Chemistry & Biology
Previews Espiritu, R.A., Matsumori, N., Murata, M., Nishimura, S., Kakeya, H., Matsunaga, S., and Yoshida, M. (2013). Biochemistry 52, 2410–2418. Harris, J.R. (2010). Cholesterol Binding and Cholesterol Transport Proteins (New York: Springer). Krause, M.R., and Regen, S.L. (2014). Acc. Chem. Res. 47, 3512–3521.
Mehbub, M.F., Lei, J., Franco, C., and Zhang, W. (2014). Mar. Drugs 12, 4539–4577.
H., Kobayashi, T., and Yoshida, M. (2013). PLoS One 8, e83716.
Nishimura, S., Arita, Y., Honda, M., Iwamoto, K., Matsuyama, A., Shirai, A., Kawasaki, H., Kakeya, H., Kobayashi, T., Matsunaga, S., and Yoshida, M. (2010). Nat. Chem. Biol. 6, 519–526.
Nishimura, S., Tokukura, M., Ochi, J., Yoshida, M., and Kakeya, H. (2014). Chem. Biol. 21, 1690–1699.
Nishimura, S., Ishii, K., Iwamoto, K., Arita, Y., Matsunaga, S., Ohno-Iwashita, Y., Sato, S.B., Kakeya,
Rivero, F. (2012). Rho GTPase (New York: Springer).
Simplifying Complexity in Metabolomics Katharina Eick1 and Georg Pohnert1,* 1Institute of Inorganic and Analytical Chemistry, Friedrich Schiller University Jena, Lessingstraße 8, 07743 Jena, Germany *Correspondence: [email protected] http://dx.doi.org/10.1016/j.chembiol.2015.05.001
Metabolomics analysis to unravel secondary metabolite dynamics in microorganisms faces the challenge of immense data sets and complex experimental setups. In this issue, Goodwin et al. present a multiple stimuli approach combined with self-organizing map-based analysis to elucidate variations in the metabolome of Streptomyces coelicolor caused by biotic and environmental perturbations. By investigating primary and secondary metabolite dynamics, it is possible to characterize the physiological states of cells and elucidate metabolite based interactions of microorganisms. These studies are enabled by different types of metabolomics techniques that have matured over the last decade and can now be employed to study different aspects of a microbe’s life. Comparative metabolomics, for example, provides a valuable tool to monitor and unravel chemically mediated intra- and interspecies interactions (Kuhlisch and Pohnert, 2015). The investigation of microbial interactions and stress responses is of great interest from a drug mining perspective, as the production of novel secondary metabolites might be induced under specific conditions. For example, Traxler et al. (2013) use a combination of different mass spectrometry and chemoinformatics approaches to illustrate how interspecies interactions stimulate diversification in the Streptomyces coelicolor metabolome. Such processes might result in upregulation of compounds with antibiotic, antifungal, or anticancer properties. The high throughput metabolomics techniques introduced in recent years
result in outstandingly complex data sets. These data sets are difficult to visualize and interpret. In practical terms, it means that the value of metabolomics analysis is determined not only on the performance of a given analytical strategy determine, but on the subsequent treatment of data sets. In-depth data analysis often requires the use of specialized data analysis tools, adjusted to the specific character of the data set. For example, multiple sample groups might be evaluated with the help of the metaanalysis software metaXCMS, which reveals differences between metabolic profiles across experimental groups, thereby overcoming limitations of pairwise comparisons (Patti et al., 2012). This platform enables also the comprehensive analysis of independently performed studies. A pattern based self-organizing map (SOM) algorithm to visualize the timedependent cellular metabolic response of lymphoblastic cell lines to ionizing radiation was introduced by Patterson et al. (2008). This SOM algorithm is based on concepts from the Gene Expression Dynamics Inspector (GEDI) software and allows creating 2D visualizations in a metabolomics context. It is evident that new insights into microbial metabolite
dynamics can benefit substantially from refined and expanded experimental perturbations as well as from the development and application of adequate data analysis strategies to categorize, prioritize, and comprehensively visualize the findings. Goodwin et al. (2015) set out to investigating the metabolic response of S. coelicolor under multiple perturbation conditions using ultra performance liquid chromatography/ion-mobility mass spectrometric analysis (UPLC-IM-MS) and a GEDI-based SOM algorithm. The experimental setup comprises 23 single perturbations of three stimuli categories known to cause changes in the metabolome: exposures to rare earth elements, resistance inducing antibiotics, and competing microorganisms. To elucidate the metabolic responses, total cellular extracts were analyzed by UPLC-IM-MS in a comparative metabolomics approach. SOMbased analysis of the data sets was employed to structure and categorize response specificity of metabolic features. Compared to established multivariate statistical analysis (MVSA), the SOM approach is optimized for the large data set. It allows a simple 2D-visualization based on the similarity of metabolite
Chemistry & Biology 22, May 21, 2015 ª2015 Elsevier Ltd All rights reserved 567
Chemistry & Biology
Previews feature intensity trends among the different conditions. As a result, topological heat maps are created, representing the metabolic phenotype of a sample or condition but still retaining the attributes of MVSA-based analytics. An additional benefit of SOM analysis is a reduced masking of low-abundant metabolites compared to MVSA. The basis of these topological heatmap visualizations are ‘‘trained maps’’ that are obtained as follows. In a first step, all metabolite features detected in the 23 perturbations are characterized by their intensity profiles among all perturbations. These intensity profiles are then ‘‘self-organized’’ in a 2D graphical representation, locating those metabolites with similar intensity profiles in close coordinates of a grid. As a result, regions of correlated metabolite intensity profiles are visualized. In order to characterize each sample or experimental condition, information about specific abundances of the metabolic features is then introduced into these ‘‘trained maps.’’ The comparison of perturbed and unperturbed profiles allows prioritization of so-called regions of interest (ROIs), specifically visualizing significant metabolic responses in the experimental condition considered. After identification of significant metabolic features with the help of ROIs, determination of their molecular identity can be pursued by standard analytical approaches. Using these protocols, Goodwin et al. (2015) identified 2154 significant meta-
bolic features, 61% of them either undetected in controls or produced in at least two-fold increased abundance in at least one perturbed system relative to control. Specific metabolic responses of S. coelicolor to particular perturbations were mainly detected in secondary metabolites. This stresses the importance of secondary metabolites in the adaption to chemical and biological stimuli. Among others, in co-cultures of S. coelicolor with competing microorganisms, a 22-fold increase of undecylprodigiosin and a 13fold increase of streptorubin B was observed. Because both metabolites exhibit antimicrobial properties, this observation underlies the function of antibiotics in the ecological context of the producer. A metabolite that was universally upregulated in response to perturbations is ectoin, an osmoprotectant with enzyme-activity-stabilizing and growthstimulating effects. Furthermore, discrete changes in primary metabolism, as in guanosine and phenylalanine pools upon lanthanide exposure, were detected. With a relatively small set of stimuli, the majority of already known secondary metabolites of S. coelicolor could be activated, and the prioritized metabolic response features were associated with 8 of 22 hitherto known gene clusters involved in secondary metabolite production (Goodwin et al., 2015). This shows the potential of the global investigation strategy to unravel a broad range of metabolite features in one experimental set-up, presenting a promising approach
568 Chemistry & Biology 22, May 21, 2015 ª2015 Elsevier Ltd All rights reserved
toward timesaving, high-throughput metabolomics. The potential of the presented SOM-based data analysis lies in the comparably unbiased prioritization, categorization, and visualization of relevant metabolite features, especially considering low-abundance metabolites of relevance. SOM-based analysis is a promising tool for big data sets and an important step toward holistic and comprehensive analysis of complex metabolic dynamics. One has to keep in mind, though, that the study of Goodwin et al. has been performed on a very well-known model organism that has been thoroughly mined for secondary metabolites. The next step is to demonstrate the true power of the workflow on organisms with hitherto unknown metabolomes.
REFERENCES Goodwin, C.R., Covington, B.C., Derewacz, D.K., McNees, C.R., Wikswo, J.P., McLean, J.A., and Bachmann, B.O. (2015). Chem. Biol. 22, this issue, 661–670. Kuhlisch, C., and Pohnert, G. (2015). Nat. Prod. Rep. Published online April 30, 2015. http://dx. doi.org/10.1039/C5NP00003C. Patterson, A.D., Li, H., Eichler, G.S., Krausz, K.W., Weinstein, J.N., Fornace, A.J., Jr., Gonzalez, F.J., and Idle, J.R. (2008). Anal. Chem. 80, 665–674. Patti, G.J., Tautenhahn, R., and Siuzdak, G. (2012). Nat. Protoc. 7, 508–516. Traxler, M.F., Watrous, J.D., Alexandrov, T., Dorrestein, P.C., and Kolter, R. (2013). MBio 4, e00459–13.
Previews Espiritu, R.A., Matsumori, N., Murata, M., Nishimura, S., Kakeya, H., Matsunaga, S., and Yoshida, M. (2013). Biochemistry 52, 2410–2418. Harris, J.R. (2010). Cholesterol Binding and Cholesterol Transport Proteins (New York: Springer). Krause, M.R., and Regen, S.L. (2014). Acc. Chem. Res. 47, 3512–3521.
Mehbub, M.F., Lei, J., Franco, C., and Zhang, W. (2014). Mar. Drugs 12, 4539–4577.
H., Kobayashi, T., and Yoshida, M. (2013). PLoS One 8, e83716.
Nishimura, S., Arita, Y., Honda, M., Iwamoto, K., Matsuyama, A., Shirai, A., Kawasaki, H., Kakeya, H., Kobayashi, T., Matsunaga, S., and Yoshida, M. (2010). Nat. Chem. Biol. 6, 519–526.
Nishimura, S., Tokukura, M., Ochi, J., Yoshida, M., and Kakeya, H. (2014). Chem. Biol. 21, 1690–1699.
Nishimura, S., Ishii, K., Iwamoto, K., Arita, Y., Matsunaga, S., Ohno-Iwashita, Y., Sato, S.B., Kakeya,
Rivero, F. (2012). Rho GTPase (New York: Springer).
Simplifying Complexity in Metabolomics Katharina Eick1 and Georg Pohnert1,* 1Institute of Inorganic and Analytical Chemistry, Friedrich Schiller University Jena, Lessingstraße 8, 07743 Jena, Germany *Correspondence: [email protected] http://dx.doi.org/10.1016/j.chembiol.2015.05.001
Metabolomics analysis to unravel secondary metabolite dynamics in microorganisms faces the challenge of immense data sets and complex experimental setups. In this issue, Goodwin et al. present a multiple stimuli approach combined with self-organizing map-based analysis to elucidate variations in the metabolome of Streptomyces coelicolor caused by biotic and environmental perturbations. By investigating primary and secondary metabolite dynamics, it is possible to characterize the physiological states of cells and elucidate metabolite based interactions of microorganisms. These studies are enabled by different types of metabolomics techniques that have matured over the last decade and can now be employed to study different aspects of a microbe’s life. Comparative metabolomics, for example, provides a valuable tool to monitor and unravel chemically mediated intra- and interspecies interactions (Kuhlisch and Pohnert, 2015). The investigation of microbial interactions and stress responses is of great interest from a drug mining perspective, as the production of novel secondary metabolites might be induced under specific conditions. For example, Traxler et al. (2013) use a combination of different mass spectrometry and chemoinformatics approaches to illustrate how interspecies interactions stimulate diversification in the Streptomyces coelicolor metabolome. Such processes might result in upregulation of compounds with antibiotic, antifungal, or anticancer properties. The high throughput metabolomics techniques introduced in recent years
result in outstandingly complex data sets. These data sets are difficult to visualize and interpret. In practical terms, it means that the value of metabolomics analysis is determined not only on the performance of a given analytical strategy determine, but on the subsequent treatment of data sets. In-depth data analysis often requires the use of specialized data analysis tools, adjusted to the specific character of the data set. For example, multiple sample groups might be evaluated with the help of the metaanalysis software metaXCMS, which reveals differences between metabolic profiles across experimental groups, thereby overcoming limitations of pairwise comparisons (Patti et al., 2012). This platform enables also the comprehensive analysis of independently performed studies. A pattern based self-organizing map (SOM) algorithm to visualize the timedependent cellular metabolic response of lymphoblastic cell lines to ionizing radiation was introduced by Patterson et al. (2008). This SOM algorithm is based on concepts from the Gene Expression Dynamics Inspector (GEDI) software and allows creating 2D visualizations in a metabolomics context. It is evident that new insights into microbial metabolite
dynamics can benefit substantially from refined and expanded experimental perturbations as well as from the development and application of adequate data analysis strategies to categorize, prioritize, and comprehensively visualize the findings. Goodwin et al. (2015) set out to investigating the metabolic response of S. coelicolor under multiple perturbation conditions using ultra performance liquid chromatography/ion-mobility mass spectrometric analysis (UPLC-IM-MS) and a GEDI-based SOM algorithm. The experimental setup comprises 23 single perturbations of three stimuli categories known to cause changes in the metabolome: exposures to rare earth elements, resistance inducing antibiotics, and competing microorganisms. To elucidate the metabolic responses, total cellular extracts were analyzed by UPLC-IM-MS in a comparative metabolomics approach. SOMbased analysis of the data sets was employed to structure and categorize response specificity of metabolic features. Compared to established multivariate statistical analysis (MVSA), the SOM approach is optimized for the large data set. It allows a simple 2D-visualization based on the similarity of metabolite
Chemistry & Biology 22, May 21, 2015 ª2015 Elsevier Ltd All rights reserved 567
Chemistry & Biology
Previews feature intensity trends among the different conditions. As a result, topological heat maps are created, representing the metabolic phenotype of a sample or condition but still retaining the attributes of MVSA-based analytics. An additional benefit of SOM analysis is a reduced masking of low-abundant metabolites compared to MVSA. The basis of these topological heatmap visualizations are ‘‘trained maps’’ that are obtained as follows. In a first step, all metabolite features detected in the 23 perturbations are characterized by their intensity profiles among all perturbations. These intensity profiles are then ‘‘self-organized’’ in a 2D graphical representation, locating those metabolites with similar intensity profiles in close coordinates of a grid. As a result, regions of correlated metabolite intensity profiles are visualized. In order to characterize each sample or experimental condition, information about specific abundances of the metabolic features is then introduced into these ‘‘trained maps.’’ The comparison of perturbed and unperturbed profiles allows prioritization of so-called regions of interest (ROIs), specifically visualizing significant metabolic responses in the experimental condition considered. After identification of significant metabolic features with the help of ROIs, determination of their molecular identity can be pursued by standard analytical approaches. Using these protocols, Goodwin et al. (2015) identified 2154 significant meta-
bolic features, 61% of them either undetected in controls or produced in at least two-fold increased abundance in at least one perturbed system relative to control. Specific metabolic responses of S. coelicolor to particular perturbations were mainly detected in secondary metabolites. This stresses the importance of secondary metabolites in the adaption to chemical and biological stimuli. Among others, in co-cultures of S. coelicolor with competing microorganisms, a 22-fold increase of undecylprodigiosin and a 13fold increase of streptorubin B was observed. Because both metabolites exhibit antimicrobial properties, this observation underlies the function of antibiotics in the ecological context of the producer. A metabolite that was universally upregulated in response to perturbations is ectoin, an osmoprotectant with enzyme-activity-stabilizing and growthstimulating effects. Furthermore, discrete changes in primary metabolism, as in guanosine and phenylalanine pools upon lanthanide exposure, were detected. With a relatively small set of stimuli, the majority of already known secondary metabolites of S. coelicolor could be activated, and the prioritized metabolic response features were associated with 8 of 22 hitherto known gene clusters involved in secondary metabolite production (Goodwin et al., 2015). This shows the potential of the global investigation strategy to unravel a broad range of metabolite features in one experimental set-up, presenting a promising approach
568 Chemistry & Biology 22, May 21, 2015 ª2015 Elsevier Ltd All rights reserved
toward timesaving, high-throughput metabolomics. The potential of the presented SOM-based data analysis lies in the comparably unbiased prioritization, categorization, and visualization of relevant metabolite features, especially considering low-abundance metabolites of relevance. SOM-based analysis is a promising tool for big data sets and an important step toward holistic and comprehensive analysis of complex metabolic dynamics. One has to keep in mind, though, that the study of Goodwin et al. has been performed on a very well-known model organism that has been thoroughly mined for secondary metabolites. The next step is to demonstrate the true power of the workflow on organisms with hitherto unknown metabolomes.
REFERENCES Goodwin, C.R., Covington, B.C., Derewacz, D.K., McNees, C.R., Wikswo, J.P., McLean, J.A., and Bachmann, B.O. (2015). Chem. Biol. 22, this issue, 661–670. Kuhlisch, C., and Pohnert, G. (2015). Nat. Prod. Rep. Published online April 30, 2015. http://dx. doi.org/10.1039/C5NP00003C. Patterson, A.D., Li, H., Eichler, G.S., Krausz, K.W., Weinstein, J.N., Fornace, A.J., Jr., Gonzalez, F.J., and Idle, J.R. (2008). Anal. Chem. 80, 665–674. Patti, G.J., Tautenhahn, R., and Siuzdak, G. (2012). Nat. Protoc. 7, 508–516. Traxler, M.F., Watrous, J.D., Alexandrov, T., Dorrestein, P.C., and Kolter, R. (2013). MBio 4, e00459–13.
