Identification of Privileged Antichlamydial Natural Products by a

The strategy consists of the application of ligand-based virtual screening to a natural product library of 502 compounds with the ChemGPS-NP chemograp...
0 downloads 0 Views 1MB Size
Article Cite This: J. Nat. Prod. 2017, 80, 2602-2608

pubs.acs.org/jnp

Identification of Privileged Antichlamydial Natural Products by a Ligand-Based Strategy Elina Karhu,† Janne Isojar̈ vi,‡ Pia Vuorela,†,§ Leena Hanski,† and Adyary Fallarero*,† †

Exploration of Anti-Infectives Research Group, Pharmaceutical Design and Discovery, Drug Research Program, Faculty of Pharmacy, University of Helsinki, Viikinkaari 5E, Helsinki FI-00014, Finland ‡ Bioinformatics, Molecular Plant Biology, Department of Biochemistry, University of Turku, Vatselankatu 2, Turku FI-20500, Finland S Supporting Information *

ABSTRACT: The obligate intracellular pathogen Chlamydia pneumoniae remains a difficult target for antimicrobial therapy. Owing to the permeability barrier placed by bacterial and host vacuolar membranes, as well as the propensity of the bacterium for persistent infections, treatment failures are common. Despite the urgent need for new antichlamydial compounds, their discovery is challenged by the technically demanding assay procedures and lack of validated targets. An alternative strategy of using naturally occurring compounds and their derivatives against C. pneumoniae is presented. The strategy consists of the application of ligand-based virtual screening to a natural product library of 502 compounds with the ChemGPS-NP chemography tool followed by in vitro antichlamydial assays. The reference set used for the 2D similarity search was constructed of 19 known antichlamydial compounds of plant origin. Based on the similarity screen, 53 virtual hits were selected for in vitro testing. Six compounds (leads) were identified that cause ≥50% C. pneumoniae growth inhibition and showed no impact on host cell viability. The leads fall into completely new antichlamydial chemotypes, one of them being mycophenolic acid (IC50 value 0.3 μM). The outcome indicates that using this flipped, target-independent strategy is useful for facilitating the antimicrobial lead discovery against challenging microbes.

and which is further modified by the pathogen in the course of the infection.6 Along with the two cell membranes typical of all Gram-negative bacteria, the inclusion membrane contributes to form an extra barrier to be penetrated by any chemical agent in order to reach the replicating machinery of the bacterium. Apart from these penetration barriers, another challenging feature of antichlamydial drug discovery is the lack of validated protein targets, which relates to the fact that tools for genetic manipulation of Chlamydiae are limited.7 The membrane layers and the fragile nature of the intracellular form of Chlamydiae (RBs) upon isolation hinder applying genetic engineering protocols to replicating intracellular forms. On the other hand, the extracellular forms are more robust but are metabolically silent and thus not applicable for genetic manipulation. In many countries, the first line treatment of C. pneumoniae infection is azithromycin.8 The widespread use of azithromycin and other macrolide antibiotics in both developed and developing countries has led to the incidence of azithromycin-resistant fecal Escherichia coli.9 In addition, reported cases of azithromycin-resistant Streptococcus pneumoniae carriage in young children have underlined the risks associated with the extensive use of macrolides.10 Collectively, these factors demonstrate the need for novel, more effective, and safe therapies against C. pneumoniae.

Chlamydia pneumoniae is a ubiquitous human pathogen that causes infections in the respiratory tract.1 The clinical picture of these infections varies from asymptomatic cases to severe community-acquired pneumonias. Acute C. pneumoniae infection responds to certain antibiotic treatments, but the firstchoice antibiotics for respiratory tract infections are not effective against this pathogen, and β-lactams may even trigger the conversion of C. pneumoniae into a persistent state.2,3 The persistent form of infection resulting from untreated or mistreated acute infections represents a major therapeutic challenge, as it cannot be eradicated with any known pharmacological agent. Moreover, C. pneumoniae belongs to the unique group of intracellular Gram-negative bacteria, which involve challenging features from the drug discovery point of view.4 The chlamydial life cycle is characterized by successive conversions of the extracellular infectious elementary bodies (EBs) into replicative reticulate bodies (RBs). In a typical epithelial cell infection, differentiation of EBs into RBs is initiated 2−8 h postinfection. Replication of RBs by binary fission occurs shortly after the differentiation and is followed by the asynchronous redifferentiation of newly formed RBs into EBs, providing the next generation of infectious bacterial cells approximately 3 days postinfection.5 During its entire time inside host cells C. pneumoniae resides inside a membranous organelle known as an inclusion, which is formed from the host cell plasma membrane components that are incorporated upon the entry of the bacterium into the cell © 2017 American Chemical Society and American Society of Pharmacognosy

Received: November 14, 2016 Published: October 18, 2017 2602

DOI: 10.1021/acs.jnatprod.6b01052 J. Nat. Prod. 2017, 80, 2602−2608

Journal of Natural Products

Article

Table 1. Stepwise Summary of the Antichlamydial Lead Mining Process in a Flipped, Target-Independent Approach step 1

ligand selection

step 2

ligand-based virtual screening

step 3

identification of virtual hits

step 4

hits selection

step 5

in vitro confirmation of selected hits potency measurements and lead identification

step 6

Reference antichlamydial natural products (NP) (19) were selected as ligands to be used in ligand-based virtual screening (VS). ChemGPS-NP-based VS was applied. A natural product library of 502 compounds was screened, using Euclidean distances (EDs). VS yielded 53 virtual hits, which were closest to the reference compounds, based on EDs and not known for chlamydial inhibitory activity. Exclusion of cytotoxic hits based on host cell viability testing of virtual hits. Twenty-six compounds caused no significant changes in host cell viability (≥75%). In vitro activity assay produced 6 lead compounds with ≥50% Chlamydia inhibiting activity. Potency measurements of the 6 lead compounds identified one new potent natural product chlamydial inhibitor (mycophenolic acid).

Even though plant-derived compounds have not been the general focus of antibacterial drug discovery, our previous studies on C. pneumoniae indicate that plant phenolic compounds are a rich source of antichlamydial compounds.4 In drug discovery, it has recently been widely recognized that the practice of replacing traditional whole-cell assays with target-based screening has failed to produce new drug candidates, especially in the field of antibacterials.11 Highthroughput screening (HTS) has enabled rapid exploration of large compound libraries. However, increasing throughput has not met the expectations in obtaining an expected increase in new drug candidates. This is said to primarily result from the biochemical screening assays being unable to reflect the complexity of living systems.12 When the target is unknown, phenotypic assays offer a solution to proceed in the drug discovery process.13 This approach is particularly useful when the biological activity outcome in the phenotypic assay is relevant and informative, as is the case most often with antimicrobial agents. Ligand-based virtual screening is not dependent on one specific target, and combining computational methods with bioactivity enables a shift from random-based testing to highcontent screening.14 The massive and costly HTS campaigns can be facilitated with preliminary in silico screening. By in silico identifying first the pharmaceutically relevant chemical space where most promising antimicrobial hit compounds are located, the screening process can be accelerated, resulting in a more cost-effective start for early drug discovery. As a result, the integration of in silico and in vitro tools has been increasingly emphasized in the realm of antimicrobial discovery. The utilization of natural and naturally inspired compounds, to discover and develop the final drug entity, is important in the area of microbes, and this area of natural products research should be expanded significantly.15 However, new strategies to identify successful antimicrobials, in this case privileged antichlamydial natural products, are to be explored, assayed, and perused. The analysis of the chemical space that is populated by bioactive compounds is an attractive alternative tool for guiding natural product drug discovery approaches particularly in cases in which target-based discovery is not a suitable strategy. Since C. pneumoniae lacks validated targets for lead discovery and it poses other challenges for in vitro drug discovery, ligand-based virtual screening can be a particularly useful strategy for the exploration of antichlamydial compounds. Here, we present as well as offer an in vitro validation of a novel, successful alternative to antichlamydial drug discovery, where we introduce the term “flipped strategy” to describe a targetindependent approach for finding new natural product lead candidates. The concept focuses on using ligand-based virtual

screening of focused natural product libraries combined with phenotypic assays. A summary of the strategy is shown in Table 1.



RESULTS AND DISCUSSION ChemGPS-NP, a chemography concept developed by Backlund and collaborators, provides a strategy to allow exploration of the entire biologically relevant chemical space, particularly suited to natural compounds.16 ChemGPS-NP, freely available online, is based on the analysis of eight principal components (PCs) (dimensions) that describe relevant physicochemical properties such as size, shape, aromaticity, lipophilicity, polarity, flexibility, and hydrogen bond capacity for a reference set of compounds. Using ChemGPS-NP, compounds are mapped into the chemical space using interpolation based on PC analysis score prediction.17,18 As reference antichlamydial compounds we searched the literature for plant-derived compounds that have been shown to have a minimum inhibitory concentration (MIC) value of 50 μM or lower on C. pneumoniae clinical isolate K7 (Kajaani 7). Compounds were selected that have been assayed using the same methods planned to be used in the in vitro experiments to permit comparison of the results gained with the reference compounds and the compounds in the library. A total of 19 reference compounds were identified, belonging to various chemical classes: gallates, coumarins, flavonoids, and lignans (Table S1, Supporting Information).19−21 The target library used for virtual screening consisted of 502 compounds, containing alkaloids, coumarins, flavones, isoflavones, macrolides, peptolides, terpenoids, and synthesized derivatives of some compounds of these groups. The chemical space analysis was made from two-dimensional descriptors (total of 35) that describe physical-chemical properties of the compounds and were calculated from their canonical SMILES. All salts, hydration information, and counterions were excluded from the SMILES annotation, and differences in stereochemistry ignored, since ChemGPS-NP only uses two-dimensional descriptors. Analysis was focused on the first four dimensions of ChemGPS-NP (PC1−4), as it has been shown that they are the most significant and explain 77% of the variance.18,22 PC1 consists of descriptors for size, shape, and polarizability, PC2 represents aromatic- and conjugationrelated properties, PC3 corresponds to lipophilicity, polarity, and H-bond capacity, and PC4 describes flexibility and rigidity. Through interpolation from the ChemGPS-NP descriptors, locations of novel compounds can be defined on this map. The 19 reference compounds (blue dots, Figure 1), despite their chemical diversity, all share relatively small molecular sizes (PC1) and high aromaticity (PC2), and they were predominantly rigid (PC4). However, they were distributed equally 2603

DOI: 10.1021/acs.jnatprod.6b01052 J. Nat. Prod. 2017, 80, 2602−2608

Journal of Natural Products

Article

the three closest compounds to each of the reference antichlamydial compounds. The computed EDs between the reference compounds and the library ranged from 0 to 17.73, while EDs of virtual hits varied between 0.08 and 1.4. These distances are graphically presented as a heatmap (Figure 2). Hit compounds were excluded in the following cases: (1) compounds coinciding with the references also present in the target library; (2) overlapping hit compounds that were closed to more than one reference; and (3) one previously known antichlamydial compound, betulin.23 Next, the resulting 53 virtual hits were assayed in vitro for cytotoxicity on the epithelial cells applied for C. pneumoniae infections (HL, human line epithelial cells of respiratory origin) using the typical C. pneumoniae life cycle (72 h) as the exposure time (Table 2.). Based on this, 27 compounds were excluded from further in vitro studies as cytotoxic. In the rest of the cases (26 compounds) the HL cell viability was maintained over 75%, and these were studied for antichlamydial activity in vitro. In view of the new approach to evaluate chemical compounds as antichlamydials, the results were compared to those emanating from a whole-cell-based assay. Thus, the 26 selected virtual hits were assayed against C. pneumoniae clinical isolate K7 at 50 μM concentration in HL cells. This yielded six compounds (“leads”) showing ≥50% C. pneumoniae growth inhibition, which were identified: jervine, mycophenolic acid, manool, peganole, 6-acetamino-6-deoxycastanospermine, and catharanthine base (Table 2.) The six lead compounds were screened for pan assay interference compounds (PAINs)24 and found to be free of functional groups listed as the most frequent hitters. Peganole was a borderline case, as it is structurally close

Figure 1. ChemGPS-NP-based analysis of the chemical space populated by the antichlamydial reference compounds (blue dots), the tested library (light green dots), and the identified lead compounds (yellow dots). PC1 (principal component 1) describes the size, shape, and polarizability; PC2, aromatic- and conjugation-related properties. (A) PC3 represents lipophilicity, polarity, and H-bond capacity (increases with lipophilicity). (B) PC4 describes flexibility and rigidity (increases with flexibility). The value of a PC increases in the direction of the arrow. Mycophenolic acid, MPA, is indicated with a red arrow.

among lipophilic and hydrophilic regions (PC3). In contrast, the selected target library for exploration (green dots, Figure 1) was widely distributed across the chemical space, including large molecules as well as rigid structures, which typically populate natural product collections. A similarity-based search on the target library was performed via the calculation of Euclidean distances (EDs) between the 502 library compounds and each of the 19 reference compounds (Figure 2.). Virtual hits were selected to include

Figure 2. Heatmap of the 502 natural product library compounds (horizontally aligned rows) vs 19 antichlamydial reference compounds (vertically aligned columns). Intensity of red color describes closer proximity in the Euclidean four-dimensional space (see Color Key). The dendrogram illustrates the arrangement of the clusters produced by the complete linkage clustering method. The six lead compounds are highlighted in the figure. The calculated EDs are presented in Table S2 in the Supporting Information. 2604

DOI: 10.1021/acs.jnatprod.6b01052 J. Nat. Prod. 2017, 80, 2602−2608

Journal of Natural Products

Article

Table 2. List of Virtual Hits Identified in the Ligand-Based Virtual Screening and Chlamydia Inhibitory Activity (%) Shown for Virtual Hits That Were Deemed Nontoxic (26), Showing Host Cell Viability ≥75% at 50 μM Concentration compound 6-acetamino-6-deoxycastanospermine aphidicolin apigenin-7-O-β-Dglucopyranoside (−)-asarinin baccatin III caffeic acid phenethyl ester caryophylleneoxide catharanthine base cevadine chartreusin chrysine cinobufagin convolvamine HCl coumestrol cycloheximide deguelin dehydrokawain digitoxin dihydroergocristine mesylate dipterocarpol epicathecin eriocitrin flavanomarein gossypol graveoline harmaline HCl hesperidin

HL cell viability % (±SEM %)

chlamydial inhibition % (±SEM %)

86.6 ± 2.0

59.8 ± 7.1

92.3 ± 4.8 48.1 ± 2.2

−3.8 ± 8.3

57.1 ± 1.3 34.6 ± 10.2 46.0 ± 4.1 106.9 ± 0.8 88.9 ± 2.8 98.6 ± 4.8 4.0 ± 9.8 cytotoxica 0.2 ± 0.7 78.8 ± 3.1 96.3 ± 5.0 66.4 ± 4.5 21.6 ± 4.9 89.8 ± 1.7 0.1 ± 0.9 2.7 ± 4.2 32.1 ± 2.2 93.4 ± 5.7 75.2 ± 6.1 82.6 ± 3.4 0.1 ± 1.1 100.4 ± 3.2 87.3 ± 5.3 100.6 ± 1.3

compound honokiol 7-hydroxyflavone hypocrellin A indirubin isoquercitrin jervine laudanosoline HBr leucomisine lupinine manool mycophenolic acid nalidixic acid nonactin norleaginine oleanolic acid oridonin osthole peganole pimaricin podocarpic acid prostaglandin B1 radicicol remerine HCl syrosingopine taxifolin vasicine

35.7 ± 4.9 68.1 ± 4.9 39.0 ± 11.2

21.1 ± 7.5 0.0 ± 7.7

35.7 ± 7.4

41.7 ± 6.6 8.6 ± 9.9 22.1 ± 8.6

HL cell viability % (±SEM %) 8.8 ± 16.7 71.1 ± 7.2 −0.3 ± 1.6 92.5 ± 2.9 95.8 ± 3.2 89.3 ± 3.8 72.8 ± 11.8 93.9 ± 3.5 98.7 ± 3.5 88.6 ± 3.3 86.6 ± 2.1 15.8 ± 5.4 23.7 ± 1.2 84.8 ± 5.0 36.3 ± 2.6 0.0 ± 0.6 cytotoxica 95.2 ± 4.1 67.1 ± 5.5 27.9 ± 6.3 109.4 ± 1.5 0.6 ± 1.9 45.2 ± 6.3 67.6 ± 4.3 95.2 ± 4.3 102.2 ± 1.6

chlamydial inhibition % (±SEM %)

2.0 ± 8.0 38.0 ± 4.3 74.8 ± 5.4 30.1 28.2 60.0 98.0

± ± ± ±

6.5 7.1 20.8 1.3

4.9 ± 6.4

66.2 ± 5.4

39.8 ± 8.4

13.4 ± 4.4 −1.4 ± 8.6

a

5.3 ± 7.9 26.0 ± 5.2 26.8 ± 5.3

Chrysine and osthole showed cytotoxicity in the Chlamydia inhibition assay.

to the PAIN functional group and might therefore be flagged as problematic in HTS screening assays. However, peganole was one of the less active compounds among the six leads. Potency was measured for the six lead compounds (Table 3 and Figures S1 and S2, Supporting Information). The potency (expressed as the 50% inhibitory concentration, IC50 values) for these compounds was found to be in a range of 310 nM to 67.2 μM. This level of inhibition can be considered relevant for inhibitors of C. pneumoniae19−21,23,25,26 particularly in the case of the most potent compound identified in the screening, mycophenolic acid. These lead compounds belong to the following chemical groups: steroidal alkaloids (jervine), phenolic acids (mycophenolic acid), diterpenoids (manool), quinazoles (peganole), indolizine alkaloids (6-acetoamido-6-castanospermine), and terpenoidal indole alkaloids (catharanthine base). Thus, the identified leads do not fall into the same chemical categories as the reference compounds, but represent entirely new chemotypes. Moreover, none of these lead compounds has thus far been reported as antichlamydial. Of note, a simple structure− activity relationship screen from the reference compounds using the same chemical groups would not have identified these leads. Four of the lead compounds, jervine, manool, mycophenolic acid, and peganole, showed full inhibition of the growth of C. pneumoniae, although peganole showed rather poor activity at the lower concentrations (Figures S1 and S2, Supporting Information).

Manool and mycophenolic acid have been reported to show antimicrobial activities,26,27 while the four other leads have not previously been identified as antibacterial agents. However, all of them have been reported to interact with eukaryotic cell signaling pathways,27−31 which might indicate that the observed antichlamydial effects are an outcome of the impact of the compound on the HL cells in a manner unfavorable to C. pneumoniae survival. It is, in fact, not clear from the current data whether the identified lead compounds or the compounds in the reference set share a similar mode of antichlamydial action or whether the identified antichlamydial hot spot in chemical space rather reflects physicochemical properties necessary for physically reaching the cellular microenvironment susceptible for the antichlamydial action. The potency (50% inhibitory concentration) value of mycophenolic acid was in the high nanomolar range. With a calculated MIC90 of 1 μM, mycophenolic acid has a higher potency against C. pneumoniae than ofloxacin, for which an MIC90 value of 1.38 μM has been reported with the same strain of C. pneumoniae, K7, and the same detection method.32 Mycophenolic acid has been found to display antimicrobial activities,26 and it has been approved as an immunosuppressant drug to prevent organ transplant rejection. It is used as a prodrug, mycophenolate, which is metabolized to mycophenolic acid, the active drug. The immunosuppressive properties of mycophenolic acid are likely to limit its development as an antichlamydial agent, but identification of such a highly potent compound despite the relatively small size of the target library 2605

DOI: 10.1021/acs.jnatprod.6b01052 J. Nat. Prod. 2017, 80, 2602−2608

Journal of Natural Products

Article

Table 3. Potency Values for the Lead Compounds, Expressed as IC50

a

In the potency assays HL cells were infected with C. pneumoniae strain K7 at a multiplicity of infection (MOI) of 0.3. Lead compounds were assayed at different concentrations, covering from 1 nM to 100 μM (5 log scale). Potencies were estimated after sigmoidal fitting of the results, using SPSS 23.0 (IBM Corporation, USA) (n = 3). Azithromycin at 20 nM concentration was used as a positive control.

successful use of ligand-based virtual screening for identifying novel inhibitors of C. pneumoniae, and it thus represents a major leap forward among technologies for developing drugs against this challenging microorganism. On the basis of the results obtained here, it is expected that in the future more virtual screening efforts could pave the way for the discovery of new antichlamydial leads. In vitro activity testing proved that the compounds closest to the references in the chemical space contained six new leads, and mycophenolic acid had a relevant potency better than even currently used antibiotics (ofloxacin). Overall, a 1.2% hit rate was achieved during the in silico lead mining process, which can be regarded as successful and further confirms that ligand-based virtual screening is efficient in finding active molecules. One of our recent drug discovery projects, using engineered antibiotic biosynthesis pathways on more readily approached bacteria (Staphylococcus spp.), did not give improved potency for the evaluated natural product library compounds.37 Thus, the aforementioned strategy lays the foundation for a drug discovery rationale that yields superior results and can open entirely new venues in antibacterial research. The rise of modern phenotypic drug discovery together with relevant antimicrobial models to guide early drug development of pharmacotherapy represents one paradigm shift in the field. Thus, we have aimed at triggering a positive

demonstrates the potential of the approach proposed here. The reference ligand to which mycophenolic acid was closest (in terms of ED) is a flavonoid, rhamnetin. The chemical space populated around rhamnetin would be worth exploring in further discovery projects. Virtual screening is an efficient traditional strategy for quick evaluation of large libraries of compounds that permits a focus on the supposedly best candidates, reducing the amount of in vitro and in vivo experiments. It has become an integral part of the drug discovery process, with proven value in several therapeutic areas. Ligand-based methods are based on the idea that similar compounds have similar physicochemical properties and, thus, similar biological activities.33,34 Therefore, the selection of reference compounds is vital for the success of a virtual screen.35 The 19 antichlamydial reference structures selected for this study are from different groups of chemical compounds in order to find a wider area in chemical space where the possible lead candidates can be found. In this case, a scatter plot of the reference compounds shows that they share relevant physicochemical properties (Figure 1), which further suggests this to be a “privileged” space for antichlamydial natural compounds. While one earlier study has reported the use of a target-based homologue modeling strategy for antichlamydial compound identification,36 the current work is the first report on the 2606

DOI: 10.1021/acs.jnatprod.6b01052 J. Nat. Prod. 2017, 80, 2602−2608

Journal of Natural Products

Article

was used, due to its permissiveness toward the infection.39 Another reason for choosing the HL cell line as the host cells was that they were used for assaying the compounds in the reference set, and the aim was to keep the antichlamydial assay conditions similar to those carried out in the antichlamydial assay of the reference compounds. In brief, HL cells were grown to confluence, incubated overnight at 37 °C, 5% CO2, and 95% humidity, and infected with C. pneumoniae K7 stock in cell culture medium supplemented with 1 μg/mL cycloheximide using a multiplicity of infection (MOI) of 0.3. The infected cells were centrifuged at 500 rpm for 1 h, and incubated at 37 °C for 1 h. After 2 h of infection, the inocula were replaced with cell culture medium supplemented with 1 μg/mL cycloheximide and containing the natural compounds at 50 μM or vehicle (DMSO) control. The plates were incubated for 70 h at 37 °C and washed with sterile phosphate-buffered saline. The C. pneumoniae inclusions and host cells were stained with Pathfinder Chlamydia culture confirmation system (BioRad, US). The coverslips were examined under a fluorescent microscope (Evos FL imaging system, Thermo Scientific, USA) using a 20× objective to count the number of chlamydial inclusions per eye field from four eye fields per coverslip. The results are expressed as mean inhibitory percentages, calculated by comparing the inclusion counts of each sample with average inclusion counts in the vehicletreated sample. The antichlamydial potency assays for the six lead compounds were done at a concentration scale from 1 nM to 100 μM (12 different concentrations covering five log-units), three times. Azithromycin at 20 nM concentration was used as a positive control for chlamydia inhibition in all antichlamydial assays. Data Analysis and Processing. Heatmaps were built using Gnuplot (version 4.5), which is a portable command-line-driven graphing utility. The mapping of the chemical space in the 3D space was done using PyMOL 1.7.0.0. Potency values (IC50) were calculated from sigmoidal fittings of the concentration−response curves (variable slope) using SPSS 23.0 (IBM Corporation, USA).

inflection point by using ligand-based design and phenotypic screening in an iterative manner. In conclusion, completely new natural chemotypes were identified as new antichlamydial compounds, which would not have been achieved if only conventional structure−activity relationship studies were conducted of the known chemical classes. In particular, identification of mycophenolic acid, with antichlamydial potency in the nanomolar range, illustrates the relevance of the identified area within the natural product chemical space as a source of drug-like molecules with high antichlamydial potency. An additional novelty of this work is that ChemGPS-NP was, for the first time, applied during the full cycle of the in silico to in vitro screening process. Brunhofer and collaborators earlier used a conservative but rather tedious approach to cluster a natural product library by chemical groups.37 A “flipped strategy” is demonstrated here to discover lead compounds, exemplified by a target library of natural products. In this model, the close proximity of compounds to the references very well predicted the biological activity as chlamydial inhibition. The chemical space for these antichlamydial natural compounds shown is worth continuing as a strategy for future drug development projects.



EXPERIMENTAL SECTION

Space Analysis. Chemical space analysis was performed using the principal component analysis (PCA)-based chemical space navigation tool ChemGPS-NP,16,18 which is freely available online at http://www. chemgps.bmc.uu.se/. Mapping of the chemical space was done in a 3D space, using PyMOL 1.7.0.0. The target library (Screen-Well natural product library version 7.1, cat. no. BML-2865-0500) was purchased from Enzo Life Sciences, Inc. (Enzo Biochem, Inc., Farmingdale, NY, USA) and contained compounds as 2 mg/mL stocks in DMSO and were used at 50 μM final concentration during the screening. Selection of Reference Compounds. The reference set that was used for the similarity search based on 2D descriptors was constructed of 19 known antichlamydial compounds with MIC values of 50 μM or lower, measured against C. pneumoniae K7 (Kajaani 7, clinical isolate) (Table S1, Supporting Information). Ligand-Based Virtual Screening. Euclidean distances were computed between the reference compounds (19) and the natural product library compounds (504). EDs were calculated between points P = (p1, p2, ..., p4) and Q = (q1, q2, ..., q4) in Euclidean 4D space provided by the ChemGPS-NP coordinates: ED =



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jnatprod.6b01052. Experimental procedures, chlamydial inhibition and host cell viability data, information about reference and lead compounds (PDF)



AUTHOR INFORMATION

Corresponding Author

(p1 − q1)2 + (p2 − q2)2 + (p3 − q3)2 + (p4 − q4)2

*Phone: +358 44 2834933. E-mail: adyary.fallarero@helsinki.fi. ORCID

Cell Viability Assays. The assay procedure for HL cell viability was described by Hakala et al., 2015.19 In brief, human epithelial HL cell (C. pneumoniae host cells) viabilities in the presence of the 53 virtual hit compounds were tested growing them overnight to confluence in 96-well plates in four parallel replicates. Usnic acid (Serva Electrophoresis GmbH, Germany) at 50 μM concentration served as a positive control. Viability percentages were compared to control cells grown with 0.25% DMSO supplemented growth medium. The cells were incubated at 37 °C for 72 h, which corresponds to the in vitro acute infection cycle of C. pneumoniae. After 2 h of exposure with 20 μM resazurin (Sigma-Aldrich, St. Louis, MO, USA) in PBS the fluorescence emitted by the reduced resazurin was measured with a Varioskan LUX multimode plate reader (Thermo Scientific, Vantaa, Finland) at 570/590 nm (ex/em). In Vitro Antichlamydial Assays. The infections were carried out using C. pneumoniae strain K7 (Kajaani 7), a clinical strain isolated from a young man suffering from pneumonia in Finland.38 The bacterium was grown and handled and the antichlamydial assays were carried out as described earlier.19,20 For growing the intracellular bacterium C. pneumoniae, a standard cell line for C. pneumoniae infections, HL cells (human line, epithelial cells of respiratory origin),

Leena Hanski: 0000-0002-3121-8542 Adyary Fallarero: 0000-0003-4127-5476 Notes

The authors declare no competing financial interest.

■ ■ ■

ACKNOWLEDGMENTS The authors thank the Academy of Finland (grants 128870, 252216, 272266) for financial support. DEDICATION This article is dedicated to the memory of Prof. Pia Vuorela (1961−2017). §

REFERENCES

(1) Blasi, F.; Tarsia, P.; Aliberti, S. Clin. Microbiol. Infect. 2009, 15, 29−35. (2) Schoborg, R. Microbes Infect. 2011, 13, 649−662.

2607

DOI: 10.1021/acs.jnatprod.6b01052 J. Nat. Prod. 2017, 80, 2602−2608

Journal of Natural Products

Article

(3) Kohlhoff, S. A.; Hammerschlag, M. R. Expert Opin. Pharmacother. 2015, 16, 205−212. (4) Hanski, L.; Vuorela, P. M. Expert Opin. Drug Discovery 2014, 9, 791−802. (5) Wolf, K.; Fischer, E.; Hackstadt, T. Infect. Immunolog 2000, 68, 2379−2385. (6) Cocchiaro, J. L.; Valdivia, R. H. Cell. Microbiol. 2009, 11, 1571− 1578. (7) Gérard, H. C.; Mishra, M. K.; Mao, G.; Wang, S.; Hali, M.; Whittum-Hudson, J. A.; Kannan, R. M.; Hudson, A. P. Nanomedicine 2013, 9, 996−1008. (8) Kohlhoff, S. A.; Hammerschlag, M. R. Expert Opin. Pharmacother. 2015, 16, 205−212. (9) Seidman, J. C.; Coles, C. L.; Silbergeld, E. K.; Levens, J.; Mkocha, H.; Johnson, L. B.; Muñoz, B.; West, S. K. Int. J. Epidemiol 2014, 43, 1105−1113. (10) Coles, C. L.; Mabula, K.; Seidman, J. C.; Levens, J.; Mkocha, H.; Muñoz, B.; Mfinanga, S. G.; West, S. Clin. Infect. Dis. 2013, 56, 1519− 1526. (11) Keller, T. H.; Shi, P.; Wang, Q. Curr. Opin. Chem. Biol. 2011, 15, 529−530. (12) Nierode, G.; Kwon, P. S.; Dordick, J. S.; Kwon, S. J. Microbiol. Biotechnol. 2015, 26, 213. (13) Dobi, K.; Hajdú, I.; Flachner, B.; Fabó, G.; Szaszkó, M.; Bognár, M.; Magyar, C.; Simon, I.; Szisz, D.; Lő rincz, Z.; Cseh, S.; Dormán, G. Molecules 2014, 19, 7008−7039. (14) Bleicher, K. H.; Böhm, H.; Müller, K.; Alanine, A. I. Nat. Rev. Drug Discovery 2003, 2, 369−378. (15) Newman, D. J.; Cragg, G. M. J. Nat. Prod. 2016, 79, 629−661. (16) Rosén, J.; Lövgren, A.; Kogej, T.; Muresan, S.; Gottfries, J.; Backlund, A. J. Comput.-Aided Mol. Des. 2009, 23, 253−259. (17) Oprea, T. I.; Gottfries, J. J. Comb. Chem. 2001, 3, 157−166. (18) Larsson, J.; Gottfries, J.; Muresan, S.; Backlund, A. J. Nat. Prod. 2007, 70, 789−794. (19) Hakala, E.; Hanski, L.; Uvell, H.; Yrjönen, T.; Vuorela, H.; Elofsson, M.; Vuorela, P. M. J. Antibiot. 2015, 68, 609−614. (20) Hanski, L.; Genina, N.; Uvell, H.; Malinovskaja, K.; Gylfe, Å.; Laaksonen, T.; Kolakovic, R.; Mäkilä, E.; Salonen, J.; Hirvonen, J.; Elofsson, M.; Sandler, N.; Vuorela, P. M. PLoS One 2014, 9, e115115. (21) Alvesalo, J.; Vuorela, H.; Tammela, P.; Leinonen, M.; Saikku, P. Biochem. Pharmacol. 2006, 71, 735−741. (22) Rosén, J.; Gottfries, J.; Muresan, S.; Backlund, A.; Oprea, T. I. J. Med. Chem. 2009, 52, 1953−1962. (23) Salin, O.; Alakurtti, S.; Pohjala, L.; Siiskonen, A.; Maass, V.; Maass, M.; Yli-Kauhaluoma, J.; Vuorela, P. Biochem. Pharmacol. 2010, 80, 1141−1151. (24) Baell, J. B.; Holloway, G. A. J. Med. Chem. 2010, 53, 2719−2740. (25) Hanski, L.; Ausbacher, D.; Tiirola, T. M.; Strøm, M. B.; Vuorela, P. M. PLoS One 2016, 11, e0157306. (26) Souza, A. B.; de Souza, M. G.; Moreira, M. A.; Moreira, M. R.; Furtado, N. A.; Martins, C. H.; Bastos, J. K.; dos Santos, R. A.; Heleno, V. C.; Ambrosio, S. R.; Veneziani, R. C. Molecules 2011, 16, 9611− 9619. (27) Regueira, T. B.; Kildegaard, K. R.; Hansen, B. G.; Mortensen, U. H.; Hertweck, C.; Nielsen. J. Appl. Environ. Microbiol. 2011, 77, 3035− 3043. (28) Ghezali, L.; Leger, D. Y.; Limami, Y.; Cook-Moreau, J.; Beneytout, J. L.; Liagre, B. Exp. Cell Res. 2013, 319, 1043−1053. (29) Jadhav, A.; Liang, W.; Papageorgiou, P. C.; Shoker, A.; Kanthan, S. C.; Balsevich, J.; Levy, A. S.; Heximer, S.; Backx, P. H.; Gopalakrishnan, V. J. Pharmacol. Exp. Ther. 2013, 345, 383−392. (30) de Oliveira, P. F.; Munari, C. C.; Nicolella, H. D.; Veneziani, R. C.; Tavares, D. C. Cytotechnology 2016, 68, 2139−2143. (31) Michael, J. P. Alkaloids Chem. Biol. 2016, 75, 1−498. (32) Tammela, P.; Alvesalo, J.; Riihimäki, L.; Airenne, S.; Leinonen, M.; Hurskainen, P.; Enkvist, K.; Vuorela, P. Anal. Biochem. 2004, 333, 39−48. (33) Johnson, M.; Lajiness, M.; Maggiora, G. Prog. Clin. Biol. Res. 1989, 291, 167−171.

(34) Patterson, D. E.; Cramer, R. D.; Ferguson, A. M.; Clark, R. D.; Weinberger, L. E. J. Med. Chem. 1996, 39, 3049−3059. (35) Kirchmair, J.; Distinto, S.; Markt, P.; Schuster, D.; Spitzer, G. M.; Liedl, K. R.; Wolber, G. J. Chem. Inf. Model. 2009, 49, 678−692. (36) Alvesalo, J. K.; Siiskonen, A.; Vainio, M. J.; Tammela, P. S.; Vuorela, P. M. J. Med. Chem. 2006, 49, 2353−2356. (37) Brunhofer, G.; Fallarero, A.; Karlsson, D.; Batista-Gonzalez, A.; Mohan, C. G.; Vuorela, P. Bioorg. Med. Chem. 2012, 20, 6669−6679. (38) Ekman, M. R.; Grayston, J. T.; Visakorpi, R.; Kleemola, M.; Kuo, C. C.; Saikku, P. Clin. Infect. Dis. 1993, 17, 420−425. (39) Kuo, C.; Grayston, J. T. J. Infect. Dis. 1990, 162, 755−758.

2608

DOI: 10.1021/acs.jnatprod.6b01052 J. Nat. Prod. 2017, 80, 2602−2608