DOI: 10.1002/cbic.201500157
Highlights
Biocontainment through Reengineered Genetic Codes Arjun Ravikumar[b] and Chang C. Liu*[a] The realized and potential impact of genetically modified microbes (GMMs) on human living conditions cannot be overstated: GMMs enable the production of therapeutic proteins, the cheap and green biosynthesis of drugs and commodity chemicals, microbiome engineering, environmental sensing, and fundamental discoveries in biology. Fortunately, concerns over the accidental release of GMMs and the synthetic genes they contain have not materialized into major incidents.[1] This is due to a combination of factors: 1) professional genetic engineers and review boards take a precautionary approach to experiments, following many of the Asilomar conference recommendations; 2) genetic engineering is expensive and specialized and thus practiced mainly in regulated academic or industrial labs; 3) most synthetic genes and GMMs are not all that functionally distinct from what natural genes and microbes already do or can do; and 4) the differences GMMs contain usually make them less fit than their wild-type counterparts. In the instances that GMMs are predicted to be dangerous or more fit, for example, the products of gain-of-function studies, they are physically contained according to strict laboratory biosafety procedures. The landscape of biotechnology, however, is rapidly changing. As the scale, scope, and frequency of genetic manipulation increase—fueled by remarkable recent improvements in genome-scale editing and engineering, computational design, directed evolution, and synthetic biology—genes and GMMs with truly novel functions will become routine. Indeed, some potentially transformative genetic technologies with clear practical applications but also clear potential risks are already here—customizable gene drives, phage-resistant genetic codes, and sequence-programmable nucleases, to name just a few.[2] Furthermore, synthetic biology and do-it-yourself bio seek to “de-skill” genetic engineering for broad use; this means that in the coming decades, synthetic genes and GMMs will be frequently created outside regulated settings.[3] Likewise, the regulated GMM products of academia and industry will inevitably diffuse into the broader hobbyist synthetic biology public, which we should expect to be less aware of the specific risks associated with genetic engineering. As these changes take hold, physical biocontainment strategies might [a] Prof. Dr. C. C. Liu Departments of Biomedical Engineering and Chemistry University of California at Irvine 3105 Natural Sciences II, Irvine, CA 92697 (USA) E-mail: [email protected] [b] A. Ravikumar Department of Biomedical Engineering University of California at Irvine 3120 Natural Sciences II, Irvine, CA 92697 (USA)
ChemBioChem 2015, 16, 1149 – 1151
be insufficient. The field needs to make GMMs and the synthetic genes they encode intrinsically unable to function outside of the lab (or limited defined environments), yet still do everything we want in the lab. The question is, how? The common approach to intrinsic biocontainment is to add metabolic deficiencies or conditionally repressible “kill” switches into GMMs so that they propagate only in the presence of a nutrient or small molecule ligand.[1, 4] This so-called trophic containment strategy is effective as long as the nutrient or ligand supplement is naturally scarce and biosynthesis genes for the supplement are difficult to obtain through horizontal gene transfer or mutation. However, even if trophic containment is perfectly achieved, there is still the problem that genetic information itself (i.e., a piece of DNA) can leak into nature. Containing a potentially dangerous gene, which is often the very reason for containing a GMM, is a more difficult problem, as it requires semantic containment, where the meaning of a gene is cloaked.[1, 4, 5] Recently, landmark papers by the Isaacs and Church groups have taken major steps in an ambitious strategy that promises extraordinarily effective trophic and semantic containment.[6] The idea is to engineer microbes with expanded and reconfigured genetic codes. With few exceptions, natural organisms use a shared genetic code to specify 20 canonical amino acids in protein synthesis. An expanded genetic code that specifies an unnatural amino acid (uAA) would therefore require a uAA supplement if any proteins containing the uAA are to be made.[7] By incorporating in-frame codons that specify the uAA in essential genes, trophic containment is achieved (Figure 1). Furthermore, as the genetic code is universal across all life, protein-coding genes written for a reconfigured synthetic code would be mistranslated in nature. As a result, these genes would be semantically contained, functioning only in the engineered organism with a reconfigured genetic code. To implement this idea, the Isaacs and Church labs previously reported the construction of an Escherichia coli strain lacking release factor 1 and all instances of the amber (UAG) stop codon.[2b] Introduction of an orthogonal uAA-specific aminoacyl-tRNA synthetase (aaRS)/tRNA pair into this strain permitted efficient site-specific uAA incorporation at multiple amber codons. Building on this accomplishment, the same two labs recently achieved trophic containment of the engineered strain. Rovner et al. inserted multiple in-frame amber codons across 22 essential genes to yield 60 uAA-dependent strains. These strains exhibited modest escape frequencies (10¢3–10¢7), meaning that they rarely developed mutations releasing them from uAA-dependent growth. To lower these escape frequencies, the authors combined amber codons from the most effectively contained strains, reaching < 4.4 Õ 10¢11 (below detecta-
1149
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Highlights
Figure 1. Recoded genetic codes enable trophic and semantic biocontainment. Trophic biocontainment is achieved by linking the expression of essential genes to exogenously supplied uAAs. Semantic biocontainment is achieved by rewiring the correspondence between codon and amino acid.
ble levels). With the addition of host modifications that reduced the occurrence of amber suppression by natural amino acids, the best strain failed to produce escape mutants even in 1 L cultures (~ 1011 cells) grown for 20 days. Mandell et al. employed the same basic strategy for trophic biocontainment but instead, computationally redesigned the cores of essential proteins to require the uAA l-4,4’-biphenylalanine (bipA). The strategic nature of their approach was threefold. First, bipA is a structurally unique uAA unlike any natural amino acid. Second, their computational design algorithm identified designs that tightly accommodated bipA in the core of essential proteins. Third, the designs were optimized to destabilize the protein if a natural amino acid was incorporated in place of bipA. In making these design choices, the authors were able to create viable protein variants that were strongly dependent on bipA. Strains containing a single bipA-dependent essential enzyme exhibited escape frequencies as low as ~ 10¢6. To achieve more effective trophic containment, Mandell et al. observed that different bipA-containing enzymes tolerated different sets of limited amino acid substitutions at bipA. Therefore, amber suppression by a specific natural amino acid, which might arise from a mutation in a natural tRNA, would be able to undermine bipA dependence in one essential enzyme but not in another. Combining such bipA-dependent enzymes into the same strain yielded an escape frequency < 10¢10 after 72 h of growth. The authors also cleverly designed a bipA-dependent bipA synthetase to reduce the level of misincorporation of natural amino acids, thereby generating a strain that failed to produce any escape mutants even after 14 days of growth. Although similarly low escape frequencies have been previously reported, the strategies highlighted here have several key advantages.[8] Cross-feeding experiments demonstrate that ChemBioChem 2015, 16, 1149 – 1151
www.chembiochem.org
commonly employed metabolic and biotin auxotrophs can be rescued in rich and diverse medias, whereas the engineered uAA-dependent strains cannot. More importantly, the uAA-dependent strains contain very few genetic perturbations, ranging from as few as three amber codons in essential genes to at most 49 base pair substitutions in the genes of redesigned proteins. Still, the mutations are well dispersed throughout the genome so that horizontal gene transfer events are unlikely to compromise containment. In the future, it might even be possible to engineer variants of essential proteins that are superior because they use uAAs, which would further prevent escape. The current tradeoff, however, for using these exceptionally biocontained uAA-regulated strains is that uAAs can be expensive. p-Azido-l-phenylalanine, the uAA used by Rovner et al. to contain their best strain, adds a cost of ~ $300 per liter of media, which is a barrier to widespread adoption. The bipA uAA used by Mandell et al. is much less at ~ $3 per liter of medium, but would still be expensive at scale. Another barrier is that the strains reported in both studies have significantly longer doubling times than their wild-type counterpart. Nevertheless, one can imagine many solutions to these problems. Looking beyond trophic biocontainment to semantic biocontainment, the same multiplex genome engineering techniques used to free the amber stop codon for uAAs apply to sense codon reassignment. In principle, the mapping of codon to amino acid can be rewritten by engineering a microbe’s codon/tRNA/aaRSs correspondences (Figure 1). By doing so, any gene encoded in the GMM would be meaningless in a natural organism. Of course, this would require that all essential genes of the GMM be recoded to adhere to the new genetic code. Although a formidable challenge, pilot studies in the Church group have already begun to explore the possibility of genome-wide sense codon reassignment.[2b] Alternative approaches relying on the design of orthogonal translation sys-
1150
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Highlights tems could lead to the same goal. If an orthogonal ribosome and associated orthogonal sets of aaRSs and tRNAs were established in a cell, then the orthogonal translation system could run a different genetic code.[9] This would achieve semantic containment of any gene targeted for translation by the orthogonal ribosome. Indeed, the universality and centrality of the genetic code makes it ripe for trophic and semantic biocontainment strategies in GMMs. The only things more universal than the genetic code are nucleic acids themselves. All life encodes genes in DNA, so an alternative genetic polymer made of xeno-nucleic acid (XNA) building blocks would achieve both trophic and semantic biocontainment as well. Efforts to engineer XNA replication systems in vitro and in vivo are currently underway.[10] Ultimately, one could envision a microbe that uses XNAs for replication and transcription, synthetic genetic codes for protein synthesis, and a wealth of genetically encoded uAAs. Such GMMs would not only represent multiple layers of biocontainment options but would fundamentally expand the world of biological function and challenge the exceptionality of life’s chemistries. It is fitting that such powerful and potentially disruptive synthetic biology goals bring equally powerful biocontainment strategies, ensuring a balance between caution and innovation.
Keywords: amino acids · biocontainment · genetic code · genome engineering · microbes · synthetic biology
Acknowledgements
[1] a) M. Schmidt, V. de Lorenzo, FEBS Lett. 2012, 586, 2199 – 2206; b) O. Wright, G. B. Stan, T. Ellis, Microbiology 2013, 159, 1221 – 1235. [2] a) J. E. DiCarlo, A. Chavez, S. L. Dietz, K. M. Esvelt, G. M. Church, bioRxiv 2015; DOI: http://dx.doi.org/10.1101/013896; b) M. J. Lajoie, A. J. Rovner, D. B. Goodman, H.-R. Aerni, A. D. Haimovich, G. Kuznetsov, J. A. Mercer, H. H. Wang, P. A. Carr, J. A. Mosberg, N. Rohland, P. G. Schultz, J. M. Jacobson, J. Rinehart, G. M. Church, F. J. Isaacs, Science 2013, 342, 357 – 360. [3] G. M. Church, E. Regis, Regenesis, Basic Books, New York, 2012, pp. 225 – 253. [4] G. H. G. Moe-Behrens, R. Davis, K. A. Haynes, Front. Microbiol. 2013, 4, 1 – 10. [5] M. Schmidt, Bioessays 2010, 32, 322 – 331. [6] a) A. J. Rovner, A. D. Haimovich, S. R. Katz, Z. Li, M. W. Grome, B. M. Gassaway, M. Amiram, J. R. Patel, R. R. Gallagher, J. Rinehart, F. J. Isaacs, Nature 2015, 518, 89 – 93; b) D. J. Mandell, M. J. Lajoie, M. T. Mee, R. Takeuchi, G. Kuznetsov, J. E. Norville, C. J. Gregg, B. L. Stoddard, G. M. Church, Nature 2015, 518, 55 – 60. [7] C. C. Liu, P. G. Schultz, Annu. Rev. Biochem. 2010, 79, 413 – 444. [8] R. R. Gallagher, J. R. Patel, A. L. Interiano, A. J. Rovner, F. J. Isaacs, Nucleic Acids Res. 2015, 43, 1945 – 1954. [9] O. Rackham, J. W. Chin, Nat. Chem. Biol. 2005, 1, 159 – 166. [10] a) V. B. Pinheiro, A. I. Taylor, C. Cozens, M. Abramov, M. Renders, S. Zhang, J. C. Chaput, J. Wengel, S.-Y. Peak-Chew, S. H. McLaughlin, P. Herdewijn, P. Holliger, Science 2012, 336, 341 – 344; b) A. Ravikumar, A. Arrieta, C. C. Liu, Nat. Chem. Biol. 2014, 10, 175 – 177.
We thank the University of California at Irvine for startup funds provided to C.C.L.
Final article published: April 27, 2015
ChemBioChem 2015, 16, 1149 – 1151
www.chembiochem.org
Manuscript received: March 24, 2015
1151
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Highlights
Biocontainment through Reengineered Genetic Codes Arjun Ravikumar[b] and Chang C. Liu*[a] The realized and potential impact of genetically modified microbes (GMMs) on human living conditions cannot be overstated: GMMs enable the production of therapeutic proteins, the cheap and green biosynthesis of drugs and commodity chemicals, microbiome engineering, environmental sensing, and fundamental discoveries in biology. Fortunately, concerns over the accidental release of GMMs and the synthetic genes they contain have not materialized into major incidents.[1] This is due to a combination of factors: 1) professional genetic engineers and review boards take a precautionary approach to experiments, following many of the Asilomar conference recommendations; 2) genetic engineering is expensive and specialized and thus practiced mainly in regulated academic or industrial labs; 3) most synthetic genes and GMMs are not all that functionally distinct from what natural genes and microbes already do or can do; and 4) the differences GMMs contain usually make them less fit than their wild-type counterparts. In the instances that GMMs are predicted to be dangerous or more fit, for example, the products of gain-of-function studies, they are physically contained according to strict laboratory biosafety procedures. The landscape of biotechnology, however, is rapidly changing. As the scale, scope, and frequency of genetic manipulation increase—fueled by remarkable recent improvements in genome-scale editing and engineering, computational design, directed evolution, and synthetic biology—genes and GMMs with truly novel functions will become routine. Indeed, some potentially transformative genetic technologies with clear practical applications but also clear potential risks are already here—customizable gene drives, phage-resistant genetic codes, and sequence-programmable nucleases, to name just a few.[2] Furthermore, synthetic biology and do-it-yourself bio seek to “de-skill” genetic engineering for broad use; this means that in the coming decades, synthetic genes and GMMs will be frequently created outside regulated settings.[3] Likewise, the regulated GMM products of academia and industry will inevitably diffuse into the broader hobbyist synthetic biology public, which we should expect to be less aware of the specific risks associated with genetic engineering. As these changes take hold, physical biocontainment strategies might [a] Prof. Dr. C. C. Liu Departments of Biomedical Engineering and Chemistry University of California at Irvine 3105 Natural Sciences II, Irvine, CA 92697 (USA) E-mail: [email protected] [b] A. Ravikumar Department of Biomedical Engineering University of California at Irvine 3120 Natural Sciences II, Irvine, CA 92697 (USA)
ChemBioChem 2015, 16, 1149 – 1151
be insufficient. The field needs to make GMMs and the synthetic genes they encode intrinsically unable to function outside of the lab (or limited defined environments), yet still do everything we want in the lab. The question is, how? The common approach to intrinsic biocontainment is to add metabolic deficiencies or conditionally repressible “kill” switches into GMMs so that they propagate only in the presence of a nutrient or small molecule ligand.[1, 4] This so-called trophic containment strategy is effective as long as the nutrient or ligand supplement is naturally scarce and biosynthesis genes for the supplement are difficult to obtain through horizontal gene transfer or mutation. However, even if trophic containment is perfectly achieved, there is still the problem that genetic information itself (i.e., a piece of DNA) can leak into nature. Containing a potentially dangerous gene, which is often the very reason for containing a GMM, is a more difficult problem, as it requires semantic containment, where the meaning of a gene is cloaked.[1, 4, 5] Recently, landmark papers by the Isaacs and Church groups have taken major steps in an ambitious strategy that promises extraordinarily effective trophic and semantic containment.[6] The idea is to engineer microbes with expanded and reconfigured genetic codes. With few exceptions, natural organisms use a shared genetic code to specify 20 canonical amino acids in protein synthesis. An expanded genetic code that specifies an unnatural amino acid (uAA) would therefore require a uAA supplement if any proteins containing the uAA are to be made.[7] By incorporating in-frame codons that specify the uAA in essential genes, trophic containment is achieved (Figure 1). Furthermore, as the genetic code is universal across all life, protein-coding genes written for a reconfigured synthetic code would be mistranslated in nature. As a result, these genes would be semantically contained, functioning only in the engineered organism with a reconfigured genetic code. To implement this idea, the Isaacs and Church labs previously reported the construction of an Escherichia coli strain lacking release factor 1 and all instances of the amber (UAG) stop codon.[2b] Introduction of an orthogonal uAA-specific aminoacyl-tRNA synthetase (aaRS)/tRNA pair into this strain permitted efficient site-specific uAA incorporation at multiple amber codons. Building on this accomplishment, the same two labs recently achieved trophic containment of the engineered strain. Rovner et al. inserted multiple in-frame amber codons across 22 essential genes to yield 60 uAA-dependent strains. These strains exhibited modest escape frequencies (10¢3–10¢7), meaning that they rarely developed mutations releasing them from uAA-dependent growth. To lower these escape frequencies, the authors combined amber codons from the most effectively contained strains, reaching < 4.4 Õ 10¢11 (below detecta-
1149
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Highlights
Figure 1. Recoded genetic codes enable trophic and semantic biocontainment. Trophic biocontainment is achieved by linking the expression of essential genes to exogenously supplied uAAs. Semantic biocontainment is achieved by rewiring the correspondence between codon and amino acid.
ble levels). With the addition of host modifications that reduced the occurrence of amber suppression by natural amino acids, the best strain failed to produce escape mutants even in 1 L cultures (~ 1011 cells) grown for 20 days. Mandell et al. employed the same basic strategy for trophic biocontainment but instead, computationally redesigned the cores of essential proteins to require the uAA l-4,4’-biphenylalanine (bipA). The strategic nature of their approach was threefold. First, bipA is a structurally unique uAA unlike any natural amino acid. Second, their computational design algorithm identified designs that tightly accommodated bipA in the core of essential proteins. Third, the designs were optimized to destabilize the protein if a natural amino acid was incorporated in place of bipA. In making these design choices, the authors were able to create viable protein variants that were strongly dependent on bipA. Strains containing a single bipA-dependent essential enzyme exhibited escape frequencies as low as ~ 10¢6. To achieve more effective trophic containment, Mandell et al. observed that different bipA-containing enzymes tolerated different sets of limited amino acid substitutions at bipA. Therefore, amber suppression by a specific natural amino acid, which might arise from a mutation in a natural tRNA, would be able to undermine bipA dependence in one essential enzyme but not in another. Combining such bipA-dependent enzymes into the same strain yielded an escape frequency < 10¢10 after 72 h of growth. The authors also cleverly designed a bipA-dependent bipA synthetase to reduce the level of misincorporation of natural amino acids, thereby generating a strain that failed to produce any escape mutants even after 14 days of growth. Although similarly low escape frequencies have been previously reported, the strategies highlighted here have several key advantages.[8] Cross-feeding experiments demonstrate that ChemBioChem 2015, 16, 1149 – 1151
www.chembiochem.org
commonly employed metabolic and biotin auxotrophs can be rescued in rich and diverse medias, whereas the engineered uAA-dependent strains cannot. More importantly, the uAA-dependent strains contain very few genetic perturbations, ranging from as few as three amber codons in essential genes to at most 49 base pair substitutions in the genes of redesigned proteins. Still, the mutations are well dispersed throughout the genome so that horizontal gene transfer events are unlikely to compromise containment. In the future, it might even be possible to engineer variants of essential proteins that are superior because they use uAAs, which would further prevent escape. The current tradeoff, however, for using these exceptionally biocontained uAA-regulated strains is that uAAs can be expensive. p-Azido-l-phenylalanine, the uAA used by Rovner et al. to contain their best strain, adds a cost of ~ $300 per liter of media, which is a barrier to widespread adoption. The bipA uAA used by Mandell et al. is much less at ~ $3 per liter of medium, but would still be expensive at scale. Another barrier is that the strains reported in both studies have significantly longer doubling times than their wild-type counterpart. Nevertheless, one can imagine many solutions to these problems. Looking beyond trophic biocontainment to semantic biocontainment, the same multiplex genome engineering techniques used to free the amber stop codon for uAAs apply to sense codon reassignment. In principle, the mapping of codon to amino acid can be rewritten by engineering a microbe’s codon/tRNA/aaRSs correspondences (Figure 1). By doing so, any gene encoded in the GMM would be meaningless in a natural organism. Of course, this would require that all essential genes of the GMM be recoded to adhere to the new genetic code. Although a formidable challenge, pilot studies in the Church group have already begun to explore the possibility of genome-wide sense codon reassignment.[2b] Alternative approaches relying on the design of orthogonal translation sys-
1150
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Highlights tems could lead to the same goal. If an orthogonal ribosome and associated orthogonal sets of aaRSs and tRNAs were established in a cell, then the orthogonal translation system could run a different genetic code.[9] This would achieve semantic containment of any gene targeted for translation by the orthogonal ribosome. Indeed, the universality and centrality of the genetic code makes it ripe for trophic and semantic biocontainment strategies in GMMs. The only things more universal than the genetic code are nucleic acids themselves. All life encodes genes in DNA, so an alternative genetic polymer made of xeno-nucleic acid (XNA) building blocks would achieve both trophic and semantic biocontainment as well. Efforts to engineer XNA replication systems in vitro and in vivo are currently underway.[10] Ultimately, one could envision a microbe that uses XNAs for replication and transcription, synthetic genetic codes for protein synthesis, and a wealth of genetically encoded uAAs. Such GMMs would not only represent multiple layers of biocontainment options but would fundamentally expand the world of biological function and challenge the exceptionality of life’s chemistries. It is fitting that such powerful and potentially disruptive synthetic biology goals bring equally powerful biocontainment strategies, ensuring a balance between caution and innovation.
Keywords: amino acids · biocontainment · genetic code · genome engineering · microbes · synthetic biology
Acknowledgements
[1] a) M. Schmidt, V. de Lorenzo, FEBS Lett. 2012, 586, 2199 – 2206; b) O. Wright, G. B. Stan, T. Ellis, Microbiology 2013, 159, 1221 – 1235. [2] a) J. E. DiCarlo, A. Chavez, S. L. Dietz, K. M. Esvelt, G. M. Church, bioRxiv 2015; DOI: http://dx.doi.org/10.1101/013896; b) M. J. Lajoie, A. J. Rovner, D. B. Goodman, H.-R. Aerni, A. D. Haimovich, G. Kuznetsov, J. A. Mercer, H. H. Wang, P. A. Carr, J. A. Mosberg, N. Rohland, P. G. Schultz, J. M. Jacobson, J. Rinehart, G. M. Church, F. J. Isaacs, Science 2013, 342, 357 – 360. [3] G. M. Church, E. Regis, Regenesis, Basic Books, New York, 2012, pp. 225 – 253. [4] G. H. G. Moe-Behrens, R. Davis, K. A. Haynes, Front. Microbiol. 2013, 4, 1 – 10. [5] M. Schmidt, Bioessays 2010, 32, 322 – 331. [6] a) A. J. Rovner, A. D. Haimovich, S. R. Katz, Z. Li, M. W. Grome, B. M. Gassaway, M. Amiram, J. R. Patel, R. R. Gallagher, J. Rinehart, F. J. Isaacs, Nature 2015, 518, 89 – 93; b) D. J. Mandell, M. J. Lajoie, M. T. Mee, R. Takeuchi, G. Kuznetsov, J. E. Norville, C. J. Gregg, B. L. Stoddard, G. M. Church, Nature 2015, 518, 55 – 60. [7] C. C. Liu, P. G. Schultz, Annu. Rev. Biochem. 2010, 79, 413 – 444. [8] R. R. Gallagher, J. R. Patel, A. L. Interiano, A. J. Rovner, F. J. Isaacs, Nucleic Acids Res. 2015, 43, 1945 – 1954. [9] O. Rackham, J. W. Chin, Nat. Chem. Biol. 2005, 1, 159 – 166. [10] a) V. B. Pinheiro, A. I. Taylor, C. Cozens, M. Abramov, M. Renders, S. Zhang, J. C. Chaput, J. Wengel, S.-Y. Peak-Chew, S. H. McLaughlin, P. Herdewijn, P. Holliger, Science 2012, 336, 341 – 344; b) A. Ravikumar, A. Arrieta, C. C. Liu, Nat. Chem. Biol. 2014, 10, 175 – 177.
We thank the University of California at Irvine for startup funds provided to C.C.L.
Final article published: April 27, 2015
ChemBioChem 2015, 16, 1149 – 1151
www.chembiochem.org
Manuscript received: March 24, 2015
1151
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
