sign in sign up
<<
Home , Ecosystem model, Ecotype

REPORTS

germ insects not only uncovers those features es- mRNA localization indeed appears to be an sential to this developmental mode but also sheds important component of long-germ embryogene- light on how the bcd-dependent anterior patterning sis, perhaps even playing a role in the transition program might have evolved. Through analysis of from the ancestral short-germ to the derived long- the regulation of the trunk gap gene Kr in Dro- germ fate. sophila and Nasonia,wehavebeenabletodem- onstrate that anterior repression of Kr is essential References and Notes for head and thorax formation and is a common 1. G. K. Davis, N. H. Patel, Annu. Rev. Entomol. 47, 669 (2002). feature of long-germ patterning. Both insects 2. T. Berleth et al., EMBO J. 7, 1749 (1988). accomplish this task through maternal, anteriorly 3. W. Driever, C. Nusslein-Volhard, Cell 54, 83 (1988). localized factors that either indirectly (Drosophila) 4. J. Lynch, C. Desplan, Curr. Biol. 13, R557 (2003). or directly (Nasonia) repress Kr and, hence, trunk 5. J. A. Lynch, A. E. Brent, D. S. Leaf, M. A. Pultz, C. Desplan, Nature 439, 728 (2006). fates. In Drosophila, the terminal system and bcd 6. J. Savard et al., Genome Res. 16, 1334 (2006). regulate expression of gap genes, including Dm-gt, 7. G. Struhl, P. Johnston, P. A. Lawrence, Cell 69, 237 (1992). that repress Dm-Kr. Nasonia’s bcd-independent 8. A. Preiss, U. B. Rosenberg, A. Kienlin, E. Seifert, long-germ embryos must solve the same problem, H. Jackle, Nature 313, 27 (1985). Fig. 4. Repression of Nv-Kr by maternal Nv-gt is but they employ a maternally localized repression 9. M. Hulskamp, C. Pfeifle, D. Tautz, Nature 346, 577 (1990). required for head and thorax formation in Nasonia. system in which maternal Nv-gt is localized to the (A)TwomodelsformaternalNv-gt function. Cu- 10. X. Wu, R. Vakani, S. Small, Development 125, 3765 (1998). oocyte’s anterior, where it represses Nv-Kr.Inthe 11. R. Kraut, M. Levine, Development 111, 611 (1991). ticular analysis (B, D,andF)andNv-hb expression zyg dipteran lineage, whereas gt retained the ability to 12. J. A. Lynch, C. Desplan, Nat. Protocols 1, 486 (2006). (C, E,andG) after knockdown for Nv-Kr [(B) and (C)], repress Kr, maternal regulation of Kr’s position 13. R. Kraut, M. Levine, Development 111, 601 (1991). Nv-gt+gfp [(D) and (E)], and Nv-gt+Kr [(F) and (G)]. — 14. M. A. Pultz et al., Development 132, 3705 (2005). was taken over by two novel features bcd,aspe- 15. G. F. Hewitt et al., Development 126, 1201 (1999). cific dipteran innovation, and the terminal pathway, 16. J. A. Lynch, E. C. Olesnicky, C. Desplan, Dev. Genes Evol. Nv-gt and green fluorescent protein (gfp) and which, although present ancestrally, appears to 216, 493 (2006). observed the expected Nv-gt : deletion function less extensively in the anterior of non- 17. R. Schroder, C. Eckert, C. Wolff, D. Tautz, Proc. Natl. Acad. Sci. U.S.A. 97, 6591 (2000). of head and thorax, as well as loss of anterior dipteran insects (16, 17). In addition to activating 18. E. C. Olesnicky et al., Development 133, 3973 (2006). Nv-hb expression (Fig. 4, D and E). Knockdown anterior patterning genes such as otd and hb, bcd 19. A. Bashirullah, R. L. Cooperstock, H. D. Lipshitz, Annu. on March 30, 2007 of Nv-gt and Nv-Kr yielded striking results. In also acquired regulation of gt, which became a Rev. Biochem. 67, 335 (1998). 92% of examined embryos, the head and thorax strictly zygotic gene with a reduced role in repress- 20. G. Bucher, L. Farzana, S. J. Brown, M. Klingler, Evol. Dev. 7, 142 (2005). (T1/T2) were restored (Fig. 4F), and the resulting ing Kr. Our findings thus identify two independent 21. The authors wish to thank members of the Desplan and cuticular were essentially identical mechanisms for long-germ anterior patterning— Small laboratories for support and advice. This project to those after Nv-Kr RNAi alone (Fig. 4B). one using two maternally localized genes, otd1 and was supported by NIH grants GM64864, awarded to C.D., Consistent with rescued head and thorax devel- gt, that respectively activate anterior zygotic pat- and GM51946, awarded to S.S. A.E.B is a Damon Runyon opment, anterior zygotic Nv-hb was also restored, terning genes and repress trunk fates, and a second Fellow, supported by the Damon Runyon Cancer Research Foundation (DRG-1870-05). although not to wild-type levels (Fig. 4G). None- using bcd for these same functions, thereby de- theless, the amount of Nv-hb present in Nv-gt+Kr moting otd and gt to zygotic gap genes. Interest- Supporting Online Material www.sciencemag.org RNAi embryos was sufficient to direct head and ingly, it appears that long-germ embryos use RNA www.sciencemag.org/cgi/content/full/315/5820/1841/DC1 thorax development, demonstrating that Nv-Kr localization for a number of different developmen- Materials and Methods SOM Text expansion impedes anterior patterning and that tal processes (5, 18, 19). By contrast, in short-germ Fig. S1 maternally localized Nv-gt confines Nv-Kr to the insects, although some localized RNAs have been References embryo’s center. Thus, whereas in Drosophila, identified, there is as yet no evidence of their con- 13 November 2006; accepted 7 March 2007 bcd-activated Dm-gt plays only a moderate role tribution to anterior-posterior patterning (20). 10.1126/science.1137528 in positioning Nv-Kr (Fig.1C),inNasonia,ma- ternal Nv-gt is sufficient to perform this func- Downloaded from tion. This distinction led us to consider whether Dm-gt’sroleinDrosophila would be enhanced Emergent of Microbial if the Drosophila embryo were reengineered to develop like Nasonia—with Dm-gt maternally provided and anteriorly localized. We found that, Communities in a Model Ocean whereas Dm-gt was sufficient to repress Dm-Kr Michael J. Follows,1* Stephanie Dutkiewicz,1 Scott Grant,1,2 Sallie W. Chisholm3 anteriorly in the absence of bcd (fig. S1B), head and thoracic structures were not rescued A marine model seeded with many phytoplankton types, whose physiological traits were (fig. S1C)—an unsurprising result given that, in randomly assigned from ranges defined by field and laboratory data, generated an emergent addition to permitting anterior development by structure and biogeography consistent with observed global phytoplankton regulating Kr-repressing gap genes, bcd also distributions. The modeled organisms included types analogous to the marine cyanobacterium functions instructively to activate genes required . Their emergent global distributions and physiological properties simultaneously for head and thorax formation. In Nasonia,by correspond to observations. This flexible representation of community structure can be used to contrast, the instructive and permissive anterior explore relations between , biogeochemical cycles, and climate change. patterning functions are discrete. Head- and thorax- specific genes are triggered by an instructive anterior determinant, maternal Nv-otd1, which is significant challenge in understanding known to regulate important biogeochemical localized independently of the permissively acting the changing earth system is to quantify pathways, including the efficiency of export of maternal repression system, Nv-gt. A and model the role of ocean ecosystems organic carbon to the deep ocean. Although A comparison of the molecular mechanisms in the global carbon cycle. The structure of there is extraordinary diversity in the oceans, employed by two independently evolved (6) long- microbial communities in the surface ocean is the of local microbial communities at

www.sciencemag.org SCIENCE VOL 315 30 MARCH 2007 1843 REPORTS

any location is typically dominated by a smaller its physiological capabilities and the values of and chemical tracers. All phytoplankton types subset of strains. Their relative fitness and eco- coefficients that control the rates and sensitiv- were initialized with identical distributions of system community structure are regulated by a ities of metabolic processes. These were pro- biomass, and the model was integrated forward variety of factors, including physical condi- vided by random drawing from broad ranges for 10 years, over which time a repeating annual tions, dispersal, , for re- guided by laboratory and field studies (table S1). cycle in ecosystem structure emerged. We re- sources, and the variability of the environment We focused these choices on light, temperature, peated the integration 10 times, each time with a (1–3). Models reflecting this conceptual view and nutrient requirements (fig. S1), the niche different random selection of phytoplankton phys- have been examined in idealized ecological dimensions for phytoplankton thought to be most iologies, forming an ensemble of 10 members. Al- settings (4) and have been applied to studies of important in regulating growth. To facilitate a test though each ensemble member produced a unique terrestrial ecosystems (5). We have used this of the approach, we also specifically addressed emergent ecosystem, the broad-scale patterns of approach in a marine that em- functions that differentiate Prochlorococcus spp. , community structure, and biogeogra- braces the diversity of microbes and their ge- from other phytoplankton, including their small phy were robust across all 10. Global patterns of nomic underpinnings, a model in which microbial size and inability to assimilate nitrate. Other open-ocean biomass (Fig. 1A), primary produc- community structure “emerges” from a wider set functions could be emphasized depending on tion, and nutrients (fig. S3) were qualitatively con- of possibilities and, thus, mimics aspects of the the aim of the study. Ecological trade-offs were sistent with in situ and remote observations. The process of natural selection. The system is flexible imposed through highly simplified allometric ensemble mean globally integrated, annual primary enough to respond to changing ocean envi- constraints [see supporting online material production was 44 gigatons C per year, with a ronments and can be used to interpret the structure (SOM)]. To reflect the extra energetic expense of standard deviation of less than 5%. This small and development of marine microbial commu- using nitrate, relative to other inorganic nitrogen standard deviation suggested that sufficient phyto- nities and to reveal critical links between marine sources, we allowed the maximum growth rate to “types” were initialized for consistent ecosystem structure, global biogeochemical increase slightly when nitrate was not the major emergent solutions and also reflects the large-scale cycles, and climate change. nitrogen source (12). Organisms incapable of regulation by the physical transport of nutrients. Recent ocean models have begun to resolve utilizing nitrate were given a slightly lower nu- After an initial adjustment, the biomass of community structure by the explicit repre- trient half-saturation. We explicitly represented some phytoplankton types fell below the thresh- sentation of three or four classes, or functional predation by two classes of grazer and, for the old of numerical noise, and these types were groups, of phytoplankton (6–9), but significant action of heterotrophic microbes, we used a assumed to have become “extinct.” In all ensem-

challenges remain (10, 11). First, the specifica- simple remineralization rate (SOM). ble members, about 20 phytoplankton types ac- on March 30, 2007 tion of functional groups and diversity of the A global ocean circulation model constrained counted for almost all of the total global biomass model ecosystem is subjective and somewhat by observations (13) provided flow fields and (fig. S2). We classified the phytoplankton types arbitrary. Second, it is difficult to evaluate the mixing coefficients that transport all biological into four broad functional groups, each a parameters controlling such models because quantitative, physiological information from laboratory cultures is extremely limited. Third, observations of microbial community structure Fig. 1. Annual mean with which to evaluate global-scale models are biomass and biogeog-

still relatively sparse. Finally, model ecosystem raphy from single in- www.sciencemag.org structures optimized to reflect today’socean tegration. (A)Total phytoplankton biomass may not be sufficiently dynamic to adapt ap- m propriately to a changing climate where radical ( MP,0to50maver- age). (B)Emergentbio- shifts in community structure might be possible. geography: Modeled To circumvent some of these difficulties, we photo- were formulated a model that repre- categorized into four sents a large number of potentially viable phyto- functional groups; color Downloaded from plankton types whose physiological characteristics coding is according to were determined stochastically. The initialized group locally dominating organism types interacted with one another and annual mean biomass. their environment, evolving into a sustainable eco- Green, analogs of Prochlo- system where community structure and diversity rococcus; orange, other were not imposed, but were emergent properties. small photo-autotrophs; The ecosystem model consisted of a set of red, diatoms; and yel- coupled prognostic equations (eqs. S1 to S5), low, other large phyto- with idealized representations of the transforma- plankton. (C)Total tions of inorganic and organic forms of phos- biomass of Prochlorococ- phorus, nitrogen, iron, and silica. Many tens of cus analogs (mMP,0to phytoplankton types (here, 78) were initialized 50 m average). Black line in each simulation, each type distinguished by indicates the track of AMT13.

1Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology, 54-1514 MIT, Cam- bridge, MA 02139, USA. 2Department of Oceanography, University of Hawaii, 1000 Pope Road, Honolulu, HI 96822, USA. 3Departments of Civil and Environmental Engineering and Biology, Massachusetts Institute of Technology, 48-419 MIT, Cambridge, MA 02139, USA *To whom correspondence should be addressed. E-mail: [email protected]

1844 30 MARCH 2007 VOL 315 SCIENCE www.sciencemag.org REPORTS composite of several types, according to aspects Prochlorococcus analogs) qualitatively and quan- “model ” based only on distinct geo- of their : (i) diatom analogs—large titatively reflected the major features of the ob- graphic , without regard to physiology, phytoplankton that require silica, (ii) other large served distribution with highest abundances in which had a qualitative resemblance to the ob- eukaryotes, (iii) Prochlorococcus analogs—small the most oligotrophic (nutrient-depleted) waters served distributions of ecotypes along AMT13. phytoplankton that cannot assimilate nitrate, and (15, 17) (Fig. 2, A to D). In any ensemble member, more than one emer- (iv) other small photo-autotrophs. The large-scale Real-world Prochlorococcus exhibit genetic gent Prochlorococcus analog may fall into a biogeography of the emergent phytoplankton diversity, which leads to differences in light and particular model- classification, and some community was plausible with respect to obser- temperature sensitivities (17–20), as well as ni- were ambiguous. Model ecotype m-e1 (Fig. 2F) vations (Fig. 1B) and consistent among the 10 trogen assimilation abilities (21). The strains, or was defined to include emergent analogs with ensemble members. The model successfully cap- ecotypes, of Prochlorococcus exhibit distinct pat- significant biomass in the upper 25 m along the tured the domination of annual biomass by large terns of along ocean gradients (15, 17), transect between 15°N and 15°S, qualitatively phytoplankton in subpolar upwelling regions, and observations on AMT13 (17) (Fig. 2, E, G, corresponding to the of real-world eco- where both light and macronutrients are season- I, and K) provide an ideal test for the stochastic type eMIT9312 (Fig. 2E). Model ecotype m-e2 ally plentiful. The subtropical oceans were domi- modeling strategy: Do the emergent model ana- (Fig. 2H) included analogs that had significant nated by small phytoplankton functional types logs of Prochlorococcus reflect the geographic biomass in surface waters polewards of 15o but (14). Large areas of the tropics and subtropics distributions, relative abundances, and physio- low biomass within 15o of the equator, broadly were dominated by several Prochlorococcus logical properties of their real-world counterparts? reflecting eMED4 (Fig. 2G). Finally, model eco- analogs (Fig. 1C), also in accord with observa- Of the Prochlorococcus analogs initialized in type m-e3 (Fig. 2J) was defined to include ana- tions (15, 16). Along the cruise track of At- each model solution, between three and six var- logs that had a subsurface maximum biomass, in lantic Meridional Transect 13 (AMT13), total iants persisted with significant abundances (fig. common with eMIT9313 and eNATL2A (Fig. 2, Prochlorococcus abundance (the sum of all S4). We grouped the analogs by defining three I and K). The observed widespread distribution of deep maxima with low abundance associated with eMIT9313 and eNATL2A was not clearly reflected in the model analogs. This might be explained by the tendency toward unrealistically complete competitive exclusion typical in eco-

system models (22, 23), precluding persistent on March 30, 2007 at low abundance. There is a deep, high biomass layer in the model made up of other, nitrate-consuming, small phytoplankton. This may partially reflect a contribution from nitrate-utilizing Prochlorococcus, which have re- cently been inferred from ocean observations (24), but which have not yet been seen in culture. www.sciencemag.org Downloaded from

Fig. 3. Optimum temperature and light inten- sity for growth, Topt and Iopt, of all initialized Prochlorococcus analogs (all circles) from the ensemble of 10 model integrations. Large circles indicate the analogs that exceeded a total biomass of 106 mol P along AMT13 in the 10th year. Colors indicate classification into Fig. 2. Observed and modeled properties along the AMT13 cruise track. Left column shows model ecotypes (see main text): Red circles, m-e1; observations (17), right column shows results from a single model integration. (A and B) Nitrate blue circles, m-e2; green circles, m-e3.Mixed- (mmol kg−1); (C and D) total Prochlorococcus abundance [log (cells ml−1)]. (E, G, I, and K) color and solid black circles denote ambiguity in Distributions of the four most abundant Prochlorococcus ecotypes [log (cells ml−1)] ranked model-ecotype classification. Bold diamonds indi- vertically. (F, H, and J) The three emergent model ecotypes ranked vertically by abundance. Model cate real-world Prochlorococcus ecotypes (red, Prochlorococcus biomass was converted to cell density assuming a quota of 1 fg P cell−1 (27). Black eMIT9312; blue, eMED4; green, eNATL2A; and lines indicate isotherms. yellow, eMIT9313).

www.sciencemag.org SCIENCE VOL 315 30 MARCH 2007 1845 REPORTS

Within each ensemble member, emergent emergent Prochlorococcus model ecotypes and 14. Y. Dandonneau, Y. Montel, J. Blanchot, model ecotypes typically followed the abun- their real-world counterparts. These parallels indi- J. Giraudeau, J. Neveux, Deep-Sea Res. I 53, 689 (2006). dance ranking of their geographically identified cate that the stochastic, self-organizing represen- 15. M. V. Zubkov, M. A. Sleigh, P. H. Burkill, R. J. G. Leakey, real-world counterparts (Fig. 2 and fig. S4): tation of marine ecosystems reflects real-world Prog. Oceanogr. 45, 369 (2000). Model ecotypes m-e1 and m-e2 ranked first and processes and is suitable for application in eco- 16. H. A. Bouman et al., Science 312, 918 (2006). second (compare these with eMIT9312 and logical and biogeochemical studies. This approach 17. Z. I. Johnson et al., Science 311, 1737 (2006). 18. L. R. Moore, G. Rocap, S. W. Chisholm, Nature 393, 464 eMED4, respectively), with m-e3 consistently at circumvents some of the obstacles facing most (1998). lower abundances (compare this with ecotypes current ocean ecosystem models, such as the a 19. G. Rocap et al., Nature 424, 1042 (2003). eNATL2A and eMIT9313). priori imposition of low diversity, the prescrip- 20. L. R. Moore, S. W. Chisholm, Limnol. Oceanogr. 44, 628 There is a simultaneous correspondence tion of dominant functional types, and the dif- (1999). between the physiological characteristics of emer- ficulty of specifying the physiological rate 21. L. R. Moore, A. F. Post, G. Rocap, S. W. Chisholm, Limnol. Oceanogr. 47, 989 (2002). gent, modeled ecotypes and cultured represent- coefficients that define them. This function- 22. G. E. Hutchinson, Am. Nat. 95, 137 (1961). atives of the wild . Each cultured based approach can naturally evolve to exploit 23. R. A. Armstrong, R. McGehee, Am. Nat. 115, 151 strain of Prochlorococcus and the emergent the growing body of genomic and metagenomic (1980). model ecotypes from all 10 ensemble members data mapping the oceans in terms of genes and 24. M. W. Lomas, F. Lipschultz, Limnol. Oceanogr. 51, 2453 (2006). were characterized by an optimal temperature their encoded physiological functionality (25, 26). 25. J. C. Venter et al., Science 304, 66 (2004). (Topt) and photon flux (Iopt) for growth, the tem- Finally, because the ecosystem structure and 26. E. F. DeLong et al., Science 311, 496 (2006). perature or light intensity at which growth rates function are, by design, emergent and not tightly 27. S. Bertilsson, O. Berglund, D. M. Karl, S. W. Chisholm, are greatest if all other limitations are set aside prescribed, this modeling approach is ideally Limnol. Oceanogr. 48, 1721 (2003). 28. Thanks to J. Marshall, R. Williams, P. Falkowski, J. Cullen, (fig. S1). Potentially viable Prochlorococcus ana- suited for studies of the relations between marine and J. Bragg for inspiration and encouragement. Thanks logs were seeded in the model over wide ranges ecosystems, , biogeochemical cycles, also to M. Coleman, R. Hood, and three anonymous of optimal temperature and photon fluxes (all and past and future climate change. reviewers for stimulating comments on the manuscript; to circles, Fig. 3), but those that maintained signif- C. Hill for computing guidance; and to P. Heimbach, C. Wunsch, and the ECCO group for ocean circulation icant abundances along the AMT transect (solid References and Notes state estimates. We are grateful for funding from the 1. D. Tilman, 58, 338 (1977). large circles, Fig. 3) were all characterized by PARADIGM consortium of the National Ocean Partnership 2. R. Margalef, Perspectives in Ecological Theory (Univ. of T > 15°C. This is consistent with the observa- Program, NSF (M.J.F., S.D.), NSF, DOE (S.W.C.), and the opt Chicago Press, Chicago, 1968). Gordon and Betty Moore Foundation (S.W.C., M.J.F.). on March 30, 2007 tions of Prochlorococcus in warmer waters and 3. C. Pedrós-Alió, Trends Microbiol. 14, 257 (2006). M.J.F. is also grateful for the MIT Global Habitat with the warm Topt of cultured strains (17). Our 4. S. L. Pimm, J. H. Lawton, Nature 268, 329 (1977). Longevity Award. We acknowledge the Atlantic Meridional model indicates that the oligotrophic conditions 5. A. Kleidon, H. A. Mooney, Glob. Change Biol. 6, 507 (2000). Transect consortium (NER/O/S/2001/00680), which confined Prochlorococcus analogs to warmer wa- 6. J. K. Moore, S. Doney, J. Kleyplas, D. Glover, I. Fung, enabled the biogeographical observations first published “ Deep-Sea Res. II 49, 403 (2001). in (17) (AMT contribution no. 107). ters and selected for warm Topt ,anemergent ad- 7. W. W. Gregg, P. Ginoux, P. S. Schopf, N. W. Casey, aptation” driven by other environmental factors. Deep-Sea Res. II 50, 3143 (2003). In the cooler waters of the model, nutrients are 8. E. Litchman, C. A. Klausmeier, J. R. Miller, Supporting Online Material typically abundant, and so larger phytoplankton, O. M. Schofield, P. G. Falkowski, Biogeosciences 3, 585 www.sciencemag.org/cgi/content/full/315/5820/1843/DC1 (2006). Materials and Methods with higher intrinsic maximum growth rates, have 9. C. LeQuere et al., Glob. Change Biol. 11, 2016 (2006). SOM Text an advantage. In the highly oligotrophic (typically 10. T. R. Anderson, J. Plankton Res. 27, 1073 (2005). Figs. S1 to S4 www.sciencemag.org warmer) regions, the Prochlorococcus analogs’ 11. R. R. Hood et al., Deep-Sea Res. II 53, 459 (2006). Table S1 lower half-saturation (consistent with their very 12. P. A. Thompson et al., Limnol. Oceanogr. 34,1014 References and Notes small size) is advantageous. (1989). 13. C. Wunsch, P. Heimbach, Physica D 10.1016/ 7 December 2006; accepted 5 March 2007 Across the ensemble of 10 integrations, the j.physd.2006.09.040 (2006). 10.1126/science.1138544 geographically defined model ecotypes were clus- tered in optimal temperature and light parameter (Fig. 3): Model ecotype m-e1 (red circles) generally occupied the warmest area of parameter Cascading Effects of the Loss of Downloaded from space over a broad, upper range of optimal pho- ton fluxes; m-e2 (blue circles) generally had a Apex Predatory from a lower Topt but a similar range of Iopt. This is con- sistent with their surface-oriented habitats and latitudinal (or temperature) separation. In con- Coastal Ocean trast, m-e3 (green circles) occupied a wider 1 1 1 range of Topt but only in the region of lowest Ransom A. Myers, Julia K. Baum, * Travis D. Shepherd, 2 3 Iopt, consistent with its expression of subsurface Sean P. Powers, Charles H. Peterson * maxima. Although there were exceptions, the clustering of geographically defined model eco- Impacts of chronic overfishing are evident in population depletions worldwide, yet indirect types in physiological parameter space indicated ecosystem effects induced by predator removal from oceanic food webs remain unpredictable. that robust ecological controls were operating As abundances of all 11 great sharks that consume other elasmobranchs (rays, skates, and small across the 10 integrations. The physiological char- sharks) fell over the past 35 years, 12 of 14 of these prey increased in coastal northwest acteristics (Topt, Iopt)ofreal-worldecotypes Atlantic ecosystems. Effects of this community restructuring have cascaded downward from the (colored diamonds, Fig. 3) are notably consistent cownose ray, whose enhanced predation on its bay scallop prey was sufficient to terminate a with the grouping of their model counterparts. This century-long scallop . Analogous top-down effects may be a predictable consequence of correspondence was not imposed, but emerged eliminating entire functional groups of predators. as a feature of the model solution. Significantly, there was simultaneous con- cological impacts of eliminating top pred- from predatory control (2) and induction of sistency between the geographical habitat, rank ators can be far-reaching (1) and include subsequent cascades of indirect trophic interac- abundance, and physiological specialization of the Erelease of prey populations tions (3–5). In the oceans, fishing has dispropor-

1846 30 MARCH 2007 VOL 315 SCIENCE www.sciencemag.org