The Chemical Characterization of Nigerian Propolis samples and Their Activity Against Trypanosoma brucei

Profiling of extracts from twelve propolis samples collected from eight regions in Nigeria was carried out using high performance liquid chromatography (LC) coupled with evaporative light scattering (ELSD), ultraviolet detection (UV) and mass spectrometry (MS), gas chromatography mass spectrometry (GC-MS) and nuclear magnetic resonance spectroscopy (NMR). Principal component analysis (PCA) of the processed LC-MS data demonstrated the varying chemical composition of the samples. Most of the samples were active against Trypanosoma b. brucei with the highest activity being in the samples from Southern Nigeria. The more active samples were fractionated in order to isolate the component(s) responsible for their activity using medium pressure liquid chromatography (MPLC). Three xanthones, 1,3,7-trihydroxy-2,8-di-(3-methylbut-2-enyl)xanthone, 1,3,7-trihydroxy-4,8-di-(3-methylbut-2-enyl)xanthone a previously undescribed xanthone and three triterpenes: ambonic acid, mangiferonic acid and a mixture of α-amyrin with mangiferonic acid (1:3) were isolated and characterised by NMR and LC-MS. These compounds all displayed strong inhibitory activity against T.b. brucei but none of them had higher activity than the crude extracts. Partial least squares (PLS) modelling of the anti-trypanosomal activity of the sample extracts using the LC-MS data indicated that high activity in the extracts, as judged from LCMS2 data, could be correlated to denticulatain isomers in the extracts.

exact geographical origin. In this current paper we report the proiling and testing of propolis samples from different regions within Nigeria, which are considered to be within a wet savannah environment. Besides chemical characterization of these samples, we focused on activity against one particular infective organism which thrives in this environment and that is the protozoan parasite that is the causative agent of African Sleeping Sickness, Trypanosoma brucei. Trypanosma species are a health threat to cattle grazing in the savannah areas, as well as to humans. he existing drugs for treating either the human 8 or veterinary disease 9 are old, in many cases quite toxic, and oten inefective because of drug resistance 10 . Moreover, the rate of drug development against this disease is very slow. he vector for T. brucei is the tsetse ly, where the parasite initially colonises its gastro-intestinal tract. Members of the protozoal genus Crithidia, which are very closely related to T. brucei, having a similar life cycle and metabolism, similarly infect bees 11 and have been mentioned as a possible cause of winter colony collapse in Europe 12 .

Results and Discussion
The Origin of the Nigerian Propolis Samples. Figure 1 and Table 1 show details of the origins of the Nigerian propolis samples.
Bioassay of crude extracts from Nigerian Propolis. he crude ethanolic extracts from the Nigerian propolis samples were tested against T. brucei and C. fasciculata. he highest activities were in the samples from Southern Nigeria but all samples displayed some degree of activity apart from sample AF23N ( Table 2). he activity against C. fasciculata supports the idea that the bees may collect propolis to protect from infection by Crithidia species which are known to be pathogens of bees 12 and are quite closely related to T. brucei. Indeed there was a surprisingly good correlation between the EC 50 values for the T. brucei control strain and for Crithidia (see Supplementary Figure S1), which further supports the idea that propolis provides a natural defence against Crithidia and could be exploited for treatments against other trypanosomatids. Moreover, the extracts were active against both the standard strain of T. brucei and two multi-drug resistant clones that displayed a high degree of resistance against pentamidine. It has been previously shown that clone B48, in particular, has very high levels of resistance to all diamidine drugs including pentamidine and diminazene aceturate, as well as the entire class of melaminophenyl arsenical drugs that includes melarsoprol and cymelarsan 13 -i.e. most of the irst-line human and veterinary treatments. Although slight diferences (less than 2-fold sensitisation or resistance) were observed between the control and resistant strains these were very minor compared to the 225-fold or 31-fold resistance to pentamidine observed in parallel with clones B48 and aqp2/aqp3 null, respectively. he crude ethanol extracts were all complex mixtures and either individual components in the mixtures or isolated components might display higher activity.
Profiling of Samples with HPLC-UV-ELSD. The chromatograms of the crude samples run on HPLC-UV-ELSD suggested a wide diversity in the chemical composition for the Nigerian samples. he samples could be divided into two groups, group I comprised of samples that were collected from the central part of Nigeria (AF2N3, P3N, KKN, S97N, S95N, KAPN, S96N) which mostly demonstrated an intense ELSD only response, with weak or absent of UV peaks suggesting a high content of terpenoids and/or fats, and the absence of any chromophore containing compounds such as lavonoids or lignans or other phenolic compounds. All of these samples were also noted to have at most weak activity against trypanosomes. In contrast, samples in group II were collected from Southern areas and showed strong UV-ELSD responses and high activity against T. brucei. Figure S1 shows a comparison of the chromatograms obtained for a Southern and a Central Nigerian sample. he mass spectra of these peaks were searched against the NIST library and were identiied as various triterpenoids, waxes and long chain fatty acids, with similarity scores of more than 700 which appeared to be the major chemical components in these samples and this explained the high responses in ELSD only. Table S1 shows some of the compounds identiied by GC-MS in the group I samples according to matching against the NIST library. Most of the components in the group I samples had retention times >40 min.
In the case of group II samples, (from Rivers State, Nigerian) i.e. BRN and RSN, most of the GC-MS peaks eluted before 35 min, while for the ION sample the retention times were mainly less than 20 min. Figure S2 shows a comparison of the GC-MS chromatograms for three representative samples. Menadione

LC-MS Analysis of Crude Extracts from Nigerian Propolis Samples. All samples were run in dupli-
cate in order to verify the precision and reproducibility of the instrument. he data collected from the LC-MS were complex and diicult to process manually, and were split into positive and negative ion data, and then processed by MZ-mine 2.14 14 as described recently by Siheri et al. 7 , and the extracted features were then searched against the Dictionary of Natural Products 15 database. he 2000 most intense features with the highest mean peak areas across the 12 samples, selected by m/z mine from the negative ion data, were used to build a Principal Components Analysis (PCA) model (Fig. 2). he data was univariate scaled and log transformed prior to PCA modelling which was carried out by using Simca P 14.0. Hierarchical cluster analysis (HCA) was used to divide the samples into 3 groups. he score plot shows that most of the samples collected from central Nigeria are clustered near the centre due to the absence of LC-HRMS features with a high MS response. he compositions of the rest of samples, even for samples that were collected from the same area such as BRN and RSN (both of which were collected from Rivers State, Nigeria), were not identical. Samples UDN and CCN were also from the South of the country and were intermediate in composition between the samples from central Nigeria and those from Rivers State. he ION samples lay slightly outside the ellipse demonstrating a more unique chemical composition; from the loading plot characteristic chemical compounds responsible for variation among samples could be assigned exclusively to single chemical formulas within a 3 ppm mass error window. Each elemental composition generally corresponded to a large number of isomers within the DNP database.  O 5 . hese components were targeted and isolation and to gave three compounds all of which were xanthones. Two of the xanthones, 1 and 2 ( Figure S3) have been isolated previously from Cudrania cochinchinensis and Cudrania tricuspidata and the NMR spectra of compound 1 and 2 isolated from UDN matched the spectral data for gerontoxanthone H and 6-deoxv-γ-mangostin in literature 16,17 . he mass spectral data and NMR data for 1 and 2 are given in S1, Tables S2 and S3. . Four methyl groups appeared as doublets (J = 1.3) at δ 1.68, 1.71, 1.78, and 1.74. Analysis of its HMQC ( 1 J) and HMBC ( 2 J and 3 J) correlations enabled assignments of the proton and carbon signals as well as the positions of the two prenyl and two hydroxy groups as follows: correlations from the chelated -OH at C-1 identiied C-1 (δ C 162.5), C-2 (δ C 97.9) and C-9a (δ C 103.9). he methyl groups at δ H 1.68 (H-14) and 1.71 (H-15) showed correlations to C-12 (δ C 124.0) while the other two methyl groups at δ 1.78 (H-19) and 1.74 (H-20) showed correlations to C-17 (δ C 119.6). he prenyloxy group must be at C-3 based on H-2, H-4 and H-16 correlations to C-3 and the non-oxygenated prenyl group must be at C-8 from correlations of H-12 and H-6 to C-8. Hence the two hydroxyl groups must be at C-1 and C-7. Comparing its NMR data to literature reports and that for calothwaiteaixanthone 18 the compound was identiied as 1,7-dihydroxy-8-(3-methylbut-2-enyl)-3-(methylbut-2-enyloxy) xanthone (Fig. 3). he structure was further conirmed by its mass ions obtained from HRESI-MS n as follows: MS 2 (% intensity) 310.0844 (100) C 18    Closely related xanthones were found for the irst time in propolis collected by Tetragonula laeviceps stingless bees from hailand. Garcinia mangostana (Mangosteen) was proposed as the most probable plant source 16 .

Structure elucidation of Compounds puri ed from Southern Nigeria propolis samples
Fractionation of the ION sample resulted in the isolation of two cyclortanes ambonic 4 and mangiferonic acid 5 and a fraction containing α-amyrin 6 ( Figure S4) with mangiferonic acid. he compounds were all identiied by comparison with literature NMR data 19,20 . he NMR data are summarised in Tables S4-S6. hese kind of compounds were isolated before from Cameroonian propolis, and mango (M. indica, Anacardiaceae) seemed to be the plant source of the resin 19 . Mango is widely used in honey production in Cameroon and throughout tropical     19 . α Amyrin and Mangiferonic were also found in Brazillian propolis 20 , and the latter has been isolated from Myanmar propolis and showed strong cytotoxic properties 21 .
Bioassay results For the Isolated Compounds. Table 4 shows the bioassay results for the compounds isolated from the ION and UDN propolis samples. Only xanthone 1 was more active than the crude ION extract and the range of activity against T. brucei was dependent on the position of the prenyl substituent. Only α-amyrin was more active than the crude UDN sample. Similar to the crude samples, the puriied compounds displayed only marginally different activities against the two drug resistant strains. Thus, the classes of compounds described herein would not be cross-resistant with the main drugs currently in use against human and animal African trypanosomiasis.

Prediction of the most active compounds in the crude propolis extracts from LC-MS data. he
extracted LC-MS data was used to build a Partial Least Squares (PLS) model for predicted against measured EC 50 values against T. brucei (Fig. 4). Two of the low activity samples AF23N and KKN were omitted from the model. he model was edited to remove variables having less impact in the prediction of anti-trypanosomal activity and Fig. 4 is based on 180 features. When the loadings plot for the model ( Figure S5) was examined it was clear that the compounds predicted to be most responsible for anti-trypanosomal were isomers of denticulatain, which we observed previously 6 . Most of the isomers had a characteristic fragment ion at m/z 241.05 as described in our previous paper 6 where a detailed elucidation of the MS 2 spectra is given. hese compounds probably are obtained from Macaranga species 22 . Figure 5 shows an extracted ion trace for the denticulatain isomers in four high activity samples indicating that the abundance of these isomers correlates with activity against T. brucei. We were interested in the occurrence of guttiferone isomers, which we had observed in Rivers State propolis samples in our earlier study 5 and speculated that these might be responsible for the high anti-trypanosomal activity of the of the samples since we had previously observed very high activity for a phloroglucinone compound isolated from Cameroonian propolis 23 . However, as can be seen in the extracted ion traces shown in Fig. 6 for guttiferone isomers sample CCN, which has lower anti-trypanosomal activity than sample.  Table 4. Bioassay of compounds isolated from Nigerian propolis against T. brucei (n = 3). EC 50 values are expressed in µM for pentamidine, and as µg/mL for the puriied compounds, and represent the average and SEM of 3 independent experiments. Statistical signiicance was determined using an unpaired two-tailed Student's t-test comparing EC 50 value of the resistant strain with that of the same sample for the control strain s427. RF: resistance factor, being the EC 50 value for the resistant strain divided by the EC50 value for the control (sensitive) strain.

7
SCIENTIFIC REPORTS 923 DOI s RSN, contains much higher levels of these isomers than sample RSN. hus it would appear that levels of denticulatain isomers correlate more closely with observed anti-trypanosomal activity. hus these compounds should be targeted for isolation and testing.

Discussion
The composition of Nigerian propolis samples varies widely according to geographical origin. Two main types were observed which were either rich in UV-absorbing phenolic compounds or contained mainly non-UV-absorbing triterpenoids. he highest anti-trypanosomal activities were displayed by the samples from southern Nigeria which had a high content of phenolic compounds. Six compounds were isolated from two of the Southern Nigerian samples comprised of three xanthones and three triterpenoids. he anti-trypanosomal acitivities of the isolated compounds in most cases were no higher than those of the crude extracts. In order to determine what the most active components in the extracts were PLS modelling was carried out using the high resolution mass spectrometry data for the extracts. he loadings plot for a PLS model pointed towards denticulatain isomers as being most strongly associated with the anti-trypansomal activities and thus a target for isolation and testing.
Propolis sample collection and preparation. he propolis samples were either supplied by BeeVital (Whitby, UK) or collected by Dr John Igoli from various regions of Nigeria (Table 1, Fig. 1). All samples were sticky and brown in colour except samples collected in the Southern regions (RSN, BRN and UDN) which were reddish in colour. For proiling, 50 mg of each was extracted with ethanol (3 × 5 mL) by sonication at 40 °C for three hours each, and then maceration overnight with ethanol (5 mL). Each extract was iltered through ilter paper, and then combined together, the iltered solutions were dried using a low of nitrogen, the amount of residue was measured and the percentage yield was calculated. he ethanolic crude extracts were stored at −20 °C for further experiments.

Pro ling of the Extracts by ELSD and High Resolution Mass Spectrometry
For proiling experiments 2 mg of crude extract was reconstituted in 1 mL of methanol and iltered using a nylon syringe ilter (Fisher Scientiic, pore size 0.45 µm); 10 µL of the iltrate was injected into an Agilent 1100 HPLC system (Agilent Technologies, Germany) to a UV two channel detector (290 and 320 nm) coupled with an Evaporative Light Scattering Detector (ELSD) (model: SEDEX75, Sedere, France). he mobile phase used was water as mobile phase A and acetonitrile (ACN) as B at a low rate of 0.3 mL/min. he gradient elution was programmed as follows: 0-15 min linear gradient from 30% to 50% of B, 15-25 min at 50% of B, 25-40 min linear gradient from 50% to   80% of B, 40-50 min at 80% of B, 50-51 min increasing to 100% of B, 51-59 min at 100% of B with the low rate increasing to 0.5 mL/min and at 61 min back to 30% of B and hold until 70 min. he LC-HRMS analysis was performed on an Accela 600 HPLC system combined with an Exactive (Orbitrap) mass spectrometer from hermo Fisher Scientiic (Bremen, Germany) using the same conditions as for the UV-ELSD analysis but with the addition of 0.1% formic acid to mobile phases A and B. he samples were run in duplicate, the MS detection range was from m/z 100-1500 and scanning was performed under ESI polarity switching mode. he needle voltages were −4.0 kV, 4.5 kV positive and the sheath and auxiliary gases were set at 50 and 17 arbitrary units, respectively. he data obtained were split into positive and negative ions and the 'negative' dataset was processed using MZMine 2.14, with the masses selected between m/z 100-1200. In all experiments the column used was an ACE C18 column (150 × 3 mm, 3 µm particle size) (HiChrom, Reading UK). he Xcalibur 2.2 mass spectrometry data system from hermo Fisher Scientiic was used to check the data manually. For fragmentation of puriied samples and extracts an LTQ-Orbitrap Classic mass spectrometer coupled with a Surveyor HPLC system from hermo Fisher Scientiic (Bremen, Germany) was employed in negative ion mode using collision-induced dissociation (CID) at 35 V. he same HPLC conditions were used as were used with the Exactive instrument. he mass axes of both MS instruments were externally calibrated according to the manufacturer's instructions just before commencing the experiments. In case of UDN two high weight fractions UDN10 (265 mg) collected at 40:60 EtOAc:hexane, and UDN14 (130 mg) eluted at 50:50 EtOAc:hexane, were chosen for further fractionation. he fractions were reconstituted in 5 mL EtOAc mixed with celite in a 1:2 proportion (v/v), dried and packed into an empty dryloader cartridge (Alltech, Carnforth, UK) to be transferred to a Grace Davison Reveleris ® lash chromatography system (Alltech, Carnforth, UK) equipped with a dual-UV wavelength detector that was set at 290 and 320 nm, an ELSD, and an automatic fraction collector which collected peaks according to the threshold set for the ELSD (set to medium). Ater method development with an analytical LC-UV-ELSD system. Both open column fractions were re-chromatographed using the Grace system in reversed phase mode using a (12 g) C18 cartridge and a low rate of 12 mL/min and isocratic conditions with 30:70 ACN:water for 30 min then increasing to 100% ACN over 30 min and holding for 5 min then back to 30% ACN in 1 min and holding for 5 min. his yielded compounds 1 (8.6 mg), 2 (11.0 mg) and 3 (7.0 mg); all were obtained as ine yellow powders.

GC-MS Analysis.
In the case of the open column fractions of the ION sample, fractions ION12 (165 mg) and ION13 (160 mg) both eluted at 60:40 hexane:EtOAc. hey were puriied by using the reversed phase C18 (12 g) cartridge. he elution method used isocratic conditions with 90:10 ACN:water for 30 min with a low rate of 9 mL/min yielding compounds 4 (13 mg), 5 (6 mg) and 6 (12.2 mg as a 3:1 mixture with 5) which were all white amorphous solids. NMR Spectra. he 1 H, 13 C and DEPT 135, and 2D 1 H, 1 H-COSY, and 13 C-1 H HSQC and HMBC NMR spectra were obtained by a JEOL-LA 400 FT-NMR spectrometer system using C 2 HCl 3 and 2 H 6 -DMSO as solvents.
Anti-trypanosomal testing. All crude extracts and puriied compounds were screened using a variant of the Alamar blue assay 24,25 , based on the reduction of the blue dye resazurin sodium salt (Sigma-Aldrich) to luorescent and pink resoruin by live trypanosomes 26 . In this screen testing was carried out against the standard drug-sensitive T. b. brucei clone and two derived drug resistant lines, in order to assess the potential for cross-resistance with existing drugs. he results were expressed as half maximal efective concentration (EC 50 ) values, each based on at least three independent determinations. he assays were performed using serial dilutions in white opaque plastic 96-well plates (F Cell Star, Greiner Bio-one GmbH, Frickenhausen, Germany), with each compound or mixture double diluted over two rows of the plate (i.e. 23 double dilutions and a no-drug control well), allowing an optimally deined EC 50 value ater plotting of the luorescence intensity against test compounds concentration and itting to a sigmoid curve with variable slope (GraphPad Prism 5.0). Bloodstream forms of the following clonal strains of T. b. brucei were utilised: Lister strain 427 (s427) 27 , the standard drug-sensitive control strain; the B48 clone that was derived by in vitro adaptation of s427 to pentamidine 28 ; and the aqp2/aqp3 null strain 13 , from which the gene encoding the High Ainity Pentamidine Transporter (HAPT1) has been deleted. For each strain, the seeding density at the start of the assay was 2 × 10 4 cells/well, and the cells were exposed for 48 h to the test compounds, at 37 °C/5% CO 2 , before the addition of the resazurin dye and a further incubation of SCIENTIFIC REPORTS 923 DOI s 24 h under the same conditions. Fluorescence was determined in a FLUOstar Optima (BMG Labtech, Aylesbury, UK) at wavelengths of 544 nm and 620 nm for excitation and emission, respectively.
All crude extracts were also tested against C. fasciculata (ATCC 50083) as described in our previous publication 7 .