Synthesis and Biological Evaluation of Antimetastatic Agents

Article pubs.acs.org/jmc

Synthesis and Biological Evaluation of Antimetastatic Agents Predicated upon Dihydromotuporamine C and Its Carbocyclic Derivatives Aaron Muth,† Veethika Pandey,‡ Navneet Kaur,† Melissa Wason,‡ Cheryl Baker,§ Xianlin Han,∥ Teresa R. Johnson,⊥ Deborah A. Altomare,‡ and Otto Phanstiel, IV*,⊥ †

Department of Chemistry, University of Central Florida, Orlando, Florida 32816, United States Burnett School of Biomedical Sciences, University of Central Florida, Orlando, Florida 32827, United States § BioCurity, Inc., New Smyrna Beach, Florida 32169, United States ∥ Sanford-Burnham Medical Research Institute, Orlando, Florida 32827, United States ⊥ Department of Medical Education, University of Central Florida, Orlando, Florida 32827, United States ‡

S Supporting Information *

ABSTRACT: The motuporamines isolated from the sea sponge Xestospongia exigua are of biological interest because of their unique antimigration and antiangiogenic properties. Key bioactive features were found to be a saturated 15membered heterocycle and a norspermidine motif. This paper describes new analogues that modulate the cytotoxicity of this compound class and have enhanced antimigration properties. By movement of the polyamine chain outside the ring, new carbocycles were discovered that doubled the antimigration potency and reduced compound toxicity by 133-fold. Mice injected with metastatic human L3.6pl pancreatic cancer cells demonstrated significant reduction in liver metastases when treated with N1-(3-aminopropyl)-N3-(cyclopentadecylmethyl)propane-1,3-diamine compared with dihydromotuporamine C. Significant changes in specific ceramide populations (N16:0 and N22:1) were noted in L3.6pl cells treated with dihydromotuporamine C but not for the cyclopentadecylmethylnorspermidine derivative, which had lower toxicity. Both compounds gave increased levels of specific low molecular weight sphingomyelins, suggesting that they may act upon sphingomyelin processing enzymes.

■

INTRODUCTION

In 1998, Andersen et al. reported the discovery of the motuporamine family of compounds from the sea sponge Xestospongia exigua off the coast of Motupore Island in Papua New Guinea.1 These compounds were of particular interest, as they each possessed a large heterocycle containing a norspermidine message (Figure 1). Biologically, dihydromotuporamine C (4a) proved to be the most interesting, as it was cytotoxic to MDA-MB-231 breast carcinoma cells and had good anti-invasive properties.1,2 Previous work by the Andersen group determined key structural aspects of dihydromotuporamine C (4a) responsible for its biological activity.2b The structure of the motuporamines can be broken into two main components: the macrocyclic ring and the norspermidine tail. The 15-membered macrocycle was preferred over other size rings, as it demonstrated the highest levels of cytotoxicity and invasion inhibition in MDA-MB-231 breast cancer cells.2a Williams et al. also showed that the degrees of unsaturation in the ring dramatically affected the biological activity of the series, where a saturated ring proved to be the most potent design in terms of antimigration properties.2a Work by our group also demonstrated that © 2014 American Chemical Society

Figure 1. Structures of motuporamine natural products (1, 2, 3, 4a) and dihydromotuporamine C mimics (5a−c) and Ant44 (5d). The number inside the macrocycle denotes ring size.

anthryl ring based mimics (5a and 5b) could reproduce the cytotoxic effects of 4a but not its antimetastatic properties.3 Received: December 11, 2013 Published: May 2, 2014 4023

dx.doi.org/10.1021/jm401906v | J. Med. Chem. 2014, 57, 4023−4034

Journal of Medicinal Chemistry

Article

The other common characteristic of the motuporamines is the polyamine tail. It was demonstrated that acetylation of the terminal amino group had no effect on the biological activity of motuporamine C (3); however, acetylation of both exocyclic amine groups showed a complete loss of activity.2b This finding suggested the importance of the central secondary amine for high biological activity. As shown in Figure 2, compounds 4a and 4b each possess a tertiary amine at the N1 position. Our early efforts to mimic 4a

Since the goal was to install a polyamine message onto the motuporamine architecture to provide cell targeting properties via the PTS, it seemed reasonable to consider derivatives that made the polyamine tail more accessible to cell surface receptors. In this report derivatives of 4a and 4b that incrementally moved the polyamine away from the macrocyclic core were investigated. The goal was to understand how additional tether length between the macrocycle and polyamine message affected PTS selectivity, cytotoxicity, and antimigration ability of new compounds 6 and 7 (Figure 2). Improvements in antimigration ability were initially unexpected because the Andersen group had synthesized over 40 motuporamine mimics and none of them had outperformed the parent system 4a in terms of its antimigration properties.2 We now report, however, the discovery of 7a, an analogue that is 133-fold less toxic and twice as potent as the parent 4a in terms of its antimigration properties. The antimigration properties of 7a were demonstrated both in vitro and in vivo. This is a significant discovery, as it demonstrates for the first time how one can reduce the toxicity of motuporamines and improve upon their antimetastatic properties. The fact that commercially available cyclic ketones can be used in lieu of macroheterocycles (generated by multistep synthesis)3a adds further value to this report.

■

RESULTS AND DISCUSSION Synthesis. The synthesis of 4a was previously described by Goldring et al.,6 Fürstner et al.,7 Williams et al.,2a and Kaur et al.,3a while the synthesis of 4b was previously described by Kaur et al.3a For installing the polyamine outside the ring, we elected to use the commercially available ketone cyclopentadecanone, 8. As shown in Scheme 1, reductive amination of ketone 8 with Boc-protected polyamines (9a or 9b)8 gave the respective conjugate compounds 10a and 10b in moderate yield (54% and 75%, respectively). These two compounds were then each treated with 4 M HCl to remove the Boc groups and provided 6a and 6b in high yield (98% and 77%, respectively). The overall yields for compounds 6a and 6b were 53% and 74% from ketone 8, respectively. In contrast, the synthesis of the extended motifs 7a and 7b proved to be more challenging as the generation of aldehyde 13 was more difficult than expected (Scheme 2). The first attempt to generate aldehyde 13 utilized (methoxymethyl)triphenylphosphonium chloride in a classic Wittig reaction to generate the vinyl ether, followed by hydrolysis. Generation of the vinyl ether failed, despite efforts using several base sources (BuLi, NaHMDS, NaH), and only the starting ketone 8 was recovered. Another attempt utilized Vilsmeier−Haack conditions with POCl3 in DMF to generate (Z)-2-chlorocyclopentadec-1-enecarbaldehyde followed by reduction with H2 and Pd/C. The alkene intermediate was generated fairly smoothly,

Figure 2. Motuporamine derivatives (4, 6, and 7) and the tricyclic antidepressants imipramine and desipramine.

resulted in the synthesis of anthracenyl polyamines 5a−d (Figure 1). The anthracenylmethyl substituent was chosen to represent the 15 carbon side chain of 4a, and the N1 position was N-ethylated to introduce a 3° amine at the N1 position as seen in 4a. The polyamine chain was altered from norspermidine (5a) to homospermidine (5c) in an effort to target this compound to cancer cells via their polyamine transport system (PTS). This strategy was based upon prior work that had shown that the homospermidine motif (5d) was a superior PTS targeting ligand.3a,4 These motuporamine mimics were then assessed for their ability to target the PTS. 3a Unfortunately, none of the analogues of 4a (5a−c) outperformed Ant44 (5d) in terms of PTS targeting.3a,5 The N-ethylated norspermidine analogue, 5a, was shown to mimic the cytotoxicity profile targeting of 4a in several cell lines.3b We speculated that the steric crowding associated with the macrocycle may have buried and partially obscured the polyamine component from binding to its putative cell surface receptor. Indeed, molecular modeling demonstrated that the Nethylation step favored orientation of the polyamine message over the anthracene ring. This phenomenon was shown experimentally via significant upfield shifts in the 1H NMR spectrum of the N-ethylated systems compared to their nonethylated counterparts.3b Scheme 1. Synthesis of 6a and 6ba

a

Reagents: (i) NaBH(OAc)3, AcOH, CH2Cl2; (ii) 4 M HCl, EtOH. 4024

dx.doi.org/10.1021/jm401906v | J. Med. Chem. 2014, 57, 4023−4034

Journal of Medicinal Chemistry

Article

Scheme 2. Synthesis of 7a and 7ba

Reagents: (i) Ph3PCH3I, BuLi, THF, 0 °C; (ii) BH3−THF, 0 °C, then H2O2, 3 M NaOH, rt; (iii) PCC, CH2Cl2; (iv) 9a8b or 9b,8a NaBH(OAc)3, AcOH, CH2Cl2; (v) 4 M HCl, EtOH.

a

albeit at 70% conversion. However, the subsequent reduction method was unsuccessful at reducing the alkene intermediate to the desired aldehyde 13. After these unsuccessful attempts at generating aldehyde 13, a longer synthetic route was developed. This route also began with ketone 8 and a standard Wittig reaction using methyltriphenylphosphonium iodide to generate alkene 11 in 69% yield. Alkene 11 was then subjected to hydroboration− oxidation conditions, followed by oxidation of the resultant alcohol with PCC to generate aldehyde 13. Attempts to purify aldehyde 13 by column chromatography provided a sharp reduction in yield. Therefore, aldehyde 13 was best used without further purification. Reductive amination of 13 with Boc-protected polyamines 9a and 9b generated 14a and 14b in 36% and 51% yield, respectively. These Boc-protected compounds were then deprotected with 4 M HCl to give 7a and 7b in 92% and 95% yields, respectively. The overall yields of 7a and 7b were 17% and 21% from ketone 8, respectively. Biological Evaluation. Once synthesized, the conjugates were screened for toxicity in Chinese hamster ovary (CHO), CHO-MG*, and L3.6pl cells. CHO cells were chosen along with a mutant PTS-deficient line (CHO-MG*) in order to comment on polyamine transport selectivity.5,9 L3.6pl cells were selected as a metastatic human pancreatic cancer cell line to assess the compounds’ cytotoxicity and antimigration properties.10 CHO and CHO-MG* Studies. With the compounds in hand, their ability to target the PTS was assessed via a published CHO assay.4 CHO cells were chosen along with a mutant cell line (CHO-MG*) to comment on how the synthetic conjugates gain access to the cells.11 The wild-type CHO cell line represents cells with high polyamine transport activity.9a,12 In contrast, the CHO-MG* cell line is polyaminetransport deficient and represents a model for alternative modes of entry (other than the PTS) including passive diffusion or use of another transporter.9 A comparison of toxicity in these two CHO cell lines allowed for a screen that would detect selective use of the polyamine transport system.11 Cytotoxic compounds with high utilization of the PTS would be very toxic to CHO cells. However, the CHO-MG* cells should be less sensitive to these compounds because of limited transport activity.4,11 Ultimately, a CHO-MG*/CHO IC50 ratio was determined to assess PTS selectivity. A high IC50 ratio (>10) would be observed for highly PTS selective compounds. The results obtained with compounds 4, 6, and 7 are shown in Table 1. As reported earlier, dihydromotuporamine C (4a) and its analogue 4b did not exhibit PTS selectivity, as their IC50 ratios were both 1.3a Low CHO-MG*/CHO IC50 ratios were

Table 1. Biological Evaluation of Motuporamine Derivatives (4, 6, and 7) in CHO and CHO-MG* Cellsa IC50 (μM) compd 4a (Motu33) 4b (Motu44) 6a (MotuN33) 6b (MotuN44) 7a (MotuCH233) 7b (MotuCH244)

CHO-MG* 2.96 4.67 5.95 5.18 87.3 45.5

(±0.1) (±0.9) (±0.5) (±0.9) (±11) (±1.4)

CHO 2.90 4.38 2.84 4.53 82.9 47.8

(±0.2) (±0.3) (±0.2) (±0.2) (±20) (±2.4)

IC50 ratiob 1 1 2.1 1 1 1

a

CHO and CHO-MG* cells were incubated with 1 mM aminoguanidine (AG) for 24 h prior to compound addition. Cells were incubated for 48 h at 37 °C with the respective conjugate. All experiments were run in triplicate. bThe ratio denotes the (CHOMG*/CHO) IC50 ratio, a measure of PTS selectivity.

observed for analogues 6a, 6b, 7a, and 7b, demonstrating a general lack of PTS selectivity even though 6b and 7b contain the PTS-targeting homospermidine motif. Despite this apparent lack of PTS selectivity, an interesting general trend in cytotoxicity was observed in both CHO and CHO-MG*. Compounds 4 and 6, which have the polyamine embedded or directly appended to the macrocycle, have similar cytotoxicity profiles. In contrast, the cytotoxicity decreased significantly when the polyamine chain was appended further away via a methylene linker from the macrocyclic core (e.g, derivatives 7a and 7b). These findings are consistent with prior investigations of anthracenylalkylhomospermidine conjugates, where both the compound toxicity and the CHOMG/CHO IC50 ratio decreased when the polyamine message was moved further away from the anthryl ring via methylene spacers.13 It is also interesting to note that the 4,4-analogue 7b was approximately twice as toxic as the 3,3-analogue 7a. In an effort to determine if the cytotoxicity seen for compound 7b was polyamine-transport-related, CHO cells were treated with a toxic dose of 7b (100 μM, 2 × IC50). Compound 7b was chosen because of its structural similarities to 5d, a known PTS ligand.5 Prior studies with 5d revealed that exogenous spermidine could be used to rescue cells treated with 5d, which was consistent with its use of the PTS. Increasing amounts of spermidine (Spd) were added to CHO cells dosed with compound 7b (100 μM, 2 × IC50). No cell rescue was observed even up to 100 μM Spd, which was not toxic to cells. These results suggested that, unlike 5d, compound 7b did not use the PTS for cell entry and is consistent with the CHO IC50 ratio result in Table 1. L3.6pl Studies. Compounds 4, 6, and 7 were also tested in the metastatic human pancreatic cancer cell line L3.6pl.10 This 4025

dx.doi.org/10.1021/jm401906v | J. Med. Chem. 2014, 57, 4023−4034

Journal of Medicinal Chemistry

Article

cell line was chosen because of its K-ras mutation which has been linked to increased polyamine uptake in other cancers.14 The increased polyamine uptake activity of L3.6pl cells provided a cell line that was sensitive to polyamine conjugates.11 Moreover, this pancreatic cell line is very aggressive in vivo and was selected for its high metastatic potential via six rounds of metastasis in mice.10 These features made this cell line an ideal candidate for evaluating antimetastatic polyamine-based compounds. Even though the CHO studies indicated that these compounds were not PTSselective, we were interested in how the series behaved in a metastatic human cancer cell line with high polyamine transport activity.11 Compound 4a was especially promising because of its reported antiangiogenic and antimetastatic properties.15 Compounds 4, 6, and 7 were tested in L3.6pl cells, and their IC50 values (Table 2) were nearly identical to the IC50 values

time. Each compound was evaluated for its dose-dependent inhibition of cell migration over 48 h. Each compound was evaluated at its own IC5 value in order to avoid cytotoxic effects, which are well-known to bias migration studies.2a In addition, each compound was also tested at the IC5 value of 4a (Motu33, 0.6 μM) to directly compare the potency of each compound at the same concentration. These findings are reported in Table 3. Table 3. Inhibition of L3.6pl Cell Migration by Compounds 4, 6, and 7a compd control 4a (Motu33) 4b (Motu44) 6a (MotuN33)

Table 2. Biological Evaluation of Motuporamine Derivatives (4, 6, and 7) in L3.6pl Cellsa

6b (MotuN44)

L3.6pl

7a (MotuCH233)

compd

IC50 (μM)

IC5 (μM)

4a (Motu33) 4b (Motu44) 6a (MotuN33) 6b (MotuN44) 7a (MotuCH233) 7b (MotuCH244)

0.99 (±0.07) 1.8 (±0.2) 2.8 (±0.2) 2.8 (±0.1) 89.4 (±5.4) 48.7 (±2.8)

0.60 (±0.04) 1.0 (±0.1) 2.0 (±0.2) 2.0 (±0.1) 80 (±4) 40 (±3)

7b (MotuCH244)

concn (μM)

% cell migration inhibition

untreated 0.6b 0.6 1b 0.6 2b 0.6 2b 0.6 80b 0.6 40b

0 (±1.5) 20.3 (±1.8) 16.7 (±0.8) 29.6 (±4.5) 19.3 (±1.3) 32.9 (±0.1) 31.4 (±4.1) 33.4 (±0.5) 38.4 (±4.2) 44.8 (±2.2) 29.6 (±1.6) 45.3 (±3.7)

L3.6pl cells were incubated with 250 μM AG for 24 h prior to compound addition. Cells were incubated for 48 h at 37 °C with the respective conjugate. bIC5 value of the respective conjugate. a

a L3.6pl cells were incubated with 250 μM aminoguanidine (AG) for 24 h prior to compound addition. Cells were incubated for 48 h at 37 °C with the respective conjugate. All experiments were run in triplicate. Steep cytotoxicity curves were observed, and as a result the IC5 value was often near the respective IC50 values. A similar steep curve was also seen with CHO cells and may reflect the importance of the putative target to cell growth.

Time-lapsed images were obtained using a 2.5× objective in order to view ∼95% of each well. The resultant images (Figures S1A−S1F shown in the Supporting Information) were analyzed via the Image J software.17 This software allowed for enhancing the contrast between areas that were occupied by cells versus areas devoid of cells. In this regard, one was able to accurately define the edges of each boundary. Images with defined boundaries between the cell-free and cell-containing zones provided relative area calculations at t = 0 and t = 48 h. These were then used to calculate the % cell migration inhibition values, which are listed in Table 3 (see Experimental Section). When the 3,3-triamine (norspermidine) containing systems 4a, 6a, and 7a were compared at 0.6 μM, compound 7a (MotuCH233) had the highest inhibition observed in the series. Thus, the insertion of a methylene spacer between the macrocycle and the N1-position of the polyamine provided increased cell migration inhibition. The corresponding 4,4triamine series at 0.6 μM also showed a significant increase in antimigration ability upon moving the polyamine message outside the ring (4b → 6b ∼ 7b). Compounds 7a (IC5 = 80 μM) and 7b (IC5 = 40 μM) were especially interesting and had IC5 values that were significantly higher than 0.6 μM (>130fold and >60 fold, respectively). In this regard, these derivatives were significantly less toxic than the parent 4a. While dosedependent responses were observed, dramatic increases in cell migration inhibition (e.g., 100% inhibition) were not observed even at significantly higher doses (e.g., 7a in Table 3), suggesting that the target of these compounds may be easily saturated at lower doses. The significant differences observed in cytotoxicity suggested that the polyamine tail, especially norspermidine, and increased distance from the macrocycle provide ways to reduce the cytotoxicity of the 4a pharmacophore.

found in CHO and CHO-MG*. The IC50 value was the concentration of compound that resulted in 50% inhibition of L3.6pl cell growth compared to untreated control. These IC50 results were interesting because they showed that as the polyamine message was moved away from the macrocyclic core, the cytotoxicity decreased, which was most evident for 7a and 7b. The IC5 value was also determined for each of these compounds (Table 2). The IC5 value was the concentration of compound that resulted in 5% inhibition of L3.6pl cell growth (i.e., cells remained ≥95% viable) compared to untreated control. This dosing information was important, as it allowed each compound’s antimigration properties to be assessed at a relatively nontoxic concentration and also at the IC5 of the parent compound 4a (L3.6pl IC5 4a = 0.6 μM) in subsequent wound healing assays for comparisons. The parent compound 4a was the most toxic of the series, and all compounds were compared at 0.6 μM (with no introduction of toxicity). Wound Healing Assay. A wound healing assay was employed to evaluate the antimigration properties of compounds 4, 6, and 7 in L3.6pl cells. This involved growing a monolayer of L3.6pl cells and then scraping the monolayer with a sterile pipet tip to create a cell free channel in the middle of the well.16 The experiment was run in triplicate in a 96-well plate format. Wound healing was measured by time-course imaging of each well using a computer-controlled microscope with an x, y, z stage that allowed for reproducible imaging over 4026

dx.doi.org/10.1021/jm401906v | J. Med. Chem. 2014, 57, 4023−4034

Journal of Medicinal Chemistry

Article

Figure 3. Inhibited liver micrometastasis in mice treated with 7a (MotuCH233) relative to vehicle or the parent compound 4a (Motu33, with representative images from each treatment group). First column shows H&E stained images of mouse liver tissue with tumor invasion indicated by black arrows. The scale bar on each H&E image indicates 500 μm. Second and third columns show nuclear staining (brown) of the proliferation marker Ki67 and the human epithelial cell marker cytokeratin-18 membrane staining (brown) in tumor cells, respectively. The scale bars for the Ki67 and cytokeratin-18 columns denotes 50 μm. Cytokeratin-18 staining could not be confirmed in the tumor cells in the liver from the mouse treated with 7a. Indeed, both the incidence and size of the liver metastases were significantly smaller in the mice treated with 7a.

Xenograft Study. Preliminary experiments in nu/nu mice indicated that the parent compound 4a could be dosed by ip injection 5 days per week for 2 weeks (or alternatively at a regimen of 3 days per week for 3 weeks) at 2.5 mg of 4a/kg of body weight with no adverse effects (see Experimental Section). In this regard, 2.5 mg/kg represented the maximum tolerated dose (MTD) of 4a in this mouse model that could be given repeatedly over the time period of the experiment with no adverse effects. Consistent with the in vitro findings, compound 7a was found to be less toxic than 4a in vivo and was dosed at an equimolar dose to 4a for comparisons. Mice received ip injections of vehicle or each compound (e.g., 2.50 mg of 4a/kg-mouse, 2.66 mg of 7a/kg-mouse for 5 days per week for 2 weeks). After a total of 10 vehicle or compound injections per L3.6pl pancreatic tumor engrafted mouse, the experiment was terminated and tumor and liver tissues were collected. The incidence and metastases of L3.6pl pancreatic tumors in the livers of the mice were then recorded. There was no significant difference observed between the vehicle-treated and the 4a treated groups, either macroscopically or upon histological analysis in terms of the severity of metastasis, with four of seven mice in each group showing detectable liver micrometastases. In contrast, only two of seven mice from the 7a (MotuCH233) group had micrometastases, and these were significantly smaller by comparison (see Figure 3). Representative sections of hematoxylin and eosin (H&E), in

combination with immunohistochemical detection of human Ki67 for cell proliferation and human cytokeratin-18 protein marker, showed markedly reduced spreading in the 7a (MotuCH233) treated group (Figure 3), compared to the vehicle or dihydromotuporamine C (4a) treated mice. Differences were especially evident in regard to both the incidence and size of micrometastasis in the liver. Rewardingly, these in vivo results were very consistent with the in vitro performance of 7a in cell culture. Sphingolipid Metabolism Targeting. Using a genome-wide haploinsufficiency screen in yeast, Baetz et al. demonstrated that dihydromotuporamine C (4a) targets sphingolipid metabolism (Figure 4).18 Specifically, two heterozygous yeast deletion strains involving genes LCB1 and TSC10 (i.e., strains lcb1Δ/LCB1 and tsc10Δ/TSC10) were shown to be supersensitive to the addition of 4a.18 These genes are involved in the biosynthesis of dihydrosphingosine (DHS), and compound 4a at 60 μM was able to fully inhibit the growth of yeast cells.19 The addition of DHS (an intermediate in the biosynthesis of ceramide pathway, Figure 4) to 4a-treated yeast cells was able to rescue this growth inhibition.18,19 Compound 4a was also shown to lower the intracellular levels of ceramide and the addition of exogenous ceramide was able to partially rescue yeast growth, suggesting that 4a directly targets sphingolipid metabolism.18,20 4027

dx.doi.org/10.1021/jm401906v | J. Med. Chem. 2014, 57, 4023−4034

Journal of Medicinal Chemistry

Article

properties.18,20b Therefore, the target of 4a remains unclear because subsequent in vitro experiments failed to show that 4a inhibits serine palmitoyltransferase (step a in Figure 4), which was one of the predicted targets.15 To gain further mechanistic knowledge of 4a, we investigated the ceramide populations of L3.6pl cells treated with 4a or 7a (versus untreated cells as a control). Specifically, compounds 4a and 7a were tested at the IC5 value of 4a (0.6 μM) for their influence on ceramide populations in L3.6pl cells, and the results are shown in Table 4. At this IC5 concentration, no toxicity from the compounds was expected based upon parallel experiments under the same conditions. Since sphingolipid metabolism was indicated as a target for 4a,18 we conducted experiments to measure the sphingolipid profiles of cells treated with vehicle, 4a, and 7a. We anticipated a case where 4a gave sphingolipid metabolite levels that were significantly different from the untreated control and 7a. Metabolites that met this criteria (case 1) may provide insight into the cellular processes that give rise to the toxicity of 4a because the IC5 of 4a (0.6 μM) was very close to the IC50 value of 4a (0.99 μM). In addition, we also anticipated a second case (case 2), where compounds 4a and 7a gave similar sphingolipid metabolic profiles but levels that were significantly different from that of the untreated control. Sphingolipid metabolites that fit this second criteria may give insight into the antimigration properties of these compounds, as both 4a and 7a inhibited cell migration at the 0.6 μM dose in vitro. After incubation of L3.6pl cells for 48 h with each compound (4a, 7a, and an untreated control), eight ceramide species containing different side chains (N16:0, N18:0, N20:0, N22:1, N22:0, N24:2, N24:1, N24:0) were measured and only three ceramide species showed significant differences between 4a and

Figure 4. Mammalian Sphingolipid metabolism: (a) serine palmitoyltransferase, (b) 3-ketosphingosine reductase, (c) ceramide synthase, (d) ceramide desaturase, (e) ceramidase, (f) sphingosine kinase, (g) sphingosine 1-phosphate lyase, (h) phosphatase, (i) ceramide kinase, (j) sphingomyelin synthase, (k) sphingomyelinase, (l) glucosylceramide synthase, (m) glycosidases, (n) sphingosine 1-phosphate phosphatase.

Despite this finding in yeast, exogenous ceramide could not completely rescue the effects of 4a.18,20b This observation was explained by 4a having multiple targets in these cells or that the rescue event was limited by the amount of ceramide that could be added, as ceramide is known to have proapoptotic

Table 4. Intracellular Ceramide Species Concentrations (pmol/mg protein) after a 48 h Treatment with 0.6 μM 4a or 7a in L3.6pl Cells

treatment

N16:0

N22:1

N20:0a

totalb

4a % contributionc 7a % contributionc control % contributionc

108.2 ± 19.2 20.3% 251.4 ± 39.5 38.4% 201.9 ± 94.8 37.3%

98.8 ± 15.9 18.6% 30.1 ± 11.4 4.6% 35.0 ± 37.1 6.5%

3.2 ± 1.2 0.6% 10.1 ± 3.7 1.5% 6.2 ± 1.2 1.1%

532.1 ± 45.1 653.0 ± 105.4 541.8 ± 134.2

a

These values also include a possible contribution from a hydroxylated derivative, OH_N19:1 ceramide, which has a similar mass to the N20:0 analogue (these two ceramides were not distinguished in the study and are potentially summed together in the table entry). bSum of all eight detected ceramide species in pmol/mg protein. c% contribution to the total ceramide population measured (example, 100 × 108.2/532.1 = 20.3%). 4028

dx.doi.org/10.1021/jm401906v | J. Med. Chem. 2014, 57, 4023−4034

Journal of Medicinal Chemistry

Article

Table 5. L3.6pl Cell Sphingomyelin Levels (Expressed in nmol/mg protein) after a 48 h Treatment with 0.6 μM 4a or 7a vs Untreated L3.6pl Cellsa

a

The green highlighted entries denote the levels of N16:0 and N22:1 sphingomyelins, which can be interconverted with the aforementioned N16:0 and N22:1 ceramides, respectively. The red highlighted entries are statistically significant compared to the untreated control. bMean values differed significantly across treatments (p = 0.001); 4a and 7a both differed significantly from the untreated control (p = 0.001 and 0.01, respectively). cMean values differed significantly across treatments (p = 0.01); 4a and 7a both differed significantly from the untreated control (p = 0.03 and 0.01, respectively). dMean values differed significantly across treatments (p = 0.01); 4a and 7a both differed significantly from the untreated control (p = 0.02 and 0.01, respectively). eMean values differed significantly across treatments (p = 0.03); 7a differed significantly from the untreated control (p = 0.03). fMean values differed significantly across treatments (p = 0.045); 7a differed significantly from the untreated control (p = 0.04).

treated with 4a had nearly half the level of saturated N16:0 ceramide and nearly triple the level of the unsaturated ceramide N22:1 compared to untreated cells (and cells treated with 7a) is relevant because alterations in ceramide populations can give rise to dramatic changes in membrane fluidity and morphology.22,23 These cellular outcomes stem from the interesting biophysical properties of ceramides containing saturated and unsaturated N-acylated fatty acid side chains. For example, Pinto et al. recently compared the biophysical properties of ceramides containing saturated C16, C18, C24 and unsaturated C16:1, C18:1, and C24:1 acyl side chains.22 Ceramides with saturated acyl side chains were found to increase membrane lateral organization and promote gel/fluid separation, while their unsaturated counterparts have a lower ability to form gel domains at 37 °C.22 Very long acyl side chains like C24:1 can create localized morphological alterations (e.g., tubule formation) via interdigitation processes, which may have a strong impact on cell function.22,24 We speculate that the increased levels of the unsaturated N22:1 ceramide and lower levels of the saturated N16:0 ceramide observed with 4a may have profound consequences in terms of membrane stability, which in turn could lead to toxic outcomes for the cell. Since ceramides and sphingomyelin are interconvertible via sphingomyelin synthase and sphingomyelinases (Figure 4), we investigated whether the specific ceramides (N16:0 and N22:1) were derived from their corresponding sphingomyelin population. The results (highlighted in green) are shown in Table 5. Unfortunately, since the levels of sphingomyelins containing N16:0 and N22:1 acyl chains in cells treated with 4a and 7a (vs

7a (see Experimental Section and Supporting Information). Mean total ceramide concentrations shown in Table 4 were not significantly different across the three treatments (p = 0.34), and compound 7a uniformly resembled the control values for each species measured (see Table 4 and Supporting Information). In contrast, there were specific ceramide populations that were noticeably altered upon treatment with 4a (Table 4). Of these three species, the two major contributors to the overall ceramide population were ceramides containing N-acylated 16:0 and 22:1 fatty acid side chains. Compound 4a resulted in a 2.8-fold increase in the intracellular amount of N22:1 ceramide, which approximately tripled its contribution to the intracellular ceramide pools versus the untreated control. As described in the Experimental Section, statistical pairwise comparisons of the measured N22:1 ceramide levels gave 4a vs control (p = 0.04), 7a vs control (p = 0.97), and 4a vs 7a (p = 0.03). Moreover, these perturbations to the N22:1 ceramide pool had a dramatic impact on the intracellular N16:0 to N22:1 ceramide ratios, where the N16:0/N22:1 ratios were the following: 4a, 1.1; 7a, 8.4; control, 5.8. Precisely how the decreased N16:0 and increased N22:1 ceramide levels influence cell viability is unclear at this time. We speculate that these alterations give rise to toxic outcomes for the cell based upon the fact that, unlike 4a, compound 7a had a high IC50 value (low toxicity) and a high N16:0/N22:1 ratio and resembled the untreated control (case 1). Indeed, ceramides have profound effects on cell signaling and are known to play roles in apoptosis.21 The fact that L3.6pl cells 4029

dx.doi.org/10.1021/jm401906v | J. Med. Chem. 2014, 57, 4023−4034

Journal of Medicinal Chemistry

Article

polyamine message from the macrocyclic ring can have a profound effect on the biological properties of the motuporamine pharmacophore. Compounds 4, 6, and 7 did not demonstrate PTS selectivity in the CHO-MG*/CHO screen. The biological evaluation in L3.6pl, CHO, and CHO-MG* showed that a methylene spacer between the N1-amine and the macrocyclic ring (i.e., 7a and 7b) dramatically reduced cytotoxicity. It was also demonstrated that a change in the polyamine tail from norspermidine to homospermidine showed a significant effect on cytotoxicity and the magnitude and specificity of this effect depended on the system investigated. The wound healing assay demonstrated that insertion of a methylene group between the macrocyclic ring and polyamine not only reduced the cytotoxicity but also greatly increased the antimigration properties. When 7a (IC5 = 80 μM) and 7b (IC5 = 40 μM) were both dosed at the IC5 value of 4a (i.e., 0.6 μM), a significant decrease in cell migration was observed for 7a and 7b relative to the parent system, 4a. Compound 7a was also shown to be well-tolerated in vivo and was effective in reducing the metastases of L3.6pl pancreatic tumor cells in vivo.10 The current study is the first of its kind to show that dihydromotuporamine C (4a) alters N16:0 and N22:1 ceramide populations in human cancer cells. This potentially toxic property was designed out of the motuporamine pharmacophore by moving the polyamine sequence outside the macrocycle (e.g., 7a). Most importantly, this structural modification provided an effective in vivo antimetastatic agent, 7a, with enhanced antimigration potency and lower toxicity than the parent compound 4a. In addition, this is the first study to demonstrate a buildup of specific sphingomyelin pools in human cancer cells treated with 4a and raises the possibility that 4a and 7a may act as selective sphingomyelin synthase agonists or sphingomyelinase inhibitors. Lastly, the discovery of 7a is significant, as it provides a relatively nontoxic, antimetastatic agent for future preclinical studies with metastatic pancreatic cancers.28

untreated cells) are in the nmol/mg protein range (and are not significantly different across treatments) and the expected molar changes in ceramide levels are in the pmol/mg protein range, one could not conclude definitively whether the sphingomyelin pools provided the specific source of N22:1 ceramide or acted as the sole repository for the storage of N16:0 acyl chains. Nevertheless, Table 5 revealed an interesting trend for both 4a and 7a. Both compounds resulted in a statistically significant buildup of specific short chain sphingomyelins (e.g., N16:1, N15:1, and N14:0) compared to the untreated control (case 2). As a result, the total sphingomyelin pools were higher in 4a and 7a treated L3.6pl cells [8.19 (±0.76) and 9.00 (±1.01) nmol of sphingomyelin/mg of protein, respectively] compared to the untreated control, 6.57 (±1.43) nmol of sphingomyelin/mg of protein. While these sum totals were not significantly different, the individual low molecular weight sphingomyelin populations were significantly different (e.g., N16:1, N15:1, and N14:0) from that of the untreated control (see footnotes in Table 5). There are two likely explanations for these findings. First, these compounds may act directly or indirectly as sphingomyelin synthase agonists and selectively convert low molecular weight ceramides into low molecular weight sphingomyelins, which may result in antimigration properties. A recent report by Asano et al. showed that sphingomyelin synthase def iciencies promoted cell migration through a CXCL12/CXCR4-dependent pathway and that added sphingomyelin repressed the dimerization of CXCR4 and reduced the responsiveness and migration induced by the CXCL12 cytokine.25 We speculate that the increased low molecular weight sphingomyelin levels induced by 4a and 7a may inhibit CXCR4 signaling in a similar manner. Alternatively, these compounds could act as selective sphingomyelinase inhibitors, which inhibit sphingomyelinases operating primarily on short chain sphingomyelin substrates. This putative mechanism of action would also explain the buildup of specific sphingomyelin pools noted in Table 5. Indeed, two known functional acid sphingomyelinase (ASMase) inhibitors, imipramine and desipramine, shown in Figure 2, also contain a large macrocycle design (with 15 atoms along its circumference) and an appended amine motif similar to 4a.26 Bases like desipramine that can be trapped in acidic lysosomal compartments are not direct inhibitors of ASMase but work by detaching ASMase from the inner lysosomal membrane. This detachment results in inactivation of ASMase presumably via proteolytic degradation.26 The fact that structurally related compounds (imipramine and desipramine) have this property suggests that 4a and 7a could, indeed, function as sphingomyelinase inhibitors. The ability to modulate sphingomyelin pools is important, as sphingomyelin has been shown to play a key role in vesicle-induced endothelial cell migration and angiogenesis.27a While more work is needed to confirm this latter hypothesis, the development of novel sphingomyelinase inhibitors that target select sphingomyelin populations is exciting and may have additional applications in liver, neurological, and cardiovascular diseases.27b Regardless of its precise antimigration mechanism, compound 7a has promising properties worthy of further study, especially in combination with other chemotherapeutic agents.

■

EXPERIMENTAL SECTION

Materials. Silica gel 32−63 μm and chemical reagents were purchased from commercial sources and used without further purification. All solvents were distilled prior to use. All reactions were carried out under an N2 atmosphere, when appropriate. 1H and 13 C spectra were recorded at 400 and 75 MHz, respectively. TLC solvent systems were listed in volume percentages, and NH4OH referred to concentrated aqueous ammonium hydroxide. All tested compounds provided satisfactory elemental analyses and were ≥95% pure. Statistical Analyses. Descriptive data were presented as the mean ± standard deviation. Between-group comparisons were conducted with analysis of variance, and all follow-up pairwise comparisons were made using Tukey post hoc tests. Tests were two-sided, and p-values of