at the longer time points. This investigation can help design alternative compounds with increased half-life. In addition, in this study we characterized the DNA damage and pro-apoptotic effects of 12 on cells at the different phases of the cell cycle. Besides the expected effects of treatment (cells arrested in G2 and increased numbers of apoptotic sub-G1 cells), we found a sub-population of cells having DNA content in between G1 and G2, which we named “sub-G2” and interpreted as being cells that underwent unequal division of their chromosomes. Cytogenetic analysis by fluorescence in situ hybridization of chromosomes would be advisable to validate this finding. Regarding the in vivo experiments, it would be interesting to investigate the effect of compound 12 against patient-derived ALL cells with multidrug-resistant phenotype. Last, we noticed increased weight loss of animals receiving compound 12 orally, suggesting either harm by the gavage process or gastrointestinal toxicity at a dose of 50 mg/kg twice a week.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.