Synthesis, Crystallography, and Anti-Leukemic Activity of the Amino Adducts of Dehydroleucodine

Dehydroleucodine is a bioactive sesquiterpene lactone. Herein, four dehydroleucodine amino derivatives were synthesized using the amines proline, piperidine, morpholine, and tyramine, and spectroscopic methods and single-crystal X-ray diffraction unambiguously established their structures. The cytotoxic activity of these compounds was evaluated against eight acute myeloid leukemia cell lines, and their toxicity to peripheral blood mononuclear cells was also determined. The proline adduct was the most active compound, it showed anti-leukemic activity, upregulated heme oxygenase 1 (HMOX1) and the primary stress-inducible isoform of the heath shock 70 kDa protein 1 (HSPA1A), and downregulated NFkB1 transcription, it was also found to be about 270 times more water soluble than dehydroleucodine.


Introduction
Sesquiterpene lactones (SLs) are a large and diverse group of biologically active plant chemicals, mainly distributed in the Asteraceae family. Parthenolide, one of the most widely studied SLs, is extracted from the medicinal plant feverfew, Tanacetum parthenium L., a herb that has been used to treat migraines and arthritis for centuries [1]. Parthenolide has been evaluated in different biological systems, including the inhibition of acute myeloid leukemia (AML) stem and progenitor cells [1]. Other SLs of the guaianolide type have also been shown to have activity against AML [2]. We previously reported that dehydroleucodine (DHL), a guaianolide SL isolated from the Ecuadorian medicinal plant Gynoxys verrucosa Wedd., shows activity against several AML cell lines, exhibits reduced toxicity to normal blood cells relative to AML cell lines, and shows activity against human leukemia cells isolated from patient samples [3]. The postulated general mechanism of the action of SLs is that the exocyclic methylene group in the lactone ring undergoes a nucleophilic attack by endogenous nucleophiles, such as a thiol in the cysteine residue of the protein, leading to the formation of covalent adducts [4,5]. In general, SLs lack water solubility, resulting in poor drug-like properties. In order to mitigate this limitation, an amino-prodrug strategy has been reported. This strategy utilizes the masking of exocyclic methylene in the lactone ring by reaction with primary or secondary amines. It has been proposed that in vivo, the exocyclic double bond is regenerated [6]. Several amino analogues of parthenolide have shown anti-leukemic activity [7], including dimethylamino parthenolide, raising the interest in using it as a potentially clinically effective anticancer compound alone or in combination with other anticancer compounds [8,9].
SLs have shown intracellular effects in various cancer cell lines, including inhibition of nuclear factor-κB (NF-κB), inhibition of sarco/endoplasmic reticulum calcium ATPase (SERCA) pump, iron-dependent free radical generation, blocking of the signal transducer and activator of transcription 3 (Stat3), p53 activation, and COX-2 inhibition [2,[10][11][12]. The molecular mechanisms of action of compound 1 for AML-specific cell death was found to be the inhibition of NF-κB and the upregulation of heme oxygenase 1 (HMOX1) and the primary stress-inducible isoform of Hsp70 (HSPA1A) [3].
NF-κB has an important role in the regulation of cell cycle progression, proliferation, and apoptosis. It can be formed by homodimers and heterodimers of a p50 and a p65 subunit. This protein is found in the cytoplasm of every cell in its inactive state, in which it is normally bound to inhibitory proteins, including the inhibitor of κB (IκB). When cells are exposed to invaders or stressors, it travels to the nucleus and binds with DNA to turn certain genes on or off. On activation, NF-κB regulates the expression of approximately 400 different genes [13]. Meanwhile, dysregulation of NF-κB can generate several diseases, including inflammation and cancer [14].
Experimental evidence suggest that SLs react by a Michael addition mechanism with the cys-38 residue in the p65 subunit of NF-κB, preventing it from binding to DNA [14][15][16].
Herein, we report the synthesis and chemical characterization of four amino derivatives of dehydroleucodine, as well as their evaluation against several AML cell lines.

Chemistry
Compound 1 was isolated from G. verrucosa following the procedure previously described [3]. Four amino adducts of compound (1), namely, dehydroleucodine-proline (2), dehydroleucodine-piperidine (3), dehydroleucodine-morpholine (4), and dehydroleucodine-tyramine (5), were synthesized by stirring a solution of 1 at room temperature (RT) with the appropriate amine in a single-step reaction using an adaption of a previously reported procedure (Scheme 1). In all of the cases, single diastereoisomeric products were obtained after purification (60-91% yield).  Table 1. The assignment of the 1 H and 13 C NMR resonances was accomplished using two-dimensional (2D) correlation spectroscopy (COSY), heteronuclear multiple quantum coherence (HMQC), rotating-frame Overhauser effect spectroscopy (ROESY), and heteronuclear multiple-bond correlation (HMBC) experiments. The proton and carbon resonances appeared at nearly the same chemical shifts and with similar multiplicities as dehydroleucodine, except for the δ 2.25 signal for H-12 and the signals corresponding to H-13, which appeared as a doublet at δ 1.27 in the 1 H NMR spectrum, suggesting the presence of a C-H bond on the lactone ring. This result is also consistent with the changes in the 13 C NMR spectrum. Thus, a large (anti) H-8/H-12 3 J coupling, as well as a ROESY peak between H-12 and H-9, were observed, indicating that these hydrogens lie on the same molecular face. In addition, a ROESY peak between H-12 and H-8 indicates that these hydrogens also lie on the same face. Finally, the structure of compound 2 was confirmed by X-ray crystallographic analysis, and an Oak Ridge thermal ellipsoid plot ORTEP perspective is shown in Figure 1.
The crystal structure of 2 shows that the H atom of the carboxyl group migrated to the N atom in the molecule. The differential electron density maps between compounds 2 and 1 clearly show the location of this H atom, as well as its absence on the COO functional group. Additionally, the H atoms of an associated water molecule were found through electron density map differences.  Table 1. Assignments of all of the proton and carbon resonances were achieved by using the 2D NMR, COSY, ROESY, and HMBC techniques.     13 C NMR assignments are given in Table 1 and were assigned by the COSY, ROESY, HMQC, and HMBC experiments. The relative configuration of compound 4 was determined by analysis of its nuclear Overhauser enhancement spectroscopy (NOESY) data. A nuclear Overhauser effect (NOE) correlation was observed between the H-8 and H-13 methylene protons, but not with H-12. H-8 shows three large (10-12 Hz) vicinal couplings to H-9, H-12, and one of the H-7 (δ 1.36) protons, indicating that it is anti to H-9, H-12, and a H-7 proton. H-9 shows strong NOE correlations with H-12 and the δ 1.36 proton. Thus, H-8 and H-12 are on opposite faces of the molecule. Ultimately, the structure of 4 was confirmed by X-ray crystallographic analysis, as shown in Figure 2A.
location of this H atom, as well as its absence on the COO functional group. Additionally, the H atoms of an associated water molecule were found through electron density map differences.
Compound 3 is a pale yellow amorphous solid with a molecular formula of C20H27NO3 based on HRMS data (m/z 330.2075 [M + H] + ). The UV spectrum in methanol (MeOH) shows two peaks at  (log є) 204 (3.66) nm and 256 (4.14) nm. The IR (ATR) spectrum shows bands at 2936 (C-H st), 1773 (C=O st γ lactone), 1683 (α,β unsaturated C=O), 1617, and 1637 (C=C st) cm −1 . The 1 H and 13 C NMR chemical shifts are shown in Table 1. Assignments of all of the proton and carbon resonances were achieved by using the 2D NMR, COSY, ROESY, and HMBC techniques.  Table 1 and were assigned by the COSY, ROESY, HMQC, and HMBC experiments. The relative configuration of compound 4 was determined by analysis of its nuclear Overhauser enhancement spectroscopy (NOESY) data. A nuclear Overhauser effect (NOE) correlation was observed between the H-8 and H-13 methylene protons, but not with H-12. H-8 shows three large (10-12 Hz) vicinal couplings to H-9, H-12, and one of the H-7 (δ 1.36) protons, indicating that it is anti to H-9, H-12, and a H-7 proton. H-9 shows strong NOE correlations with H-12 and the δ 1.36 proton. Thus, H-8 and H-12 are on opposite faces of the molecule. Ultimately, the structure of 4 was confirmed by X-ray crystallographic analysis, as shown in Figure 2A. The crystal unit cell of 4 is composed of two different conformers, as shown in Figure 2A; one conformation is trans at the C-13-N and C-H-12 bonds, and the other is gauche cis ( Figure 2B). A systematic conformational analysis of the rotation around the bond between C-12 and C-13 of compound 4 showed three minima and three maxima ( Figure 2C). The lowest two minima, conformations II and VI with energies of 14.60 and 14.61 kcal/mol, respectively, correspond to the The crystal unit cell of 4 is composed of two different conformers, as shown in Figure 2A; one conformation is trans at the C-13-N and C-H-12 bonds, and the other is gauche cis ( Figure 2B). A systematic conformational analysis of the rotation around the bond between C-12 and C-13 of compound 4 showed three minima and three maxima ( Figure 2C). The lowest two minima, conformations II and VI with energies of 14.60 and 14.61 kcal/mol, respectively, correspond to the conformations shown in the crystal structure unit cell. The three maxima, conformations I, III, and V with energies of 21.00, 21.50, and 25.40 kcal/mol, respectively, correspond to the eclipsed conformations.  Table 1. Assignments of proton and carbon resonances were completed by using 2D NMR, COSY and HMBC techniques. The J8-12~13Hz, indicates that H-8 and H-12 are anti and therefore on opposite faces. In the nuclear Overhauser effect (NOE) experiment on compound 5, a NOE correlation was observed between H-9 and H-12 protons indicating that they are on the same face, this is supported by moderate NOESY correlations between H-8 and both H-13 protons. Table 2 shows the cytotoxic activity of the amino derivatives against eight AML leukemia cell lines after 48 h of treatment. Compounds 2-4 showed in vitro cytotoxicity at a more moderate level than compound 1. Compound 2 was the most potent derivative and compound 5 was found to be inactive. Table 2. Cytotoxic effects of compounds 1-5, expressed as EC 50 (± SD) values *.

DHL ** (1) DHL-Proline (2) DHL-Piperidine (3) DHL-Morpholine (4) DHL-Tyramine (5)
HL In addition, we evaluated the intracellular effects upon treatment with the amino derivatives 2-4 using quantitative PCR ( Figure 3B). Leukemia cells were treated with 20 µM of each test compound for 6 h before RNA extraction. HMOX1 and HSPA1A were measured to determine the activation of an oxidative stress response upon drug treatment. Previously, we reported that compound 1 upregulated HMOX1 and HSPA1A, suggesting that this compound activates oxidative stress responses [3]. We performed the same experiment for the amino derivatives of compound 1; the results show that compounds 2-4 moderately upregulated HMOX1 and HSPA1A. Interestingly, compound 2 displayed a higher upregulation of HMOX1 and HSPA1A compared to 3 and 4, and it also displayed greater cytotoxic effects on the AML cell lines, as shown in Table 2 and Figure 3A. The NF-κB transcript levels were measured upon treatment with compounds 2-4 ( Figure 3B), and in all cases we observed a downregulation of NFkB1 transcription.
The activity of compounds 2-4 was also tested on normal bone marrow (n = 1) and peripheral blood (n = 3) mononuclear cells. Figure 3C displays the viability of the normal samples after 48 h of treatment with the test compounds. At 25 µM, compound 2 treated normal cells showed an average viability of 90%; at 50 µM, cells treated with compounds 2, 3, or 4 showed an average viability of 78, 80, and 79%, respectively; at 100 µM, cells treated with compounds 3 or 4 showed an average viability of 74 and 73%, respectively. Comparison of these results with the average EC 50 values for the leukemia cell lines (Table 2) shows that the tested compounds exhibited less toxicity against normal cells than against leukemia cell lines.
The four amino derivatives of dehydroleucodine were synthetized by a regioselective and stereoselective conjugate addition reaction and tested against eight AML cell lines. All of the derivatives showed less activity compared to the parent compound. Among the derivatives, DHL-proline (2) showed the most activity and displayed higher upregulation of HMOX1 and HSPA1A.
The strategy of synthetizing amino derivatives has previously been applied to different sesquiterpene lactones, including helenalin, ambrosin, costunolide, alantolactone, and parthenolide [6]. In some cases, the amino derivatives of SLs showed similar activity to their respective parent compounds, for example, the pyrrolidine and piperidine analogues of ambrosin, the diethylamine derivatives of alantolactone and isoalantolactone, the dimethylamine, pyrrolidine, and piperidine adducts of custonolide, and the proline amino derivative of dehydrocostus lactone [6]. However, as has been reported for helenalin [6], and as we have shown in this paper for dehydroleucodine, it is clear that in some cases, the amino derivatives of these SLs can be less active than their parent compounds. The activity of compounds 2-4 was also tested on normal bone marrow (n = 1) and peripheral blood (n = 3) mononuclear cells. Figure (Table 2) shows that the tested compounds exhibited less toxicity against normal cells than against leukemia cell lines.
The four amino derivatives of dehydroleucodine were synthetized by a regioselective and stereoselective conjugate addition reaction and tested against eight AML cell lines. All of the derivatives showed less activity compared to the parent compound. Among the derivatives, DHLproline (2) showed the most activity and displayed higher upregulation of HMOX1 and HSPA1A.
The strategy of synthetizing amino derivatives has previously been applied to different sesquiterpene lactones, including helenalin, ambrosin, costunolide, alantolactone, and parthenolide [6]. In some cases, the amino derivatives of SLs showed similar activity to their respective parent compounds, for example, the pyrrolidine and piperidine analogues of ambrosin, the diethylamine derivatives of alantolactone and isoalantolactone, the dimethylamine, pyrrolidine, and piperidine adducts of custonolide, and the proline amino derivative of dehydrocostus lactone [6]. However, as has been reported for helenalin [6], and as we have shown in this paper for dehydroleucodine, it is The solubility (±SD) of the most active derivative, compound 2, was 64.62 ± 5.47 mg/mL when measured in buffer at pH 7.4, and the solubility of compound 1 was 0.24 ± 0.02 mg/mL when measured in the same conditions. These results show that the proline derivative is about 270 times more water soluble than the parent compound.
Compound 1 has potent activity against leukemia cell lines and leukemia primary cells, is less toxic to primary mononuclear cells than to cancer cells [3], is very abundant in nature, easy to isolate and relatively stable. We have also determined, experimentally, that compound 1 has an octanol/water Log P (±SD) of 2.33 ± 0.16, which is well within the range of the optimum partition for cell permeation (Log P = 2-3.5) [17]. Unfortunately, compound 1, has very limited water solubility and that limits its applicability, in this context the high solubility of the proline adduct 2 bodes well for its potential as dehydroleucodine prodrug.

General Experimental Procedures
Melting points were measured with a capillary melting point apparatus, and the values were uncorrected. Optical rotations were measured on an Autpol III (Rudolph) polarimeter. The UV spectra were recorded on a Hewlett-Packard 8452A diode array spectrophotometer, while the IR spectra were measured using a Magna IR 550 spectrometer with attenuated total reflectance (ATR). NMR spectra were recorded in CDCl 3 at 25 • C on a Varian Unity NMR spectrometer equipped with a Nalorac 3 mm inverse detection probe with a z-axis gradient coil, operating at 500 MHz for 1 H and 125 MHz for 13 C spectra. Each sample consisted of approximately 5 mg of the desired compound dissolved in 1 mL CDCl 3 containing TMS as an internal reference. All of the one-dimensional (1D) and 2D (i.e., gradient COSY, ROESY, gradient HSQC, and gradient HMBC) spectra were acquired and processed with standard Varian software. Flash chromatography was performed using a Teledyne ISCO CombiFlash Companion Automated Flash Chromatography system with pre-packed silica gel columns. Column chromatography was performed using 60-230 mesh silica gel. All of the reagents were used as purchased.

General Procedure for the Preparation of Amino Adducts
A solution of 1 (0.22 mmol) in EtOH (6 mL) was treated with the corresponding amines (4.4 mmol) in the presence of Et 3 N, and the mixture was stirred at room temperature overnight. The reaction mixture was evaporated under reduced pressure, and the residue was further purified by silica gel flash chromatography to yield products 4-5. All of the reactions were performed under a dry nitrogen atmosphere.
Dehydroleucodine-proline 2. Single-crystal X-ray diffraction analysis and crystallographic data for compounds 2 and 4. Crystallographic data for the structures were deposited in the Cambridge Crystallographic Data Centre, and the deposition number is indicated below. The data can be obtained free of charge at www.ccdc.cam.ac.uk (or from the Cambridge Crystallographic Data Centre, 12 Union Road, Cambridge CB2 1EZ, U.K.; fax (+44) 1223-336-033 or email desposit@ccdc.cam.ac.uk).
Crystallographic data of 2. Crystallization from ethyl acetate yielded colorless crystals of 2. The X-ray data were collected on a Bruker-Nonius X8 Proteum CCD diffractometer using CuKα radiation. The structures were solved using SHELXT and were refined using SHELXL. Molecular fragment editing was performed using the XP program of SHELXTL [18]. Crystal data: C 20 H 25 NO 5

Determination of Partition Coefficient and Solublity
The experimental octanol/water partition coefficient of compound 1 was determined by quadruplicate measurements at two concentrations (1 and 0.25 mg compound/mL of octanol) using the shake-flask method [20], in which the compound was fractionated between octanol (1 mL) and 100 mL of deionized water into glass bottles of 150 mL. Both phases were pre-saturated before the determination. The glass bottles were placed in a shaker for 1 h and then centrifuged for 30 min at 20 • C and 5000 rpm. For the quantitation 1 mL aliquotes from the aqueous phase were evaporated under nitrogen the residue was then reconstituted in a 200 µL glass insert with methylene chloride.
To determine the solubility compounds 1 and 2, supersaturated mixtures of the compounds, in phosphate buffered (11.9 mM) saline (pH 7.4) were sonicated (30 min at 25 • C), vortexed (90 min at 25 • C), and centrifuged (30 min at 13,600 G). Aliquots of the supernatants were passed through a syringe filter, and the amounts of compounds 1 and 2 were determined. The experiments were conducted in triplicate.
The analytical determination of the partition coefficient compound 1 and the solubility of compounds 1 and 2 were performed using single ion monitoring gas chromatography/mass spectrometry (GC/MS) with an Agilent 5975 GC/MSD instrument. The solubility determinations were conducted using a 30 m DB-5MS column (0.250 mm 10 µm, J&W, Folson, CA, USA) with a helium flow rate of 1.8 mL/min, the injector temperature was 200 • C, the transfer line was maintained at 280 • C, and column temperature was maintained at 170 • C for 1 min followed by a gradient of 35 • C/min to 270 • C, and then a gradient of 100 • C/min to 300 • C. The mass spectrometer conditions were electron impact, ion source temperature 230 • C, and ionization voltage 70 eV. The injection volume was 0.2 uL for compound 1, and 0.5 uL for compound 2. A calibration curve for the analytical method was developed with five points, plotting the average ratio of the peak area of the respective compound to the peak area of parthenin internal standard (R 2 = 0.999).
The partition coefficient determination was conducted using a 15 m DB-1 column (0.250 mm 10 µm, J&W, Folson, CA, USA) with a helium flow rate of 1.8 mL/min, the injector temperature was 200 • C, the transfer line were maintained at 300 • C, the column temperature was maintained at 120 • C for 5 min followed by a gradient of 17 • C/min to 300 • C. The mass spectrometer conditions were electron impact, ion source temperature 200 • C, and ionization voltage 70 eV. A five-point calibration curve (R 2 = 0.999) was used to calculate the concentration of 1 in the aqueous phase using calaxin as the internal standard. The milligrams of compound 1 in the octanol phase were then indirectly calculated and the concentrations of each compound in each phase were used to determine the partition coefficient. The average of the measurements is reported as the logarithm of the octanol/water partition coefficient (Log P).