Journal Pre-proof Structure activity relationship analysis of antiproliferative cyclic C5-curcuminoids without DNA binding: Design, synthesis, lipophilicity and biological activity

The chemical susceptibility of the ß-diketone linker between the two aromatic rings in the structure of curcumin to hydrolysis and metabolism has made it crucial to investigate structurally modified analogs of curcumin without such shortcomings. The synthesis of twenty cyclic C 5 -curcuminoids is described in this study in order to gain more insight into their anticancer structure-activity relationship (SAR). The design of their synthesis included four different cyclanones and five substituted aromatic aldehydes to form four, five-membered subgroups. These model compounds were evaluated in vitro for antiproliferative activity in an XTT cell viability assay against MCF-7 human non-invasive breast adenocarcinoma cancer cells and Jurkat human T lymphocyte leukemia cells in five different concentrations (10 nM, 100 nM, 1 µ M, 10 µ M and 20 µ M). The majority of the compounds investigated have shown remarkable cytotoxicity with IC 50 values in the range of 120 nM and 2 µ M with very high relative toxicity values to curcumin. The SAR conclusions are drawn and summarized. A method was developed and applied in a TLC based experimental log P measurement, which is new for such C 5 curcuminoids. The log P data and structural modifications have shown a strong correlation. The correlation of these experimental log P and the corresponding IC 50 values of the model-compounds were calculated according to the Pearson and Kendall correlation coefficient and showed weak concordance. The physicochemical behaviors of the and computed Clog P values have revealed good agreement. These cyclic C 5 -curcuminoids in contrast to curcumin do not bind to natural DNA based on their CD spectra.


Introduction
Plants are unbeatable sources of nutraceuticals. They are also considered to be one of the major sources of lead molecules and drug candidates [1]. The intense yellow rhizome of herbs in the Zingiberaceae (ginger) family provides turmeric. Species like Curcuma longa L. and/or Curcuma domestica L. for example, are used to prepare that well-known spice nowadays. The major component of turmeric is its secondary metabolite, curcumin (Fig 1). The preventive and therapeutic applications of curcumin are extremely diverse. Many clinical trials have been conducted to evaluate their pharmacokinetics, safety and efficacy. In cancer, cardiovascular diseases, diabetes, inflammatory diseases and communicable diseases, chronic arsenic exposure and alcohol intoxication, curcumin has been analyzed [2].

Fig. 1. Natural curcumins and synthetic homocyclic or heterocyclic C 5 -curcuminoids
This compound, C 5 -curcumin and its related derivatives proved to be more potent anticancer molecules compared to curcumin [7,8]. Their interesting reversible thia-Michael reaction, which is principal in binding to the biological place of action of such curcuminoids is described in detail [8].
In view of the discovery of C 5 -curcumin showing superior anticancer activity to curcumin, a number of new cyclic C 5 -curcuminoids have been synthesized. Structural modifications of C 5 -curcuminoids focusing on enhancing their bioactivities have been investigated intensively during the last couple of decades. Curcumin, C 5 -curcumin and cyclic C 5 -curcuminoids were subjected to structure-activity relationship (SAR) studies [3,[6][7][8][9] in order to find details on the most appropriate structural changes for the best cytotoxic effect.
The synthetic, cyclic C 5 -curcuminoid with a cyclohexanone core ( Fig. 1) for example proved to be more active against castration-resistant prostate cancer compared to curcumin both in vitro and in vivo [9]. Numerous studies have been conducted also on the (3E,5E)-3,5-dibenzylidene-4-piperidone synthetic, heterocyclic C 5 -curcuminoid family [10][11][12][13][14][15]. It became clear about these 4-piperidone derivatives that they exhibit higher cytotoxicity than curcumin towards different tumor cell lines like breast, prostate, cervix, melanoma, etc. [3]. Compound EF31 for example ( Fig. 1), similarly to related derivatives is shown to be a pleiotropic inhibitor of kinases (relevant to many forms of cancer), that operate at multiple points along cell signaling pathways. In addition to superior cytotoxicity to curcumin, these 4-piperidone derivatives show differential cytotoxicity; they are less toxic to noncancer cells when compared to cancer cells [10,13,16,17].
Another crucial advantage of these synthetic C 5 -curcuminoids is that a number of them are able to revert multi-drug resistance (MDR) [14,16,17]. Moreover, there are reports on the in vivo tolerability and lack of acute toxicity of these curcuminoids on rodents [18]. A number of C 5 -curcuminoids have several different modes of action, such as inducing apoptosis, cell cycle arrest, inhibiting the biosynthesis of polypeptides fundamental to tumor-progression, affecting mitochondrial respiration and stimulation/inhibition of certain enzymes playing a role in tumor growth [19].
These cyclic C 5 -curcuminoids bear a ß-dienon cross-conjugated moiety essential for cytotoxicity [22] in their structure. This ß-dienon function is a reactive and selective Michael-acceptor [8]. Selectivity appears in the fact that the Michael-reaction takes place exclusively with the nucleophilic thiol groups of biological macromolecules. It seems to be logical, that these curcuminoids cannot react with DNA in contrast to curcumin [20,21]. To gain details and proof of their mode of action in this respect we decided to perform CD spectroscopic investigations. Such CD spectroscopic data about the cyclic and heterocyclic C 5 -curcuminoids are not available in the corresponding literature to our knowledge. In view of the considerations above as a continuation of our previous SAR studies [22,23], and to gain more insight into the influence of structural modifications on the cytotoxic activity of such homocyclic and heterocyclic C 5 -curcuminoids we resynthesized twenty previously described molecules (see experimental part) for a systematic SAR study. These C 5 -curcuminoids are less susceptible to metabolism compared to curcumin [24,25]. We selected the following two cell-lines to perform XTT cell viability assays: MCF-7 (human non-invasive breast cancer cells) and Jurkat (human T lymphocyte leukemia cells). We applied an RP-TLC method to measure the experimental logP values of the twenty selected molecules. Some physicochemical parameters of the molecules were also computed in order to obtain the corresponding calculated ClogP values for example.

Physicochemical calculations:
For the prediction of some physicochemical parameters of our compounds, we have used the

Cytotoxicity test
Following the protocol of XTT cell viability assay (Biotium, Hayward, CA), the concentrations were 50.000 cells/ml (MCF-7) and 100.000 cells/ml (Jurkat) in the above described supplemented media. Finally, the absorbance was measured at a wavelength of 490 nm with DiaReader ELx800 (Dialab, Vienna, Austria) microplate reader. Results of curcumin and its analogs were compared to the DMSO controls with Student's t-test and statistical significance (p<0.05) was calculated.

Experimental (logP):
RP-TLC logP determinations were performed by a slight modification of the previously optimized RP-TLC method used for determination of some chalcones and cyclic chalcone analogs [28,29].
Compounds of the calibration set (selected chalcones and cyclic chalcone derivatives, see the structures in "Supportive material") were synthesized and purified as described [37]. Their logP values were determined earlier [28,29]. Two compounds of the validation set (progesterone, diazepam) were of pharmacopoeial grade. All other reagents used were of analytical grade. Three TLC determinations were performed for each substance.

RP
Some of the investigated molecules contain weakly basic functional groups. It was found, however, that they occur as non-ionized, neutral species under the experimental conditions.

Validation of the RP-TLC system:
The optimized chromatographic system underwent validation prior to logP measurements. For this purpose, four molecules (progesterone, diazepam, chalcone and a chalcone derivative (Q-693)) with known logP values were tested. Comparison of their logP TLC values obtained in this work with previously published experimental logP data [28] resulted in rather good agreement. Thus, these four compounds were also added to the calibration set. The logP TLC data obtained by our optimized and validated RP-TLC method are listed in Table 6. The logP data can provide a good basis for evaluation of structure-lipophilicity relationship for the examined compounds 6-9. See "Supplementary data" for more details.

Chemistry
The synthesis of our model compounds 6-9 for this SAR study was performed according to known methods involving the conditions of a Claisen-Schmidt condensation as depicted in Fig. 2. The nitrogen containing heterocyclic derivatives 7 and 8 have been prepared according to our new "onepot" synthesis [22] starting from the corresponding ß-ketoesters and aldehydes. Condensation reactions were carried out under basic or acidic conditions in good yields in a range of 75-90%.
Formally, four different cyclic ketones, like 4-methylcyclohexanone (1), 4-piperidone (2), N-methyl-4piperidone (3) or 4-hydroxycyclohexanone (4) and always the same five benzaldehydes (5a-e) have been transformed in these cross-aldol condensation reactions to form the desired model compounds 6-9. Cyclanone 4 was prepared accordingly to our previously published method [23]. Our goal was on the one hand to prepare target compounds 6-9 with a small substituent in the central ring (X=CH-CH 3 , N-H, N-CH 3 or CH-OH) due to the fact, that derivatives 7b and 7c (Fig.3) turned out to be very promising antiproliferative agents. They have even been selected as standard lead compounds to be compared to newer promising agents [22 and references therein]. We selected five substituents onto the aromatic rings (nitro, chloro, hydrogen, methoxy, dimethylamino) on the other hand, in order to find the most optimal structure from the SAR analysis with the two substituents in the central and aromatic rings. In Fig. 3 we can see the structures of the twenty cyclic C 5 -curcuminoid derivatives prepared for this study. Compounds 7b, 7c, 8b, 8c [22] and 9a-e [23] were prepared by us previously together with the others (6a-e, 7a, 7d, 7e, 8a, 8d and 8e, see the experimental part) are in good accordance with the literature data in terms of their melting points and 1 H NMR spectra.

Fig. 3.
Structures of the twenty homo-and heterocyclic C 5 -curcuminoids prepared for this SAR study

Cytotoxicity
The synthesized compounds were evaluated for their in vitro antiproliferative activity against MCF-7 human non-invasive breast adenocarcinoma cancer cells [26] and Jurkat human T lymphocyte leukemia cells [27] in an XTT cell viability assay. Inhibition of cell proliferation by these active compounds at various concentrations (10 nM, 100 nM, 1 µM, 10 µM and 20 µM) was measured, and their IC 50 (the concentration that causes a 50% cell proliferation inhibition) values were calculated and are summarized in Table 1.
It was observed generally, that Jurkat leukemia cells were more susceptible to cytotoxic treatment in almost all cases compared to MCF-7 cells (Table 1). We have observed in eight cases, that the given compound was practically ineffective. Compounds 6e and 8e were ineffective on both cell lines.
Although we did not use non-cancer cells for cytotoxicity tests, the difference in susceptibility of the two cell lines and of ineffective cases of these compounds is proof of their selective toxicity. We can state, however, that the majority of these compounds proved very good activity on cytotoxicity/viability tests. Many of them like 6a, 7a, 7b, 7c, 7e, 8a, 8b and 9a (6a, 7a, 7c and 8b on both cell lines) were active even in the nanomolar range.
The relative potential, which shows the rate of dominance over curcumin of compounds in Table 1, is higher on MCF-7 cell lines in general. For example, 7e on Jurkat and 7a on MCF-7 cells have a similar IC 50 value (0.14 and 0.12 respectively), but the relative potential is different. It is 6.79 for 7e on Jurkat and 81.33 for 7a on MCF-7. This means, that 7a is the most potent congener compared to curcumin. It is important to emphasize, that derivatives 6a, 7a, 7c and 8b possess cytotoxicity on both cell lines.
We can draw a very positive conclusion that our model compounds are showing selective toxicity with different relative potentials. The ß-dienone moiety, which serves as the primary pharmacophore function (Fig. 4, A) in the structure of our compounds 6-9 is the same in all derivatives. In order to compare changes in activity after structural modifications on the central ring and on the aromatic benzylidene rings we have created a "heat-map" including the structures of the compounds with substituents on the central and aromatic rings, IC 50 values of our cyclic C 5 -curcuminoids and the two cell lines ( Table 2). The smaller IC 50 values (higher antiproliferative activity) are in red whilst the bigger IC 50 values (lower activity) are in blue. The others in between are in yellow and green.

Table 2
The IC 50 "heat-map" of compounds 6-9. Curcumin IC 50 : Jurkat -0.95; MCF-7 -9.76 µM Our first conclusion is that the heterocyclic derivatives 7 and 8 exhibited more pronounced antiproliferative activities compared to their homocyclic counterparts 6 and 9. This fact promotes our earlier findings [22,23] on such heterocyclic C 5 -curcuminoids including compounds 7b, 7c, 8b and 8c [22] or 9a-e [23]. The reference standard was cisplatin in both of these two studies. The position opposite to the ketone carbonyl on the central cyclanone ring, being an auxiliary binding ability (secondary pharmacophore, Fig. 4, B) to the in vivo biological site of action, has a great influence on the cytotoxicity of these compounds. In other words, the binding strength of these cyclic C 5curcuminoids to the biological site of action is stronger when there is a suitable substituent and/or a nitrogen heteroatom at this position. This moiety on the central cyclanone ring (Fig. 4, B) is able to enhance the interaction between the cyclic C 5 -curcuminoids and their biological place of action in the cancer cell. This is visible on the example of 6b and 8b: the cytotoxic activity increases on both cell lines (see the corresponding IC 50 values) after the nitrogen atom "appears" in the homocycle of 6b to form the heterocyclic 8b counterpart (Table 2). A similar example is the exchange of the NH functional group of compound 7c to an OH in 9c.
The second conclusion is that the electron withdrawing p-chloro and especially p-nitro substituents on the benzylidene parts are dominating the p-methoxy or p-dimethylamino electron donor substituents.  Experimental and calculated data in Table 1 and 2, in relation to the two moieties A and B (Fig. 4) suggest that there is an electron withdrawing substituent on the benzylidene groups and a nitrogen heteroatom in the optimal structure for the optimal antiproliferative activity. From this SAR analysis, it is noteworthy that we could not find (apart from a few exceptions like 6a, 7a or 6e, 8e) a very strong relationship between the IC 50 values of the two cell lines. The two cell lines correlate neither the Pearson correlation nor the Kendall coefficient of concordance. Our results on the corresponding calculations showed a Kendall coefficient match bigger than 1. It was 1.75, which may mean interestingly two different ways of biological action on the two different cancer cell lines.

Experimental logP (logP):
On the basis of our former results [28,29], Results from the RP-TLC measurements for experimental logP data are summarized in Table 3.  This tendency is visible in general (see the average values) and in parts (see the rows in Table 3). On the other hand, lipophilicity is growing from nitro toward dimethylamino substituent (see the columns in Table 3) in the sequence of a, c, d, b and e along with exceptions like 7d or 9a. The exceptions mentioned before are due to a possible self-dimerization with H-bonds of compounds 7 or 9 respectively, through their polar functional groups (NH or OH) [30,31]. This possible dimerization makes them less polar and so more lipophilic under the conditions of the measurement. The most lipophilic substance is 7e with an extreme high logP value.
Our goal was also to determine if there was a correlation between the obtained logP data and the pharmacological activity of model compounds 6-9. Therefore, we collected the experimental logP and the corresponding IC 50 values into Table 4. In order to have a more detailed insight, the data from   Table 4 Table 4). The majority of these compounds are in good accordance with Lipinski's rule [32].
To visualize the relationship between experimental logP and IC 50 data we have plotted them against each other. The diagrams on the two cancer cell lines are in Fig 5. The most promising compounds are in blue. They have lower lipophilicity with higher antiproliferative activity (smaller IC 50 values).
Derivatives represented with red spots are more lipophilic with lower activity, except 6c, 6d and 7e on Jurkat cells. The most promising compound is 7a, which is the most potent on both cell lines with the lowest lipophilicity, which is also important from a pharmacokinetic point of view.

Predictive physicochemical calculations (ClogP, solubility in water and 3D shape):
For an evaluation of some physicochemical properties (compared to curcumin), the parameters of the synthesized compounds were computed. Just as in our previous studies [22,23] ChemAxon's Marvin Suite Plugins were used for all of the calculations [33]. Calculated logP was determined according to three different methods of Marvin's calculator plus the average value (   Table 5 Computed physicochemical parameters of the investigated compounds were undertaken. Calculated partition coefficient (ClogP) values were determined according to three methods (VG, KLOP, PHYS) of Marvin Suite [33]. Predicted solubility at pH=7.4 of the compounds is considered to be "Low" in this table, when it does not exceed 0.01 and is moderate if it was 0.01-0.06 mg/ml in water.
Number of hydrogen bond forming ability of the structures, including donor or acceptor, and solubility in water (mg/ml) were also calculated. We couldn't find any rationale among the predicted physicochemical data in Table 5 to explain the difference between the antiproliferative activity of curcumin and our model compounds 6-9. There are only some slight variations in the values of the data listed in Table 5. This can not be the cause of the great difference in pharmacological activity between curcumin and our cyclic C5-curcuminoids. The reason for this must be somewhere else, which leads us to investigate further in this direction.
Data in Table 5 together with our earlier findings [22,23] however show that the software [33] we used in our research is useful and beneficial.
Computed structures were also cleaned into 3D shape by the software used [33]. Compounds 8c and 9c as examples are available in their "ball and stick" model without their hydrogen atoms in Fig. 6.
What we can see in general is that the two benzene rings in a molecule are not in the same plane. It is also visible, that the central ring (4-piperidone or 4-hydroxycyclohexanone respectively) is located in a third plane. These three planes occupy a "close-to-planar" molecular shape. It is important to note that the central rings are almost planar. Only the atom opposite to the carbonyl function is out of the plane in an envelope-like form. Finally, the cyclic ß-dienone moiety in this cross-conjugated system keeps the conformation of these molecules fixed.

DNA binding
CD is a reliable tool for the detection of DNA binding of ligands, like drugs or any other different molecules. The appearance of the induced circular dichroism signal (ICD) is definitive proof of the interaction. In addition, shifts in the DNA bands are also indicative of the binding [34,35].
There is little known about the DNA binding of cyclic C 5 -curcuminoids. However, we know more about curcumin in this respect. It is known from in vitro experiments for example, that curcumin appears in the nucleus of cultured glioma cells after incubation. It was revealed that nuclear homing is not a result of curcumin's DNA binding [36]. The temporal relationship of curcumin's apoptotic induction effect and its nuclear homing is under investigation to acquire details about the mechanism of action. This fact among others prompted us to initiate measurements to see possible interactions between C 5curcuminoids and DNA. For this reason, we conducted circular dichroism spectroscopic investigations on the synthesized cyclic C 5 -curcuminoid derivatives in series 6-9 using natural DNA.
Curcumin, as well as 20 cyclic C 5 -curcuminoid derivatives 6-9 were tested on chicken erythrocyte DNS. In the CD spectrum of curcumin, a large ICD band appeared in the ligand's absorption region, showing a strong interaction between the ligand molecule and the chicken erythrocyte polynucleotide (Fig. 7). It also caused some minor shift in the DNA band, meaning the double helical structure is slightly distorted (Fig. 7). Those derivatives containing either aliphatic or aromatic nitrogen showed signs of interaction -a weak ICD sign appeared in the recorded spectra for example in the case of dimethylamino substituted 6e, 7e, 8e and 9e (Fig. 7). DNA bands did not change significantly at the same time, indicating that the polynucleotide remains in its native B-form. The binding is most probably the result of the weak ionic interaction of the nitrogen atoms from the cyclic C 5 -curcuminoid structure with the phosphate groups of DNA. In the case of other derivatives with no nitrogen atom in 8c 9c their structure neither an ICD signal, nor a shift in the DNA bands was detected, indicating that no interaction occurs between the molecules and DNA (Fig. 7). Although it would be consistent with an ability of these curcuminoids 6-9 to adopt a "close-to-planar" molecular shape (Fig. 6), in contrast to curcumin, none of them showed stronger interaction with the DNA used in this study. It would be conceivable that the very active inhibitors in Table 2 show diverse interaction compared to the practically ineffective counterparts (Table 1). However, compounds in these series showed similar properties under CD conditions that make it possible to generalize: based on these data we conclude that these derivatives do not bind to DNA in vitro. Due to this finding, we can disclose that the antiproliferative activity of these cyclic C 5 -curcuminoids is not due to their interaction with DNA.

9e
Curcumin 9a 8a The CD spectra of compounds 6-9 can be found in the "Supplementary data" section.

Conclusions
The chemical susceptibility of the ß-diketone linker between the two aromatic rings in the structure of curcumin to hydrolysis and metabolism [24,25] has made it a crucial point to investigate structurally modified analogs of curcumin without such shortcomings. These well-known shortcomings at the same time are the possible reasons for the drawbacks in the bioavailability of curcumin [3]. On this basis and as a continuation of our former SAR studies [22,23] we have designed and synthesized twenty cyclic C 5 -curcuminoids (6-9), the truncated forms of curcumin in order to have a deeper insight into the impact of structural modifications over cytotoxic activity and/or physicochemical parameters of our model compounds 6-9. The ß-dienone linker of these C 5 -curcuminoids is stable and has the same role as the ß-diketone moiety of curcumin such as the primary pharmacophore function. We have modified the polarity/lipophilicity of this ß-dienone linker by introducing five different substituents (nitro, chloro, hydrogen, methoxy and dimethylamino, respectively) onto the aromatic rings dividing the group (6)(7)(8)(9) into five subgroups (a, b, c, d and e). The four different cyclanons (1-4 in Fig. 2) divided the group of these model compounds further into four additional subgroups (6)(7)(8)(9). The four subgroups have selected moieties in the central cyclanone ring in the position opposite the carbonyl function, and these moieties are filling the role of the secondary pharmacophore of these compounds (Fig. 4). As a conclusion, we can state that the structural modifications in the primary and secondary pharmacophore of these model molecules resulted in clear correlations in our SAR analysis.
Experimental and calculated data in Table 1 and 2, in relation to the two moieties A and B (Fig. 4) suggest that there is an electron withdrawing substituent on the benzylidene groups and a nitrogen heteroatom in the optimal structure for the optimal antiproliferative activity. The IC 50 values of the antiproliferative activity dropped to the minimum compared to curcumin, even to submicromolar in cases close to the optimal structure. The physicochemical parameters such as molecular weight, polarity, logP and solubility of these compounds are in good accordance with Lipinski's rule [32]. The most promising compound is 7a, which is the most effective (IC 50 = 0.12-0.32 µM), most potent (80 times of curcumin) with the lowest lipophilicity (experimental logP = 3.22) which is important also from a pharmacokinetic point of view. There was no sign (or very weak if at all) of interaction between cyclic C 5 -curcuminoids 6-9 and DNA in the CD spectra. Therefore, we can also conclude that there is no risk for such possible and serious side effects from this source in the case of these curcuminoids. This fact is an advantage over curcumin if we compare the results of our circular dichroism (CD) investigations.
These findings increase the knowledge about such cyclic C 5 -curcuminoids in order to find the optimal structure in terms of antiproliferative activity and potential.

Appendix: Supplementary data
Details of the logP measurement method and CD spectra are provided.