Bringing molecules back into molecular evolution.
Review Bringing Molecules Back into Molecular Evolution Claus O. Wilke* Section of Integrative Biology, Center for Computational Biology and Bioinformatics, and Institute of Cell and Molecular Biology, The University of Texas at Austin, Austin, Texas, United States of America In this article, I focus on the evolution of protein-coding genes, the area I am most familiar with myself. However, my overall message, that it is time to bring the molecules back into molecular evolution, similarly applies to other genetic sequences, such as intergenic regions, RNA genes, or the various forms of short RNAs. I will consider three broad areas, corresponding to three distinct research goals: (i) reconstructing and interpreting past evolutionary events; (ii) identifying fundamental principles of molecular evolution; and (iii) predicting probable evolutionary trajectories. Abstract: Much molecular-evolution research is concerned with sequence analysis. Yet these sequences represent real, three-dimensional molecules with complex structure and function. Here I highlight a growing trend in the field to incorporate molecular structure and function into computational molecular-evolution work. I consider three focus areas: reconstruction and analysis of past evolutionary events, such as phylogenetic inference or methods to infer selection pressures; development of toy models and simulations to identify fundamental principles of molecular evolution; and atom-level, highly realistic computational modeling of molecular structure and function aimed at making predictions about possible future evolutionary events. Reconstructing and Interpreting Past Evolutionary Events A major goal of comparative sequence analysis is to reconstruct and/or interpret past evolutionary events. For example, we may have sequences from multiple species and want to know how they relate to each other, which specific sequence changes caused them to diverge, and whether certain sites were under particularly strong selective pressure. The standard analysis pipeline for such questions is to align sequences, build trees, and run scans for positive or other types of selection, and/or for recombination. This analysis pipeline uses nothing but sequences as input. Only once the analysis is completed may the researchers take sites of interest they have identified, map them back onto the structure of the protein they are studying, and carry out further experimentation. (However, increasingly the initial sequence analysis is only the prerequisite for a successful study, and the value of the study is defined by the follow-up work; see e.g., [1,2].) The standard analysis approach has been highly successful. Yet it ignores most of the biochemistry that ultimately determines the fitness landscape in which sequences evolve. Thus, methods that combine sequence data with additional information, such as protein structure, should yield more sensitive and more accurate estimates than methods based on sequence data alone. On the basis of this premise, a few groups have started to develop such methods. For example, some authors have developed models of coding-sequence evolution that incorporate interactions among sites mediated by protein structure [3–5]. (See also this review: This is an ‘‘Editors’ Outlook’’ article for PLoS Computational Biology. Introduction The field of molecular evolution investigates how genes and genomes evolve over time. It has its origin in the late 1960s, when the first DNA and protein sequences were becoming available. With rapid progress in sequencing technologies came ever increasing demand for computational tools to study molecular evolution. Today, molecular evolution is among the largest subfields of evolutionary biology, and arguably one of the most computationally advanced. Thousands of person years have gone into developing sophisticated alignment algorithms, phylogenetictree reconstruction methods, or statistical tests for positive selection. A side effect of the strong emphasis on developing sophisticated methods for sequence analysis has been that the underlying biophysical objects represented by the sequences, DNA molecules, RNA molecules, and proteins, have taken a back-seat in much computational molecular-evolution work. The vast majority of algorithms for sequence analysis, for example, incorporate no knowledge of biology or biochemistry besides that DNA and RNA sequences use an alphabet of four letters, protein sequences use an alphabet of 20, and the genetic code converts one into the other. The choice to treat DNA, RNA, and proteins simply as strings of letters was certainly reasonable in the late 20th century. Computational power was limited and many basic aspects of sequence analysis were still relatively poorly understood. However, in 2012 we have extremely powerful computers and a large array of highly sophisticated algorithms that can analyze strings of letters. It is now time to bring the molecules back into molecular evolution. Several groups have embarked on this path, and I will highlight some of the work that has been done and speculate on future developments we may see. PLoS Computational Biology | www.ploscompbiol.org Citation: Wilke CO (2012) Bringing Molecules Back into Molecular Evolution. PLoS Comput Biol 8(6): e1002572. doi:10.1371/journal.pcbi.1002572 Editor: Sergei L. Kosakovsky Pond, University of California San Diego, United States of America Published June 28, 2012 Copyright: ß 2012 Claus O. Wilke. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: Funding was received from NIH grant R01 GM088344 and NSF Cooperative Agreement No. DBI-0939454. The funders had no role in the preparation of the manuscript. Competing Interests: The author has declared that no competing interests exist. * E-mail: firstname.lastname@example.org 1 June 2012 | Volume 8 | Issue 6 | e1002572 .) Similarly, some authors have incorporated knowledge of protein structure in methods of ancestral state reconstruction . Finally, in phylogenetic-tree inference, evidence is accumulating that independence of sites may not be a good assumption  for protein-coding and even more so for RNA-coding sequences. Thus, future methods of phylogenetic tree reconstruction may also incorporate structural information in some form. Coarse-grained models of protein-sequence evolution are being developed that may be useful for this purpose . The development of methods that integrate molecular structure into sequence analysis is still in its infancy. While several groups are exploring a variety of approaches, none of these approaches is well established at this time. Comparative analyses that use nothing but sequence data remain state of the art. My expectation for the near future is that we will continue to see efforts to extend comparative analyses beyond sequence data alone. Eventually, some of these efforts will prove sufficiently useful that it will become commonplace to combine sequence data with structural, functional, or other molecular data in comparative analyses. [9,14–16,18–22] or protein–protein interactions [22–24]. Models may represent individual evolving proteins [18,20,21,23] or entire cells [19,22,24]. Finally, some groups elucidate the molecular fitness landscapes that underlie adaptive events [25–27]. Works such as these aim to identify the biophysical mechanisms that drive molecular evolution. I believe that we have only scratched the surface of what is possible with simple, biophysically inspired models of molecular evolution. I expect that we are going to see more of this modeling approach in the coming years, and that it will help us to develop a deeper understanding of fundamental principles of molecular evolution. Predicting Probable Evolutionary Trajectories For many real-world applications, it would be useful to be able to predict future evolutionary events. For example, we know that H5N1 avian influenza could potentially cause a deadly pandemic if it ever evolved the ability to effectively spread between humans. What we do not know  is the likelihood that it will evolve this ability, nor whether it might possibly become less pathogenic as it evolves more effective human-to-human transmission capabilities. As a second example, some authors have proposed treating infectious diseases with interfering particles (e.g., ). Because of the potential for transmission of these particles among infected patients, the safety of such treatments stands and falls with our ability to accurately predict how such therapeutic particles might evolve once released. Since evolution is a stochastic process, we cannot expect to ever predict which specific mutations will accumulate in a given lineage. However, at least in principle, we should be able to make probabilistic predictions of the form ‘‘Outcome A is the most likely, and has a 37% probability of occurring; outcome B is the second most likely, and has a 24% probability of occurring.’’ It would be tremendously useful if we could make such predictions reliably, in particular for rapidly evolving pathogens. Therefore, there is growing interest among evolutionary biologists to develop predictive frameworks [30–33]. In my opinion, successful approaches in this area will most likely involve realistic, atomlevel computational modeling of the system of interest. With rapid increases in computational power over the last two decades, realistic modeling of biological systems is becoming increasingly feasible. At the molecular level, obvious applications of realistic modeling are atom-level predictions of protein structure  or protein-folding dynamics [35,36], and computational enzyme design [37–39]. The accuracy of these computational models, when they work, is getting quite good. For example, in computational enzyme design, where the goal is to design catalytically active enzymes de novo, crystal structures of successfully designed enzymes are often very close to the computationally predicted ones . However, it is common that only a small number of the computational designs actually work as expected. In a recent study, 84 computationally designed enzymes were evaluated experimentally . Of those, 50 were soluble and only two catalyzed the desired reaction. At present, atom-level modeling of proteins is not commonly used in applications of evolutionary biology (but see ). However, it seems to me that as our modeling capability improves, a logical next step will be to apply these models to predicting evolution. If we can predict computationally which mutants will be able to carry out specific functions, then we should also be able to predict which mutants are likely to arise under specific, welldefined selection pressures. While I cannot imagine that we will ever be able to solve open-ended problems, such as, for example, Identifying Fundamental Principles of Molecular Evolution Besides understanding and interpreting specific evolutionary events, evolutionary biologists also aim to identify fundamental principles of molecular evolution. Fundamental principles are insights that apply to many different biological systems; a classical example would be the finding that codon usage bias correlates with gene expression level [10,11]. The search for fundamental principles tends to require somewhat different computational approaches than the analysis of past evolutionary events. It often involves developing toy models (either in the form of mathematical equations or of simulations) to explore possible system dynamics under different modeling assumptions or parameter choices. The specific toy models to be explored are usually inspired by observations from past evolutionary events. To give an example from my own research, starting about 10 years ago many groups found that highly expressed proteins evolve slowly . This observation prompted several authors to develop models of varying complexity that might explain the pattern [13–17]. Toy models of evolution have been studied for over a century. And much of this work has not considered the underlying biochemistry of the evolving organism. For example, the population-genetics literature contains plenty of abstract, mathematical models (such as two-locus, two-allele models) that make absolutely no assumptions about the mechanisms that connect different allelic states with different fitness values. These abstract mathematical models are valuable, of course, yet they can get us only so far. Most importantly, they cannot explain how, mechanistically, genotype maps to phenotype and fitness. As we try to get a better understanding of the genotypephenotype map, we have to build more realistic models. For example, virtually all the models trying to explain the relationship between evolutionary rate and expression level make concrete assumptions about mechanisms of protein folding and function [13–17]. Many implement an actual (though simplified) proteinfolding model in which actual amino-acid sequences are computationally folded, using either a lattice [14,15] or an off-lattice  approach. More generally, we are seeing an increased trend towards integrating some biophysical realism into toy models of evolution. Several groups are regularly working with models that incorporate some aspect of protein biochemistry, such as protein fold stability PLoS Computational Biology | www.ploscompbiol.org 2 June 2012 | Volume 8 | Issue 6 | e1002572 to predict all the sequence changes an invasive species will undergo as it is introduced into a new environment, we should have reasonable success for well-defined problems, such as to find the mutations an animal virus would require to bind to the human form of the receptor it uses for cell entry in its host species. An alternative to atom-level modeling can be statistical inference of biophysically important sites from large sequence alignments. For example, in a recent paper Bloom and Glassman  proposed a method to infer the effect of point mutations on protein stability from the distribution of mutations in a dense phylogeny. This method performed better in predicting measured DDG values than alternative methods based on protein structure and atomic force fields. Bloom and co-workers then used this method to identify mutations that were likely to be involved in the evolution of oseltamivir resistance in influenza . Regardless of whether one uses atom-level modeling or statistical approaches, computational predictions are not going to be perfect. Thus, computational methods to predict evolution are most likely going to be useful in generating candidate scenarios. These candidate scenarios will include many false positives and will have to be screened experimentally to separate false from true positives. Author’s Biography Claus O. Wilke is an Associate Professor in the Section of Integrative Biology at The University of Texas at Austin. He is also a member of the Center for Computational Biology and Bioinformatics and the Institute for Cell and Molecular Biology at The University of Texas at Austin. Dr. Wilke received his PhD in theoretical physics from the RuhrUniversity Bochum, Germany, in 1999. From 2000 to 2004, he was a postdoc at the California Institute of Technology, working on viral evolution and artificial life. After his postdoc, he spent a year as a Research Assistant Professor at the Keck Graduate Institute of Applied Life Sciences, Claremont, before joining The University of Texas in the fall of 2005. His current research interests are in molecular evolution, structural biology, and biostatistics. Dr. Wilke is the author of approximately 100 scientific papers. He serves as Associate Editor for PLoS Computational Biology and PLoS Pathogens, and as Section Editor for BMC Evolutionary Biology. In 2011, Dr. Wilke was recognized as a Leading Texas Innovator by The Academy of Medicine, Engineering, and Science of Texas. Summary impediment to more routine use of such models is likely the lack of widely available, easy-to-use implementations. Hopefully, we will see progress in this area soon. Methods to predict future evolutionary trajectories do not really exist at this time. However, there is a growing interest in developing them. I believe that the computational methods required for this type of prediction are falling into place in the protein-design field; we may soon see a first, small-scale demonstration that computational prediction of evolutionary trajectories is actually possible. There is a growing trend in widely differing subfields of molecular evolution to increase biophysical realism in computational models of sequence evolution. Some subfields are further along this path than others. Among groups developing simple toy models of evolution, models incorporating some biophysical realism have been quite popular in recent years. By contrast, statistical models of sequence evolution incorporating biophysical realism are being developed by some groups but are not being routinely applied in sequence-analysis applications. A major References 1. Sawyer SL, Wu LI, Emerman M, Malik HS (2005) Positive selection of primate TRIM5a identifies a critical species-specific retroviral restriction domain. Proc Natl Acad Sci U S A 102: 2832–2837. 2. Balakirev ES, Anisimova M, Ayala FJ (2011) Complex interplay of evolutionary forces in the ladybird homeobox genes of Drosophila melanogaster. PLoS ONE 6: e22613. doi:10.1371/journal.pone.0022613 3. Robinson DM, Jones DT, Kishino H, Goldman N, Thorne JL (2003) Protein evolution with dependence among codons due to tertiary structure. Mol Biol Evol 20: 1692–1704. 4. Lartillot NN, Bryant D, Philippe H (2005) Site interdependence attributed to tertiary structure in amino acid sequence evolution. Gene 347: (uma glicolipoproteína essencial para a manutenção dos ovócitos) necessária para o ciclo reprodutivo (KIME, 1999). Esse fato pode ser corroborado por Sato et al (2005), que mostraram um comprometimento na maturação dos ovócitos de fêmeas de P. argenteus coletadas no trecho JB. O aspecto estrutural geral do fígado em P. argenteus mostra diferenças marcantes nas características histológicas e histoquímicas entre os diferentes locais de coleta, indicando assim que a abordagem histológica aqui utilizada para o fígado é altamente sensível às condições ambientais à jusante da barragem Três Marias. No fígado dos animais coletados em JB, alterações morfológicas, tais como vacuolização, núcleos deslocados, hiperemia, inflamações e necrose foram menos frequentes que em JA. A ocorrência dos CMMs também pode ser um bom biomarcador de exposição a agentes de estresse diversos (RABITTO et al, 2005). Alguns autores demonstraram que CMMs em órgãos como fígado, rins, baço e intestino aumentam de tamanho e/ou frequência quando expostos a metais pesados e alcalóides (FOURNIER et al, 2001; AGIUS e ROBERTS, 2003; GLENCROSS et al, 2006; FISHELSON, 2006). Entretanto, não houve diferenças relativas a estes aspectos no presente estudo. Executando a diferença entre a composição dos pigmentos dentro de centros de diâmetros diferentes. Esse dado pode apontar para a dissimilaridade entre os CMMs de diferentes órgãos que é ressaltada no ANEXO I. Nesse contexto, nosso estudo analisou a morfologia de CMMs do fígado e do baço e mostra que as variações no número, morfologia, distribuição dos pigmentos composição elementar e atividade NAG encontradas fornem evidências dessa dissimilaridade metabólica dos CMMs nos diferentes órgãos. Estudos anteriores - 54 - (WOLKE, 1992; AGIUS e ROBERTS, 2003; FOURNIE et al, 2001; FISHELSON, 2006) têm sugerido que os CMMs podem aumentar em número e tamanho sob condição de estresse, mas a possível relação entre fígado e baço CMMs não é discutida. Outro aspecto que aponta para essa diferença entre os órgãos são as características dos grânulos que contem os pigmentos. A ocorrência dos grânulos contendo pigmentos, tais como a melanina, a hemossiderina e a lipofuscina nos CMMs pode ser uma indicação de sua função metabólica (WOLKE, 1992; AGUIS e ROBERTS, 2003). A lipofuscina está associada ao desequilíbrio na renovação celular, envolvendo atividades lisossômicas e peroxidação lipídica de membrana (PICKFORD, 1953), enquanto hemossiderina está associada à degradação de hemácias e processos de armazenagem de ferro (AISEN et al, 2001). Em nosso estudo, nos CMMs do fígado a lipofuscina foi predominante nos CMMs menores do que 0,010 m2, enquanto que os CMMs maiores que 0,010 mm2 foram ricos em hemossiderina. Estes resultados mostram que no fígado de P. argenteus, os CMMs parecem não ter a mesma função, o que pode refletir uma diferença entre os CMMs do fígado e baço. Nossos resultados demonstram também diferenças na distribuição dos pigmentos dos CMMs nos diferentes órgãos. Esses dados são corroborados pelo estudo histoquímico de Leknes (2004) que mostrou uma grande quantidade de grânulos contendo precipitados de ferro no fígado de Xiphophorus maculatus. A prevalência de hemossiderina, indicada pela presença de ferro detectado pela histoquímica de Perls, microanálise de raios-x e a associação entre eritrócitos e macrófagos em CMMs do fígado, demonstra que essas estruturas presente no fígado estão diretamente envolvidas no metabolismo do ferro, e, portanto estão associadas com atividade fagocítica. Isto contrasta com a baixa ocorrência deste pigmento em CMMs do baço, sugerindo uma diminuição na capacidade fagocitária destes macrófagos. Rabitto et al (2005) associou a presença de hemossiderina no fígado de Hoplias malabaricus com eventos fagocitários. Além disso, Bucke et al (1992) mostraram que a diminuição em áreas ocupadas por hemossiderina em CMMs do baço de Limanda limanda foi associada com uma redução na atividade de fagocitose. Outro dado importante que demonstra claramente as diferenças na atividade fagocítica entre fígado e baço é a presença de grandes áreas desprovidas de pigmento nas amostras de baço, que não foram encontradas no fígado. Os processos de fagocitose dos CMMs podem estar relacionados também à presença de lipofuscina (BRUSLÈ e ANADÓN, 1996), uma vez que a presença deste - 55 - pigmento está diretamente associada com a peroxidação lipídica (ou seja, o estresse oxidativo) de organelas membranosas (AGIUS e ROBERTS, 2003; DREVNICK et al, 2008; DING et al, 2010). Assim, a maior ocorrência de lipofuscina em CMMs do fígado do que a encontrada no baço indica a maior atividade fagocitária do primeiro em comparação com o último em P. argenteus. Em contraste aos mamíferos, a melanogênese de peixes pode ocorrer em outras células estacionárias, tais como os macrófagos presentes em CMMs (HAUGARVOLL et al, 2006). Portanto, o aumento significativo de melanina em CMMs do baço pode indicar um papel na resposta imunológica. Na verdade, ROSAS et al (2002) mostraram que a melanina ativada pode mediar respostas imunológicas através da deposição de componentes C3 do sistema complemento, resultando em respostas inflamatórias. As diferenças entre MMCs do fígado e baço foram ainda confirmadas, considerando os resultados com o MET e MEV neste estudo. Em CMMs de fígado, foi observada a presença de grandes áreas com depósitos de partículados de ferritina associada com porções de material hidrofóbico, que foram evidentes em picos de ferro e enxofre, caracterizadas por microanálise de raios-X. Utilizando métodos histoquímicos, Leknes (2004) descreveu a prevalência de ferro em grânulos nos CMMs hepáticos de Xiphophorus maculatus, contrastando com sua ausência nos CMMs do baço de Trichogaster leeri e X. maculatus (LEKNES, 2007). Estes dados, juntamente com os nossos resultados de MET e microanálise de raios-X, reforçam nossa hipótese de que, além de armazenamento de ferro, os CMMs do fígado estão envolvidos em processos catabólicos celulares. Além disso, a identificação de enxofre e ferro nos grânulos dos CMMs encontrados no fígado indica a presença de complexos ferro-enxofre, que medeia processos de oxirredução (JOHNSON, 1998; AISEN et al, 2001). Este fato demonstra claramente o importante papel dos CMMs hepáticos no metabolismo do ferro. Estes dados são consistentes com os resultados da análise NAG mostrando uma maior taxa de absorbância no fígado quando comparado com o baço. Esses resultados sugerem a maior atividade fagocitária dos macrófagos presentes nos CMMs do fígado, uma vez que o ensaio de NAG é específico para macrófagos ativados (RUSSO et al, 2010). A estreita associação física entre macrófagos e linfócitos vista nos CMMs esplênicos de P. argenteus baço indica sua relação com reações imunológicas tais como eventos de apresentação de antígenos. Vigliano et al (2006) apresentaram evidências moleculares em sua análise imuno-histoquímica - 56 - mostrando marcação positiva para anticorpos de células dendríticas foliculares nos CMMs do baço de Cyprinus carpio, Odontesthes bonariensis e Solea senegalensis. Concluímos que as diferenças na composição elementar dos CMMs que estão associados com as dissimilaridades em suas características morfológicas e de grânulo pigmento contendo pigmentos claramente apontam para diferenças metabólicas em CMMs do fígado e do baço de P. argenteus. A partir desses resultados, propomos que estudos utilizando CMMs de P. argenteus como um biomarcador de estresse ambiental devam ser levados em consideração em situações distintas: estudo de estressores ambientais que levem ao estabelecimento de repostas imunológicas devem focar em CMMs do baço como agente de estudo; e por outro lado estressores que promovam alterações no estado metabólico normal das células em especial na renovação celular e processos de armazenamento devem ser desenvolvidos em CMMs presentes no figado. - 57 - CONCLUSÕES - 58 - A caracterização d melhor estado de saúde do peixe Prochilodus argenteus pode ser avaliada pela análise do fator de condição de Fulton (K). Os P. argenteus coletados à jusante da barragem de Três Marias apresentam diferenças quanto aos aspectos histopatológicos e ultra-estruturais do fígado e do baço e em relação àqueles à jusante do Rio Abaeté Os resultrados indicam que as alterações teciduais e ultra-estruturais no fígado de P. argenteus são resultantes das variações ambientais relacionadas à temperatura, à oxigenação e à concentração de metais (Cd, Zn, Cu e As) A degeneração hidrópica dos hepatócitos de P. argenteus pode ser utilizada como biomarcador à jusante da represa de Três Marias. Os CMMs presentes no fígado e no baço de P. argenteus apresentam diferenças na morfologia, características ultra-estruturais dos grânulos. Os CMMs hepáticos estudados estão principalmente envolvidos em processos relacionados à fagocitose e à armazenagem do ferro, diferindo dos CMMs do baço que têm seu seating systems for consumers, clinicians, and manufacturers. Further research is necessary to discover the optimal backward degree for BTS and the optimal period for regulating the backward and forward alternation of the BTS during prolonged sitting. Acknowledgments The authors are very grateful to Cho-Lin Wang, Chih-Hsien Chen, and Ting-Wei Huang for technical assistance. PLOS ONE | DOI:10.1371/journal.pone.0113644 November 26, 2014 11 / 14 Backward Adjustable Thoracic Support Author Contributions Conceived and designed the experiments: CTL YNC CHC KHT. Performed the experiments: CTL YNC CHC KHT. Analyzed the data: CTL YNC CHC KHT. Contributed reagents/materials/analysis tools: CTL YNC CHC KHT. Wrote the paper: CTL YNC CHC KHT. References 1. Samuelsson K, Larsson H, Thyberg M, Gerdle B (2001) Wheelchair seating intervention. Results from a client-centred approach. Disabil Rehabil 23: 677–682. 2. Zacharkow D (1988) Posture: sitting, standing, chair design and exercise. Springfield: Charles C Thomas. 3. Lin F, Parthasarathy S, Taylor SJ, Pucci D, Hendrix RW, et al. (2006) Effect of different sitting postures on lung capacity, expiratory flow, and lumbar lordosis. Arch Phys Med Rehabil 87: 504–509. 4. Makhsous M, Lin F, Hendrix RW, Hepler M, Zhang LQ (2003) Sitting with adjustable ischial and back supports: biomechanical changes. Spine 28: 1113–1122. 5. Makhsous M, Lin F, Bankard J, Hendrix RW, Hepler M, et al. (2009) Biomechanical effects of sitting with adjustable ischial and lumbar support on occupational low back pain: evaluation of sitting load and back muscle activity. BMC Musculoskelet Disord 10: 17. 6. Bennett DL, Gillis DK, Portney LG, Romanow M, Sanchez AS (1989) Comparison of integrated electromyographic activity and lumbar curvature during standing and during sitting in three chairs. Phys Ther 69: 902–913. 7. Guimond S, Massrieh W (2012) Intricate Correlation between Body Posture, Personality Trait and Incidence of Body Pain: A Cross-Referential Study Report. PLoS One 7: e37450. 8. Wang HJ, Giambini H, Zhang WJ, Ye GH, Zhao C, et al. (2012) A Modified Sagittal Spine Postural Classification and Its Relationship to Deformities and Spinal Mobility in a Chinese Osteoporotic Population. PLoS One 7: e38560. 9. Adams MA, Bogduk N, Burton K, Dolan P (2002) The biomechanics of back pain. 2nd ed. Edinburgh: Churchill Livingstone Elsevier. 10. Kasahara S, Miyamoto K, Takahashi M, Yamanaka M, Takeda N (2008) Lumbar-pelvic coordination in the sitting position. Gait Posture 28: 251–257. 11. Congdon R, Bohannon R, Tiberio D (2005) Intrinsic and imposed hamstring length influence posterior pelvic rotation during hip flexion. Clin Biomech 20: 947–951. 12. Dewberry MJ, Bohannon RW, Tiberio D, Murray R, Zannotti CM (2003) Pelvic and femoral contributions to bilateral hip flexion by subjects suspended from a bar. Clin Biom 18: 494–499. 13. Andersson GB, Murphy RW, Ortengren R, Nachemson AL (1979) The influence of backrest inclination and lumbar support on lumbar lordosis. Spine 4: 52–58. 14. Bridger RS, Von Eisenhart-Rothe C, Henneberg M (1989) Effects of seat slope and hip flexion on spinal angles in sitting. Hum Factors 31: 679–688. 15. Makhsous M, Rowles DM, Rymer WZ, Bankard J, Nam EK, et al. (2007) Periodically relieving ischial sitting load to decrease the risk of pressure ulcers. Arch Phys Med Rehabil 88: 862–870. 16. Huang HC, Yeh CH, Chen CM, Lin YS, Chung KC (2011) Sliding and pressure evaluation on conventional and V-shaped seats of reclining wheelchairs for stroke patients with flaccid hemiplegia: a crossover trial. J Neuroeng Rehabil 8: 40. 17. Janssen-Potten YJ, Seelen HA, Drukker J, Huson T, Drost MR (2001) The effect of seat tilting on pelvic position, balance control, and compensatory postural muscle use in paraplegic subjects. Arch Phys Med Rehabil 82: 1393–1402. 18. May LA, Butt C, Kolbinson K, Minor L, Tulloch K (2004) Wheelchair back-support options: functional outcomes for persons with recent spinal cord injury. Arch Phys Med Rehabil 85: 1146–1150. PLOS ONE | DOI:10.1371/journal.pone.0113644 November 26, 2014 12 / 14 Backward Adjustable Thoracic Support 19. Hastings JD, Fanucchi ER, Burns SP (2003) Wheelchair configuration and postural alignment in persons with spinal cord injury. Arch Phys Med Rehabil 84: 528–534. 20. Reed MP, Schneider LW (1996) Lumbar support in auto seats: conclusions from a study of preferred driving posture. SAE Technical Paper Series 960478: 19–28. 21. De Carvalho DE, Callaghan JP (2012) Influence of automobile seat lumbar support prominence on spine and pelvic postures: a radiological investigation. Appl Ergon 43: 876–882. 22. Folsch C, Schlogel S, Lakemeier S, Wolf U, Timmesfeld N, et al. (2012) Test-retest reliability of 3D ultrasound measurements of the thoracic spine. PM & R: the journal of injury, function, and rehabilitation 4: 335–341. 23. Uritani D (2013) Reliability of Upper Quadrant Posture Analysis Using an Ultrasound-based Threedimensional Motion Analyzer. Journal of physical therapy science 25: 1181–1184. 24. Moutzouri M, Billis E, Strimpakos N, Kottika P, Oldham JA (2008) The effects of the Mulligan Sustained Natural Apophyseal Glide (SNAG) mobilisation in the lumbar flexion range of asymptomatic subjects as measured by the Zebris CMS20 3-D motion analysis system. BMC Musculoskelet Disord 9: 131. 25. De Foa JL, Forest W, Biedermann HJ (1989) Muscles fibre direction of longissimus, iliocostalis and multifidus: landmark-derived reference lines. J Anat 163: 243–247. 26. Callaghan JP, Dunk NM (2002) Examination of the flexion relaxation phenomenon in erector spinae muscles during short duration slumped sitting. Clin Biomech 17: 353–360. 27. Danneels LA, Cagnie BJ, Cools AM, Vanderstraeten GG, Cambier DC, et al. (2001) Intra-operator and inter-operator reliability of surface electromyography in the clinical evaluation of back muscles. Man Ther 6: 145–153. 28. Dankaerts W, O’Sullivan PB, Burnett AF, Straker LM, Danneels LA (2004) Reliability of EMG measurements for trunk muscles during maximal and sub-maximal voluntary isometric contractions in healthy controls and CLBP patients. J Electromyogr Kinesiol 14: 333–342. 29. O’Sullivan PB, Dankaerts W, Burnett AF, Farrell GT, Jefford E, et al. (2006) Effect of different upright sitting postures on spinal-pelvic curvature and trunk muscle activation in a pain-free population. Spine 31: E707–712. 30. Sabol TP, Haley ES (2006) Wheelchair evaluation for the older adult. Clin Geriatr Med 22: 355–375. 31. Cooper RA (1998) Wheelchair selection and configuration. New York: Demos Medical Publishing. 32. Stojadinovic O, Minkiewicz J, Sawaya A, Bourne JW, Torzilli P, et al. (2013) Deep Tissue Injury in Development of Pressure Ulcers: A Decrease of Inflammasome Activation and Changes in Human Skin Morphology in Response to Aging and Mechanical Load. PLoS One 8: e69223. 33. Poole DC, Musch TI, Kindig CA (1997) In vivo microvascular structural and functional consequences of muscle length changes. Am J Physiol 272: H2107–2114. 34. Welsh DG, Segal SS (1996) Muscle length directs sympathetic nerve activity and vasomotor tone in resistance vessels of hamster retractor. Circ Res 79: 551–559. 35. Pope MH, Goh KL, Magnusson ML (2002) Spine ergonomics. Annu Rev Biomed Eng 4: 49–68. 36. Solberg G (2008) Postural disorders and musculoskeletal dysfunction: diagnosis, prevention and treatment. 2nd ed. Edinburgh: Churchill Livingstone Elsevier. 37. Barrett E, McCreesh K, Lewis J (2014) Reliability and validity of non-radiographic methods of thoracic kyphosis measurement: a systematic review. Man Ther 19: 10–17. 38. Nordin M, Frankel V (2001) Basic biomechanics of the musculoskeletal system. 3rd ed. Philadelphia: Lippincott Williams & Wilkins. 39. Andersson BJ, Ortengren R (1974) Lumbar disc pressure and myoelectric back muscle activity during sitting. II. Studies on an office chair. Scand J Rehabil Med 6: 115–121. 40. Newman AB, Lee JS, Visser M, Goodpaster BH, Kritchevsky SB, et al. (2005) Weight change and the conservation of lean mass in old age: the health, aging and body composition study. Am J Clin Nutr 82: 872–878. PLOS ONE | DOI:10.1371/journal.pone.0113644 November 26, 2014 13 / 14 Backward Adjustable Thoracic Support 41. Newman AB, Haggerty CL, Goodpaster B, Harris T, Kritchevsky S, et al. (2003) Strength and muscle quality in a well-functioning cohort of older adults: the health, aging and body composition study. J Am Geriatr Soc 51: 323–330. 42. Panjabi MM (1992) The stabilizing system of the spine, part I: function, dysfunction, adaptation, and enhancement. J Spinal Disord 5: 383–389. 43. Toosizadeh N, Nussbaum MA, Bazrgari B, Madigan ML (2012) Load-Relaxation Properties of the Human Trunk in Response to Prolonged Flexion: Measuring and Modeling the Effect of Flexion Angle. PLoS One 7: e48625. PLOS ONE | DOI:10.1371/journal.pone.0113644 November 26, 2014 14 / 14