Evidence for a Correlation Between Systematics and Bioactivity in New Caledonian Cunoniaceae and Its Implications for Screening and Conservation1

Evidence for a Correlation between Systematics and Bioactivity in New Caledonian Cunoniaceae and Its Implications for Screening and Conservation1

Yohan Pillon2,3 and Bruno Fogliani3

Abstract: It is generally assumed that there is a good correlation between systematics and the secondary compounds found in plants. However because of the frequent homoplasy of chemical characters this has been difficult to test using statistical methods. Here we applied two nonparametric tests on a published data set, where 50 species of New Caledonian Cunoniaceae were screened for bioactivity against several pathogenic strains. Using Moran’s I index we showed that in two of nine tests against pathogenic strains there was a significantly higher similarity than expected in bioactivities between species belonging to the same genus and a significantly higher than expected dissimilarity in bioactivity between species belonging to different tribes. When considering the bioactivities against all pathogenic strains with Mantel tests, we also found significant correlation between bioactivity and phylogenetic distance in two of four tests. This has implications in screening and conservation. Searches for new molecules and bioactivity should preferentially be made on species spread across the tree of life. There is also a need to preserve as much phylogenetic diversity as possible to make sure that the widest reservoir of natural compounds remains available for future generations.

Biochemistry has played an important hamed 2003, Albach et al. 2005). Some of role in plant systematics (see, e.g., Grayer these problems may also be explained by sam- et al. 1999) by providing insights into the pling issues because a single taxon may not relationships between species (Albach et al. always be representative of a higher group 2005, Petrakis et al. 2005), within families (Grayer et al. 1999) or by variation associated such as the Cunoniaceae (Bate-Smith 1977), with geography (van Heerden et al. 2005) or and between families (Grayer et al. 1999). ontogeny (Ç irak et al. 2006). However, chemical compounds are characters Secondary compounds are economically that are often affected by homoplasy because important because they can be used as, or be similar compounds can appear independently the base for, active molecules in pharmaceu- several times in the course of evolution or tical chemistry (Young 1999, Butler 2004); might be lost secondarily (Wink and Mo- they can also be important for the produc- tion of pesticides, perfumes, and other compounds. It is quite possible that the tropical 1 Manuscript accepted 27 March 2008. floras being screened for new molecules today 2 Corresponding author: Laboratoire de Botanique, could provide the basis for tomorrow’s new Institut de Recherche pour le De´veloppement, BP A5, medicines (Butler 2004). Because it is now ac- 98848 Noumeá Cedex, New Caledonia (phone: (þ687) cepted that biodiversity is facing an important 26 09 28; fax: (þ687) 26 43 26; e-mail: pillon@noumea .ird.nc). crisis, the economic value of that diversity 3 Laboratoire Insulaire du Vivant et de l’Environne- may be a strong argument in favor of in- ment, Universite´ de la Nouvelle-Cale´donie, BP R4, creasing conservation efforts on wild species 98851 Noumeá Cedex, New Caledonia. (Balmford et al. 2002). If there is a significant correlation between chemical composition of Pacific Science (2009), vol. 63, no. 1:97–103 species and their relatedness, then conserva- : 2009 by University of Hawai‘i Press tion strategies should be implemented to pre- All rights reserved serve the widest spectrum of the tree of life

97 98 PACIFIC SCIENCE . January 2009

(Mace et al. 2003) to ensure that the greatest Independence of Variables diversity of molecules remains available. The correlation between biochemistry and The independence of variables needed to be systematics has generally been accepted, but assessed before further testing. Correlation there has been little supporting evidence that of activities between bark and leaf extracts it is statistically significant, probably because and ethyl acetate and methanol were tested of the homoplasy of chemical characters. by the mean of a chi-squared test. We tested Few studies are available that have investi- if species showing activity in their bark were gated a large number of taxa with multiple more likely to show activity in their leaves tests using a standard protocol. In this study for a given solvent, and similarly we tested if we conducted statistical analysis on the re- species showing antimicrobial properties with the methanol extract of their leaves or bark sults obtained by Fogliani et al. (2002a), who tested the activity of 50 species of Cunonia- were more likely to show properties with the ceae against 10 pathogenic strains, and show ethyl acetate of the same part of the plant. that there is a significant and positive correla- Correlation between activities against the dif- tion between species bioactivity and their re- ferent pathogenic strains was also assessed latedness. Specifically, we used nonparametric (for an extract with a given solvent from a given part of the plant). Chi-squared tests tests such as Moran’s I index that have proven their efficiency at revealing taxonomic pat- were only conducted on a contingency table terns (Lockwood et al. 2002). New Caledo- where all expected occurrences exceeded five. nian Cunoniaceae have been the subject of extensive taxonomic work over the past de- Test of the Correlation between Systematics and cades (Hoogland et al. 1997, Hopkins 2005, Bioactivity 2006, 2007, Hopkins et al. 2007, Pillon et al. in press), and a tribal classification (Bradford Moran’s I index was used to test if there was a et al. 2004) based on morphological (Hufford significant similarity in the bioactivity of spe- and Dickison 1992) and molecular (Bradford cies belonging to the same taxonomic group and Barnes 2001) phylogenies is available. (genus or tribe). Moran’s I index was calcu- This taxonomic scheme was used to provide lated for each test against pathogenic strains rough estimates of phylogenetic distances be- for leaf or bark extracts with ethyl acetate tween species. or methanol. Only tests where at least 10 species showed activities were included. We es- materials and methods tablished three distance classes (1, 2, and 3) between species depending on their phyloge- Data netic distance: 1, if they belong to the same Fogliani et al. (2002a) tested the bioactivity of genus; 2, if they belong to the same tribe; 3, 50 of the ca. 90 species of Cunoniaceae en- if they belong to different tribes. The Mor- demic to New Caledonia, representing six an’s I index tests whether species that are genera and four tribes following the most re- closely related phylogenetically have similar cent classification of Cunoniaceae (Bradford activity. and Barnes 2001, Bradford et al. 2004). Activ- One single test against one pathogen pro- ities against eight bacteria strains and two vides little statistical power to detect signifi- fungi were assessed. In most cases the activity cant correlation for several reasons. In most of ethyl acetate and methanol extract of leaf cases, few species showed activity against a and bark were tested. A disk containing the given pathogen, and this activity does not extract was placed in a petri dish containing follow a normal distribution. To increase the a culture of each strain. In the presence of sensitivity of the tests we used another ap- antimicrobial activities, an inhibition zone proach. We took into consideration the 10 can be observed around the disk. The diame- pathogenic strains in a single statistical test ter of this zone is used as a measure of the known as the Mantel test. Activities were cen- strength of the activity. tered and reduced (i.e., for each species we Systematics and Bioactivity . Pillon and Fogliani 99 calculated the activity of the species minus tate extract from bark against Staphylococcus the average over all species divided by the aureus and Erwinia carotovora, activity of standard deviation over all). The aim of this methanol extract from bark against Coryne- procedure was to give a similar weight to bacterium accolans and Fusarium oxysporum, each pathogenic strain. ‘‘Bioactive’’ distance and activity of methanol extract from leaves was calculated for each pair of species as the against Corynebacterium accolans and Fusarium geometric distance between their activities oxysporum. In only one case we found a sig- against all 10 pathogenic strains: nificant correlation of activity against two rXffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi microbial strains: activity of ethyl acetate 2 extract from bark against Corynebacterium BD ¼ ða a Þ ij ik jk accolans and Staphylococcus aureus (P ¼ :006). k Therefore we assumed that the different mi- BD is the ‘‘bioactive’’ distance between spe- crobial strains could be reasonably considered as independent measures of bioactivity. cies i and j. aik and ajk are the activity of species i and j, respectively, against the pathogenic strain k. Test of the Correlation between Systematics and We used the same phylogenetic distance Bioactivity as for the Moran’s I index. The Mantel test assesses the correlation between ‘‘bioactive’’ Nine tests against pathogenic strains could be distance and phylogenetic distance. One tested with the Moran’s I index; results for thousand replicates were used to evaluate the each test and the average over the nine tests significance of the test. are displayed in Figure 1. In the case of the Both Moran’s I index and the Mantel test activity of the methanol extract from leaves were computed with the software PASSAGE against Corynebacterium accolans, the value of (Rosenberg 2002). the Moran’s I index (I ¼ 0:059) was significantly higher (P ¼ :008) than expected by chance for species belonging to the same ge- results nus. Thus activities between congeneric species were closer than expected by chance. Independence of Variables Concerning the activity of the ethyl acetate The conditions necessary to apply the chi- extract from bark against Corynebacterium ac- squared test (expected occurrence of at least colans, we observed Moran’s I index values five) to test the correlation between activities that were significantly higher than expected of leaf and bark extracts or of the two solvent for species belonging to the same genus (I ¼ extracts were satisfied in only a few cases. The 0:201; P <:001) or to the same tribe (I ¼ chi-squared test always indicated a significant 0:081; P ¼ :028). If a Bonferroni correction positive correlation between the activities of (Rice 1989) is applied to account for multiple bark and leaf extract (test of methanol extract tests (nine in total), the second I value would against Fusarium oxysporum, P <:0001; and still be significant. In turn, species belonging Corynebacterium accolans, P ¼ :003) but no sig- to different tribes showed significantly differ- nificant correlation between ethyl acetate and ent activities in two cases (Figure 1), with or methanol extract in the single case tested (test without a Bonferroni correction. The seven of bark extract against Corynebacterium acco- other tests did not show values that were sig- lans). Therefore activities of bark and leaves nificantly different from expectation. Average were considered as nonindependent variables. Moran’s I index for the nine tests showed that Correlation between activities against the overall, I values are positive for species be- different pathogenic strains was also assessed longing to the same genus (similar activities), (for an extract with a given solvent from a negative but close to zero for species belong- given part of the plant). In four cases this ing to the same tribe, and lower for species could be tested; in three cases there was no belonging to different tribes (dissimilar activ- significant correlation: activity of ethyl ace- ities).

Systematics and Bioactivity . Pillon and Fogliani 101

Because the correlation between activities possible by the large number of compounds against different strains is limited, we consid- included in their matrix (400). ered that pooling them was reasonable and Although several of our tests were non- that potential addition effects would be lim- significant, the few significant ones indicated ited. This justifies the use of the ‘‘bioactive’’ a positive correlation between bioactivity and distance, which takes into account the activ- systematics. We believe that the lack of sig- ities against the different pathogenic strains nificant values is due to insufficient data, altogether. On the other hand we tested sep- particularly the small number of species dis- arately the activities of bark and leaf extracts playing activity against a given pathogenic using ethyl acetate and methanol, considering strain, which limits the statistical power of that activities of bark and leaf extracts were at the tests. The use of the Bonferroni correc- least partially correlated. tion shows that the few significant values ob- Mantel tests revealed a significant correla- served are not due to chance as we multiplied tion between phylogenetic distance and ‘‘bio- statistical tests but to a genuine pattern. Also, active’’ distance for methanol extract of bark because activities against pathogenic strains and leaves (P ¼ :001 in both cases) and no are not fully independent, there may be significant correlation in the two other cases some addition effects that could enhance sig- (ethyl acetate extract of bark and leaves). nificance of the Mantel test for which we The first two tests were still statistically sig- pooled the data. However, each activity was nificant after applying a Bonferroni correc- considered separately when we used the tion. Moran’s I index; therefore this cannot solely explain the significant correlation we ob- discussion served. Species that are closely related are more It is generally accepted that biochemistry is likely to have similar bioactivity. This can well correlated with systematics. However probably be explained by the fact that closely this has been difficult to test in a statistical related species have the same active mole- manner because of the inadequacy of the cules. Nevertheless we cannot exclude other data available. Chemical characters are often possibilities (e.g., species with similar activ- binary (presence/absence) and bear little in- ities may have similar amounts of the active formation. It is relatively rare that a large molecules or have different molecules with number of characters are available for a large the same bioactivity or other explanations). number of taxa. Data generated by different In turn, species that are not closely related studies may not be comparable, because dif- are more likely to have different activities; ferent protocols were used. Therefore few this bears strong implications in screening cladistics analyses have been carried out and conservation. on chemical characters. Nevertheless when Most of the world’s biodiversity is found Bininda-Emonds et al. (2001) built cladistics in the tropics, and the tropical floras are in- using the lipids of the anal sac secretion of deed being intensively screened for new ac- felids, they found good congruence with the tive molecules (Butler 2004). Because many existing taxonomic scheme. This was made plant species are still to be discovered (Prance

H Figure 1. Correlograms obtained for activity of different extracts (methanol or ethyl acetate extract from bark or leaves) against different pathogenic strains. Moran’s I index is a measure of autocorrelation (i.e., it indicates whether species tend to have activities that are more similar [positive autocorrelation] or more different [negative autocorrelation] than expected by chance). It is calculated for species belonging to the same genus (phylogenetic distance ¼ 1), for species belonging to the same tribe but different genera (phylogenetic distance ¼ 2), and for species belonging to different tribes (phylogenetic distance ¼ 3). Moran’s I index expected value is 0.023; positive values for I indicate positive autocorrelation and negative values for I indicate negative autocorrelation. Asterisks indicate I values that are significantly different from those expected by chance (*, P <:05; **, P <:01; ***, P <:001). 102 PACIFIC SCIENCE . January 2009 et al. 2000), most species have never been markers. Biochem. Syst. Ecol. 33:1167– studied for their bioactivity, and some are 1177. likely to disappear before they can be suffi- Balmford, A., A. Bruner, P. Cooper, R. Cos- ciently surveyed. This is true for New Cale- tanza, S. Farber, R. E. Green, M. Jenkins, donian Cunoniaceae in particular, where P. Jefferis, V. Jessamy, J. Madden, K. many species remain undescribed and new Munro, N. Myers, S. Naeem, J. Paavola, genera are still being discovered (e.g., Mc- M. Rayment, S. Rosendo, J. Roughgarden, Pherson and Lowry 2004). The flora of the K. Trumer, and R. K. Turner. 2002. Eco- island is also being screened for new bioac- nomic reasons for conserving wild nature. tivities (Bosisio et al. 2000, Fogliani et al. Science (Washington, D.C.) 297:950– 2002a,b) and is threatened by fires, intro- 953. duced species, and the continued develop- Bate-Smith, E. C. 1977. Chemistry and tax- ment of nickel mining. onomy of the Cunoniaceae. Biochem. Thus, species included in screening should Syst. Ecol. 5:95–105. be spread out across the tree of life to in- Bininda-Emonds, O. R. P., D. M. Decker- crease the likelihood of finding new and Flum, and J. L. Gittleman. 2001. The util- useful bioactive molecules. Also, because the ity of chemical signals as phylogenetic sampling for testing bioactivity is sometimes characters: An example from the Felidae. relatively destructive or because the discovery Biol. J. Linn. Soc. 72:1–15. of interesting molecules may lead to unsus- Bosisio, E., D. Mascetti, and P. Cabalion. tainable exploitation, red-listed and rare spe- 2000. Screening of plants from New Cale- cies could be avoided for screening purposes donia and Vanuatu for inhibitory activity and instead be replaced by closely related of xanthine oxydase and elastase. Pharm. and more abundant species. Biol. 38:18–24. Vane-Wright et al. (1991) argued that Bradford, J. C., and R. W. Barnes. 2001. Phy- each species should not be given equal weight logenetics and classification of Cunonia- in conservation (e.g., the single species of ceae (Oxalidales) using chloroplast DNA Welwitschia versus one of the many species sequences and morphology. Syst. Bot. of Taraxacum). The fact that phylogenetically 26:354–385. isolated taxa are more likely to have original Bradford, J. C., H. C. F. Hopkins, and R. W. bioactivity may be a good argument to pri- Barnes. 2004. Cunoniaceae. Pages 91–111 oritize them in conservation planning. The in K. Kubitzki, ed. The families and genera potentially high economic value of some nat- of vascular plants. Springer, Berlin. ural products should be kept in mind, because Butler, M. S. 2004. The role of natural prod- it is a strong argument in favor of the preser- uct chemistry in drug discovery. J. Nat. vation of biodiversity (Young 1999). Prod. (Lloydia) 67:2141–2153. Ç irak, C., B. Saglam, A. Kemal Ayan, and K. Kevseroglu. 2006. Morphogenetic and acknowledgments diurnal variation of hypericin in some Hypericum species from Turkey during We thank Jean-François Arnaud for assis- the course of ontogenesis. Biochem. Syst. tance with the software. We are grateful to Ecol. 34:1–13. an anonymous reviewer and especially Zach- Fogliani, B., S. Bouraı¨ma-Madjebi, V. Mede- ary Rogers for carefully reviewing an early vielle, and R. Pineau. 2002a. Screening of version of the manuscript. 50 Cunoniaceae species from New Cale- donia for antimicrobial properties. N. Z. Literature Cited J. Bot. 40:511–520. Fogliani, B., S. Bouraı¨ma-Madjebi, R. Pineau, Albach, D. C., R. J. Grayer, G. C. Kite, and and P. Cabalion. 2002b. Screening of fifty S. R. Jensen. 2005. Veronica: Acylated Cunoniaceae species from New Caledonia flavone glycosides as chemosystematic for inhibitors of xanthine oxidase and scav- Systematics and Bioactivity . Pillon and Fogliani 103

engers of superoxide anions. Pharm. Biol. Cunoniaceae from New Caledonia. Ann. 40:526–533. Mo. Bot. Gard. 91:260–265. Grayer, R. J., M. W. Chase, and M. S. J. Petrakis, P. V., M. Couladis, and V. Roussis. Simmonds. 1999. A comparison between 2005. A method for detecting the bio- chemical and molecular characters for the systematic significance of the essential oil determination of phylogenetic relation- composition: The case of five Hellenic ships among plant families: An apprecia- Hypericum L. species. Biochem. Syst. Ecol. tion of Hegnauer’s ‘‘Chemotaxonomie 33:873–898. der Pflanzen.’’ Biochem. Syst. Ecol. Pillon, Y., H. C. F. Hopkins, and J. C. Brad- 27:369–393. ford. in press. Two new species of Cunonia Hoogland, R. D., J. Je´re´mie, and H. C. F. (Cunoniaceae) from New Caledonia. Kew Hopkins. 1997. Le genre Cunonia (Cuno- Bull. niaceae) en Nouvelle-Cale´donie: Descrip- Prance, G. T., H. Beentje, J. Dransfield, and tion de cinq espe`ces nouvelles. Adansonia R. Johns. 2000. The tropical flora remains 19:7–19. undercollected. Ann. Mo. Bot. Gard. Hopkins, H. C. F. 2005. Nomenclature and 87:67–71. typification in the endemic genus Codia Rice, W. R. 1989. Analyzing tables of statisti- (Cunoniaceae) from New Caledonia. cal tests. Evolution 43:223–225. Adansonia 27:243–254. Rosenberg, M. S. 2002. PASSAGE: Pattern ———. 2006. Nomenclature and typification analysis, spatial statistics, and geographi- in Geissois (Cunoniaceae) in the South- cal exegesis, version 1.0. Department West Pacific. Adansonia 28:311–327. of Biology, Arizona State University, ———. 2007. Geissois bradfordii, a new species Tempe. of Cunoniaceae from New Caledonia. Vane-Wright, R. I., C. J. Humphries, and Kew Bull. 62:275–280. P. H. Williams. 1991. What to protect?— Hopkins, H. C. F., B. Fogliani, and Y. Pillon. Systematics and the agony of choice. Biol. 2007. Four new species in the endemic ge- Conserv. 55:235–254. nus Codia (Cunoniaceae) from New Cale- van Heerden, F. R., B.-E. van Wyk, A. M. donia. Kew Bull. 62:259–274. Viljoen, and P. A. Stoekamp. 2005. Pheno- Hufford, L., and W. C. Dickison. 1992. A lic variation in wild populations of Aspala- phylogenetic analysis of Cunoniaceae. thus linearis (rooibos tea). Biochem. Syst. Syst. Bot. 17:181–200. Ecol. 31:885–895. Lockwood, J. L., G. J. Russell, J. L. Gittle- Wink, M., and G. I. A. Mohamed. 2003. Evo- man, C. C. Daehler, M. L. McKinney, lution of chemical defense traits in the and A. Purvis. 2002. A metric for analyzing Leguminosae: Mapping of distribution taxonomic patterns of extinction risk. patterns of secondary metabolites on a Conserv. Biol. 16:1137–1142. molecular phylogeny inferred from nu- Mace, G. M., J. L. Gittleman, and A. Purvis. cleotide sequences of the rbcL gene. Bio- 2003. Preserving the tree of life. Science chem. Syst. Ecol. 31:897–917. (Washington, D.C.) 300:1707–1709. Young, R. N. 1999. Importance of biodiver- McPherson, G., and P. P. Lowry II. 2004. sity to the modern pharmaceutical indus- Hooglandia, a newly discovered genus of try. Pure Appl. Chem. 71:1655–1661.