Antiviral Compounds from Codiaeum peltatum

Article pubs.acs.org/jnp

Cite This: J. Nat. Prod. XXXX, XXX, XXX−XXX

Antiviral Compounds from Codiaeum peltatum Targeted by a Multiinformative Molecular Networks Approach Florent Olivon,† Simon Remy,† Gwendal Grelier,† Ceć ile Apel,† Ceć ilia Eydoux,‡ Jean-Claude Guillemot,‡ Johan Neyts,§ Leen Delang,§ David Touboul,† Fanny Roussi,† and Marc Litaudon*,† †

Institut de Chimie des Substances Naturelles, CNRS-ICSN, UPR 2301, Université Paris-Saclay, 91198, Gif-sur-Yvette, France Aix Marseille University, CNRS, AFMB, AD2P, 163 Avenue de Luminy, 13288 Marseille Cedex 09, France § Laboratory for Virology and Experimental Chemotherapy, Rega Institute for Medical Research, KU Leuven, 3000 Leuven, Belgium J. Nat. Prod. Downloaded from pubs.acs.org by IOWA STATE UNIV on 01/25/19. For personal use only.

‡

S Supporting Information *

ABSTRACT: From a set of 292 Euphorbiaceae extracts, the use of a molecular networking (MN)-based prioritization approach highlighted three clusters (MN1−3) depicting ions from the bark extract of Codiaeum peltatum. Based on their putative antiviral potential and structural novelty, the MS-guided purification of compounds present in MN1 and MN2 afforded two new daphnane-type diterpenoid orthoesters (DDO), codiapeltines A (1) and B (2), the new actephilols B (3) and C (4), and four known 1,4-dioxane-fused phenanthrene dimers (5−8). The structures of the new compounds were elucidated by NMR spectroscopic data analysis, and the absolute configurations of compounds 1 and 2 were deduced by comparison of experimental and calculated ECD spectra. Codiapeltine B (2) is the first daphnane bearing a 9,11,13-orthoester moiety, establishing a new major structural class of DDO. Compounds 1−8 and four recently reported monoterpenyl quinolones (9− 12) detected in MN3 were investigated for their selective activities against chikungunya virus replication and their antipolymerase activities against the NS5 proteins of dengue and zika viruses. Compounds 3−8 exhibited strong inhibitory activities on both dengue and zika NS5 in primary assays, but extensive biological analyses indicated that only actephilol B (3) displayed a specific interaction with the NS5 targets.

R

symptoms renders their differential diagnosis difficult.11,12 In order to identify novel CHIKV inhibitors, several Euphorbiaceae species have been investigated in the past few years using bioassay-guided purification procedures. This has resulted in the isolation of tigliane, jatrophane, flexibilane, daphnane, and myrsinane diterpenoids showing micro- to nanomolar range activities.13−19 However, bioassay-guided fractionation has several limitations. As a noteworthy example, its workflow, which relies on bioactivity concerns only, does not take full advantage of the inherent structural diversity embedded in crude extract collections. Thus, this approach frequently leads to the reisolation of known bioactive compounds or analogues

ecent outbreaks of chikungunya (CHIKV), dengue (DENV), and zika (ZIKV) massive epidemics highlighted the worldwide public health threat that emerging and reemerging viruses represent.1 These viruses, which belong to the Alphavirus (CHIKV) and Flavivirus (ZIKV and DENV) genera, are responsible for severe infectious diseases associated with fever, rashes, and headache.2,3 ZIKV is also associated with neurological disorders such as Guillain-Barre syndrome and other congenital syndromes in newborns such as microcephaly.4−6 In Flaviviruses, the nonstructural protein 5 (NS5) polymerase plays an essential role in viral replication.7−9 Since DENV and ZIKV NS5 proteins harbor highly similar RdRp domains, these enzymes are considered as targets of particular interest in the prospect of discovering dual inhibitors.10 The discovery of such dual inhibitors could be particularly relevant given that the similarity of dengue and zika © XXXX American Chemical Society and American Society of Pharmacognosy

Received: September 20, 2018

A

DOI: 10.1021/acs.jnatprod.8b00800 J. Nat. Prod. XXXX, XXX, XXX−XXX

Journal of Natural Products

Article

Figure 1. Extract and compound prioritization strategy using multi-informative molecular networking: (A) Generation of a molecular network from LC-MS2 analyses of 292 EtOAc extracts of New Caledonian Euphorbiaceae. (B) Detection of bioactive clusters by mapping CHIKV, DENV, and ZIKV activities over networks. (C) Selected MN1, MN2, and MN3 with a taxonomical mapping consisting in overlaying the 20 Euphorbiaceae genera represented in the extract collection to detect genus- or extract-specific clusters. Node sizes are proportional to MS1 peak area values.

dissimilarity within a taxonomically homogeneous set of samples could imply chemical uniqueness, the generation of these multi-informative maps unifying structural data, taxonomical information, and bioassay results allows bioactivities to be associated with taxon-specific scaffolds. From a network made of approximately 1200 clusters and constructed from 292 LC-MS2 analyses of EtOAc Euphorbiaceae extracts, this approach highlighted the original chemical composition of Codiaeum peltatum, a species endemic from New Caledonia. A

and slows down the natural products (NPs) drug discovery process. Recently, we developed a molecular networking (MN)based strategy consisting in deciphering the relationship between spectral networks and biological activities and further exploiting it to prioritize the isolation of bioactive NPs.20,21 The core concept of this approach is based on the cross-linking of various information layers within a massive molecular network to spot bioactive scaffolds. Assuming that spectral B



Article

Chart 1

percentage of inhibition at 50 μg·mL−1. For each target, the bioactivity results were first split up according to their activity levels. The division into various bioactivity levels is much more informative than a simple active/nonactive categorization, as has been shown in our previous study.20 The bioactive clusters were highlighted by mapping these bioactivity degrees over networks using a color gradient applied to all nodes (Figure 1B). CHIKV EC50 values were layered into five activity ranges (0 to 5; 5 to 10; 10 to 20; 20 to 100; and >100 μg·mL−1), DENV NS5 percentages of inhibition were layered into five groups (0 to 40; 40 to 50; 50 to 60; 60 to 80; and >80%), whereas ZIKV NS5 percentages of inhibition were sliced into five ranges (0 to 60; 60 to 70; 70 to 80; 80 to 90; and >90%). The significance levels of 40% for DENV NS5 and 60% for ZIKV NS5 roughly correspond to values above average plus one standard deviation. Starting from the whole Euphorbiaceae network showing 1231 constellations made of at least three nodes (Figure 1A), this first filtering step allowed the number of constellations of interest to be reduced to 27 for CHIKV, 28 for DENV, and 30 for ZIKV. The second step of the prioritization strategy consisted in targeting spectral uniqueness to detect uncommon secondary metabolites. Indeed, genus- or species-specific compounds can be distinguished from other ubiquitous metabolites in networks by using a taxonomical mapping.20,21 Each Euphorbiaceae genus was assigned a specific color tag (see Figure 1C), and this layout was used to highlight clusters of nodes that would be restricted to a given taxon and thus a given color. This methodology stressed the bioactivity potential and the original chemical composition of the EtOAc bark extract of Codiaeum peltatum (see Figure 1C). This extract was active

first chemical investigation of this extract afforded four unusual chlorinated monoterpenyl quinolones (9−12).21 A careful analysis of the networks with CHIKV, DENV, and ZIKV bioactivity overlays revealed two additional clusters of interest (MN1 and MN2) associated with this C. peltatum bark extract. This paper deals with the MS-guided purification of these compounds that led to the isolation of new bioactive daphnane-type diterpenoid orthoesters (DDO) (1 and 2) and 1,4-dioxane-fused phenanthrene dimers (3−8).

■

RESULTS AND DISCUSSION A collection of 292 EtOAc extracts of New Caledonian Euphorbiaceae was first filtered over polyamide to remove polar compounds that might interact with biological assays and induce false-positive responses. This set of samples was prepared from 107 different species, of which 93% are endemic. The extracts were prepared from leaves (47%), bark or twigs (48%), whole plants (3%), or fruits (2%). The Euphorbiaceae family was chosen for its large structural diversity, characterized by unique classes of structurally diverse diterpenoids endowed with a wide range of biological activities, such as antiviral activity against human immunodeficiency virus or modulation of P-glycoprotein and protein kinase C.22−29 The 292 samples were profiled by LC-HRMS2 and MS2 data were organized as molecular networks to map the spectral space of the extract collection (Figure 1A). Concurrently, this set was screened over three biological targets: (i) a CHIKV cytopathogenic effect (CPE) inhibition assay affording EC50 values ranging from 0.02 to 100 μg.mL−1; (ii) a DENV NS5 inhibition assay affording a percentage of inhibition at 50 μg· mL−1; and (iii) a ZIKV NS5 inhibition assay affording a C



Article

Figure 2. Key COSY (bold), HMBC (blue arrows), and NOESY or ROESY (red arrows) correlations for compounds 1 and 3.

5-OH and 20-OH were supported by the downfield resonances of the C-5 methine (δH 4.18, s; δC 74.5) and C-20 methylene groups. Correlations from 11-OH (δH 3.48, s) to C-11 (δC 77.0) and C-18 (δC 24.6) revealed the presence of an unusual oxidation at C-11. Two benzoyloxy groups were located at C-3 and C-12, as supported by HMBC correlations from H-3 to C1‴ (δC 168.1) and H-12 (δH 5.46, s) to C-1″ (δC 166.6). The three remaining oxygen atoms were assigned to C-9 (δC 84.3), C-13 (δC 86.1), and C-14 (δC 82.2). Eventually, the presence of the 9,13,14-orthobenzoate moiety was confirmed by the downfield shifts of the latter together with HMBC correlations from H-14 (δH 4.73, d) and H-3′/H-7′ (δH 7.79, m) to C-1′ (δC 118.2). The relative configuration of 1 was confirmed by 1H NMR coupling constant values and key NOESY correlations as depicted in Figure 2. The α-orientation of the 11-hydroxy group was supported by the strong NOESY correlation between H3-18 (δH 1.54, s), H-12, and H-8 (δH 3.08, d). For establishing the absolute configurations, electronic circular dichroism (ECD) calculations were carried out for compound 1 at the B3LYP/6-31G* level (Figure 3). Comparison of the experimental and calculated ECD spectra supported the absolute configuration of codiapeltine A (1) (2S,3S,4S,5S,6R,7S,8S,9R,10R,11R,12S,13R,14R). Compound 2 showed a protonated molecular ion at m/z 727.2749 (calcd 727.2749) in the HRESIMS, corresponding to the molecular formula C41H42O12, the same as compound 1. From this formula and similar spectroscopic data, it was apparent that 1 and 2 shared the same DDO core, but with a different C-ring substitution. Indeed, when compared with compound 1, the upfield resonances of the oxygenated secondary carbon C-12 (δC 68.8) and oxygenated tertiary carbons C-13 (δC 81.8) and C-14 (δC 75.6) and the downfield resonances of the 8-methine and oxygenated tertiary carbons C-9 (δC 89.1) and C-11 (δC 84.1) of compound 2 suggested that the orthobenzoate moiety is attached to different C-ring positions. From the HMBC correlation of H-12 (δH 5.24, s) to carbonyl C-1″ (δC 166.6) it was confirmed that a benzoyloxy group is attached at C-12, as was the case in compound 1. In addition, strong COSY correlations between H-14 (δH 4.20, dd) and H-8 at δH 3.40 (d, J = 5.8 Hz) and a proton at δH 3.11 (d, J = 12.5 Hz) showing no correlation in the HSQC spectrum, together with HMBC correlations from the latter and C-8, C-13, and C-14, suggested the presence of a hydroxy group at C-14 (δC 75.6). These results and HMBC correlations from the 18-methyl protons to the deshielded carbons C-9 and C-11 suggested that the orthobenzoate unit is attached at C-9, C-11, and C-13. Coupling constants and NOESY correlations

toward the three targets studied and was thus selected for an in-depth investigation. Among the high chemical diversity of this extract (Figure S1, Supporting Information) three constellations were spotted either for bioactivity (MN1) or bioactivity associated with a putative structural originality (MN2 and MN3). Besides sample prioritization, this multi-informative MN approach allows a targeted isolation of compounds of interest. After a first semipreparative HPLC fractionation step directed toward the purification of MN1−3 associated compounds traceable in the crude extract, all fractions obtained were analyzed in LC-MS2. As the amount of starting extract was relatively low (600 mg), only fractions showing ions detected in high abundances and belonging to one of the targeted clusters were selected for further isolation efforts. The MN1 cluster displayed ions shared between bark extracts of Codiaeum peltatum and C. oligogynum and bark and wood extracts of Trigonostemon cherrieri. This information together with the presence of chlorinated ions in the Trigonostemon subunit of MN1 allowed this cluster to be associated with DDO analogues.13,30,31 Compound 1 was obtained as a white, amorphous powder. Its HRESIMS showed a protonated molecular ion peak at m/z 727.2752 (calcd 727.2749), which, in conjunction with the 13C NMR data, was consistent with the molecular formula C41H42O12. The 1H and 13C NMR data, along with the HSQC spectrum, revealed three methyl groups, eight methine groups, of which five are oxygenated, three aromatic rings, two carbonyls, an exomethylene and an olefinic carbon, two methylene groups (one oxygenated), five oxygenated tertiary carbons, and one deshielded sp3 trioxygenated carbon at δC 118.2 typical of an orthobenzoate moiety.32 Additionally, two protons showing no correlation in the HSQC spectrum were attributable to the exchangeable protons of two hydroxy groups. The A−B−C rings of 1 can be constructed by comparison of NMR literature data and analysis of COSY and HMBC correlations (Figure 2).33−35 The spin system connecting H2-1 (δH 2.07, m; 2.27, ddd), H-2 (δH 1.93, m), H-3 (δH 5.09, d), H-10 (δH 3.12, dd), and H3-19 (δH 1.08, d) together with HMBC correlations from H3-19 to C-1 (δC 34.7), C-2 (δC 35.9), and C-3 (δC 81.9) and from 4-OH (δH 3.00, s) to C-4 (δC 80.6) and C-10 (δC 48.2) indicated the presence of a saturated cyclopentane oxygenated at C-3 for ring A. HMBC correlations from H-7 (δH 3.44, brs) to C-6 (δC 64.4), C-8 (δC 34.8), C-9 (δC 84.3), and C-20 (δC 66.6) and from H2-20 (δH 3.82, d; 3.87, d) to C-6 and C-7 (δC 64.4) confirmed the presence of an epoxide moiety at C-6/C-7.34 The presence of D



Article

Table 1. NMR Data for Compounds 1 and 2 (500 and 125 MHz, CDCl3) 1

Figure 3. Experimental (black) and calculated (blue) ECD spectra of codiapeltines A (1) and B (2) in MeOH. Vertical bars represent the experimental Δε in L·mol−1·cm−1 (left) and the B3LYP/6-31G*computed rotational strengths in 10−40 cgs units (right).

observed for compound 2 were comparable to 1, allowing its relative configuration to be depicted as shown. TDDFT calculation of the ECD spectrum of 2 was conducted at the B3LYP/6-31G* level. The absolute configuration of codiapeltine B (2) was defined as (2S,3S,4S,5S,6R,7S,8R,9R,10R,11R,12R,13R,14R) by comparison of its experimental and calculated ECD curves (Figure 3). Unlike all reported diterpenoids sharing the 5−7−6 tricyclic ring system, i.e., daphnane- and tigliane-type diterpenoids, codiapeltines A (1) and B (2) bear an unusual but unique Cring 9,11,12,13,14-oxygenation pattern. Furthermore, compound 2 is the first daphnane possessing a 9,11,13-orthoester linkage, establishing a new major structural class of DDO.23,32,36 The cluster MN2 was composed of ions found exclusively in bark extracts of the endemic C. oligogynum and C. peltatum species. Based on the molecular formula deduced from the four protonated ions at m/z 581.219 and 609.253 of MN2, the known actephilol A (5), epi-actephilol A (6), fimbricalyx C (7), and fimbricalyx D (8), which shared the original dioxane phenanthrene dimeric scaffold, were dereplicated and subsequently isolated.37−39 Interestingly, these phenanthrene dimers showed characteristic absorption bands at 224, 247, and 290 nm in their UV spectra. This typical UV profile has also been observed for a third analogue (compound 3) that was not included in MN2 but appeared in a separated cluster in the massive MN representation. As shown in Figure S2, Supporting Information, some spectral differences indicated that 3 was not part of MN2. Compound 3 was isolated as a brownish-red resin and was assigned the molecular formula C34H28O9 by analysis of its positive HRESIMS (m/z 581.1796 [M + H]+, calcd for C34H29O9, 581.1806). The spectroscopic data of 3 were closely

a

2

pos.

δC

δH (J in Hz)

δC

1α 1β 2 3 4 4-OH 5 5-OH 6 7 8 9 10 11 11-OH 12 13 14 14-OH 15 16-a 16-b 17 18 19 20-a 20-b 1′ 2′ 3′−7′ 4′−6′ 5′ 1″ 2″ 3″−7′′ 4″−6′′ 5″ 1‴ 2‴ 3‴−7″′ 4‴−6″′ 5‴

34.7

2.07 m 2.27 ddd (14.6, 13.2, 11.5) 1.93 m 5.09 d (4.7)

34.0

35.9 81.9 80.6 74.5 61.4 64.4 34.8 84.3 48.2 77.0 75.4 86.1 82.2 141.6 114.6 19.9 24.6 13.5 66.6 118.2 134.8 126.6 128.5b 130.2 166.6 129.4 130.4 128.6b 133.5 168.1 129.4 129.8 128.9b 133.9

3.00 s 4.18 s

73.9

3.44 brs 3.08 d (2.7) 3.12 dd (13.2, 6.8) 3.48 s 5.46 s 4.73 d (2.7)

4.98 5.24 1.78 1.54 1.08 3.82 3.87

brs brs s s d (6.7) d (12.3) d (12.3)

7.79 m 7.42 m 7.42 m

8.09 d (8.1) 7.40 m 7.55 m

7.97 d (8.1) 7.40 m 7.55 m

Overlapping signals.

36.0 82.7 80.2

61.5 62.9 40.8 89.1 46.8 84.1 68.8 81.8 75.6 142.8 114.4 19.5 18.1 13.6 67.0 116.0 135.4 126.3 128.5 130.1 166.6 129.5c 130.3 128.6d 133.6 168.4 129.6c 129.8 128.9d 133.9

δH (J in Hz) 1.97 2.20 1.99 4.98

ma m ma ma

3.06 s 4.19 sa 3.94 ma 3.58 s 3.40 d (5.8) 3.39 ma

5.24 s 4.20 ma 3.11 d (12.5) 4.97 5.19 1.75 1.65 1.12 3.77 3.93

brs brs s s d (6.4) d (12.2) d (12.2)

7.82 m 7.45 m 7.45 m

8.04 d (8.0) 7.40 m 7.55 m

7.96 d (8.1) 7.40 m 7.55 m

b,c,d

Values are interchangeable.

comparable to those of actephilol A (5) but revealed the presence of only one methoxy group and two additional carbonyls at δC 178.6 and 182.4. The structure of compound 3 may be divided in two different monomeric units, A and B (Figure 1). While the 1H and 13C NMR chemical shifts of subunit A were close to those of compound 5, those of subunit B showed marked differences. The ortho-quinone motif was deduced through the observation of the key HMBC correlations from H-8′ (δH 7.71, s) to C-6′ and a carbonyl at δC 178.6 on one hand and from H3-11′ (δH 2.65, s) to C-1′, C-2′, C-10a′, and a carbonyl at δC 182.4, which established the 9′,10′-dione function. Other HMBC correlations fully supported the proposed structure depicted in Figure 2 for subunit B. E



Article

Table 2. NMR Data for Compounds 3 (600 and 150 MHz, DMSO-d6) and 4 (600 and 150 MHz, Methanol-d4) 3 (A unit) position

δC

1 2 3 3-OMe 4 5 6 7 8 9 10 11 12 12-OMe 13 14 15 16 17

47.2 204.2 96.7 52.1 72.5 112.1 142.8 103.3 155.2 118.1 123.8 126.0 156.4 104.0 133.6 28.7 27.9 16.7

3 (B unit) δH

3.42 s 5.71 s

6.74 s

7.93 s

7.38 s 1.35 s 1.58 s 2.32 s

position

δC

1′ 2′ 3′ 4′ 4′a 4′b 5′ 6′ 7′ 8′ 8′a 9′ 9′-OMe 10′ 10′a 11′ 12′

132.1 140.6 146.3 110.1 121.7 135.9 109.7 164.1 126.3 132.1 131.2 178.6

4 (A unit) δH

7.23 s

7.20 s

7.71 s

182.4 124.8 13.4 15.5

2.65 s 2.13 s

position

δC

1 2 3 3-OMe 4 5 6 7 8 9 10 11 12 12-OMe 13 14 15 16 17

48.8 207.0 98.5 53.0 74.5 116.2 144.6 105.8 156.3 120.5 125.5 128.2 159.9 56.5 101.4 135.4 30.0 28.9 17.1

4 (B unit) δH

3.49 s 5.80 s

6.79 s

8.01 s

3.84 s 7.35 s 1.46 s 1.65 s 2.34 s

position

δC

δH

1′ 2′ 3′ 4′ 4′a 4′b 5′ 6′ 7′ 8′ 8′a 9′ 9′-OMe 10′ 10′a 11′ 12′

120.3 141.6 141.4 108.2 120.7 132.6 107.2 156.8 126.5 125.7 122.2 154.3 56.2 96.8 129.5 11.8 16.9

7.74 s

7.64 s

7.93 s

4.02 s 6.83 s 2.61 s 2.34 s

Table 3. Antimetabolic and Antiviral Activities of Compounds 1−12 in Vero Cells against CHIKV and Their Inhibitory Activities against RNA-Dependent RNA NS5 of Dengue and Zika Viruses CHIKV compound

CC50 (Vero)

1 2 3 4 5 6 7 8 9 10 11 12 chloroquine 3′dATP

52 ± 12 49 ± 17 >172 ndc >172 >129 nd nd >208 >288 >239 >277 89 ± 28

a

EC50a

b

SI

10.0 ± 2.3 4.4 ± 0.5 >172 nd >86 >172 nd nd >277 >288 >239 >277 11 ± 7

5 11

ZIKV

DENV

IC50a

IC50a

>69 >69 8.1 10.9 4.8 8.1 13.6 12.7 >138 nd >138 nd

± ± ± ± ± ±

0.9 2.5 0.9 1.0 1.2 1.3

>69 >69 12.1 14.5 4.5 9.1 21.7 17.8 >119 nd >119 nd

± ± ± ± ± ±

1.9 2.2 0.7 1.2 2.0 3.1

8 0.14 ± 0.01

0.07 ± 0.005

Data in μM. Values are the median ± absolute deviation calculated from at least three independent assays. bSI or window for antiviral selectivity is calculated as CC50 Vero/EC50 CHIKV. cnd = not determined.

a

ROESY correlation between 3-OMe (δH 3.49, s) and H-4 (δH 5.80, s). As previously reported, the MS-guided purification of compounds displayed in MN3 afforded the first naturally occurring monoterpenyl quinolin-4-one isochloroaustralasine A (12) along with three exceptionally rare chlorinated monoterpenyl quinolin-2-ones, chloroaustralasines A−C (9− 11, respectively).21 Compounds 1−12 were evaluated for their antiviral activity in a CHIKV cell-based assay (CPE inhibition) and against DENV- and ZIKV-NS5 (Table 3). In the CHIKV assay, compounds 1 and 2 showed anti-CHIKV activities compared to the reference compound, chloroquine, with EC50 values of 10 ± 2.3 μM (SI = 5) and 4.4 ± 0.5 μM (SI = 11), respectively. Although the anti-CHIKV activity of DDO derivatives had already been reported,13,30,31 these results

The relative configuration of compound 3 was established from the comparison of NMR data with those of compound 5 and ROESY analysis (Figure 2). The cis-configuration of the C3−C-4 ring junction was confirmed by a strong ROESY correlation between 3-OMe (δH 3.42, s) and H-4 (δH 5.71, s). Actephilol B (3) is the seventh member of the 1,4-dioxanefused phenanthrene dimer family.37 Compound 4, displayed in MN2 (Figure 1C), was isolated as a brownish-red resin, and its HRESIMS data showed a protonated ion at m/z 595.2316 (calcd for C36H35O8, 595.2326). The spectroscopic data of compound 4 were close to those of actephilol A (5) and epi-actephilol A (6). However, a strong HMBC correlation between a methyl group, MeO-12 (δH 3.84, s), and C-12 (δC 159.9) indicated the presence of an additional methoxy in position 12. The relative configuration of actephilol C (4) was supported by the strong F



■

confirmed the predictions made from the multi-informative MN approach used in this study. In the DENV and ZIKV assays, results also met our expectations, as compounds 3−8 exhibited strong inhibitory activities, although ∼100 times less potent than the reference compound, 3′-dATP, on both enzymes with IC50 values in the μM range. To confirm these results, two series of experiments were conducted. First, the IC50 values of compounds 3−8 were determined in the presence of detergent, Triton X-100, at 0.01% final concentration. Using a detergent in the assay below the critical micellar concentration indicates whether a compound is a potential aggregator for one or more partners of the assay. Thus, a nonspecific inhibition would induce an increase of the IC50 value.40 Under these conditions compounds 4−8 lost their inhibitory potential (IC50 > 100 μM) on both NS5 proteins, while compound 3 showed IC50 values of 36.0 ± 4.7 and 67.2 ± 20.0 μM on ZIKV-NS5 and DENV-NS5, respectively. Using an orthogonal assay based on fluorescence polarization, the apparent Kd values of these compounds on ZIKVNS5 and DENV-NS5 were determined. This assay shows whether a compound can have a specific interaction with the enzyme by measuring its potential to shift the interaction between the GTP bodipy (boron-dipyrromethene) and the protein. Compounds 4−8 showed no fixation on the enzyme (Kdapp > 100 μM). These results, correlated with the loss of inhibition in the detergent assays, confirmed that these compounds act as aggregators and not inhibitors. However, compound 3 showed a Kdapp of 32.1 ± 4.5 and 9.3 ± 3.6 μM on ZIKV-NS5 and DENV-NS5, respectively. It agreed with the detergent assays and indicated a specific interaction toward the NS5 targets. Taken together, these results indicate that actephilol B (3) is the first natural dual inhibitor of ZIKV and DENV NS5. Time and money spent in tedious bioassay-guided fractionation and isolation processes represent major bottlenecks in NP drug discovery. Here, we reported an example of the multi-informative MN methodology set up to reduce workload and associated costs of the NP hit detection workflow. This strategy facilitated the extract prioritization step by highlighting the original chemical composition of C. peltatum and also permitted a straightforward MS-guided purification of targeted compounds. The study of the C. peltatum bark extract led to the characterization of three different chemical series. Among the compounds isolated, codiapeltine B (2) is the first daphnane bearing a 9,11,13orthoester moiety, establishing a new major structural class of daphnane-type diterpenoid orthoesters. Moreover, actephilol B (3), a member of the rare 1,4-dioxane-fused phenanthrene dimer family, has been found as the first natural dual inhibitor of DENV and ZIKV NS5 yet characterized. Without any prior chemotaxonomic knowledge or bibliographic preconceived assumptions, the multi-informative MN prioritization approach allows bioactive scaffolds to be spotted. As a complement to bioactivity observations, the taxonomical input pinpoints genus- or species-specific metabolites. Combining biological and chemical profiling results in an MN environment enables a clear view of the structural breadth and distribution within taxonomically homogeneous extract collections and drastically increases the chances of discovering new lead compounds.

Article

EXPERIMENTAL SECTION

General Experimental Procedures. Optical rotations were measured at 25 °C on an Anton Paar MCP 300 polarimeter. UV spectra were recorded on a Varian Cary 100 UV−vis spectrophotometer. ECD spectra were measured at 25 °C on a JASCO J-810 spectropolarimeter. NMR spectra were recorded on a Bruker 500 MHz instrument (Avance 500) for compounds 1, 2, 7, and 8, on a Bruker 600 MHz instrument (Avance 600) for compounds 3 and 4, and on a Bruker 300 MHz instrument (Avance 300) for compounds 5 and 6. Chemical shifts are in ppm, and coupling constants are in Hz. Kinetex analytical and semipreparative C18 columns (250 × 4.6 mm and 250 × 10 mm; 5 μm Phenomenex) and a Nucleodur PFP semipreparative column (250 × 10 mm; 5 μm Macherey-Nagel) were used for HPLC separations using a Dionex system equipped with a sample manager (Gilson 215 liquid handler), a column fluidics organizer, a binary pump (Dionex HPG-3200BX), a UV−vis diode array detector (190−600 nm, Dionex UVD340U), and a PL-ELS 1000 ELS Polymer Laboratory detector. HPLC separations were also undertaken using a Waters Alliance system equipped with a binary pump (Waters 2525), a UV−vis diode array detector (190−600 nm, Waters 2996), and an ELS detector (Waters 2424). Disconnected from ELSD, this system was also used in preparative mode to purify small amounts of products. All solvents were purchased from Carlo Erba (France) and SDS (Peypin, France). Data-Dependent LC-ESI-HRMS2 Analysis. LC analyses were performed with a Dionex Ultimate 3000 RSLC system equipped with an Accucore C18 column (2.1 × 100 mm; 2.6 μm, ThermoScientific). The mobile phase consisted of H2O−MeCN with 0.1% formic acid (20:80) during 5 min, then a gradient from 20:80 to 100:0 in 20 min, at a flow rate of 350 μL·min−1. The temperature of the column oven was set at 30 °C and the injection volume at 5 μL. LC-ESI-HRMS2 analyses were achieved by coupling the LC system to an Agilent 6540 hybrid quadrupole TOF mass spectrometer (Agilent Technologies, Massy, France) equipped with an ESI dual source, operating in positive-ion mode. Source parameters were set as follows: capillary temperature at 325 °C, source voltage at 3500 V, sheath gas flow rate at 7 L.min−1, nebulizer pressure at 30 psi, dryinggas flow rate at 10 L·min−1, drying-gas temperature at 350 °C, stealthgas temperature at 350 °C, skimmer voltage at 45 V, fragmentor voltage at 150 V, and nozzle voltage at 500 V. MS scans were operated in full-scan mode from m/z 100 to 1000 (0.1 s scan time) with a mass resolution of 20 000 at m/z 922. An MS1 scan was followed by MS2 scans of the five most intense ions above an absolute threshold of 3000 counts. Selected parent ions were fragmented with collision energy fixed at 35 eV and an isolation window of 1.3 amu.41 Calibration solution, containing two internal reference masses (purine, C5H4N4, m/z 121.0509, and HP-921 [hexakis(1H,1H,3Htetrafluoropentoxy)phosphazene], C18H18O6N3P3F24, m/z 922.0098), routinely led to mass accuracy below 2 ppm. LC-UV data were analyzed with Chromeleon software (Dionex), and MS data acquisition and processing were performed using MassHunter Workstation software (Agilent Technologies, Massy, France). MZmine 2 Data-Preprocessing Parameters. Networks were generated using the feature-finding-based methodology previously described.42 The 292 MS2 data files were converted from the .d (Agilent) standard data format to .mzXML format using the MSConvert software, part of the ProteoWizard package.43 All .mzxml were processed using MZmine 2 v31.44 The mass detection was realized keeping the noise level at 0. The ADAP chromatogram builder was used using a minimum group size of scans of 4, a group intensity threshold of 3000, a minimum highest intensity of 4000, and m/z tolerance of 0.005 (or 20 ppm).45 The ADAP wavelets deconvolution algorithm was used with the following standard settings: S/N threshold = 8, minimum feature height = 4000, coefficient/area threshold = 20, peak duration range 0.05−1 min, tR wavelet range 0.01−0.07 min. However, these parameters needed to be reoptimized for a few specific samples in this data set. MS2 scans were paired using an m/z tolerance range of 0.02 Da and tR tolerance G



Article

[M + H] + (calcd for C 34 H 29 O 9 , 581.1806); GNPS ID: CCMSLIB00004681457 ([M + H]+). Actephilol C (4): brownish-red resin; [α]24D +8 (c 0.1, MeOH); UV (MeOH) λmax (log ε) 250 (5.13), 289 (4.86), 317 (4.49) nm; 1H and 13C NMR, see Table 2; HRESIMS m/z 595.2316 [M + H]+ (calcd for C36H35O8, 595.2326); GNPS ID: CCMSLIB00004681458 ([M + H]+). Actephilol A (5): brownish-red resin; HRESIMS m/z 581.2179 [M + H ] + ( c a lc d f o r C 3 5 H 3 3 O 8 , 5 8 1 .2 1 70 ) ; G N P S I D : CCMSLIB00004681459 ([M + H]+). Epi-actephilol A (6): brownish-red resin; HRESIMS m/z 581.2176 [M + H] + (calcd for C 35 H 33 O 8 , 581.2170); GNPS ID: CCMSLIB00004681460 ([M + H]+). Fimbricalyx C (7): brownish-red resin; HRESIMS m/z 609.2467 [M + H] + (calcd for C 37 H 37 O 8 , 609.2483); GNPS ID: CCMSLIB00004681461 ([M + H]+). Fimbricalyx D (8): brownish-red resin; HRESIMS m/z 609.2466 [M + H] + (calcd for C 37 H 37 O 8 , 609.2483); GNPS ID: CCMSLIB00004681462 ([M + H]+). ECD Calculation. Computational calculations were performed using the Gaussian ’09 software package, revision D01.47 Conformational analyses and geometry optimizations were performed at the B3LYP/6-31G* level in MeOH. The ECD calculations were performed using TDDFT at the B3LYP/6-31G* level in MeOH. The ECD spectra were obtained by weighing the Boltzmann distribution rate of each geometric conformation with a bandwidth σ of 0.3 eV. Conformer structures of 1 and 2 and their optimized coordinates at the B3LYP/6-31G* level in MeOH are available in Figures S3 and S4, Supporting Information. Virus-Cell-Based Anti-Chikungunya Assay. The antiviral activity of the extracts and pure compounds against Chikungunya virus was determined using an MTS [3-(4,5-dimethylthiazol-2-yl)-5(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium]-based cytopathic effect reduction assay. Chikungunya virus (Indian Ocean strain 899), kindly provided by Prof. C. Drosten (Institute of Virology, University of Bonn, Germany), was grown on Vero (African green monkey kidney) cells. Antiviral assays were performed in DMEM medium [MEM Rega3 (cat. no. 19993013; Invitrogen), 2% FCS (Integro), 5 mL of 200 mM L-glutamine, and 5 mL of 7.5% NaHCO3]. Cells were seeded in a 96-well plate (Falcon, BD Biosciences) and allowed to adhere overnight. Serial dilutions of the test compounds were made in 100 μL of assay medium in a 96-well microtiter. Subsequently, 50 μL of a 4× virus dilution in assay medium was added (MOI 0.01), followed by 50 μL of a cell suspension. This suspension, with a cell density of 25 000 cells/50 μL, was prepared from a Vero A cell line subcultured in cell growth medium (MEM Rega3, supplemented with 10% FCS, 5 mL of Lglutamine, and 5 mL of NaHCO3) in a 1:4 ratio. The assay plates were returned to the incubator (37 °C, 5% CO2, 95−99% relative humidity) for 5 days, a time at which maximal virus-induced cell death or CPE is observed in untreated, infected controls. Subsequently, the assay medium was aspirated, replaced with 75 μL of a 5% MTS (Promega) solution (MTS-phenazinemethosulfate (MTS/PMS) stock solution [(2 g·L−1 MTS (Promega, Leiden, The Netherlands) and 46 mg·L−1 PMS (Sigma−Aldrich, Bornem, Belgium) in PBS at pH 6−6.5)], diluted 1/20 in MEM (Life Technologies, Gent, Belgium) in phenol-red-free medium, and incubated for 1.5 h. Absorbance was measured at a wavelength of 498 nm (Safire2, Tecan), with the optical densities (OD values) reaching 0.6−0.8 for the untreated, uninfected controls. The optical density values were converted to control percentages, and logarithmic interpolation was used to calculate the EC50 and CC50. The EC50 was defined as the compound concentration that protected 50% of cells from virusinduced CPE. Adverse effects of the drug on the host cell were also assessed by means of the MTS method, by exposing uninfected cells to the same concentrations of the compounds. The CC50 was defined as the compound concentration that reduced the number of viable cells by 50%. All assay conditions producing an antiviral effect that exceeded 50% were checked microscopically for signs of a cytopathic effect or adverse effects on the host cell (i.e., altered cell or monolayer

range of 0.5 min. Isotopologues were grouped using the isotopic peak grouper algorithm with an m/z tolerance of 0.005 (or 20 ppm) and a tR tolerance of 0.2 min. Peak alignment was performed using the join aligner module [m/z tolerance = 0.004 (or 10 ppm), weight for m/z = 2, weight for tR = 1, absolute tR tolerance 0.5 min]. The peak list was gap-filled with the same tR and m/z range gap filler module [m/z tolerance of 0.004 (or 10 ppm)]. Eventually, the .mgf preclustered spectral data file and its corresponding .csv metadata file (for tR, areas, and formulas integration) were exported using the dedicated “Export for GNPS” and “Export to CSV file” built-in options. See further documentation and tutorials at https://ccms-ucsd.github.io/GNPSDocumentation/ featurebasedmolecularnetworking/. Molecular Network Analysis. The molecular networks were created using the online workflow at Global Natural Products Social molecular networking (http://gnps.ucsd.edu).46 MS2 spectra were window-filtered by choosing only the top six peaks in the ±50 Da window throughout the spectrum. The data were not clustered with MS-Cluster. A network was created where edges were filtered to have a cosine score above 0.7 and at least four matching peaks. Further edges between two nodes were kept in the network if and only if each of the nodes appeared in each other’s respective top 10 most similar nodes. The library spectra were filtered in the same manner as the input data. All matches kept between network spectra and library spectra were required to have a score above 0.7 and at least 4 matched peaks. Plant Material. Stem and bark of Codiaeum peltatum were collected at “Mont Aoupinié” (New Caledonia) at an altitude of 487 m in November 2001 and authenticated by V. Dumontet. A voucher specimen (DUM-0109) has been deposited at the Herbier IRD de Nouméa, New Caledonia. Extraction and Isolation. The plant material (200 g, dry wt) was extracted with EtOAc (3 × 0.5 L each). The EtOAc extract was concentrated in vacuo at 40 °C to yield 1.5 g of residue. This extract was dissolved in MeCN (0.25 L) and subjected to a liquid/liquid partition with n-heptane (2 × 0.25 L) to afford 0.6 g of an MeCNsoluble fraction. This fraction was subjected to a semipreparative HPLC (Kinetex C18, H2O−MeCN 35:65 + 0.1% formic acid to 0:100 in 45 min at 4.5 mL·min−1) to afford 20 fractions of decreasing polarity, F1−F20. Purification of fraction F12 (9.2 mg) by semipreparative HPLC (Nucleodur PFP, H2O−MeCN 45:55 + 0.1% formic acid at 4.5 mL·min−1) afforded compounds 1 (2.5 mg) and 2 (1.5 mg) (tR: 17.5 and 15.5 min, respectively). Fraction F7 (9.1 mg) was purified by semipreparative HPLC (Nucleodur PFP, H2O− MeCN 50:50 + 0.1% formic acid at 4.5 mL·min−1) to afford compound 3 (1.1 mg) (tR: 18.0 min). Fraction F5 (17.4 mg) was purified by semipreparative HPLC (Nucleodur PFP, H2O−MeCN 30:70 + 0.1% formic acid at 4.5 mL.min−1) to afford compound 4 (1.4 mg) (tR: 13.4 min). Fraction F9 (20.9 mg) was subjected to a semipreparative HPLC (Nucleodur PFP, H2O−MeCN 45:55 + 0.1% formic acid at 4.5 mL.min−1) to afford compound 5 (2.9 mg) and compound 6 (4.1 mg) (tR: 18.5 and 20.5 min, respectively). Fraction F4 (27.8 mg) was purified by semipreparative HPLC (Nucleodur PFP, H2O−MeCN 18:82 + 0.1% formic acid at 4.5 mL·min−1) to afford compounds 7 (0.62 mg) and 8 (0.47 mg) (tR: 11.3 and 12.0 min, respectively). Codiapeltine A (1): white, amorphous powder; [α]24D −53 (c 0.1, MeOH); UV (MeOH) λmax (log ε) 230 (4.44) nm; 1H and 13C NMR, see Table 1; HRESIMS m/z 727.2752 [M + H]+ (calcd for C41H43O12, 727.2749); GNPS ID: CCMSLIB00004681455 ([M + H]+). Codiapeltine B (2): white, amorphous powder; [α]24D −43 (c 0.1, MeOH); UV (MeOH) λmax (log ε) 230 (4.46) nm; 1H and 13C NMR, see Table 1; HRESIMS m/z 727.2749 [M + H]+ (calcd for C41H43O12, 727.2749); GNPS ID: CCMSLIB00004681456 ([M + H]+). Actephilol B (3): brownish-red resin; [α]24D −9 (c 0.1, MeOH); UV (MeOH) λmax (log ε) 224 (4.45), 247 (4.40), 290 (4.31), 337 (3.86) nm; 1H and 13C NMR, see Table 2; HRESIMS m/z 581.1796 H



Article

(av(neg) − av(pos)). Compounds leading to a 70% inhibition or more at 50 μg·mL−1 and presenting a dose/response effect were qualified as hits. Selected hits were then confirmed in triplicate, and IC50 values determined only on pure compounds. IC50 Determination Based on Fluorescent Picogreen Assay. The concentrations of the compounds leading to 50% inhibition of NS5mediated RNA synthesis were determined in IC50 buffer (50 mM HEPES pH 8.0, 10 mM KCl, 2 mM MnCl2, 2 mM MgCl2, 10 mM DTT) containing 100 or 40 nM homopolymeric uridine RNA template for Zika NS5 and DENV-2 NS5, respectively, seven various concentrations of compound, and 40 nM of enzyme. For the IC50 determination in the presence of detergent, Triton X100 was added at 0.01% final concentration in the assay. Five ranges of inhibitor were available (0.01 to 5 μM; 0.1 to 50 μM; 0.5 to 50 μM; 1 to 100 μM; 5 to 400 μM). According to the inhibitory potency of the compound tested, a range was selected to obtain a more accurate IC50 based on the best repartition of the points surrounding the range. Reactions were conducted in 40 μL volume on a 96-well Nunc plate. All experiments were robotized by using a BioMek 4000 Automate (Beckman). A 2 μL amount of each diluted compound in 100% DMSO was added to the wells to the chosen concentration (5% DMSO final concentration). For each assay, the enzyme mix was distributed in wells. Reactions were started by the addition of the nucleotide mix (100 and 80 μM ATP for Zika NS5 and DENV-2 NS5, respectively) and were incubated at 30 °C for 10 min. Reaction assays were stopped by the addition of 20 μL of EDTA (100 mM). Positive and negative controls consisted respectively of a reaction mix with 5% DMSO final concentration or EDTA (100 mM) instead of compounds. Reaction mixes were then transferred to a Greiner plate using a Biomek I5 Automate (Beckman). Picogreen fluorescent reagent was diluted to 1/800 in TE buffer according to the manufacturer’s data, and 60 μL of reagent was distributed into each well of the Greiner plate. The plate was incubated for 5 min in the dark at room temperature, and the fluorescence signal was read at 480 nm (excitation) and 530 nm (emission) using a TecanSafire2. The IC50 value was determined using the following equation: % of active enzyme = 100/(1 + I2/IC50), where I is the concentration of inhibitor and 100% activity is the fluorescence intensity without inhibitor. IC50 was determined from curve-fitting using Prism software. For each value, results were obtained using triplicates in a single experiment. Apparent Kd Determination of Compounds by a Fluorescence Polarization Assay. The concentration of compounds leading to a shifting of the fixation of GTP bodipy to the D2 or Zika NS5 protein was determined in the binding buffer (50 mM Tris pH 7.5, 50 mM NaCl, 0.01% NP40, 1 mM DTT) containing 50 nM GTP bodipy, 300 nM D2 NS5, or 150 nM Zika NS5 and seven various concentrations of compound. Reactions were conducted in 50 μL volume on a black 96-half-well Greiner plate. Each diluted compound (2.5 μL) in 100% DMSO was added in plate wells to the chosen concentration (5% DMSO final concentration). For each assay, 10 μL of a 5× concentration of the enzyme was distributed in plate wells. Reactions were started by the addition of the GTP bodipy (10 μL of a 5× concentration) and were incubated at room temperature for 15 min. Total fluorescence and fluorescence polarization were measured using a microplate reader (Pherastar) with a fluorescence polarization module (485 nm excitation and 520 emission wavelengths). Instrument settings were fixed as follows: 100mP target mP, 200 flashes per well, 0.5 s settling time. Focus and gain values were adjusted prior to measurement. The positive control consisted of a reaction mix with 5% DMSO final concentration in place of the compound. GTP bodipy alone (n = 8) was used to determine the minimal fluorescence polarization intensity, and the average of this background value was subtracted from each value of the assay. The fluorescence polarization of the positive control is 100% fixation, and the residual fluorescence polarization signal (mP %) is determined for each concentration of the compound. The apparent Kd (Kdapp) of the compound was determined from curve-fitting using Prism software. For each value, results were obtained using triplicates in a single experiment.

morphology). A sample was considered to elicit a selective antiviral effect on virus replication only when, following microscopic quality control, at least at one concentration no CPE or any adverse effect was observed (image resembling untreated, uninfected cells). The antiviral experiments have been performed in a biosafety screening facility that has been validated for handling of Chikungunya virus as well as the manipulation of molecules of unknown chemical safety risk. All studies have been performed by trained staff. DENV and ZIKV Assays. Products and Reagents. Homopolymeric uridine (Poly U) RNA template was obtained from GE Healthcare. Natural extracts or pure compounds were resuspended at 10 or 1 mg·mL−1 in 100% DMSO and stored at −20 °C. The Quant-it Picogreen dsDNA assay kit (ref P11496) was obtained from Molecular Probes-Invitrogen. Expression and Purification of Zika and Dengue 2 NS5 Recombinant Proteins. The viral NS5 proteins of Zika and Dengue serotype 2 (DENV-2) virus coding sequences were cloned in fusion with a N-terminus hexahistidine tag in gateway plasmids (pQE30). The proteins were expressed in E. coli cells and purified following a previously described protocol.48 Determination of the Inhibitory Potential of Compounds by a Picogreen Assay on Zika and Dengue NS5. Principle of the Assay. A picogreen kinetic assay was based on polymerase activity of NS5 protein, which catalyzes de novo reaction of a poly(U) template and adenosine triphosphate (ATP). The reaction was carried out at 30 °C with 50 mM Hepes pH 8, 10 mM DTT, 10 mM KCl, 2 mM MgCl2, 2 mM MnCl2, and the selected concentrations in substrate (Poly U and ATP) and enzyme. Each reaction assay was robotized on a Biomek 4000 in 40 μL final volume for screening assay and IC50. Control plates, using only DMSO at 5% final concentration instead of compounds, were used to determine the Z′ factor, the background, and the gain value of the assay. Quality of measurements was assessed by calculating the Z′ factor for each plate: Z′ factor = 1 − ([3SD(neg) + 3SD(pos)]/[av(neg) − av(pos)]), where SD(neg) and SD(pos) stand for the standard deviation obtained for negative and positive controls, respectively, and where av(neg) and av(pos) are averages for negative and positive controls, respectively. Assay plates contained positive and negative controls distributed in the first and the last columns, respectively. Positive and negative controls consisted of the reaction mixture supplemented by ATP and 5% DMSO or 20 mM EDTA, respectively. In the case of the screening assay, 3′dATP was also used as an internal control in the last column at 10 μM final concentration. Time incubation (from 10 to 30 min), substrate concentrations (poly U and ATP), and enzyme concentration were optimized, based on the previous experimental results to obtain the best reproducibility (Z′ > 0.6; background

Antiviral Compounds from Codiaeum peltatum

Recommend Documents