Synthesis of Novel Lipophilic Polyamines via Ugi Reaction and Evaluation of Their Anticancer Activity

Natural polyamines (PAs) are involved in the processes of proliferation and differentiation of cancer cells. Lipophilic synthetic polyamines (LPAs) induce the cell death of various cancer cell lines. In the current paper, we have demonstrated a new method for synthesis of LPAs via the multicomponent Ugi reaction and subsequent reduction of amide groups by PhSiH3. The anticancer activity of the obtained compounds was evaluated in the A-549, MCF7, and HCT116 cancer cell lines. For the first time, it was shown that the anticancer activity of LPAs with piperazine fragments is comparable with that of aliphatic LPAs. The presence of a diglyceride fragment in the structure of LPAs appears to be a key factor for the manifestation of high anticancer activity. The findings of the study strongly support further research in the field of LPAs and their derivatives.


Introduction
According to the latest statistics, about 19.3 million cancer cases and 10 million cancerassociated deaths are annually reported worldwide [1]. Currently, the search for new chemotherapeutic agents inhibiting invasion and metastasis faces the problem of resistance of cancer cells due to their somatic changes [2,3]. In this regard, modern biomedical approaches require new therapeutic strategies and development of anticancer agents to overcome these challenges.
Natural polyamines (PAs) putrescine, spermidine, and spermine that are present in significant amounts in all eukaryotic cells are essential for various underlying cellular processes such as proliferation, differentiation, and apoptosis [4]. They are formed inside the cell but can also be obtained from exogenous sources. Exogenous PAs penetrate into the cell by active transport and, once inside, are distributed in all cellular compartments due to their high solubility [5]. In eukaryotic cells, the intracellular concentration of PAs is strictly controlled by the mechanisms of their biosynthesis, catabolism, transport, and excretion. Uptake and biosynthesis of PAs grows up in response to proliferation stimuli. At the same time, catabolism and secretion of PAs, as well as inhibition of their biosynthesis and transport, are induced when higher PA concentrations are reached in the cell [6]. The levels of PAs in cancer cells are higher than in normal cells, and this phenomenon is associated Most of the known approaches regarding the synthesis of PA derivatives and their conjugates have several disadvantages [9], namely: (1) multistage synthetic procedures, (2) the introduction of orthogonal protective groups to block internal and terminal nitrogen atoms, (3) the low overall yield of the desired molecules, and (4) complicated purification procedures that are required for highly polar compounds. Following this approach, a synthetic scheme for the preparation of a family of PAs (6) containing an alkyl diglyceride fragment and an ethyl residue attached to terminal nitrogen atoms was developed in our laboratory [10]. Within this structure, the long-chain alkyl substituent (C10-C18) was placed at the C(1) atom of glycerol, whereas the short-chain ethyl substituent was placed at C (2). The presence of an alkyl group at the terminal nitrogen atom of polyamine slightly increases cytotoxicity compared to analogues with a free NH2 terminal group. This effect may be related to the fact that the terminal alkyl group prevents potential acylation and further oxidation of the compound, which increases its stability in cells [11]. The key step of the synthesis regarded the interaction of alkyl diglyceride bromides with regioselectively protected PAs under Fukuyama reaction conditions [12]. The low yields of compound 6 and the multistage nature of the synthetic scheme revealed the disadvantages of the proposed method.
The lipophilic PA may effectively inhibit PA transport into the cell due to its effective incorporation into the transmembrane channel located on the cell membrane. These data have been previously reported for AMXT-1501 with the palmitic acid residue [13]. In addition, we have previously shown that lipophilic PAs, where the lipophilic part is presented by a diglyceride fragment, also exhibit high anticancer activity [10]. Considering the results of the mentioned above studies, conjugation of PAs with the diglyceride fragment may have beneficial pharmacological potential.
The multicomponent Ugi reaction [14] can be used as an effective tool for the rapid preparation of modified PAs. One of the modifications of this reaction (N-split-Ugi [15,16]) Most of the known approaches regarding the synthesis of PA derivatives and their conjugates have several disadvantages [9], namely: (1) multistage synthetic procedures, (2) the introduction of orthogonal protective groups to block internal and terminal nitrogen atoms, (3) the low overall yield of the desired molecules, and (4) complicated purification procedures that are required for highly polar compounds. Following this approach, a synthetic scheme for the preparation of a family of PAs (6) containing an alkyl diglyceride fragment and an ethyl residue attached to terminal nitrogen atoms was developed in our laboratory [10]. Within this structure, the long-chain alkyl substituent (C 10 -C 18 ) was placed at the C(1) atom of glycerol, whereas the short-chain ethyl substituent was placed at C (2). The presence of an alkyl group at the terminal nitrogen atom of polyamine slightly increases cytotoxicity compared to analogues with a free NH 2 terminal group. This effect may be related to the fact that the terminal alkyl group prevents potential acylation and further oxidation of the compound, which increases its stability in cells [11]. The key step of the synthesis regarded the interaction of alkyl diglyceride bromides with regioselectively protected PAs under Fukuyama reaction conditions [12]. The low yields of compound 6 and the multistage nature of the synthetic scheme revealed the disadvantages of the proposed method.
The lipophilic PA may effectively inhibit PA transport into the cell due to its effective incorporation into the transmembrane channel located on the cell membrane. These data have been previously reported for AMXT-1501 with the palmitic acid residue [13]. In addition, we have previously shown that lipophilic PAs, where the lipophilic part is presented by a diglyceride fragment, also exhibit high anticancer activity [10]. Considering the results of the mentioned above studies, conjugation of PAs with the diglyceride fragment may have beneficial pharmacological potential.
The multicomponent Ugi reaction [14] can be used as an effective tool for the rapid preparation of modified PAs. One of the modifications of this reaction (N-split-Ugi [15,16]) is based on the interaction of a secondary diamine, a carbonyl compound, a carboxylic acid, and an isocyanide, which together form an α-acylaminoamide, whose amide groups can be further reduced to form a PA. This modification makes it possible to obtain PAs of different structures in two steps from simple compounds [17].
In this work, we implemented the multicomponent N-split Ugi reaction for the synthesis of novel alkylated PAs containing aliphatic and cyclic diamines and evaluated their anticancer activity.

Results and Discussion
The synthesis of lipophilic PAs using the N-split Ugi reaction is usually carried out in two steps. On the first step, α-acylaminoamide is formed by the condensation of four components. On the second step, the reduction of amide groups is carried out followed by the removal of protective groups. In this work, the commercially available tert-butyl isocyanide (7a) or the previously obtained octadecyl isocyanide (7b) [18] were used as isonitrile components, glacial acetic acid (8a) or N-acetylglycine (8b) as the carboxyl component, and N,N'-dibenzylalkanediamine (9a-c) as the diamine component, while paraformaldehyde (10) was used as the carbonyl component. The reaction was refluxed in methanol in an equimolar ratio of starting reagents for 16 h (Scheme 1). Usually the N-split Ugi reaction proceeds under room conditions, but we used refluxing to dissolve the lipophilic isocyanide and to break paraformaldehyde completely. is based on the interaction of a secondary diamine, a carbonyl compound, a carboxylic acid, and an isocyanide, which together form an α-acylaminoamide, whose amide groups can be further reduced to form a PA. This modification makes it possible to obtain PAs of different structures in two steps from simple compounds [17].
In this work, we implemented the multicomponent N-split Ugi reaction for the synthesis of novel alkylated PAs containing aliphatic and cyclic diamines and evaluated their anticancer activity.

Results and Discussion
The synthesis of lipophilic PAs using the N-split Ugi reaction is usually carried out in two steps. On the first step, α-acylaminoamide is formed by the condensation of four components. On the second step, the reduction of amide groups is carried out followed by the removal of protective groups. In this work, the commercially available tert-butyl isocyanide (7a) or the previously obtained octadecyl isocyanide (7b) [18] were used as isonitrile components, glacial acetic acid (8a) or N-acetylglycine (8b) as the carboxyl component, and N,N'-dibenzylalkanediamine (9a-c) as the diamine component, while paraformaldehyde (10) was used as the carbonyl component. The reaction was refluxed in methanol in an equimolar ratio of starting reagents for 16 h (Scheme 1). Usually the Nsplit Ugi reaction proceeds under room conditions, but we used refluxing to dissolve the lipophilic isocyanide and to break paraformaldehyde completely. A noticeably increased yield of the N-split Ugi reaction from 10-11% to 35-48% was observed when the length of the methylene linker between the central nitrogen atoms of diamines was increased from two (compounds 11a,b) to three carbon atoms (compounds 11c,d). On the contrary, further increase in its length to one additional methylene group decreases the yield of α-acylaminoamide 11e to 30%. The obtained yields correlated with the previously published data [15]. In the NMR spectra of the α-acylaminoamides 11a-e, appearance of double sets of signals which correlate with the formation of rotamers around the amide bonds was detected [19].
Piperazine is one of the widely used structural fragments in numerous biologically active compounds. Various piperazine derivatives demonstrated a high antiproliferative activity against different cancer cell lines [20][21][22][23][24]. The replacement of aliphatic diamines with piperazine results in increased conformational rigidity and lipophilicity, altering the proteolytic [25,26] and biological activity of PAs. Compounds 12a-e with a piperazine fragment were obtained as described above for compounds 11a-e. The replacement of aliphatic diamines 9a-c with piperazine (9d) increases the yields of α-acylaminoamides 12a-e and reduces the reaction time from 16 to 12 h (Scheme 2). A noticeably increased yield of the N-split Ugi reaction from 10-11% to 35-48% was observed when the length of the methylene linker between the central nitrogen atoms of diamines was increased from two (compounds 11a,b) to three carbon atoms (compounds 11c,d). On the contrary, further increase in its length to one additional methylene group decreases the yield of α-acylaminoamide 11e to 30%. The obtained yields correlated with the previously published data [15]. In the NMR spectra of the α-acylaminoamides 11a-e, appearance of double sets of signals which correlate with the formation of rotamers around the amide bonds was detected [19].
Piperazine is one of the widely used structural fragments in numerous biologically active compounds. Various piperazine derivatives demonstrated a high antiproliferative activity against different cancer cell lines [20][21][22][23][24]. The replacement of aliphatic diamines with piperazine results in increased conformational rigidity and lipophilicity, altering the proteolytic [25,26] and biological activity of PAs. Compounds 12a-e with a piperazine fragment were obtained as described above for compounds 11a-e. The replacement of aliphatic diamines 9a-c with piperazine (9d) increases the yields of α-acylaminoamides 12a-e and reduces the reaction time from 16 to 12 h (Scheme 2). The highest yield (compound 12b) was achieved using octadecyl isocyanide (7b) and acetic acid (8a). Although the four-component N-split-Ugi reaction seems to be insensitive to steric hindrances [27], the yields of compounds 12c,d obtained from the diglyceride 7c were significantly lower, suggesting that steric hindrance caused by the ethyl substituent at the C(2) atom of glycerol might be the reason for the observed effect. Additionally, the low yield of compounds 12c,d can be linked with lower stability of isocyanide 7c and its partial transformation into formamide, as evidenced by the presence of the corresponding spot on TLC and NMR data of isolated formamide. The use of N-acetylglycine (9b) as a carboxyl component resulted in a decreased yield of α-acylaminoamides 12a,c, which may be due to reduced nucleophilicity of the carbonyl carbon atom that undergoes the Mumm rearrangement [28].
In the 13 C NMR spectra of compounds 12c,d, no signal of carbonyl carbon at the diglyceride fragment was observed when CDCl3 was used as a solvent. At the same time, a strongly broadened signal of the corresponding NH-proton was detected in the 1 H NMR spectra. Apparently, the intermolecular exchange of amide protons led to the strong broadening of the 13 C signal of the corresponding carbonyl atom and its merging with the base line of the spectrum. To avoid those problems, the spectra of compounds 12c,d were recorded in DMSO-d6, a solvent that somewhat suppresses the exchange of mobile protons.
Since cancer cells generally overexpress carbohydrate receptors, we attempted to prepare PAs that contain a diglyceride moiety at one terminal nitrogen atom and a carbohydrate moiety at the other via a two-step strategy using 2-(hydroxymethyl)benzoic acid [29]. This approach allows to prepare aminoamide, which acts as one of the components in the N-split Ugi reaction. Isonitrile 7c reacted with equimolar amounts of 2-(hydroxymethyl)benzoic acid (8c), piperazine (9d), and formaldehyde (10) (Scheme 3) to give monosubstituted aminoamide 13a in a 35% yield. The low yield of the desired product 13a was due to the formation of the symmetrical adduct of aminodiamide 13b with 22% yield. Subsequent treatment of 13a with isocyanide 14a-c, 2-(hydroxymethyl)benzoic acid (8c), and formaldehyde (10) led to formation of the disubstituted piperazines 15a-c in 75%, 80%, and 30% yields, respectively. The low yield of D-glucose containing compound 15c is supposedly associated with a partial deacetylation during reflux. The highest yield (compound 12b) was achieved using octadecyl isocyanide (7b) and acetic acid (8a). Although the four-component N-split-Ugi reaction seems to be insensitive to steric hindrances [27], the yields of compounds 12c,d obtained from the diglyceride 7c were significantly lower, suggesting that steric hindrance caused by the ethyl substituent at the C(2) atom of glycerol might be the reason for the observed effect. Additionally, the low yield of compounds 12c,d can be linked with lower stability of isocyanide 7c and its partial transformation into formamide, as evidenced by the presence of the corresponding spot on TLC and NMR data of isolated formamide. The use of N-acetylglycine (9b) as a carboxyl component resulted in a decreased yield of α-acylaminoamides 12a,c, which may be due to reduced nucleophilicity of the carbonyl carbon atom that undergoes the Mumm rearrangement [28].
In the 13 C NMR spectra of compounds 12c,d, no signal of carbonyl carbon at the diglyceride fragment was observed when CDCl 3 was used as a solvent. At the same time, a strongly broadened signal of the corresponding NH-proton was detected in the 1 H NMR spectra. Apparently, the intermolecular exchange of amide protons led to the strong broadening of the 13 C signal of the corresponding carbonyl atom and its merging with the base line of the spectrum. To avoid those problems, the spectra of compounds 12c,d were recorded in DMSO-d6, a solvent that somewhat suppresses the exchange of mobile protons.
Since cancer cells generally overexpress carbohydrate receptors, we attempted to prepare PAs that contain a diglyceride moiety at one terminal nitrogen atom and a carbohydrate moiety at the other via a two-step strategy using 2-(hydroxymethyl)benzoic acid [29]. This approach allows to prepare aminoamide, which acts as one of the components in the N-split Ugi reaction. Isonitrile 7c reacted with equimolar amounts of 2-(hydroxymethyl)benzoic acid (8c), piperazine (9d), and formaldehyde (10) (Scheme 3) to give monosubstituted aminoamide 13a in a 35% yield. The low yield of the desired product 13a was due to the formation of the symmetrical adduct of aminodiamide 13b with 22% yield. Subsequent treatment of 13a with isocyanide 14a-c, 2-(hydroxymethyl)benzoic acid (8c), and formaldehyde (10) led to formation of the disubstituted piperazines 15a-c in 75%, 80%, and 30% yields, respectively. The low yield of D-glucose containing compound 15c is supposedly associated with a partial deacetylation during reflux. The most common method to reduce the amide group to the corresponding amine utilizes LiAlH4 [30] or BH3 and its derivatives [31,32]. Unfortunately, in the case of our compounds, stable boron-amine complexes were formed which could not be further hydrolyzed to the desired amines 15a-f under basic or acidic conditions. Therefore, we applied phenylsilane and NiCl2(dme) [33] for a chemoselective aminoamides reduction, using 2 equivalents of phenylsilane and 0.1 equivalent of NiCl2(dme) per each amide group. As reported previously [34], the utilization of benzamide-type substrates is one of the key limitations for the most nickel-catalyzed amide reduction reactions. Indeed, using the abovementioned strategy, benzylamide-derivative 16a was obtained in a good yield of 66%, whereas the yields of piperazyl derivatives 16b-f were significantly lower (Scheme 4). Treatment of carbohydrate containing aminoamide 15c with phenylsilane did not provide the formation of the desired amine (Scheme 5) due to partial deacetylation of the D-glucose. To overcome this problem, the acetyl groups of aminoamide 15c were initially removed by sodium methoxide in methanol to yield compound 17 (80%). The following reduction of the amide 17 was unsuccessful, and the desired amine was not isolated from the reaction mixture. Thus, the reduction of the amide groups of aminodiamide 15c containing D-glucose requires additional efforts to find alternative synthetic approaches. The most common method to reduce the amide group to the corresponding amine utilizes LiAlH 4 [30] or BH 3 and its derivatives [31,32]. Unfortunately, in the case of our compounds, stable boron-amine complexes were formed which could not be further hydrolyzed to the desired amines 15a-f under basic or acidic conditions. Therefore, we applied phenylsilane and NiCl 2 (dme) [33] for a chemoselective aminoamides reduction, using 2 equivalents of phenylsilane and 0.1 equivalent of NiCl 2 (dme) per each amide group. As reported previously [34], the utilization of benzamide-type substrates is one of the key limitations for the most nickel-catalyzed amide reduction reactions. Indeed, using the abovementioned strategy, benzylamide-derivative 16a was obtained in a good yield of 66%, whereas the yields of piperazyl derivatives 16b-f were significantly lower (Scheme 4). The most common method to reduce the amide group to the corresponding amine utilizes LiAlH4 [30] or BH3 and its derivatives [31,32]. Unfortunately, in the case of our compounds, stable boron-amine complexes were formed which could not be further hydrolyzed to the desired amines 15a-f under basic or acidic conditions. Therefore, we applied phenylsilane and NiCl2(dme) [33] for a chemoselective aminoamides reduction, using 2 equivalents of phenylsilane and 0.1 equivalent of NiCl2(dme) per each amide group. As reported previously [34], the utilization of benzamide-type substrates is one of the key limitations for the most nickel-catalyzed amide reduction reactions. Indeed, using the abovementioned strategy, benzylamide-derivative 16a was obtained in a good yield of 66%, whereas the yields of piperazyl derivatives 16b-f were significantly lower (Scheme 4). Treatment of carbohydrate containing aminoamide 15c with phenylsilane did not provide the formation of the desired amine (Scheme 5) due to partial deacetylation of the D-glucose. To overcome this problem, the acetyl groups of aminoamide 15c were initially removed by sodium methoxide in methanol to yield compound 17 (80%). The following reduction of the amide 17 was unsuccessful, and the desired amine was not isolated from the reaction mixture. Thus, the reduction of the amide groups of aminodiamide 15c containing D-glucose requires additional efforts to find alternative synthetic approaches. Treatment of carbohydrate containing aminoamide 15c with phenylsilane did not provide the formation of the desired amine (Scheme 5) due to partial deacetylation of the D-glucose. To overcome this problem, the acetyl groups of aminoamide 15c were initially removed by sodium methoxide in methanol to yield compound 17 (80%). The following reduction of the amide 17 was unsuccessful, and the desired amine was not isolated from the reaction mixture. Thus, the reduction of the amide groups of aminodiamide 15c containing D-glucose requires additional efforts to find alternative synthetic approaches. The most common method to reduce the amide group to the corresponding amine utilizes LiAlH4 [30] or BH3 and its derivatives [31,32]. Unfortunately, in the case of our compounds, stable boron-amine complexes were formed which could not be further hydrolyzed to the desired amines 15a-f under basic or acidic conditions. Therefore, we applied phenylsilane and NiCl2(dme) [33] for a chemoselective aminoamides reduction, using 2 equivalents of phenylsilane and 0.1 equivalent of NiCl2(dme) per each amide group. As reported previously [34], the utilization of benzamide-type substrates is one of the key limitations for the most nickel-catalyzed amide reduction reactions. Indeed, using the abovementioned strategy, benzylamide-derivative 16a was obtained in a good yield of 66%, whereas the yields of piperazyl derivatives 16b-f were significantly lower (Scheme 4). Treatment of carbohydrate containing aminoamide 15c with phenylsilane did not provide the formation of the desired amine (Scheme 5) due to partial deacetylation of the D-glucose. To overcome this problem, the acetyl groups of aminoamide 15c were initially removed by sodium methoxide in methanol to yield compound 17 (80%). The following reduction of the amide 17 was unsuccessful, and the desired amine was not isolated from the reaction mixture. Thus, the reduction of the amide groups of aminodiamide 15c containing D-glucose requires additional efforts to find alternative synthetic approaches. Given the fact that Pas with a piperazyl domain have never been reported before, we chose piperazyl derivatives 16b and 16c-f with different hydrophobic domain structures; with short-chain substituents (ethyl (16c), isopropyl (16e), and pentyl (16f)) and different numbers of amino groups. To evaluate the effect of lipophilic PA structure on its anticancer activity, other aliphatic lipophilic PAs, which were obtained in this study in much lower yields, were not considered for further evaluation. Aminoamide 17 was used as the negative control.
The cytotoxicity of the lipophilic PAs 16a-f was determined using the MTT-test (see Supplementary Materials) in breast cancer (MCF7), human lung adenocarcinoma (A549), colon cancer (HCT116) cell lines ( Table 1). The cytotoxicity data showed that the presence of a diglyceride fragment as a hydrophobic domain (PAs 16c-f) increases their anticancer activity compared with the octadecyl substituent (PA 16b). Compound 16d with three amino groups showed the highest anticancer activity within all cell lines tested. Compounds with four amino groups 16c,e,f revealed similar anticancer activity. BENSpm (3) and the widely used anticancer agent cisplatin were selected as a positive control. Their IC 50 values obtained in this study were close to or of the same value as those obtained in previous reports [35][36][37]. Comparison of IC 50 values obtained suggests that new lipophilic PAs 16c-f have a cytotoxicity that is comparable to that of BENSpm, and several times higher than that of cisplatin.

Conclusions
Lipophilic Pas manifest excellent preliminary biological activity in cancer cell lines. However, the chemical synthesis of such compounds is complicated. In this paper, we demonstrated the efficient approach for the synthesis of new LPAs, which were obtained using the N-split Ugi multicomponent reaction. The application of this method allowed us to decrease the synthetic steps and to increase the total yield of LPAs from 7% to 28%. The application of PhSiH 3 and NiCl 2 (dme) effectively permitted us to reduce several amide groups in the PA precursors and has proven to be a very reliable and efficient method.
The obtained results demonstrate that the biological activity of the novel LPAs is several times higher than that of cisplatin, which is used in medical practice. At the same time, comparison with the clinically tested BENSpm showed similar cytotoxicity, which makes LPAs promising targets for further studies. More detailed biological evaluation will be carried out in the follow up study.

General
Commercially available solvents were used in this study. All the experiments were carried out under argon atmosphere with the use of the HPLC grade methanol. The reactions were monitored by thin-layer chromatography (TLC) on Silica gel 60 F 254 plates (Merck, Germany). The substances were identified in UV light (254 nm) by the treatment with Dragendorff's reagent, or by treatment with a solution of phosphomolybdic acidcerium sulfate (IV) with subsequent heating. Column chromatography was performed on Kieselgel 60 silica gel (0.040-0.063 or 0.063-0.200 mm, Merck, Germany). The 1 H and 13 C NMR spectra were recorded on Bruker DPX-300, Bruker Avance II 400, or Bruker Avance II 600 Fourier spectrometers (Bruker, Germany) in CDCl 3 , DMSO-d6 or acetone-d6. Chemical shifts (δ) were expressed in ppm relative to the peak of the residual proton of the solvent. The spin-spin interaction constants (J) are reported in Hz. The high-resolution mass spectra were recorded on a LCQ Deca XP Plus mass spectrometer with ESI ionization (Thermo Finnigan, San Jose, CA, USA) or FT ICR Apex Ultra 7 T (Bruker, Germany), mass spectra were recorded on the Agilent spectrometer. LCMS spectra were recorded on the LC Agilent Infinity 1260 II (Agilent, Beijing, China) and MSD Agilent IQ (Agilent, Singapore). Column PoroShell 120 EC-C18, 100 mm × 4.6 mm × 3 µm, constant flow 800 µL/min, linear gradient from 90% water + 0.1% FA-0-2 min to 90% ACN + 0.1% FA-15-25 min, voltage of ion capillary 3500 V, fragmenter 100 V.
To eliminate minor impurities compounds 16b-f that have been evaluated in cell models were additionally purified prior to use on silica gel and their purity (≥96%) was confirmed by LCMS method.

General Procedure for the Synthesis of Compounds 15a-c
The equimolar solution of corresponding isocyanide (14a-c), 2-(hydroxymethyl)benzoic acid (8c), amine (13a), and paraformaldehyde (10) in 0.5 M methanol was refluxed for 12 h. The solvent was evaporated, and the crude reaction mixture was purified by column chromatography.