Letting go of mucus Jeffrey J. Wine Science 345, 730 (2014); DOI: 10.1126/science.1258493
If you wish to distribute this article to others, you can order high-quality copies for your colleagues, clients, or customers by clicking here. Permission to republish or repurpose articles or portions of articles can be obtained by following the guidelines here. The following resources related to this article are available online at www.sciencemag.org (this information is current as of August 14, 2014 ): Updated information and services, including high-resolution figures, can be found in the online version of this article at: http://www.sciencemag.org/content/345/6198/730.full.html A list of selected additional articles on the Science Web sites related to this article can be found at: http://www.sciencemag.org/content/345/6198/730.full.html#related This article cites 13 articles, 6 of which can be accessed free: http://www.sciencemag.org/content/345/6198/730.full.html#ref-list-1 This article appears in the following subject collections: Medicine, Diseases http://www.sciencemag.org/cgi/collection/medicine
Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright 2014 by the American Association for the Advancement of Science; all rights reserved. The title Science is a registered trademark of AAAS.
Downloaded from www.sciencemag.org on August 14, 2014
This copy is for your personal, non-commercial use only.
INSIGHTS | P E R S P E C T I V E S
How and to what extent, if any, are we willing to live with constraints on intelligencegathering if their consequence is a less effective intelligence process? THE LIMITED ROLE OF TECHNOLOGY.
Fundamental scientific and technological realities underlie these issues, and it is tempting to think that problems originating in technology can be solved by judicious application of more technology. For example, the technology of information accountability has been proposed as a way to enforce use-based privacy restrictions. Data-mining systems based on this technology capture the logic underlying queries and inferences and record who is making such queries (5). Similarly, homomorphic encryption—a method for performing operations such as “sort” and “find” on encrypted data—is sometimes offered as a way to protect privacy of those whose data are stored in searchable databases (6). If and when these technologies become available on a large scale, they may reassure some who are skeptical of U.S. government intentions. But to do so, appropriate policies and procedures must be in place to govern their use. For example, audits of query logs must actually be performed and users held accountable if transparent accountable data mining is to enhance privacy. Whether such policies and procedures will be followed is another matter entirely. Technologies, always embedded within a broader milieu of policy and procedures, will never be the sole—or necessarily most important—element of a solution to such dilemmas. In January 2014, President Obama said that the United States would reduce the scope of phone metadata collection and called for explicit judicial authorization for nonemergency queries to the existing database of metadata information. He also extended some privacy protections to nonU.S. persons and forbade eavesdropping on the leaders of friendly nations. Legislation is pending in the U.S. Congress that would bar government collection of phone metadata in bulk. But these and other changes, adopted or proposed, leave many important issues unaddressed, and thus the policy debate is far from over. ■ REFERENCES
1. A. E. Marimow, C. Timberg, Washington Post, 24 April 2014; http://wapo.st/1o8L3Zf. 2. N. Wingfield, New York Times, 31 July 2014; http://bits.blogs. nytimes.com/2014/07/31/judge-rules-that-microsoftmust-turn-over-data-stored-in-ireland/. 3. President’s Council of Advisors on Science and Technology, Big Data and Privacy: A Technological Perspective (White House, Washington, DC, 2014); http://1.usa.gov/1s79RBo. 4. E. Nakashima, Washington Post, 26 July 2014; http:// wapo.st/1rTh2Qc. 5. D. J. Weitzner et al., Commun. ACM 51, 82–87 (2008). 6. C. Gentry, Commun. ACM 53, 97 (2010). 10.1126/science.1251485
730
MEDICINE
Letting go of mucus Defective release of mucus from airway glands is present at birth in cystic fibrosis By Jeffrey J. Wine
T
he genetic disease cystic fibrosis (CF) results from mutations that disable an anion channel called cystic fibrosis transmembrane conductance regulator (CFTR). Loss of CFTR’s anion channel functions triggers a cascade of defects that reduces the lung’s innate mucosal defenses, including the critical component of mucus clearance (1, 2). The lungs of individuals with CF, which appear grossly normal at birth, quickly become infected with bacteria and fungi localized within the lung mucus, which, like mucus in other CF organs, is thicker and more difficult to clear (3). The pathogenesis of CF airways has been difficult to determine because appropriate animal models have been lacking. On page 818 in this issue, Hoegger et al. (4) find that lung disease in a piglet model of CF is caused by abnormal tethering of mucus to the submucosal glands that produce it.
“Maybe we will fix cystic fibrosis before we fully understand it.” Hoegger et al. used computed tomography imaging of intact, anesthetized pigs to track the movements of 0.35-mm tantalum disks (which are visible with this type of imaging) in trachea and mainstem bronchi of pigs with and without CF. Imaging showed normal basal clearance by beating cilia. Columnar epithelial cells that line pulmonary airways bear cilia that sweep mucus out of the trachea and bronchi. By contrast, strong cholinergic stimulation, which elicits mucus secretion, accelerated clearance in normal pigs but not in pigs with CF, where many disks stopped moving, even though beating of cilia increased normally. No difference in the depth of periciliary liquid (which contacts the surface epithelial cells and lies below the superficial mucus layer) was detected. In another approach, Hoegger et al. removed the tracheas from piglets, cut them along their ventral surface, and pinned them out so that mucociliary clearance
over the dorsal mucosal surface could be imaged. CFTR, which conducts chloride (Cl–) and bicarbonate (HCO3–) ions to mediate fluid secretion (5), is apically localized in both surface epithelia and in submucosal glands. To disrupt ion and fluid transport by surface epithelia, the authors submerged the tracheas in saline and then tracked mucus transport in the flooded airways. The key finding was that in flooded CF tracheas, many tantalum disks were still immobile. To observe mucus movements more directly, Hoegger et al. dispersed fluorescent nanospheres in the saline, which stick to mucus. In submerged tracheas, even from piglets without CF, the nanospheres revealed strands of mucus emanating from submucosal gland ducts and stretching for hundreds of micrometers before they broke free and were transported up the airway. Sometimes the mucus strands broke apart, and the portion that remained tethered to the glands contracted back toward the duct. Because some tethering of mucus to gland ducts occurred even in non-CF tracheas, Hoegger et al. employed another clever method to quantify it. A series of 32 images of a 10.2-mm2 field were taken over a 4-s period and averaged, so that static beads produced bright signals whereas moving beads did not. A scanning stage was used to image tracheal areas of 163 to 358 mm2. In submerged, non-CF tracheas, fewer than 6 “adherent entities” accumulated on average over a 45-min period, versus 15 to 50 entities in CF tracheas. Importantly, all of the adherent strands could be traced to submucosal gland ducts. To assess the contributions of HCO3– and Cl– anions to mucus secretion, Hoegger et al. used the same visualization methods to study mucus accumulation in submerged tracheas of pigs without CF after transport of each anion was blocked individually and together. Blocking either Cl– or HCO3– alone, which reduces gland volume secretion by about half (5), had no appreciable effect on the amount of mucus tethering in submerged airways, whereas blocking both anions, which reduces volume secretion by >90% (6), resulted in tethering as severe as in the CF tracheas. This result fits neatly with the finding that mucin accumulation sciencemag.org SCIENCE
15 AUGUST 2014 • VOL 345 ISSUE 6198
Published by AAAS
Cystic fbrosis Airway lumen
Tethered mucus
Trachea
Cilated epithelial cells
Lungs
Cl⫺, HCO3⫺, H2O
Tubuloacinar mucus gland
ILLUSTRATION: C. BICKEL/SCIENCE
Clearance. Mucus-producing glands that line the human airway are multiply branched tubuloacinar structures. In cystic fibrosis, mucus remains anchored in the gland and is not swept out of the airway by ciliated epithelial cells.
in gland ducts following cholinergic stimulation required blocking of both anions (7, 8). The finding of Hoegger et al. reveals a new aspect of airway submucosal gland dysfunction in CF. It has long been hoped that the contentious problem of how CF lung disease starts could be addressed in CF animal models. Indeed, when five pigs with CF were kept alive for up to 5 months, they developed CF pulmonary disease with infection, inflammation, and mucus obstruction similar to that seen in humans with CF (9). These experiments are unlikely to be repeated widely because exhaustive effort is required to raise pigs with CF-related disruptions of their digestive systems (10). This has enforced a focus on newborn piglets, but this constraint has had the salutary effect of producing abundant new evidence about the chicken-and-egg questions that have bedeviled the CF field for decades; that is, whether, defective airway mucus clearance is a primary cause of the disease or is a secondary consequence of infection and CF Research Laboratory, Department of Psychology, Stanford University, Stanford, CA 94305–2130, USA. E-mail: wine@ stanford.edu
inflammation. The findings of Hoegger et al. suggest that it is a primary cause. By calling attention to mucus tethering in CF airways and supplying new methods to study it, Hoegger et al. initiate a new paradigm with new questions. These include the mechanisms by which mucus is normally released from glands (see the figure); why cholinergic stimulation of fluid secretion, which occurs mainly by non-CFTR pathways (11, 12), produces such profound consequences in pigs with CF; and whether physiological amounts of acetylcholine from the vagus nerve will produce the same result as the strong specific stimulus (for the muscarinic acetylcholine receptor) that was used in their study. Observing mucus transport in submerged airways was a brilliant stroke that effectively clamped the surface composition and volume. But could this approach have allowed the scant (6), high-solids mucin aggregates (7) that form when anion secretion is blocked to expand in the bathing fluid, as they do normally in other organs and on the surfaces in aquatic animals? If so, this may have amplified the subtle differences in clearance between CF and control animals
SCIENCE sciencemag.org
in vivo. A more precise estimate of the tethering phenomenon should prove useful going forward. Hoegger et al. used a four-point “tethered mucus score” because fluorescence from beads attached to mucus from regions outside the imaged field made more exact measurement difficult; so an average of “2” for pigs with CF means that 15 to 50 adherent entities accumulated over 163 to 358 mm2. With ~10 glands per mm2 in newborn pig tracheas (12), this would represent tethering to 0.4 to 3% of the ducts. Most individuals with CF eventually succumb to chronic pulmonary infections. Pathogens are localized in the mucus, so increased mucus tethering could trigger CF pulmonary disease and would exacerbate it. Do the findings of Hoegger et al. suggest new ways in which CF lung disease should be treated? Submucosal glands are an abundant source of the mucin MUC5B in human upper airways. Reducing the amount of mucus secreted seems ill advised, given the devastating effects of eliminating MUC5B from mouse airways (13). But strategies to improve the release of mucus from glands might be considered. The exciting findings of Hoegger et al. bolster evidence that mucus clearance, submucosal glands, anion secretion, and mucus maturation are worthy of the attention being devoted to them by CF researchers. As the evidence mounts that these defects are seen within hours of birth, failure to intervene early is increasingly indefensible. However, the study also underscores the complexities of airway innate defenses and the multiple defects that follow loss of CFTR function. Given the recent progress in attacking CF at its source, it may be time to resurrect a saying that has been out of fashion since the early heady days of gene therapy: Maybe we will fix cystic fibrosis before we fully understand it. ■ REFERENCES
1. M. R. Knowles, R. C. Boucher, J. Clin. Invest. 109, 571 (2002). 2. J. J. Wine, N. S. Joo, Proc. Am. Thorac. Soc. 1, 47 (2004). 3. P. M. Quinton, Am. J. Physiol. Cell Physiol. 299, C1222 (2010). 4. M. J. Hoegger et al., Science 345, 818 (2014). 5. R. A. Frizzell, J. W. Hanrahan, Cold Spring Harb. Perspect. Med. 2, a009563 (2012). 6. N. S. Joo, Y. Saenz, M. E. Krouse, J. J. Wine, J. Biol. Chem. 277, 28167 (2002). 7. S. K. Inglis, M. R. Corboz, S. T. Ballard, Am. J. Physiol. 274, L762 (1998). 8. S. T. Ballard, D. Spadafora, Respir. Physiol. Neurobiol. 159, 271 (2007). 9. C. S. Rogers et al., Science 321, 1837 (2008). 10. D. A. Stoltz et al., J. Clin. Invest. 123, 2685 (2013). 11. R. J. Lee, J. K. Foskett, Am. J. Physiol. Lung Cell. Mol. Physiol. 298, L210 (2010). 12. N. S. Joo, H. J. Cho, M. Khansaheb, J. J. Wine, J. Clin. Invest. 120, 3161 (2010). 13. M. G. Roy et al., Nature 505, 412 (2014).
10.1126/science.1258493 15 AUGUST 2014 • VOL 345 ISSUE 6198
Published by AAAS
731
If you wish to distribute this article to others, you can order high-quality copies for your colleagues, clients, or customers by clicking here. Permission to republish or repurpose articles or portions of articles can be obtained by following the guidelines here. The following resources related to this article are available online at www.sciencemag.org (this information is current as of August 14, 2014 ): Updated information and services, including high-resolution figures, can be found in the online version of this article at: http://www.sciencemag.org/content/345/6198/730.full.html A list of selected additional articles on the Science Web sites related to this article can be found at: http://www.sciencemag.org/content/345/6198/730.full.html#related This article cites 13 articles, 6 of which can be accessed free: http://www.sciencemag.org/content/345/6198/730.full.html#ref-list-1 This article appears in the following subject collections: Medicine, Diseases http://www.sciencemag.org/cgi/collection/medicine
Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright 2014 by the American Association for the Advancement of Science; all rights reserved. The title Science is a registered trademark of AAAS.
Downloaded from www.sciencemag.org on August 14, 2014
This copy is for your personal, non-commercial use only.
INSIGHTS | P E R S P E C T I V E S
How and to what extent, if any, are we willing to live with constraints on intelligencegathering if their consequence is a less effective intelligence process? THE LIMITED ROLE OF TECHNOLOGY.
Fundamental scientific and technological realities underlie these issues, and it is tempting to think that problems originating in technology can be solved by judicious application of more technology. For example, the technology of information accountability has been proposed as a way to enforce use-based privacy restrictions. Data-mining systems based on this technology capture the logic underlying queries and inferences and record who is making such queries (5). Similarly, homomorphic encryption—a method for performing operations such as “sort” and “find” on encrypted data—is sometimes offered as a way to protect privacy of those whose data are stored in searchable databases (6). If and when these technologies become available on a large scale, they may reassure some who are skeptical of U.S. government intentions. But to do so, appropriate policies and procedures must be in place to govern their use. For example, audits of query logs must actually be performed and users held accountable if transparent accountable data mining is to enhance privacy. Whether such policies and procedures will be followed is another matter entirely. Technologies, always embedded within a broader milieu of policy and procedures, will never be the sole—or necessarily most important—element of a solution to such dilemmas. In January 2014, President Obama said that the United States would reduce the scope of phone metadata collection and called for explicit judicial authorization for nonemergency queries to the existing database of metadata information. He also extended some privacy protections to nonU.S. persons and forbade eavesdropping on the leaders of friendly nations. Legislation is pending in the U.S. Congress that would bar government collection of phone metadata in bulk. But these and other changes, adopted or proposed, leave many important issues unaddressed, and thus the policy debate is far from over. ■ REFERENCES
1. A. E. Marimow, C. Timberg, Washington Post, 24 April 2014; http://wapo.st/1o8L3Zf. 2. N. Wingfield, New York Times, 31 July 2014; http://bits.blogs. nytimes.com/2014/07/31/judge-rules-that-microsoftmust-turn-over-data-stored-in-ireland/. 3. President’s Council of Advisors on Science and Technology, Big Data and Privacy: A Technological Perspective (White House, Washington, DC, 2014); http://1.usa.gov/1s79RBo. 4. E. Nakashima, Washington Post, 26 July 2014; http:// wapo.st/1rTh2Qc. 5. D. J. Weitzner et al., Commun. ACM 51, 82–87 (2008). 6. C. Gentry, Commun. ACM 53, 97 (2010). 10.1126/science.1251485
730
MEDICINE
Letting go of mucus Defective release of mucus from airway glands is present at birth in cystic fibrosis By Jeffrey J. Wine
T
he genetic disease cystic fibrosis (CF) results from mutations that disable an anion channel called cystic fibrosis transmembrane conductance regulator (CFTR). Loss of CFTR’s anion channel functions triggers a cascade of defects that reduces the lung’s innate mucosal defenses, including the critical component of mucus clearance (1, 2). The lungs of individuals with CF, which appear grossly normal at birth, quickly become infected with bacteria and fungi localized within the lung mucus, which, like mucus in other CF organs, is thicker and more difficult to clear (3). The pathogenesis of CF airways has been difficult to determine because appropriate animal models have been lacking. On page 818 in this issue, Hoegger et al. (4) find that lung disease in a piglet model of CF is caused by abnormal tethering of mucus to the submucosal glands that produce it.
“Maybe we will fix cystic fibrosis before we fully understand it.” Hoegger et al. used computed tomography imaging of intact, anesthetized pigs to track the movements of 0.35-mm tantalum disks (which are visible with this type of imaging) in trachea and mainstem bronchi of pigs with and without CF. Imaging showed normal basal clearance by beating cilia. Columnar epithelial cells that line pulmonary airways bear cilia that sweep mucus out of the trachea and bronchi. By contrast, strong cholinergic stimulation, which elicits mucus secretion, accelerated clearance in normal pigs but not in pigs with CF, where many disks stopped moving, even though beating of cilia increased normally. No difference in the depth of periciliary liquid (which contacts the surface epithelial cells and lies below the superficial mucus layer) was detected. In another approach, Hoegger et al. removed the tracheas from piglets, cut them along their ventral surface, and pinned them out so that mucociliary clearance
over the dorsal mucosal surface could be imaged. CFTR, which conducts chloride (Cl–) and bicarbonate (HCO3–) ions to mediate fluid secretion (5), is apically localized in both surface epithelia and in submucosal glands. To disrupt ion and fluid transport by surface epithelia, the authors submerged the tracheas in saline and then tracked mucus transport in the flooded airways. The key finding was that in flooded CF tracheas, many tantalum disks were still immobile. To observe mucus movements more directly, Hoegger et al. dispersed fluorescent nanospheres in the saline, which stick to mucus. In submerged tracheas, even from piglets without CF, the nanospheres revealed strands of mucus emanating from submucosal gland ducts and stretching for hundreds of micrometers before they broke free and were transported up the airway. Sometimes the mucus strands broke apart, and the portion that remained tethered to the glands contracted back toward the duct. Because some tethering of mucus to gland ducts occurred even in non-CF tracheas, Hoegger et al. employed another clever method to quantify it. A series of 32 images of a 10.2-mm2 field were taken over a 4-s period and averaged, so that static beads produced bright signals whereas moving beads did not. A scanning stage was used to image tracheal areas of 163 to 358 mm2. In submerged, non-CF tracheas, fewer than 6 “adherent entities” accumulated on average over a 45-min period, versus 15 to 50 entities in CF tracheas. Importantly, all of the adherent strands could be traced to submucosal gland ducts. To assess the contributions of HCO3– and Cl– anions to mucus secretion, Hoegger et al. used the same visualization methods to study mucus accumulation in submerged tracheas of pigs without CF after transport of each anion was blocked individually and together. Blocking either Cl– or HCO3– alone, which reduces gland volume secretion by about half (5), had no appreciable effect on the amount of mucus tethering in submerged airways, whereas blocking both anions, which reduces volume secretion by >90% (6), resulted in tethering as severe as in the CF tracheas. This result fits neatly with the finding that mucin accumulation sciencemag.org SCIENCE
15 AUGUST 2014 • VOL 345 ISSUE 6198
Published by AAAS
Cystic fbrosis Airway lumen
Tethered mucus
Trachea
Cilated epithelial cells
Lungs
Cl⫺, HCO3⫺, H2O
Tubuloacinar mucus gland
ILLUSTRATION: C. BICKEL/SCIENCE
Clearance. Mucus-producing glands that line the human airway are multiply branched tubuloacinar structures. In cystic fibrosis, mucus remains anchored in the gland and is not swept out of the airway by ciliated epithelial cells.
in gland ducts following cholinergic stimulation required blocking of both anions (7, 8). The finding of Hoegger et al. reveals a new aspect of airway submucosal gland dysfunction in CF. It has long been hoped that the contentious problem of how CF lung disease starts could be addressed in CF animal models. Indeed, when five pigs with CF were kept alive for up to 5 months, they developed CF pulmonary disease with infection, inflammation, and mucus obstruction similar to that seen in humans with CF (9). These experiments are unlikely to be repeated widely because exhaustive effort is required to raise pigs with CF-related disruptions of their digestive systems (10). This has enforced a focus on newborn piglets, but this constraint has had the salutary effect of producing abundant new evidence about the chicken-and-egg questions that have bedeviled the CF field for decades; that is, whether, defective airway mucus clearance is a primary cause of the disease or is a secondary consequence of infection and CF Research Laboratory, Department of Psychology, Stanford University, Stanford, CA 94305–2130, USA. E-mail: wine@ stanford.edu
inflammation. The findings of Hoegger et al. suggest that it is a primary cause. By calling attention to mucus tethering in CF airways and supplying new methods to study it, Hoegger et al. initiate a new paradigm with new questions. These include the mechanisms by which mucus is normally released from glands (see the figure); why cholinergic stimulation of fluid secretion, which occurs mainly by non-CFTR pathways (11, 12), produces such profound consequences in pigs with CF; and whether physiological amounts of acetylcholine from the vagus nerve will produce the same result as the strong specific stimulus (for the muscarinic acetylcholine receptor) that was used in their study. Observing mucus transport in submerged airways was a brilliant stroke that effectively clamped the surface composition and volume. But could this approach have allowed the scant (6), high-solids mucin aggregates (7) that form when anion secretion is blocked to expand in the bathing fluid, as they do normally in other organs and on the surfaces in aquatic animals? If so, this may have amplified the subtle differences in clearance between CF and control animals
SCIENCE sciencemag.org
in vivo. A more precise estimate of the tethering phenomenon should prove useful going forward. Hoegger et al. used a four-point “tethered mucus score” because fluorescence from beads attached to mucus from regions outside the imaged field made more exact measurement difficult; so an average of “2” for pigs with CF means that 15 to 50 adherent entities accumulated over 163 to 358 mm2. With ~10 glands per mm2 in newborn pig tracheas (12), this would represent tethering to 0.4 to 3% of the ducts. Most individuals with CF eventually succumb to chronic pulmonary infections. Pathogens are localized in the mucus, so increased mucus tethering could trigger CF pulmonary disease and would exacerbate it. Do the findings of Hoegger et al. suggest new ways in which CF lung disease should be treated? Submucosal glands are an abundant source of the mucin MUC5B in human upper airways. Reducing the amount of mucus secreted seems ill advised, given the devastating effects of eliminating MUC5B from mouse airways (13). But strategies to improve the release of mucus from glands might be considered. The exciting findings of Hoegger et al. bolster evidence that mucus clearance, submucosal glands, anion secretion, and mucus maturation are worthy of the attention being devoted to them by CF researchers. As the evidence mounts that these defects are seen within hours of birth, failure to intervene early is increasingly indefensible. However, the study also underscores the complexities of airway innate defenses and the multiple defects that follow loss of CFTR function. Given the recent progress in attacking CF at its source, it may be time to resurrect a saying that has been out of fashion since the early heady days of gene therapy: Maybe we will fix cystic fibrosis before we fully understand it. ■ REFERENCES
1. M. R. Knowles, R. C. Boucher, J. Clin. Invest. 109, 571 (2002). 2. J. J. Wine, N. S. Joo, Proc. Am. Thorac. Soc. 1, 47 (2004). 3. P. M. Quinton, Am. J. Physiol. Cell Physiol. 299, C1222 (2010). 4. M. J. Hoegger et al., Science 345, 818 (2014). 5. R. A. Frizzell, J. W. Hanrahan, Cold Spring Harb. Perspect. Med. 2, a009563 (2012). 6. N. S. Joo, Y. Saenz, M. E. Krouse, J. J. Wine, J. Biol. Chem. 277, 28167 (2002). 7. S. K. Inglis, M. R. Corboz, S. T. Ballard, Am. J. Physiol. 274, L762 (1998). 8. S. T. Ballard, D. Spadafora, Respir. Physiol. Neurobiol. 159, 271 (2007). 9. C. S. Rogers et al., Science 321, 1837 (2008). 10. D. A. Stoltz et al., J. Clin. Invest. 123, 2685 (2013). 11. R. J. Lee, J. K. Foskett, Am. J. Physiol. Lung Cell. Mol. Physiol. 298, L210 (2010). 12. N. S. Joo, H. J. Cho, M. Khansaheb, J. J. Wine, J. Clin. Invest. 120, 3161 (2010). 13. M. G. Roy et al., Nature 505, 412 (2014).
10.1126/science.1258493 15 AUGUST 2014 • VOL 345 ISSUE 6198
Published by AAAS
731
