Chemotaxonomy of three genera of the annonaceae family using self-organizing maps and 13C NMR data of diterpenes

ARTIGO

Chemotaxonomy of three genera of the annonaceae family using self-organizing maps and ¹³C NMR data of diterpenes^#

Luciana Scotti^I; Josean Fechine Tavares^I; Marcelo Sobral da Silva^I; Emanuela Viana Falcão^II; Luana de Morais e Silva^II; Gabriela Cristina da Silva Soares^II; Marcus Tullius Scotti^II,*

^IDepartamento de Ciências Farmacêuticas, Centro de Ciências da Saúde, Universidade Federal da Paraíba, Campus I, 58051-970 João Pessoa - PB, Brasil

^IIDepartamento de Engenharia e Meio Ambiente, Universidade Federal da Paraíba, Campus IV, 58297-000 Rio Tinto - PB, Brasil

ABSTRACT

The Annonaceae family is distributed throughout Neotropical regions of the world. In Brazil, it covers nearly all natural formations particularly Annona, Xylopia and Polyalthia and is characterized chemically by the production of sources of terpenoids (mainly diterpenes), alkaloids, steroids, polyphenols and, flavonoids. Studies from ¹³C NMR data of diterpenes related with their botanical occurrence were used to generate self-organizing maps (SOM). Results corroborate those in the literature obtained from morphological and molecular data for three genera and the model can be used to project other diterpenes. Therefore, the model produced can predict which genera are likely to contain a compound.

Keywords: chemotaxonomy; self-organizing maps; ¹³C NMR.

INTRODUCTION

The Annonaceae family is composed of about 2000 species in 129 genera found throughout Neotropical regions of the world. In Brazil, the family encompasses around 29 genera with approximately 260 species occurring in all natural formations predominantly Xylopia, Annona (genus native) and Polyalthia (genus input). This family is known to produce many edible fruits, and many of its plants are commonly used in folk medicine. The Annonaceae and all woody plants of the Magnoliales has very rich chemical characteristics and are recognized sources of terpenoids (mainly diterpenes), alkaloids (in large amounts, especially the core isoquinoline derivatives), steroids, polyphenols and, flavonoids.¹

Xylopia comprises about 150 species.² Some fruits from this genus are used for culinary purposes as condiments while others serve as a source of fiber for rope manufacture. The timber produced is light, durable and displays medical properties.³ Several species have broad applications, particularly in folk medicine as vermifuges and antimicrobial agents.⁴ Numerous chemical compounds have been isolated from Xylopia, including biologically active acetogenins,⁵ kauranes and labdanes (diterpene types), sesquiterpenes, alkaloids and, flavonoids. Notably, diterpenes exhibit metabolic characteristics of the Xylopia genus.⁶

The genus Annona has about 120 species (A. squamosa, A. cherimola Mill., A. reticulate L., A. muricata L., A. dioica among others) found in Central and South America, Africa, Asia and Australia. Some have been investigated for their chemical compounds and pharmacological activities. Many of the species are used in traditional medicines for the treatment of a variety of diseases. Several annonacaeous species have been found to contain acetogenins, a class of natural compounds with a wide variety of biological activities.^7-12

Polyalthia contains several species: P. angustissima, P. chrysotricha, P. elmeri, P. glabra, P. hirtifolia, P. hookeriana, among others. It is found natively in India and Sri Lanka and was introduced into gardens in many tropical countries around the world. These plants produce a great diversity of substances that could be of therapeutic significance in many areas of medicine. These have shown marked antimicrobial, anti-inflammatory, cytotoxic, immunosuppressive, antibacterial, antifungal and other pharmacological properties.^13-15

Phylogenetic studies have contributed to the knowledge of evolution relationships among organisms. However, to date few studies have investigated Annonaceae phylogeny, and those conducted have tended to focus on the relationship among sister families or phylogenetic studies on a genus level.^16-19 Concerning the three genera of the present study, Polyalthia is nearest to the tree's basal, while Xylopia and Annona have closed relationship (Figures 1 and 2).^16-19 These previous studies were based on molecular data sets and morphological analyses. However, chemistry data can also contribute to phylogenetic analysis. Secondary metabolites have specific botanic origin, which can be used as a chemotaxonomic marker and thus, represent a bridge in phylogeny between genetics and morphology.²⁰ In addition, studies of natural chemical products have made a significant contribution to the bioprospection/development of new drugs, as well as to associating compounds, properties and botanic occurrence.²¹

The use of chemical data for classification has been employed in several studies and secondary metabolites have been proposed and used by Gottlieb and co-workers.^22-28 Chemotaxonomic studies have been applied at several levels using different classes of secondary metabolites: superorders of angiosperms;^20,24,25 families such as Asteraceae,²⁹ Meliaceae,²⁷ Apocynaceae,²⁸ Lamiaceae;³⁰ tribes of Asteraceae.^23,31-35

ANNs (Artificial Neural Networks) are a method or, more precisely, a set of methods, used extensively since the 1990s. Since ANNs are not restricted to linear correlations and can also take into account non-linear data correlations, they can be efficiently applied for modeling, prediction and classification. The book by Zupan and Gasteiger remains the main text on this area.³⁶ Self-organizing maps (SOMs) are widely used by ANNs for pattern recognition and classification as proposed by Kohonen who called his algorithm a "self-organizing network".^37,38 This procedure can map multivariate data onto a two dimensional grid, grouping similar patterns near each other. Therefore, Kohonen learning is best suited for mapping of data. In this projection, the similarity relationship between objects is conserved.³⁸ Thus, in principle, Kohonen networks can be used for clustering of objects. It is important to note that the training of these networks (SOMs) is unsupervised, but there are some algorithms that combine the original SOM with supervised methods.³⁶

In the original SOM procedure, the investigated property is not used during the training process. Each neuron in the grid is associated with a weight, and similar patterns stimulate neurons with similar weight, so that similar patterns are mapped near one another.³⁶ In chemistry for example, there are several applications of ANNs such as in HPLC, reactivity, and the classification of olive oils.³⁶ Another very useful application is in the prediction and classification of spectra such as infrared,³⁹ mass,⁴⁰ and nuclear magnetic resonance^41-43, including some QSAR studies.⁴⁴ In natural products chemistry, there are a few studies available showing applications of ANNs, such as the classification of Asteraceae tribes,^31,32,35 and the prediction of skeletal types.^45,46

¹³C NMR (Nuclear Magnetic Resonance) data yield rich information about the molecular structure and are sufficiently sensitive to detect small differences in the molecule.⁴⁷ These differences are measured by the variation in chemical shifts, and the values can be used to associate the chemical structure with the respective botanical occurrence. This association can help to understand the influence of the chemical environment of secondary metabolite as a molecular descriptor, and, for this reason, ¹³C NMR data may be useful in chemotaxonomic studies.

In the present study, ¹³C NMR chemical shift values of 20 carbons of skeletal structures of 137 diterpenes, selected from the literature⁴⁸ and Self-Organizing Maps (SOMs), were used to perform chemotaxonomic studies of three genera of the Annonaceae family, namely: Annona, Xylopia and Polyalthia and to compare with previous studies using morphological and molecular data.

EXPERIMENTAL AND CHEMIOINFORMATIC STUDIES

We selected 118 diterpenes from the literature (Table 1) together with their ¹³C NMR chemical shifts and respective botanical occurrence in three genera: Annona, Polyalthia and Xylopia, of the Annonaceae tribe. The respective skeletal types are listed in Table 2. The ¹³C NMR chemical shifts of the diterpenes were introduced as input data as showed in Table 3. Each diterpene can appear n times within a delimited taxon, in this case a genus (Annona, Polyalthia, Xylopia). The number of occurrences for a taxon was defined by counting how many times a compound appeared in a given species belonging to that taxon (genus). The 118 diterpenes have 169 botanical occurrences. Therefore, the input data constitutes a matrix of 169 samples and 20 variables where each variable corresponds to a chemical shift. The chemical shifts are sorted by sequence number of the diterpene skeletal atoms as shown in Table 2. The samples are labeled according to botanical occurrence in the three genera used in this study, namely Annonaceae: Annona, Polyalthia and Xylopia.

A Kohonen-ANN was trained using the Matlab 6.5 computing environment by Mathworks and SOM Toolbox 2.0.⁴⁹ SOM toolbox is a set of Matlab functions that can be used to develop and implement SOM neural networks, and which contains functions for creation, visualization and analysis of self-organizing maps. A SOM grid with square geometry 13 x 5 in size was created and trained. The training was conducted through the Batch-training algorithm. In this algorithm, the whole dataset is presented to the network before any adjustment is made. In each training step, the dataset is partitioned according to the regions of the map weight vectors. Within the algorithm, the new weight vector is based on simple averages and there is no learning rate.³⁸ This feature allows missing values to be ignored by the net. The number of epochs is automatically chosen by the Toolbox, i.e., the neural network is trained until its convergence to minimal error.

All SOMs were generated with the same topology: for the local lattice structure, the rectangular grid was used, while a sheet was used to indicate the global map shape, using the Gaussian neighborhood function. The literature shows that the determination of the size of the SOMs is an empirical process.³⁸ Initially, a heuristic formula of m = 5(n)^0.5 is used for total number of map units, where n is the number of samples. The ratio of side lengths is based on the two biggest eigenvalues of the covariance matrix of the given data. Some different maps sizes were prepared, based on the initial map, generated as described earlier. The SOM toolbox automatically labels the map based on the previously labeled data. The label with most instances is added to the map unit. In the case of a match, the first encountered label is used. A hit is a sample which has the same label as the map unit where it is located. For each map, 10 cross-validations were performed, splitting 10% of the data by dividing into training and test sets, which consisted of approximately 80% (137 samples) and 20% (32 samples) of total samples, respectively (Table 4).

RESULTS AND DISCUSSION

The results are summarized in Table 5. Training and test sets have overall higher values of matches: 90.9 and 86.3%, respectively. The Polyalthia genus showed the highest match values of botanical occurrence for both training (100%) and test sets (98.9%), while Xylopia exhibited the lowest match results: 79.8% for training and 70% for test set.

The SOM (Figure 3) show a clear separation among the botanical occurrences of the three genera. Three distinct regions are evidenced: northwest of the map with black squares (Annona region), gray squares (Xylopia region) and, light gray squares (Polyalthia region). The ¹³C chemical shifts and SOM are able to clearly distinguish diterpenes from Annona (top of map), and from Polyalthia (bottom of SOM). Diterpenes of Xylopia are situated in the middle of the map, between Annona and Polyalthia regions, and share some neurons with diterpenes mainly from Annona (top of map) and Polyalthia (bottom region), thus explaining the poorer match results of this genus.

Figure 3

These results corroborate previous phylogenetic studies, in which these two genera (Annona and Xylopia) were more closely related and Polyalthia farther from these in cladograms using mainly molecular and morphological data. Thus, for these three genera, the chemistry regarding the diterpenes is closely related to the molecular and morphological data used in the previous studies on phylogeny. The carbons responsible for the division in botanical occurrence of the diterpenes of the genera used in this study can be seen in Figure 5, which shows the weight that each descriptor has in this Kohonen map. Generally, the most representative descriptors are those that have two main characteristics: the greatest weights in the genus predominant region; a considerable difference between the highest and lowest descriptors' (¹³C NMR chemical shifts) values.

The following carbons have highest ¹³C chemical shift values for the Polyalthia region: 2, 3, 4, 6, 10, 11, 13, 14 and 20. For example, compound 57 (Figure 6) polyalthialdoic acid, a clerodane which has a double bond between carbons 13 and 14. On the other hand, Annona diterpene carbons have higher chemical shift values mainly for atoms 1, 7, 9, 17 and 19, for example Annoglabasin A and B, compounds 26 and 27 respectively (Figure 6), two diterpene kauranes that have ester and carboxyl groups on carbon 17, plus aldehyde and acetoxyl groups on carbon 19, respectively. These higher chemical shifts are less representative for diterpenes from the Xylopia genus but remain significant. Some diterpenes from this genus have higher chemical shift values for carbons 15 and 18, such as those found in intrachyloban-19-oic acid.

CONCLUSIONS

SOMs and ¹³C NMR data of skeletal carbons of diterpenes produce clusters according to their botanical occurrence. Moreover, their similarities corroborate previous studies using morphological and molecular data and can predict the botanical occurrence of similar compounds. Some ¹³C chemical shift values are specific for skeletal carbons of diterpenes from these three genera of Annonaceae. Therefore, ¹³C NMR can be used as a molecular descriptor, including in QSAR studies, enabling the methodology to find new structures with potential biological activities.

ACKNOWLEDGEMENTS

The authors would like to thank the Brazilian National Research Council (CNPq) for financial support.

Recebido em 4/6/12; aceito em 17/7/12; publicado na web em 26/10/12

• 1. Chatrou, L. W.; Rainer, H.; Maas, P. J. M. In Flowering Plants of Neotropics; Smith, N.; Mori, S. A.; Henderson, A.; Stevenson D. W.; Heald, S. V., eds.; Botanical Garden: New York, 2004, p. 18-20.
• 2. Brummitt, R. K.; Powel, C. E.; Authors of Plant Names, Royal Botanic Gardens: Kew, 1988.
• 3. Silva, J. B.; Rocha, A. B.; Rev. Cienc. Farm. 1981,3,33.
• 4. Yiadom, B. K.; Fiagbe, N. I. Y.; Ayim, J. S. K.; J. Nat. Prod. 1977,40,543.
• 5. Alfonso, D.; Colman, S. T.; Zhao, G. X.; Shi, G.; Ye, Q.; Schwedler, J. T.; Mclaughlin, J. L.; Tetrahedron 1996,52,4215.
• 6. Vilegas, W.; Felício, J. D.; Roque, N. F.; Gottlieb, H. E.; Phytochemistry 1991,30,1869.
• 7. Santos, A. S.; Andrade, E. H. A.; Zoghbi, M. G. B.; Maia, J. G. S.; Flavour Frag. J. 1998,13,148.
• 8. Costa, E. V.; Pinheiro, M. L. B.; Xavier, C. M.; Silva, J. R. A.; Amaral, A. C. F.; Souza, A. D. L.; Barison, A.; Campos, F. R.; Ferreira, A. G.; Machado, G. M. C.; Leon, L. L. P.; J. Nat. Prod. 2006,69,292.
• 9. Joly, A. B.; Botânica: Introdução à taxonomia vegetal, 13Şed., Companhia Editora Nacional: São Paulo, 2002.
• 10. Leboeuf, M.; Cavé, A.; Bhaumik, P. K.; Mukherjee, B.; Mukherjee, R.; Phytochemistry 1982,21,2783.
• 11. Manica, I.; Icuma, I. M.; Junqueira, K. P.; Oliveira, M. A. S.; Cunha, M. M.; Oliveira Jr., M. E.; Junqueira, N. T. V.; Alves, R. T.; Frutas Anonáceas: Ata ou Pinha, Atemoia, Cherimólia e Graviola - Tecnologia de Produção, Pós-colheita, Mercado, Cinco Continentes: Porto Alegre, 2003.
• 12. Pontes, A. F.; Barbosa, M. R. V.; Maas, P. J. M.; Acta Bot. Bras. 2004,18,281.
• 13. Chen, C. Y.; Chang, F. R.; Shih, Y. C.; Hsieh, T. J.; Chia, Y. C.; Tseng, H. Y.; Chen, H. C.; Chen, S. J.; Hsu, M. C.; Wu, Y. C.; J. Nat. Prod 2000,63,1475.
• 14. Ravikumar, Y. S.; Mahadevan, K. M.; Kumaraswamy, M. N.; Vaidya, V. P.; Manjunatha, H.; Kumar, V.; Satyanarayana, N. D.; Environ. Toxicol. Pharmacol. 2008,26,142.
• 15. Verma, M.; Singh, S. K.; Bhushan, S.; Sharma, V. K.; Datt, P.; Kapahi, B. K.; Saxena, A. K.; Chem. Biol. Interact. 2008,171,45.
• 16. Johnson, D. M.; Syst. Bot. 2003,28,503.
• 17. Sauquet, H.; Doyle, J. A.; Scharaschkin, T.; Borsch, T.; Hilu, K. W.; Chatrou, L. W.; Le Thomas A.; Bot. J. Linn. Soc. 2003,142,125.
• 18. Scharaschkin, T.; Doyle, J. A.; Syst. Bot. 2005,30,712.
• 19. Doyle, J. A.; Bygrave, P.; Le Thomas, A. In Pollen and spores: morphology and biology; Harley, M. M.; Morton, C. M.; Blackmore, S., eds.; Royal Botanic Gardens: Kew, 2000.
• 20. Emerenciano, V. P.; Kaplan, M. A.; Gottlieb, O. R.; Biochem. Syst. Ecol 1985,13,145.
• 21. Queiroz, E. F.; Wolfender, J. L.; Hostettmann, K.; Curr. Drug Targets 2009,10,202.
• 22. Gottlieb, O. R.; Phytochemistry 1989,28,2545.
• 23. Emerenciano, V. P.; Ferreira, Z. S.; Kaplan, M. A.; Gottlieb, O. R.; Phytochemistry 1987,26,3103.
• 24. Da Silva, M. F. G. F.; Gottlieb, O. R.; Biochem. Syst. Ecol 1987,15,85.
• 25. Kaplan, M. A.; Gottlieb, O. R.; Biochem. Syst. Ecol 1982,10,329.
• 26. Emerenciano, V. P.; Kaplan, M. A.; Gottlieb, O. R.; Bonfanti, M. R. D. M.; Ferreira, Z. S.; Comegno, L. M. A.; Biochem. Syst. Ecol 1986,14,585.
• 27. Da Silva, M. F. G. F.; Gottlieb, O. R.; Dreyer, D. L.; Biochem. Syst. Ecol 1984,12,299.
• 28. Bolzani, V. S.; Da Silva, M. F. G. F.; Rocha, A. I.; Gottlieb, O. R.; Dreyer, D. L.; Biochem. Syst. Ecol 1984,12,159.
• 29. Emerenciano, V. P.; Militão, J. S. L. T.; Campos, C. C.; Romoff, P.; Kaplan, M. A.; Zambon, M.; Brant, A. J. C.; Biochem. Syst. Ecol 2001,29,947.
• 30. Alvarenga, S. A. V.; Gastmans, J. P.; Rodrigues, G. V.; Moreno, P. R. H.; Emerenciano, V. P.; Phytochemistry 2001,56,583.
• 31. Da Costa, F. B.; Terfloth, L.; Gasteiger, J.; Phytochemistry 2005,66,345.
• 32. Hristozov, D.; Da Costa, F. B.; Gasteiger, J.; J. Chem. Inf. Model. 2007,47,9.
• 33. Calabria, L. M.; Emerenciano, V. P.; Ferreira, M. J. P.; Scotti, M. T.; Mabry, T. J.; Nat. Prod. Commun. 2007,2,277.
• 34. Emerenciano, V. P.; Cabrol-Bass, D.; Ferreira, M. J. P.; Alvarenga, S. A.; Brant, A. J. C.; Scotti, M. T.; Barbosa, K. O.; Nat. Prod. Commun. 2006,1,495.
• 35. Scotti, M. T.; Emerenciano, V. P.; Ferreira, M. J. P.; Scotti, L.; Stefani, R.; Da Silva, M. S.; Mendonça, F. J. B.; Molecules 2012,17,4684.
• 36. Zupan, E.; Gasteiger, J.; Neural Networks for Chemists- An Introduction, VCH Verlagsgesellschaft: Weinheim, 1993.
• 37. Kohonen, T.; Biol. Cybern. 1982,43,59.
• 38. Kohonen, T.; Self-Organizing Maps, 1^st ed., Springer: Berlin, 2001.
• 39. Cleva, C.; Cachet, C.; Cabrol-Bass, D.; Analusis 1999,27,81.
• 40. Eghbaldar, A.; Forrest, T. P.; Cabrol-Bass, D.; Anal. Chim. Acta 1998,359,283.
• 41. Binev, Y.; Aires, S. J.; J. Chem. Inf. Comput. Sci. 2004,44,940.
• 42. Ivanciuc, O.; Rabine, J. P.; Cabrol-Bass, D.; Panaye, A.; Doucet, J. P.; J. Chem. Inf. Model. 1997,37,587.
• 43. Meiler, J.; Maier. W.; Meusinger, R.; J. Magn. Reson. 2002,157,242.
• 44. Fernandes, M. B.; Scotti, M. T.; Ferreira, M. J. P.; Emerenciano, V. P.; Eur. J. Med. Chem. 2008,43,2197.
• 45. Fraser, L. A.; Mulholland, D. A.; Fraser, D. D.; Phytochem. Anal. 1997,8,301.
• 46. Emerenciano, V. P.; Alvarenga, S. A.; Scotti, M. T.; Ferreira, M. J.; Stefani, R.; Nuzillard, J. M.; Anal. Chim. Acta 2006,579,217.
• 47. Da Costa, F. B.; Scotti, M. T. In Revisões em Processos e Técnicas Avançadas de Isolamento e Determinação Estrutural de Ativos de Plantas; De Souza, G. H. B.; De Mello, J. C. P.; Lopes, N. P., eds.; Ed. UFOP: Ouro Preto, 2012, p. 73-118.
• 48. Andrade, N. C.; Cunha, E. V. L.; Silva, M. S.; Agra, M. F.; Barbosa-Filho, J. M. In Recent Research Developments in Phytochemistry; Pandalai, S. G., ed.; Research Signpost: Kerala, 2003.
• 49. Vesanto, J.; Himberg, J.; Alhoniemi, E.; Parhankangas, J.; SOM Toolbox for Matlab, available at http://www.cis.hut.fi/projects/somtoolbox, accessed May 2011.

#

Artigo em homenagem ao Prof. Otto R. Gottlieb (31/8/1920-19/6/2011)

Publication Dates

• Publication in this collection
  30 Nov 2012
• Date of issue
  2012

History

• Received
  04 June 2012
• Accepted
  17 July 2012