Using the Cancer Dependency Map to Identify the Mechanism of Action of a Cytotoxic Alkenyl Derivative from the Fruit of Choerospondias axillaris

Abstract

An extract prepared from the fruit of Choerospondias axillaris exhibited differential cytotoxic effects when tested in a panel of pediatric cancer cell lines [Ewing sarcoma (A-673), rhabdomyosarcoma (SJCRH30), medulloblastoma (D283), and hepatoblastoma (Hep293TT)]. Bioassay-guided fractionation led to the purification of five new hydroquinone-based metabolites, choerosponols A-E (1-5), bearing unsaturated hydrocarbon chains. The structures of the natural products were determined using a combination of 1D and 2D NMR, HRESIMS, ECD spectroscopy, and Mosher ester analyses. The purified compounds were evaluated for their antiproliferative and cytotoxic activities revealing that 1, which contains a benzofuran moiety, exhibited over 50-fold selective antiproliferative activity against Ewing sarcoma and medulloblastoma cells with growth inhibitory (GI₅₀) values of 0.19 and 0.07 μM, respectively. The effects of 1 were evaluated in a larger panel of cancer cell lines and these data were used in turn to interrogate cancer dependency databases leading to the identification of the MCT1 transporter as a functional target of 1. These data highlight the utility of publicly available cancer dependency databases like Project Achilles to facilitate the identification of the mechanisms of action of compounds with selective activities among cancer cell lines, which can be a major challenge in natural products drug discovery.

Graphic Abstract

Choerospondias axillaris (Roxb.) Burtt et Hill, also known as the Nepali hog plum, is a deciduous tree in the Anacardiaceae. The fruit, which are consumed throughout many parts of Asia, are used as a folk treatment for cardiovascular conditions and these traditional uses have inspired pharmacological investigations of its chemical components.^1,2 A flavonoid-enriched extract was reported to attenuate ischemia-induced apoptosis in a rat model of acute myocardial infarction and prevent ischemia-induced activation of p38 MARK and Jun.¹ In addition, C. axillaris fruit components have been shown to improve cardiac function and reduce myocardial interstitial fibrosis in a rat model of chronic myocardial infarction.²

Further chemical studies of C. axillaris unrelated to its cardiovascular benefits resulted in the purification of the alkylated polycyclic cyclohexanones choerosponins A and B. While choerosponin A exhibited weak activity against cancer cell lines with IC₅₀ values over 100 μM, choerosponin B inhibited the growth of multiple cancer cell lines (HCT-15, HeLa, and A2780) with IC₅₀ values ranging from 3 – 13 μM.³ Structurally-similar cytotoxic metabolites have been reported from other Anacardiaceous species^4–9 including lanneaquinol and (R)-2′-hydroxylanneaquinol from Lannea welwitschii, 2-alkenyl-1,4-dihydroxybenzene from Tapirira guianensis, and additional hydroquinone-containing unsaturated hydrocarbon molecules from Pleiogynium timoriense and Tapirira obtuse. To date, such metabolites have been reported to cause weak-to-moderate non-selective activity against a range of human cancer cell lines. For those reasons, our team was intrigued when tests of plant extracts from the NCI Natural Products Repository collection uncovered a sample prepared from C. axillaris fruit that exhibited selective activity against a subset of pediatric cancer cell lines. This report details the structure determination and in vitro biological testing of new C. axillaris metabolites, choerosponols A-E (1-5), and their differential antiproliferative and cytotoxic effects against pediatric cancer cells.

One major limitation of natural products drug discovery efforts has been the identification of the mechanisms of action of compounds when mechanism-blind screens are employed. The differential sensitivity of pediatric cancer cell lines toward choerosponol A (1) prompted us to interrogate the genetic dependencies of the sensitive versus the resistant cell lines using the Project Achilles database hosted on the Cancer Dependency Map (DepMap) Portal of the Broad Institute.¹⁰ This analysis led to the identification of the MCT1 transporter as a candidate target for the new benzofuran-containing hydroquinone-based metabolite, choerosponol A (1), which was further supported through additional biological testing. Our work highlights the value of utilizing the DepMap databases to identify molecular targets of natural products based on the genetic vulnerabilities of diverse cancer cell lines.

RESULTS AND DISCUSSION

The residue from an organic extract of C. axillaris fruit was subjected to HP-20ss flash column chromatography using a step-gradient of MeOH in H₂O. The fractions that eluted with 90% and 100% MeOH exhibited selective antiproliferative effects against Ewing sarcoma (A-673) and medulloblastoma (D283) cell lines. Bioassay-guided fractionation afforded five new hydroquinone-based metabolites, choerosponols A-E (1-5) (Scheme 1), bearing unsaturated hydrocarbon chains.

Scheme 1.

Structures of metabolites derived from C. axillaris.

Compound 1 was isolated as a colorless, amorphous solid and assigned the molecular formula C₂₅H₃₈O₂, based on HRESIMS data revealing a [M + H]⁺ ion at m/z 371.2929 (calcd 371.2945). The ¹H NMR spectrum of 1 (Table 1) showed signals for an ABX spin system at δ_H 7.53 (d, J = 8.7 Hz, H-2), 7.18 (dd, J = 8.7, 2.5 Hz, H-3), and 7.41 (d, J = 2.5 Hz, H-5), together with an aromatic singlet proton at δ_H 6.49 (H-1′) that indicated the presence of a functionalized arene moiety. The ¹H–¹³C HSQC and HMBC correlation data associated with the aromatic proton and carbon signals (Figure 1) supported the presence of a benzofuran group [δ_C 150.0 (C-1), 111.9 (C-2), 113.3 (C-3), 155.4 (C-4), 106.7 (C-5), 131.2 (C-6), 103.0 (C-1′), and 161.1 (C-2′) (Table 2)]. In addition, a phenolic proton resonance was detected at δ_H 11.21 (s, OH-4), which showed ¹H–¹H ROESY correlations with the H-3 and H-5 protons (Figure S4, Supporting Information). Whereas many of the proton resonances associated with the unsaturated hydrocarbon chain produced a series of overlapping (δ_H 1.22–1.44) signals, the chemical shifts for several critical proton spins [δ_H 2.74 (t, J = 7.5 Hz, H-3′), 1.72 (p, J = 7.5 Hz, H-4′), 2.10 (m, H-13′), 2.09 (m, H-16′), and 0.88 (t, J = 6.9 Hz, H-19′)] were discernable enabling us to determine the position of the embedded cis olefin group [δ_H 5.51 (dd, J = 11.1, 6.3 Hz, H-14′) and 5.46 (dd, J = 11.1, 6.3 Hz, H-15′)].^11,12 Thus, the structure of 1 was established and the metabolite was given the trivial name choerosponol A in recognition of its biogenic source.

Table 1.

¹H NMR (500MHz) Spectroscopic Data for Compounds 1-5 in Pyridine-d₅, δ_H, Multiplicity (J in Hz)

position	1	2	4	position	3	5
1	-	-	-	1	-	-
2	7.53, d (8.7)	7.09, dd (8.5, 3.0)	6.27, dd (10.1, 1.7)	2	7.09, dd (8.5, 2.9)	6.27, dd (10.2, 1.7)
3	7.18, dd (8.7, 2.5)	7.19, d (8.5)	7.19, dd (10.1, 3.4)	3	7.19, d (8.5)	7.19, dd (10.2, 3.4)
4	-	-	4.93, m	4	-	4.93, m
5	7.41, d (2.5)	-	-	5	-	-
6	-	7.33, d (3.0)	3.33, d (15.6) 3.02, d (15.6)	6	7.33, d (2.9)	3.33, d (15.7) 3.02, d (15.7)
1′	6.49, s	3.25, dd (13.5, 4.5) 3.22, dd (13.5, 6.6)	2.44, dd (14.8, 10.7) 2.12, m	1′	3.24, dd (13.5, 4.5) 3.23, dd (13.5, 6.5)	2.44, dd (14.8, 10.7) 2.12, m
2′	-	4.40, m	4.52, m	2′	4.39, m	4.52, m
3′	2.74, t (7.5)	1.79, m	1.74, 1.59, m	3′	1.78, m	1.73, 1.58, m
4′	1.72, p (7.5)	1.75, 1.56, m	1.59, 1.45, m	4′	1.75, 1.55, m	1.58, 1.45, m
5′−11′	1.22–1.44, ma	1.21–1.42, ma	1.21–1.41, ma	5′−9′	1.20–1.40, ma	1.20–1.40, ma
12′	1.39, ma	1.39, ma	1.39, ma	10′	1.37, ma	1.36, ma
13′	2.10, m	2.10, m	2.10, m	11′	2.09, m	2.09, m
14′	5.51, dd (11.1, 6.3)	5.50, dd (11.1, 6.0)	5.50, dd (11.0, 6.0)	12′	5.48, dd (10.8, 5.8)	5.49, dd (11.0, 6.0)
15′	5.46, dd (11.1, 6.3)	5.47, dd (11.1, 6.3)	5.46, dd (11.0, 6.0)	13′	5.46, dd (10.8, 5.8)	5.46, dd (11.0, 6.0)
16’	2.09, m	2.09, m	2.09, m	14′	2.08, m	2.08, m
17′	1.33, ma	1.33, ma	1.33, ma	15′	1.32, ma	1.32, ma
18′	1.32, ma	1.32, ma	1.32, ma	16′	1.31, ma	1.32, ma
19′	0.88, t (6.9)	0.88, t (7.1)	0.88, t (7.1)	17’	0.87, t (7.0)	0.87, t (7.1)
OH-1	-	10.76, s	-	OH-1	10.76, br s	-
OH-4	11.21, s	10.44, br s	7.50, d (5.7)	OH-4	10.44, br s	7.50, d (5.7)
OH-5	-	-	6.85, s	OH-5	-	6.85, s
OH-2′	-	6.98, br s	7.04, d (4.1)	OH-2′	6.98, br s	7.04, d (4.1)

^aOverlapping signals; chemical shifts were detemined from ¹H–¹³C HSQC and HMBC correlations.

Figure 1.

Key ¹H–¹H COSY and ¹H–¹³C HMBC correlations used for determining the structures of compounds 1, 2, and 4.

Table 2.

¹³C NMR (125MHz) Spectroscopic Data for Compounds 1–5 in Pyridine-d₅, δ_C, C-Type

position	1	2	4	position	3	5
1	150.0, C	152.2, C	199.3, C	1	152.2, C	199.3, C
2	111.9, CH	115.3, CH	129.8, CH	2	115.3, CH	129.7, CH
3	113.3, CH	117.8, CH	151.6, CH	3	117.8, CH	151.6, CH
4	155.4, C	150.3, Cb	73.8, CH	4	150.4, Cb	73.8, CH
5	106.7, CH	128.9, C	78.0, C	5	128.9, C	77.9, C
6	131.2, C	120.0, CH	49.9, CH2	6	120.0, CH	49.9, CH2
1′	103.0, CH	41.0, CH2	41.4, CH2	1′	41.0, CH2	41.4, CH2
2′	161.1, C	72.8, CH	68.4, CH	2′	72.8, CH	68.4, CH
3′	29.2, CH2	38.4, CH2	40.1, CH2	3′	38.4, CH2	40.1, CH2
4′	28.5, CH2	26.9, CH2	26.4, CH2	4′	26.9, CH2	26.4, CH2
5′−10′, 12′	29.9 (C-5′), 30.5, 30.38, 30.36, 30.30, 30.27, 30.1, CH2a	30.6, 30.53, 30.52, 30.41×2, 30.40, 30.3, CH2c	30.54, 30.49, 30.43, 30.41, 30.39, 30.37, 30.3, CH2d	5′−8′, 10′	30.6, 30.52, 30.50, 30.4, 30.3, CH2e	30.51, 30.47, 30.4, 30.32, 30.27, CH2f
				9′	30.0, CH2	30.0, CH2
				11′	28.0, CH2	28.0, CH2
11′	30.0, CH2	30.0, CH2	30.0, CH2	12′	130.7, CH	130.7, CH
13′	28.0, CH2	28.0, CH2	28.0, CH2	13′	130.6, CH	130.6, CH
14′	130.7, CH	130.7, CH	130.7, CH	14′	27.7, CH2	27.7, CH2
15′	130.6, CH	130.6, CH	130.6, CH	15′	32.6, CH2	32.6, CH2
16′	27.7, CH2	27.7, CH2	27.7, CH2	16′	23.0, CH2	23.0, CH2
17′	32.7, CH2	32.6, CH2	32.7, CH2	17′	14.6, CH3	14.6, CH3
18′	23.0, CH2	23.0, CH2	23.0, CH2
19′	14.6, CH3	14.6, CH3	14.6, CH3

^a,c-fInterchangeable.

^bData extracted from ¹H–¹³C HMBC NMR spectra due to signal overlapping with solvent.

Choerosponol B (2) was obtained as a colorless, amorphous solid, and its molecular formula was established as C₂₅H₄₂O₃ based on HRESIMS data, which provided a [M − H]⁻ ion at m/z 389.3072 (calcd 389.3061). The ¹H and ¹³C NMR data for 2 (Tables 1 and 2) were comparable to those obtained for 1; however, signals for an additional oxygenated methine [δ_C 72.8 (C-2′)] and new methylene group were observed [41.0 (C-1′)] in place of the Δ^1′,2′ olefin. Moreover, the ¹H NMR data for 2 contained a total of three hydroxy group spins [δ_H 10.76 (OH-1), 10.44 (OH-4), and 6.98 (OH-2′)], whereas 1 possessed only a single hydroxy group. Taking these differences into consideration, along with data from the ¹H–¹H COSY and ¹H–¹³C HMBC correlation experiments (Figure 1), it was determined that compound 2 lacked a furan ring and instead it contained a C-2′ alcohol set in the unsaturated hydrocarbon chain, which was attached to a monosubstituted hydroquinone. Thus, the structure of metabolite 2 was determined as shown. Compound 2 bears many structural features similar to (R)-2′-hydroxylanneaquinol;⁷ however, the new compound contains two additional methylenes and the position of the double bond is shifted closer to the terminal methyl (C-19′) position.

Choerosponol C (3) was obtained as a colorless, amorphous solid. The HRESIMS data indicated that compound 3 had the molecular formula C₂₃H₃₈O₃ based on the presence of a [M − H]⁻ ion at m/z 361.2760 (calcd 361.2748). While the ¹H and ¹³C NMR data for 3 were virtually superimposable upon those of 2, it was determined that metabolite 3 lacked signals attributable to two methylenes occurring in the unsaturated hydrocarbon chain (Tables 1 and 2). Thus, compound 3 was identified as a truncated analogue of compound 2.

We were initially surprised by the lack of measurable specific rotation values for compounds 2 and 3 leading us to consider the notion that both metabolites might be racemic or scalemic mixtures. However, chiral HPLC analysis employing two sorbents (Cellulose 2 and Cellulose 3) under normal (n-hexane/isopropanol) and reversed phase (both methanol and acetonitrile/water) conditions were consistent with both samples being composed of single enantiomers. This was in general agreement with the prior observation that (R)-2′-hydroxylanneaquinol generated a small specific rotation value ([α]_D +0.8).⁷ To determine the absolute configuration of the both metabolites, Mosher-ester-derivatization experiments were performed affording the tri-MTPA esters of 2 and 3. The ¹H NMR data of the (R)- and (S)-MTPA esters (Figure 2 and Figures S12 and S20, Supporting Information) gave further evidence that these compounds existed as pure enantiomers since the products each generated single sets of ¹H NMR signals. The resulting data supported 2′S configurations for both compounds (Figure 2). However, concern arose that the presence of multiple, bulky chiral MTPA esters might interfere with the conformations and, therefore, the shielding and deshielding properties of the nearby ester groups. To test this concern, the C-1 and C-4 hydroxy groups of 2 were protected by acetylation to yield 2a, followed by Mosher monoester formation (Figure 2), which again supported a 2′S absolute configuration. Based on those observations and considering the biogenic origins of the compounds, 2′S configurations were proposed for both for 2 and 3.

Figure 2.

Δδ (δ_S − δ_R) values (ppm) for (S)- and (R)-MTPA esters of compounds 2, 2a, 4, and 4a.

Compound 4 was obtained as a colorless, amorphous solid with the molecular formula C₂₅H₄₄O₄ based on the occurrence of a [M − H]⁻ ion at m/z 407.3181 (calcd 407.3167) in the HRESIMS data. The ¹H NMR data for 4 (Table 1) revealed that the unsaturated hydrocarbon system in the new metabolite was virtually identical to that in 2; however, significant differences were found in the carbon and proton spins belonging to what had been the monosubstituted hydroquinone moiety. Specifically, the ABX spin system in 2 had been replaced by signals attributable to an endocyclic double bond [δ_H 6.27 (dd, J = 10.1, 1.7 Hz)/δ_C 129.8 (C-2) and 7.19 (dd, J = 10.1, 3.4 Hz)/δ_C 151.6 (C-3)], an unsaturated ketone [δ_C 199.3 (C-1)], an oxygenated methine [δ_H 4.93 (m)/δ_C 73.8 (C-4)], a methylene [δ_H 3.33 and 3.02 (d, J = 15.6 Hz)/δ_C 49.9 (C-6)], and a downfield tetrasubstituted carbon [δ_C 78.0 (C-5)]. Two hydroxy protons were observed in the ¹H NMR spectrum of 4 appearing at δ_H 7.50 (d, J = 5.7 Hz) and 6.85 (s). Taking these changes into consideration, an analysis of the ¹H–¹H COSY and ¹H–¹³C HMBC correlation data (Figure 1) led to the determination that compound 4 contained a 4,5-dihydroxy-2-cyclohexen-1-one moiety. This new metabolite was given the trivial name choerosponol D (4).

Choerosponol E (5) was assigned the molecular formula C₂₃H₄₀O_4, based on data from a HRESIMS experiment that revealed a [M − H]⁻ ion at m/z 379.2864 (calcd 379.2854). The ¹H and ¹³C NMR data for 5 exhibited many similarities to those obtained for 4; however, the new compound was missing signals corresponding to two methylene carbons occurring in the unsaturated hydrocarbon chain (Tables 1 and 2). This was corroborated by the data derived from the HRESIMS experiment confirming the difference between 4 and 5 was the loss of -CH₂CH₂-. Thus, compound 5 was proposed to be a truncated analogue of metabolite 4.

Compounds 4 and 5 exhibited specific rotation values of [α]²⁵_D −38 (c 0.1, MeOH) and −28 (c 0.1, MeOH), respectively. The relative and absolute configurations of the metabolites were subsequently investigated using a combination of ¹H–¹H ROESY experiments, Mosher’s ester analyses, and ECD spectroscopy. The ¹H–¹H ROESY spectra of 4 and 5 showed correlations between H₂-1′ ↔ H-4 and H₂-6 , which indicated the cylcohexenone system bore a 4R*,5S* configuration (Figure S37, Supporting Information).¹³ In parallel to the ROESY analyses, Mosher esters were prepared for the monoacetate of 4 (4a), as well as 4 and 5, which secured 2′S configurations for the compounds (Figure 2 and Figures S27, S29, and S36, Supporting Information). Next, ECD spectroscopy was employed to probe the absolute configurations of the metabolites. To limit the flexibility of the compounds and simplify the conformational analysis, the unsaturated hydrocarbon chains of the molecules were truncated at the C-5′ positions. This decision was made since the extended “tail” regions of the metabolites were reasonably anticipated to offer negligible effects on the resulting experimental or calculated ECD spectra.¹⁴ The low-energy conformers were obtained through a conformation search with Compute VOA and optimized at the DFT/B3LYP/6–31G** level in the gas phase using Gaussian 09.¹⁵ The ECD curves for each stable conformer were calculated at TDDFT/B3LYP/6–31+G** (CPCM, solvent=methanol) (Figure S37, Supporting Information) and averaged using SpecDis 1.71 to generate a Boltzmann-weighted ECD spectrum.¹⁶ The experimental ECD spectra of 4 and 5 showed strong matches with the calculated ECD spectra of the truncated model compound bearing a 4R,5S,2’S configuration (Figure 3). Thus, the absolute configurations of the metabolites were proposed as illustrated for compounds 4 and 5.

Figure 3.

Comparison of the experimental and calculated ECD curves for compounds 4, 4b, and 5.

Metabolites 1-5 were individually tested for their antiproliferative and cytotoxic effects in a panel of four types of solid pediatric cancer cells: Ewing sarcoma (A-673), rhabdomyosarcoma (SJCRH30), type 3/4 medulloblastoma (D283), and hepatoblastoma (Hep293TT). The Ewing sarcoma and medulloblastoma cells were remarkably sensitive to the growth inhibitory effects of compound 1 with GI₅₀ values of 0.19 and 0.07 μM, respectively, as compared to the rhabdomyosarcoma and hepatoblastoma cells (GI₅₀ >10 μM) (Table 3 and Figure 4). Despite its potent growth inhibitory effects, 1 did not cause robust cytotoxicity in any of the cell lines at concentrations up to 10 μM. Very different concentration response profiles were observed with 2-5 (Figure 4). The SJCRH30 rhabdomyosarcoma cells were the most sensitive to 2, 4 and 5 whereas the Hep293TT hepatoblastoma cells were resistant to 2 and 3 as compared to the other pediatric cancer cells. Each of these compounds was able to inhibit proliferation and cause significant toxicity among the cancer cell lines (Figure 4 and Tables S1 and S2, Supporting Information). The concentration response curves and differential sensitivities of the cells to these metabolites strongly suggested that the mechanism of action of 1 differed from that initiated by 2-5 and that the presence of an oxidized headgroup containing a furan system was important for conveying selective activity against Ewing sarcoma and group 3/4 medulloblastoma cells. These observations were in general consensus with prior bioassay results showing that variations in the cyclic headgroups of structurally similar natural products derived from Anacardiaceous species were dependent on the type of functionalization present within the arene/cyclohex-2-en-1-one moieties.^4–9

Table 3.

GI₅₀ Values (μM) of Compounds 1-5 in Pediatric Cancer Lines^a

	cell line
compound	A-673	SJCRH30	D283	Hep293TT
1	0.19 ± 0.07	> 10	0.07 ± 0.02	> 10
2	0.32 ± 0.07	0.12 ± 0.02	0.24 ± 0.03	0.70 ± 0.06
3	0.16 ± 0.03	0.17 ± 0.03	0.08 ± 0.01	1.02 ± 0.06
4	0.58 ± 0.15	0.18 ± 0.01	0.40 ± 0.03	0.39 ± 0.02
5	0.65 ± 0.12	0.24 ± 0.02	0.41 ± 0.06	0.50 ± 0.06

^aConcentration (μM) that caused 50% growth inhibition (GI₅₀) for each pediatric cancer cell line as determined by the SRB assay. n = 3; mean ± SEM.

Figure 4.

Concentration-response data from the SRB assay for compounds 1-5 against the pediatric cancer cell lines: A-673, SJCRH30, D283, and Hep293TT. n = 3; mean ± SEM.

To identify the mechanism responsible for the selective growth inhibitory activity of 1 against the A-673 Ewing sarcoma and D283 medulloblastoma cells, we expanded our cell line panel to include a second Ewing sarcoma line (SK-ES-1), two medulloblastoma lines that are classified as belonging to the group 3 (D341) or sonic hedgehog (DAOY) subgroups,¹⁷ and a triple-negative breast cancer cell line (HCC1806). The SK-ES-1 and D341 cell lines were sensitive to 1 with GI₅₀ values of 0.12 and 0.02 μM, respectively, whereas the HCC1806 and DAOY cell lines were not sensitive to 1 up to a concentration of 10 μM (Figure 5a). The effects of 1 were consistent among both Ewing sarcoma cell lines and the two group 3/4 medulloblastoma cell lines; however, the sonic hedgehog medulloblastoma cells were insensitive, thus, providing a potential opportunity to identify the shared vulnerabilities of the sensitive cells. The DepMap Project Achilles database¹⁰ was interrogated for shared genetic dependencies among these sensitive cell lines as compared to the resistant cell lines. Dependency values are expressed as log₂ transformed values where a more negative number denotes a greater dependency on a gene product. The Achilles gene effect values for the four sensitive lines and three resistant cell lines (the Hep293TT cell line was not included in the Project Achilles database) were each averaged and rank-sorted according to their respective differences from the averaged values (Figure 38, Supporting Information). The top differentially dependent gene identified by this methodology was SLC16A1, which encodes the MCT1 transporter. The SLC16A1 dependency scores for the sensitive cell lines ranged from −0.87 to −0.55, demonstrating a high degree of dependency, whereas the resistant cell lines were only slightly negative or positive (−0.08 to 0.21) (Figure 5b).

Figure 5.

Differential sensitivity and dependency mapping identify SLC16A1/MCT1 as a functional target of 1. A. Concentration response curves of 1 in a panel of seven pediatric cancer cell lines and a breast cancer cell line. B. Dependency scores for SLC16A1/MCT1 in seven cancer cell lines from the Project Achilles, Cancer Dependency Map (DepMap). C. Effect of 1 on the cytotoxic activity of the MCT1 substrate BrPA. D. MCT1 expression in cell lines senstive and resistent to 1.

To validate the MCT1 product of SLC16A1 as a functional target of the selective activity of 1, the toxic MCT1 substrate, bromopyruvic acid (BrPA), was utilized since inhibition of MCT1 transport protects against BrPA-induced toxicity.¹⁸ Indeed, we found that 1 inhibited the cytotoxic effects of BrPA on the A673 cell line, consistent with the effects of the known MCT1 inhibitor AZD3965 (Figure 5c). Additionally, the four cell lines sensitive to 1 were found to have elevated expression of the MCT1 protein as compared to the resistant cell lines (Figure 5d). Interestingly, the MCT1 inhibitor AZD3965 had no effects on cell growth of A673 cells at concentrations up to 10 μM, consistent with its lack of efficacy in cells that additionally express MCT4.¹⁹ Together, these data suggest that the selective activity of 1 is a result of cell line dependency on MCT1 and the ability of 1 to inhibit MCT1 and potentially other members of the MCT family. Our results are consistent with previous reports that have identified miR-124-dependent suppression of SLC16A1/MCT1 as a novel therapeutic target for medulloblastoma.²⁰ Additionally, this may also underlie the demonstrated role of miR-124 suppression in Ewing sarcoma,²¹ which would be expected to lead to increased MCT1 levels and dependency in these tumors. Our data demonstrated that compound 1 has selective antiproliferative effects in Ewing sarcoma and group 3/4 medulloblastoma cell lines that express MCT1. Furthermore, we show that 1 inhibited the transport activity of MCT1 and likely other MCT family members, which are validated anticancer targets.¹⁸

In conclusion, evaluating the differential cytotoxic effects of natural products against cancer cell lines combined with interrogation of the different genetic dependencies exhibited by the sensitive and resistant cell lines using the publicly available DepMap Project Achilles database is a powerful tool for identifying potential new targets of selectively bioactive compounds. This study highlights the impact that a simple, yet well-integrated approach employing chemistry, pharmacology, and bioinformatics has as a transformational paradigm for evolving conventional bioactive-compound-discovery research into a powerful weapon in the fight against cancers and other diseases. Most importantly, this approach is readily and immediately accessible to the natural products community as it involves little-to-no investment in new equipment or complex technologies, but instead it is wholly reliant on pragmatic assay design aimed at uncovering the differential sensitivities of cells to bioactive molecules.

EXPERIMENTAL SECTION

General Experimental Procedures.

Optical rotations were measured on a Rudolph Research Autopol III automatic polarimeter. ECD measurements were performed using a JASCO J-715 spectropolarimeter. 1D and 2D NMR data were recorded on Varian Unity Inova 500 MHz FT-NMR instruments, and the data were processed using MestReNova 10.0 software (Mestrelab Research SL, Santiago de Compostela, Spain). Mass spectrometry experiments were performed on an Agilent 6538 HRESI QTOF MS system coupled to an Agilent 1290 HPLC system. HP-20ss (Sorbent Technologies, Norcross, GA, USA) and Sephadex LH-20 (GE Healthcare Bio-Science AB, Uppsala, Sweden) were used for open-column chromatography. Thin-layer chromatographic (TLC) analyses were conducted using Kieselgel 60 F254 plates (silica gel, 0.25 mm layer thickness, Merck, Germany) with visualization under UV light (254 and 365 nm). HPLC separations were performed on a Shimadzu system using a SCL-10A VP pump and system controller with a Phenomenex Gemini 5 μm C₁₈ column (110 Å, 250 × 21.2 mm, 10 mL/min) for preparative HPLC, and on a Waters system using a Waters 1525 binary pump and Waters 2998 photodiode array detector with a Phenomenex Gemini 5 μm C₁₈ column (110 Å, 250 × 10.0 mm, 4 mL/min) for semi-preparative HPLC. All solvents were of ACS grade or better.

Plant Material.

Fruit of the plant Choerospondias axillaris were collected in Vietnam by Dr. Djaja D. Soejarto (University of Illinois at Chicago) under a contract with the Natural Products Branch for the National Cancer Institute on April 30, 2008. A voucher specimen for this collection [Voucher #0X0H1222] is housed at the Smithsonian Institution, Washington DC. An organic extract of this collection, N137673 was produced as reported previously²² and provided by NCI Natural Products Branch, (Frederick, MD, USA).

Extraction and Bioassay-Guided Isolation.

The extract (N137673, 5.0 g) was subjected to HP-20ss flash column chromatography by elution with a step gradient of MeOH-H₂O (3:7, 1:1, 7:3, 9:1, and 1:0), to afford five fractions (F1–F5). Fraction F4, eluted with MeOH-H₂O (9:1) was fractionated by preparative HPLC with gradient mixtures of MeOH-H₂O (9:1 → 1:0) as the eluent to give 10 subfractions (F4.01-F4.10). Subfraction F4.04 (18 mg) was further processed by semi-preparative HPLC with CH₃CN-H₂O containing 0.1% HCOOH (4:1), to yield 5 (5 mg, t_R 14.7 min) and 3 (3.2 mg, t_R 18.7 min). Compounds 4 (12 mg) and 2 (10 mg) were purified from subfraction F4.07 (85 mg) by semi-preparative HPLC using CH₃CN-H₂O containing 0.1% HCOOH (17:3, t_R 4: 19.0, 2: 22.5 min). Fraction F5, which eluted with 100% MeOH, was further separated over Sephadex LH-20 with 100% MeOH yielding three subfractions (F5.1-F5.3). Subfraction F5.2 (180 mg) was subjected to semi-preparative HPLC (CH₃CN-H₂O containing 0.1% HCOOH, 9:1 → 1:0) to give 1 (10 mg, t_R 21.2 min). Bioassays were performed on all fractions resulting from each chromatographic step to ensure that the bioactive molecules were tracked and obtained.

Choerosponol A (1):

Colorless, amorphous solid; ¹H and ¹³C NMR data, Tables 1 and 2; HRESIMS m/z 371.2929 [M + H]⁺ (calcd for C₂₅H₃₉O₂, 371.2945).

Choerosponol B (2):

Colorless, amorphous solid; [α]²²_D ±0 (c 0.5, MeOH or CH₂Cl₂); ¹H and ¹³C NMR data, Tables 1 and 2; HRESIMS m/z 389.3072 [M − H]⁻ (calcd for C₂₅H₄₁O₃ 389.3061).

Choerosponol C (3):

Colorless, amorphous solid; [α]²²_D ±0 (c 0.2, MeOH or CH₂Cl₂); ¹H and ¹³C NMR data, Tables 1 and 2; HRESIMS m/z 361.2760 [M − H]⁻ (calcd for C₂₃H₃₇O₃, 361.2748).

Choerosponol D (4):

Colorless, amorphous solid; [α]²⁵_D −38 (c 0.1, MeOH); ECD (MeOH) λ_max (Δε) 208 (+3.01), 237 (−3.40) nm; ¹H and ¹³C NMR data, Tables 1 and 2; HRESIMS m/z 407.3181 [M − H]⁻ (calcd for C₂₅H₄₃O₄, 407.3167).

Choerosponol E (5):

Colorless, amorphous solid; [α]²⁵_D −28 (c 0.1, MeOH); ECD (MeOH) λ_max (Δε) 207 (+2.71), 237 (−2.96) nm; ¹H and ¹³C NMR data, Tables 1 and 2; HRESIMS m/z 379.2864 [M − H]⁻ (calcd for C₂₃H₃₉O₄, 379.2854).

Acetylation and Mosher’s Ester Analysis of 2.

A 10 μL aliquot of acetic anhydride was added to a solution of 2 (1.0 mg) in pyridine (500 μL). The mixture was stirred for 1 h at room temperature and dried under a stream of N₂. The reaction mixture was purified by semi-preparative HPLC using a Phenomenex Gemini 5 μm C₁₈ column and eluted with CH₃CN-H₂O containing 0.1% HCOOH (19:1) to afford the diacetate (2a) (0.7 mg, t_R 16.1 min) together with the triacetate (0.3 mg, t_R 22.2 min). For the modified Mosher ester reaction, 0.6 mg of 2a was dissolved in pyridine-d₅ and split equally among two oven-dried NMR tubes together with 2 μL of (R)- and (S)-MTPA-Cl and catalytic amounts of DMAP to afford the (S)- and (R)-MTPA-ester products, respectively. The sealed NMR tubes were incubated at room temperature and the reactions were monitored by ¹H NMR analysis. ¹H–¹H COSY experiments were used to assign the ¹H NMR signals. The molecular formulas of the MTPA esters were confirmed by ESIMS (both products presented [M + Na]⁺ ions m/z at 713).

1,4-O-Diacetyl-choerosponol B (2a):

¹H NMR (500 MHz, pyridine-d₅) δ 7.53 (1H, d, J = 2.8 Hz, H-6), 7.31 (1H, d, J = 8.7 Hz, H-3), 7.20 (1H, dd, J = 8.7, 2.8 Hz, H-2), 6.25 (1H, br s, OH-2′), 5.50 (1H, dd, J = 11.3, 6.0 Hz, H-14′), 5.46 (1H, dd, J = 11.3, 6.5 Hz, H-15′), 4.19 (1H, m, H-2′), 3.01 (1H, dd, J = 13.8, 7.3 Hz, H-1′a), 2.97 (1H, dd, J = 13.8, 5.8 Hz, H-1′b), 2.28 (3H, s, Ac), 2.18 (3H, s, Ac), 2.10 (1H, m, H-13′), 2.09 (1H, m, H-16′), 1.75 (2H, m, H₂-3′), 1.72, 1.56 (each 1H, m, H₂-4′), 1.20–1.42 (20H, m, H₂-5′, H₂-6′, H₂-7′, H₂-8′, H₂-9′, H₂-10′, H₂-11′, H₂-12′, H₂-17′, and H₂-18′), 0.88 (3H, t, J = 6.9 Hz, H₃-19′); ESIMS m/z 497 [M + Na]⁺, 473 [M − H]⁻, 519 [M + HCOOH − H]⁻.

Acetylation and Mosher Ester Analysis of 4.

Compound 4 (1.5 mg) was dissolved in pyridine (500 μL) and reacted with acetic anhydride (10 μL) for 12 h at room temperature. The reaction mixture was dried under a stream of N₂ and purified by semi-preparative HPLC using a Phenomenex Gemini 5 μm C₁₈ column and eluted with CH₃CN-H₂O containing 0.1% HCCOH (19:1) to obtain the monoacetate (4a) (1.0 mg, t_R 13.1 min) together with the diacetate (0.6 mg, t_R 14.6 min). In a similar manner to that described for 2a (vide supra), compound 4a (0.6 mg) was treated with (R)- and (S)-MTPA-Cl, to give the (S)- and (R)-MTPA-esters, respectively, which were detected by ESIMS at m/z 689 ([M + Na]⁺) and 711 ([M + HCOOH − H]⁻).

4-O-Acetyl-choerosponol D (4a):

¹H NMR (500 MHz, pyridine-d₅) δ 7.08 (1H, s, OH-5), 6.98 (1H, dd, J = 10.3, 3.5 Hz, H-3), 6.80 (1H, br d, J = 3.5 Hz, OH-2′), 6.30 (1H, d, J = 10.3 Hz, H-2), 6.06 (1H, d, J = 3.5 Hz, H-4), 5.50 (1H, dd, J = 11.1, 6.0 Hz, H-14′), 5.46 (1H, dd, J = 11.1, 6.3 Hz, H-15′), 4.41 (1H, m, H-2′), 2.26 (1H, dd, J = 14.6, 10.7 Hz, H-6a), 2.14 (3H, s, Ac), 2.02 (1H, dd, J = 14.6, 2.0 Hz, H-6b), 2.10 (1H, m, H-13′), 2.09 (1H, m, H-16′), 1.72, 1.59 (each 1H, m, H₂-3′), 1.60, 1.45 (each 1H, m, H₂-4′), 1.22–1.41 (20H, m, H₂-5′, H₂-6′, H₂-7′, H₂-8′, H₂-9′, H₂-10′, H₂-11′, H₂-12′, H₂-17′, and H₂-18′), 0.88 (3H, t, J = 6.7 Hz, H₃-19′); ESIMS m/z 451 [M + H]⁺, 473 [M + Na]⁺, 449 [M − H]⁻, 495 [M + HCOOH − H]⁻.

Computational Details.

Conformational analyses were performed using Compute VOA. Geometry and ECD calculations were carried out with Gaussian 09 at the DFT level.¹⁵ SpecDis 1.71 was utilized to obtain Boltzmann-averaged ECD spectra.¹⁶

Pediatric Cancer Cell Lines.

The Ewing sarcoma (A-673, SK-ES-1), rhabdomyosarcoma (SJCRH30), medulloblastoma (D283, D341, DAOY), and breast cancer (HCC1806) cell lines were purchased from the American Type Culture Collection, (Manassas, VA, USA). The hepatoblastoma cell line (Hep293TT) was obtained directly from Dr. Gail Tomlinson.²³ A-673, SK-ES-1, SJCRH30, HCC1806, and Hep293TT cells were grown in RPMI-1640 medium (Corning, Corning, NY, USA) with 5% fetal bovine serum (Corning) and 50 μg/mL gentamicin (Gibco/Thermo Fisher Scientific, Waltham, MA, USA). The D283, D341, and DAOY cells were grown in IMEM (Gibco) with 5% fetal bovine serum and 25 μg/mL gentamicin. Cells were grown at 37 °C with 5% CO₂ in humidified incubators. The cells were used within four months of retrieval from liquid nitrogen. The A-673, SJCRH30, and D283 cells lines were validated by STR profiling (Genetica DNA Laboratories, Burlington, NC, USA).

Sulforhodamine B Assay.

The antiproliferative and cytotoxic activities of the compounds were evaluated using the sulforhodamine B (SRB) assay^24,25 as previously described.²⁶ Cells were plated at densities of 4,000–6,000 cells per well in 96-well tissue culture plates and allowed to adhere and grow overnight. The compounds were dissolved in DMSO and added to triplicate wells with the final volume of DMSO held at 0.5% (v/v). Cells were treated for 48 h, the medium removed, and the cells fixed with 10% TCA. Due to the low adherence of D283 cells, 180 μL aliquots of 20% TCA were added directly to the tissue culture medium in each well. At the time of compound addition, time 0 plates were fixed. The SRB absorbance was measured at 560 nM and GI₅₀, TGI, and LC₅₀ values calculated for each experiment using the non-linear regression function in Prism 6.01 (GraphPad Software, La Jolla, CA, USA). For BrPA inhibition experiments, 1 and AZD3965 were applied one hour prior to the addition of BrPA with DMSO used as the vehicle control.

DepMap Project Achilles Database Analyses.

The genetic dependencies of the cell lines evaluated for sensitivity to 1 were obtained from the Broad Institute’s DepMap portal at https://depmap.org/portal/download/. Specifically, we accessed the CRISPR (Avana) public 19Q4 dataset (Achilles_gene_effect.csv), which contains data on the relative dependency of 18,333 genes for the survival each of 689 cell lines, including seven of the eight cell lines used in our study (with the exception of the Hep293TT cell line). Dependency data for the four sensitive cell lines (A-673, D283, SK-ES-1, and D341) and the three available resistant cell lines (DAOY, HCC1806, and SJCRH30) were identified by cross referencing the ACH Depmap ID for each cell line and these data were copied into a new Excel worksheet. The average gene dependencies among the four sensitive and three resistant cell lines were calculated as separate measures and sorted by the differential between sensitive and resistant populations. SLC16A1 was identified as the gene with the highest degree of differential dependence between the cell lines that were sensitive and resistant to 1, suggesting that the inhibition of the SLC16A1 gene product, MCT1, was a putative target for 1.

Immunoblotting.

Cells were lysed with RIPA buffer (50 mM Tris [pH 7.4], 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS) supplemented with protease inhibitors (Sigma Aldrich, Cat. P2714), PMSF (Sigma Aldrich), and 2-mercaptoethanol at 20%. For each sample, twenty micrograms of protein were resolved by PAGE and evaluated by immunoblotting. The antibodies against MCT1 were purchased from Abcam (ab85021, used in Figure 5d) and Invitrogen (MA5–18288, used to confirm staining pattern observed with ab85021) and used at 1:1,000. GAPDH (D4C6R) was from Cell Signaling Technology (Danvers, MA 01923) and used at 1:1,000. IRDye 680 or 800 goat anti-rabbit secondary antibodies were used (LI-COR Biosciences) and immunoblots imaged on an Odyssey FC (LI-COR Biosciences).