Anticancer Activity and Structure Activity Relationship of Non-Symmetrical Choline Kinase Inhibitors

Choline kinase inhibitors are an important class of cytotoxic compounds useful for the treatment of different forms of cancer since aberrant choline metabolism is a feature of neoplastic cells. Here we present the characterization and the structure activity relationship of a series of non-symmetrical choline kinase inhibitors characterized by a 3-aminophenol moiety, bound to 4-(dimethylamino)- or 4-(pyrrolidin-1-yl)pyridinium cationic heads through several linkers. These derivatives were evaluated both for their inhibitory activity on the enzyme and for their antiproliferative activity in a panel of six human tumor cell lines. The compounds with the best inhibitory results were those connected to the linker by the N-atom (4a-h) and these results are supported by docking studies. The compounds with the best antiproliferative results were those connected to the linker by the O-atom (3a-h). On the other hand, as was predictable in both families, the inhibitory effect on the enzyme is greater the shorter the length of the linker, while in tumor cells, lipophilicity and choline uptake inhibition could play a decisive role. Interestingly compounds 3c and 4f, selected for both their ability to inhibit the enzyme and good antiproliferative activity, are endowed with a low toxicity in non-tumoral cells (e.g human peripheral lymphocytes) respect to cancer cells. These compounds were also able to induce to induce apoptosis in Jurkat leukemic cells without causing significative variations of cell cycle. It is worth to mention that these derivatives, beside their inhibitory effect on choline kinase, displayed a modest ability to inhibit choline uptake thus suggesting that this mechanism may also contribute to the observed cytotoxicity.

Due to its vital and widely studied role in cell division as well in tumor formation, ChoK emerged as potential target for various cancers [7][8][9] particularly Ras-induced carcinogenesis. The Ras effectors serine/threonine kinase (Raf-1), the Ral-GDP dissociation stimulator (Ral-GDS) and the phosphatidylinositol 3-kinase (PI3K) all are involved in ChoK activation during tumorigenesis [3,5,14]. We have previously reported the synthesis and the biological evaluation of a new family structurally related to hemicolinium-3 (HC-3) of non-symmetrical monocationic compounds ( Figure 1) endowed with antitumor activity including a 3-amino-phenol moiety bound to 4-pyrrolidinopyridinium or 4-dimethylaminopyridinium groups through several linkers [15]. In the last decade there have been numerous studies on ChoKα1 inhibitors published and their great applicability in different diseases. For this reason, we have been encouraged to further develop our compound library. The residues that constitute the ATP binding site in the enzyme is quite different to those residues that form the Cho binding site, thus suggesting the idea of synthesizing non-symmetric monocationic inhibitors which should have one cationic head that could be inserted into the Cho binding site and another fragment that could mimic the ATP adenine moiety, having into account that the size of symmetrical biscationic inhibitors was appropriated to bind simultaneously in both the ATP and choline putative binding sites of the protein model [15][16][17][18][19][20]. On the other hand, we have recently reported a new series of small monocationic molecules where the inhibition of choline uptake has emerged as a major contributor to the antiproliferative outcome of this class of compounds [9]. The results provided in the present study complement the outcomes earlier reported since docking studies have been done in more appropriate crystal structures, the inhibitory activity have been described over the isolate ChoKα1 enzyme and the antiproliferative effects of derivatives have been tested against a panel of six human cancer cell lines in order to complete and validate the earlier experiments reported.

Chemistry
The compounds 3a-h and 4a-h were synthesized following the protocols described previously [15]. Details about cloning and purification of human ChoKα1 and ChoKβ have been previously reported [17].
The effect of compounds 3a-h and 4a-h on ChoKα1 was assayed in purified ChoKα1 as previously described [17,19,20], by determining the rate of incorporation of 14 C from [methyl-14 C]choline into phosphocholine, both in the absence (control) or presence of different inhibitor concentrations.
Briefly, 20 ng of purified ChoKα1 were incubated with 1 mM [methyl-14 C]choline (4500 dpm/nmol) in 100 mM Tris-HCl (pH 8.5), 10 mM MgCl2, 10 mM ATP, and incubated at 37 ºC for 10 min. The reaction was stopped by immersing the reaction tubes in boiling water for 3 min. Aliquots of the reaction were applied to the origin of silica gel plates in the presence of phosphocholine (0.1 mg) and choline (0.1 mg) as carriers. The chromatography was developed in methanol/0.6% NaCl/28% NH4OH in water (50:50:5, v/v/v) as solvent, and phosphocholine was visualized under exposure to iodine vapor. The corresponding spot was scraped and transferred to scintillation vials for measurement of radioactivity by a Beckman 6000-TA (Madrid, Spain) liquid-scintillation counter. The 50% inhibitory concentrations (IC50 value) were determined from the % activity of the enzyme at different concentrations of synthetic inhibitor by using a sigmoidal dose-response curve (the ED50plus v1.0 software).

Antiproliferative assays in cancer cells
Human cervix carcinoma (HeLa) and human breast cancer (MCF-7) were grown in DMEM medium (Gibco, Milan, Italy). B-acute lymphoblastic leukemia (RS4;11), T-acute lymphoblastic leukemia (CCRF-CEM and Jurkat), human promyelocytic cells (HL-60), and human colon adenocarcinoma (HT-29) cells were grown in RPMI medium (Gibco, Milan, Italy). Both media were supplemented with 115 units/mL of penicillin G (Gibco, Milan, Italy), 115 μg/mL of streptomycin (Invitrogen, Milan, Italy) and 10% FBS (Invitrogen, Milan, Italy). Cell lines were tested for mycoplasma contamination every 6 month by RT-PCR analysis. Stock solutions (10 mM) of the different compounds were obtained by dissolving them in DMSO. Individual wells of 96-well tissue-culture microtiter plates were inoculated with 100 μL of complete medium containing 5 × 10 3 adherent cells or 25 × 10 3 leukemia cells. The plates were incubated at 37°C in a humidified 5% CO2 incubator for 18 h prior to the experiments. After medium removal, 100 μL of fresh medium containing the test compound at different concentrations was added to each well and incubated at 37°C for 72 h. Cell viability was assayed by the (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) test as previously described [3].The GI50 was defined as the compound concentration required to inhibit cell proliferation by 50%, in comparison with cells treated with the maximum amount of DMSO (0.25%) and considered as 100% viability.

Antiproliferative activity in Peripheral Blood lymphocytes (PBL)
Additional experiments were conducted with peripheral blood lymphocytes (PBL) from healthy donors obtained as described previously [21]. For cytotoxicity evaluations in proliferating PBL cultures, non-adherent cells were resuspended at 5 x 10 5 cells/mL in growth medium containing 2.5 μg/mL PHA (Irvine Scientific). Different concentrations of the test compounds were added, and viability was determined 72 h later by the MTT test. For cytotoxicity evaluations in resting PBL cultures, non-adherent cells were resuspended (5 x 10 5 cells/mL) and treated for 72 h with the test compounds, as described above.

Cell cycle analysis
Jurkat cells were treated with the test compounds for 24 h. Cells were harvested by centrifugation and fixed with 70% (v/v) cold ethanol. Cells were lysed with 0.1% (v/v) Triton X-100 containing RNase A and stained with PI. A Beckman Coulter Cytomics FC500 instrument and MultiCycle for Windows software from Phoenix Flow Systems were used to analyze the cells.

Measurement of apoptosis by flow cytometry
Jurkat cells were treated with the test compounds and after different times stained with both PI, to stain DNA, and annexin V-fluorescein isothiocyanate, to stain membrane PS exposed on the cell surface, following the instructions of the manufacturer (Roche Diagnostics) of the Annexin-V-Fluos reagent.

Choline uptake assay
Choline uptake was determined as previously reported [9]. Briefly, HepG2 cells (200,000 cells/well) were incubated for 10 min at 37 ºC either in a medium containing different concentrations of ChoKα1 inhibitors or only with medium as controls. The medium was then removed and the cells immediately exposed to a medium containing [methyl-14 C]choline (16 μM, 31 Ci/mol) for 10 min at 37 ºC. The incorporation of choline was stopped by medium aspiration followed by two washes with ice-cold PBS containing 580 μM choline. The cells were solubilized in NaOH 0.1 N and an aliquot used to determine the total amount of radiolabel taken up by the cells.

Preliminary docking studies
Human choline kinase α1 (hChoKα1; PDB id: 3G15) was chosen for the docking studies due to several reasons: (i) when 3D structures of both ChoKα1 (PDB ID: 3G15) and ChoKβ (PDB ID: 3FEG) isoforms were crystalized in complex with the first known inhibitor of ChoK, HC-3, HC-3 suffered a phosphorylation on the morpholinium moiety inserted into the choline binding site of the ChoKβ isoform, thus the obtained crystal structure of the complex was constituted by ChoKβ, ADP, and phosphohemicolinium-3 (PHC-3). This situation did not occur with ChoKα, where was possible to study the enzyme co-crystallized in complex with both, not phosphorylated HC-3 and ADP, therefore (ii) ChoKα1 co-crystallized with both ADP and HC-3 is more appropriated for docking studies in both binding sites than ChoKβ and (iii) described biological results indicate that is ChoKα1 and not ChoKβ is the most suitable target to study and design new anticancer drugs [8]. We published the crystal structure of the ChoKα1 isoenzyme with two inhibitors characterized by a cationic head; the first one named Compound 5: 1-[4-(4-{4-[(6-amino-9H-purin-9-yl) methyl] phenyl} butyl)benzyl]-4-(dimethylamino)-pyridinium (PDB ID: 3ZM9; see Figure 2A) [16][17] and a second one called Compound 6: 1-(4-{4-[(6-amino-3H-purin-3-yl)methyl]phenyl}benzyl)-4-(dimethylamino)-pyridinium (PDB ID: 4BR3; see Figure 2B) [17]. The resemblances between compounds 5 and 6 with the compounds 3a-h and 4a-h vindicate the use of these novel crystal structures obtained instead of the one provided with HC-3 to perform docking studies given that these 3a-h and 4a-h structures are characterized in like manner by one cationic head ( Figure 3 and Figure S1 and S2). Table 1 summarizes the clogP, the inhibitory effect on purified human ChoKα 1 activity and the growth inhibitory effects against a panel of six different human tumor-cell lines. Of all the compounds tested, in general terms, those where the aminophenol system is connected by the N-atom (4a-h) are the ones that offer the best results in terms of enzyme inhibition. Among them, compound 4f stands out with an IC50: 0.99μM values very similar to the reference compound MN48b.

Inhibition of ChoKα1 by compounds 3a-h and 4a-h and docking studies
As for the length of the spacer, it does not seem to affect the enzyme inhibition; however a significant increase of the inhibitory activity is observed when the spacer is longer in compounds connected by the O-atom the aminophenol group, see compounds (3a-b vs. 3c-d and 3e-f vs. 3g-h). This difference is not apparent when the compounds are by N-atom connected the aminophenol group (4a-h) where the spacer does not seem to exert much influence. On the other hand, the activity is favored by the influence of the substituent in position 4 of the pyridinium ring, where the compounds with dimethylamino (3e-h) and (4e-h) show an appreciable improvement of the inhibition values. As explained in paragraph 3.1, obtained crystal structures of the ChoKα1 isoenzyme with monocationic compounds 5 and 6 were considered more appropriate for docking studies of compounds 3a-h and 4a-h which only have a cationic head than the crystal structure of the enzyme with HC-3. Figure 2 shows a representation of the interactions of molecules 5 and 6 individually with the enzyme. The higher compound (PDB ID: 3ZM9, previously called number 5, see Figure 2A) [16] has an adenine moiety and a 4-(dimethylamino)pyridinium cationic head, connected by a long linker [1,4-diphenylbutane]. This compound occupies a long active site from the ATP to the choline (Cho) binding sites (Figure 2A). The linker is connected to the N-9 adenine atom similarly to the ribose-adenine connection in ATP, and for this reason the adenine moiety can mimic the connexion of the ATP cofactor to ChoK, being inserted into the ATP binding site [16]. The adenine moiety is stabilized by means of hydrophobic interactions with Leu144, Phe208, Ile209 and Leu313; and by two H-bonds with Glu207 and Ile209. The benzyl fragment connected to the adenine moiety is also stabilized by hydrophobic interactions with Arg117, Arg213 and Leu124 side chains. Finally, the 4-(dimethylamino)pyridinium fragment of this compound 5 is inserted into the ATP binding site and stabilized by π-cations interactions with Tyr333, Tyr354, Trp420, Tyr423 and Trp440.
The compound number 6 (PDB ID: 4BR3, see Figure 2B) also have an adenine moiety and a 4-(dimethylamino)pyridinium fragment, and the linker is smaller being connected to the N-3 adenine atom, and for these reasons one molecule of this compound cannot occupy simultaneously both ATP and Cho binding sites similarly to compound 5 [17]. In the crystal structure ( Figure 2B) a partial density corresponding to the 3-benzyl adenine fragment of this compound (carbon atoms in green colours) was detected into the ATP binding site, and another whole molecule (carbon atoms in yellow colours) was observed into the Cho binding site. The adenine moiety inserted into the ATP binding site is also stabilized by hydrophobic interactions with Leu144, Phe208, Ile209 and Leu313; and by H-bonds with Glu207, Ile209, similarly to compound 5. An additional H-bond between the adenine N-9 atom and the Arg213 side chain is also observed, due to the high flexibility of this amino acid. The rest of this compound 6 inserted in the ATP binding site were not detected probably due to two reasons: i) The 1-benzyl-4-(dimethylamino)pyridinium fragment is situated outside of the protein, showing a high flexibility and a really poor density and ii) The interaction of compound 6 with this region of the protein is not efficient, as was described recently [11]. Finally, the molecule 6 (carbon atoms in yellow colours) inserted into the Cho binding site interacts with the protein very efficiently, being stabilized by π-cations interactions with Tyr333, Tyr354, Tyr440, Trp420, Trp423, Trp435 and Phe435. In particular, the biphenyl group shows optimal hydrophobic stacking interactions with Tyr354, and the 4-(dimethylamino)pyridinium moiety interacts trough parallel π-cation interaction with Trp420. This new orientation of this compound 6 inside the Cho binding site is accommodated by a conformational change of Tyr333 that have move back and made an extra space in relation to compound 5. The adenine fragment of this molecule inserted into the Cho binding site is outside of the enzyme and does not show interaction with the protein (see Supplementary Figure S1B), being the 1-(biphenyl-4-ylmethyl)-4-(dimethylamino)pyridinium the key fragment of this compound for the interaction into the ATP binding site [17].
Docking studies have been performed in both crystal structures and the analysis of the obtained poses indicates which compounds could be similar to compound 5 or to compound 6. In fact, compounds 3a, 3b, 3e, 3f, 4a, 4b, 4e and 4f (Figure 3 and Figure S1 of the Supporting Information) have shown good poses in the crystal structure of compound 6, and the correct poses of compounds 3c, 3d, 3g, 3h, 4c, 4d, 4g and 4h (Figure 3 and Figure S2 of the Supporting Information) were obtained in the crystal structure of compound 5. The reason for these poses is the length of the linker. In fact, all compounds that have a benzene or biphenyl linker are inserted into the ChoK binding site similarly to compound 6, and compounds that have a 1,2-diphenylethane or a 1,4-diphenylbutane linker are inserted into the ATP and ChoK binding sites, similarly to compound 5. As an example, Figure 3 (Panel A and B) shows the obtained poses of compounds 3f (carbon atoms in white colour) and 4f (carbon atoms in orange colour). Both compounds have a 4-(dimethylamino)pyridinium cationic head and a biphenyl linker. The cationic head in these compounds is also stabilized by π-cation interaction and the biphenyl group shows optimal hydrophobic stacking interactions, being very similar to compound 6. Nevertheless, in compound 3f the linker is connected to the 3-aminophenol O-atom, and in compound 4f the linker is connected to the 3-aminophenol N-atom. The 3-aminophenol fragment of compound 4f is stabilized by two H-bonds, with Tyr354 and Glu434, while in compound 3f is stabilized by only one H-bond with Tyr354, and for this reason compound 4f has lower IC50 (0,99 ± 0,17μM) compared to 3f (6,39 ± 0,46 μM) regarding the inhibition of ChoKα1. Concerning compounds 3a, 3b, 3e, 4a, 4b, and 4e, the obtained poses highlights that the increased size of the 4-(pyrrolidin-1-yl)pyridininium cationic head (3a and 3b) negatively affects the π-cations interactions ( Figure S1), resulting in a larger IC50 compared to the equivalents compounds with a 4-(dimethylamino)pyridinium (3e). On the other hand, the biphenyl linker connected to the 3-aminophenol N-atom (compound 4b) allows two H-bonds with Tyr354 and Glu434 and hence increases the inhibitory activity. Figure 3 also shows the obtained pose of compounds 3g (carbon atoms in cyan colours) and 4g (carbon atoms in magenta colours), which also confirm the previous hypotheses. With respect to the 3h compounds, it does not seem to affect in this case the volume of the substituent in position 4 of the cationic head in the case of the family containing the -O-atom of the aminophenol connected to rest of the molecule, while in their isomers (4d vs. 4h) an unexplained decrease of the enzymatic activity of 4h is observed ( Figure  S2) Compound 3g Compound 4g

In vitro antiproliferative activities
The derivative 3c was the most active compound identified in this study, inhibiting the growth of HT-29, HeLa, MCF-7, CCRF-CEM, HL-60 and RS4;11cells with GI50 values ranging from 630 to 11 nM, resulting 30-fold more potent than MN-48b against HL-60 leukemic cells.
With respect to the effect of the rest of final compounds, it can be observed that: the compounds that offer inhibition values at the nanomolar values correspond to 3c-d and 4c-d, i.e. those with the longest spacer (2 or 4 carbons). In addition, these compounds are also those that possess a pyrrolidine as a constituent in position 4 of the pyridinium ring. There are no major differences in activity between the compounds linked by oxygen and those linked by nitrogen, although the most active compound corresponds to 3c. The fact that the compounds with the best IC50 values of enzyme inhibition do not correspond to those with the best antiproliferative values may be due to differences in clogP, which in turn lead to an increase in membrane uptake. Indeed, 3c-d and 4c-d have higher clogP values than their counterparts with shorter spacers (3a-b and 4a-b). On the other hand, compounds with dimethylamino substituents on the 4 of the pyridinium ring show a significant reduction in enzyme inhibition, which could be due to the larger volume of the 4-(pyrrolidin-1-yl)pyridinium cationic head produces a lower π-cation interaction, and these molecules show a lower IC50 of the enzyme relative to the equivalent compounds with a 4-(dimethylamino)pyridinium mentioned above.

Effects of compound 3c and 4f in non-tumor cells.
To investigate the cytotoxic potential of these compounds in normal human cells, the two compounds 3c and 4f were evaluated in vitro against peripheral blood lymphocytes (PBL) collected from healthy donors. Compound 3c showed an GI50 greater than 10 μM, both in quiescent lymphocytes and in proliferating lymphocytes stimulated with phytohemagglutinin (PHA) ( Table  2), suggesting that this compound specifically targets tumoral cells. On the other hand compound 4f exhibits a minimal toxicity with GI50 of 7.6 μM and 3.6 μM in quiescent and PHA-stimulated lymphocytes respectively. Nevertheless these values were almost 120 times higher than that observed against the T-lymphoblastic cell line (CCRF-CEM).

Cell cycle analysis
To study in detail the mechanism of action of these compounds we used a T-acute lymphoblatic leukemia cell line (Jurkat) against which we demonstrated a particular efficacy of the ChoKα1 inhibitors [12]. We first studied the effects on cell cycle following treatment with the two compounds and the results are shown in Figure 3. As can be seen, both compounds cause only modest and not significative changes in the cell cycle even at the highest concentration used (5 μM)

Choline uptake assay
Since recently we has been previously demonstrated that some ChoKα1 inhibitors, in addition to their effects on hChoKα1 activity, are able to reduce choline uptake into the cell [9], we decided to investigate whether these compounds were also able of inhibiting choline uptake in HepG2 cells.
The two compounds chosen were 3d ( Table 1 and Table 2); which shows moderate enzyme inhibition but very good antiproliferative values. On the other hand, the other compound chosen, 4f, has a good inhibitory IC50 but modest GI50 values on cell growth ( Table 1 and Table 2). Compound 3d inhibits choline uptake moderately, so its high antiproliferative activity cannot be directly attributed only to choline uptake but to a dual effect between choline uptake and enzyme inhibition.
In contrast, the compound with poor effect on cell growth inhibition 4f, has good enzyme and choline uptake inhibition values, 4f shows low lipophilicity, which perhaps does not allow the compound to pass through the plasma membrane. Therefore, it follows that a good value for choline uptake and enzyme inhibition is not sufficient, and the need to control the lipophilicity of inhibitors to allow the molecules to pass through the plasma membrane comes into play in the equation.

Conclusions
The initial hypothesis suggested that the derivative compounds analyzed in this work with a longer linker would potent hChoKα1 inhibitors, since they could be inserted in both ATP and Cho binding sites of the hChoKα1 enzyme simultaneously. According to this premise, compounds 3c, 3d, 3g, 3h, 4c, 4d, 4g and 4h whose results obtained after docking studies and poses analyzing indicated that they could be similar to model compound so-called number 5 (PDB ID: 3ZM9) should be the more potent. However the compound 4f (a compound that shown good poses in the crystal structure of compound 6 (PDB ID: 4BR3), the one with smaller linker) shows the better IC50 (0.99±0.17μM) for the inhibition of ChoKα1. This can be explained because not one molecule occupying the ATP and Cho binding sites all at once was found but two wholes molecules, one in the ATP binding site and another into the Cho binding site were detected. The reason which explained a higher IC50 for its counterpart 3f (6.39±0.46 μM) would be that the linker is connected to the 3-aminophenol O-atom and this moiety is stabilized by only one H-bond with Tyr354 while the 3-aminophenol fragment of compound 4f is stabilized by two H-bonds, with Tyr354 and Glu434 provided by both hydrogens from -NH-and OH groups (Figure 3; 3A and 3B). Nevertheless antiproliferative effects must be explained taking into account not only the inhibition of the enzyme but also the clogP, a measure of lipophilicity of a compound. In accordance with these two parameters, the better antiproliferative activities are obtained with compounds endowed with high clog P (3c, 3d, 4c, 4d), a consequence of a longer linker and two additional carbons in the pyrrolidine fragment. Interestingly these compounds appear to have low toxicity as they have no significant effect on either quiescent or PHA-stimulated human lymphocytes. They also induce apoptosis in a dose dependent manner in Jurkat cells. Choline uptake assays also highlight the dual target of these compounds, where lipophilicity plays an essential role in the antitumour capacity of the compounds without ruling out that other mechanisms may contribute to antiproliferative SAR.

Conflicts of Interest:
The authors declare no conflict of interest.