and transcript numbers could provide authentic
constraints on cost-benefits for modeling. For
example, in chemosynthetic environments for
which the Gibbs free energies of reactions are
known, a model of gene abundance and production rate has been tuned to replicate observations (6). The costs and benefits of these gene
processes could potentially be determined from
their best-fit model parameters.
Different communities, common
Microbial consortia do not align themselves passively along preexisting ocean biogeochemical
gradients; rather, they shape the biogeochemical
environment in a tight feedback loop that couples the physical environment to biological community structure. Thus, observations of taxa and
gene transcription across large gradients in nutrients following the path of the Amazon River
plume (17, 19, 24, 39) provided context for testing the ability of the model to reproduce observed patterns. Three potential outcomes of
GENOME model simulations were envisioned.
Initially, we considered it possible that random
allocation of genes to organisms would not result in correct and repeatable biogeochemical
gradients because the organisms needed to exploit specific conditions might be absent. However, this was not the case (Fig. 4 and fig. S2).
Second, it was possible that highly similar emergent communities developed in each model
simulation and led to common biogeochemical
gradients. Third, it was possible that different
emergent communities developed in each model
simulation, but their common metabolic functions created similar biogeochemical gradients
through space and time (i.e., that metabolic capabilities rather than specific taxa regulated ocean
biogeochemistry). The second and third outcomes
were evaluated by comparing the similarity of
community genomes between model simulations
with the similarity of community metabolic expression between model simulations.
The similarity between metagenomes (Fig. 6A)
and normalized metatranscriptomes (Fig. 4C) for
each model run was computed at locations along
the salinity gradient (Fig. 6, B and D) in the
model Amazon plume and clustered hierarchically. Salinity is a proxy for biogeochemical and
ecological community gradients that are attributable to dilution and biogeochemical dynamics.
In general, as salinity increases, inorganic and
organic nutrients decrease, and communities
shift from groups representing Gammaproteo-bacteria, Flavobacteria, and diatoms to those
containing oligotrophic bacteria, such as SAR11
and autotrophic cyanobacteria (17, 22, 40–42).
In the metagenome heatmap (Fig. 6A), different locations across a broad biogeochemical
gradient in a single model run were more similar than equivalent locations across model runs,
indicating that the emergent model communities differed widely in genome content between
runs. However, the metatranscriptome heatmap,
which indicates the actual metabolisms in use at
each location (Fig. 6B), clustered across runs,
with similar metabolisms expressed at equivalent
salinities. For example, at the lowest salinities,
genes encoding for heterotrophic processes involving degradation of dissolved and particulate
organic matter (pcaH, AMA, and AMA-det) and
processes reducing mortality (motG, mot-P, sil,
and cheA) were highly expressed across runs,
whereas at intermediate salinities, photosynthesis
and nitrogen uptake genes (pcb, pbs, nrt, and
amtB) were highly expressed. Thus, the genetic
composition of each GENOME assemblage differed between runs, but the metabolic functions
were similar for a given environment across
Randomness in community assembly has also
been observed in space-limited communities
such as the biofilms on macroalgae (43) and
in communities of coral reef fish (44), where
genetically distinct communities that perform
similar metabolic functions are found in adjacent environments. It is hypothesized that random events determine the arrival of the first
organism with a given metabolic function, resulting in highly divergent communities coexisting
side by side. By analogy, each model run represents a different surface. However, there is no
direct competition for space in the water column.
Instead, niche space is hypothesized to be jointly
developed by the interaction of physical factors
and the metabolic activity of the members of the
community that may act as ecosystem engineers
(45). In the model, all simulations developed similar biogeochemical gradients that acted as niche
space, indicating that the evolution of metabolic
capacity was more important than any specific
ecosystem engineer. Global studies of marine dispersal and evolution suggest that microbes disperse more slowly than they evolve, leading to
formation of biogeographic provinces (46). Thus,
a testable hypothesis emerging from this work is
that the global pool of available metabolic functions, rather than the distribution of functions
among organisms, drives community assembly
and formation of biogeochemical gradients in
REFERENCES AND NOTES
1. J. P. Zehr, M. T. Mellon, S. Zani, Appl. Environ. Microbiol. 64,
2. O. Béjà et al., Science 289, 1902–1906 (2000).
3. L. R. Moore, G. Rocap, S. W. Chisholm, Nature 393, 464–467
4. R. M. Morris et al., Nature 420, 806–810 (2002).
5. L. M. Proctor, J. A. Fuhrman, Nature 343, 60–62 (1990).
6. D. C. Reed, C. K. Algar, J. A. Huber, G. J. Dick, Proc. Natl. Acad.
Sci. U.S.A. 111, 1879–1884 (2014).
7. M. J. Follows, S. Dutkiewicz, S. Grant, S. W. Chisholm, Science
315, 1843–1846 (2007).
8. P. G. Falkowski, T. Fenchel, E. F. Delong, Science 320,
9. E. Marañón, Annu. Rev. Mar. Sci. 7, 241–264 (2015).
10. K. F. Edwards, M. K. Thomas, C. A. Klausmeier, E. Litchman,
Limnol. Oceanogr. 57, 554–566 (2012).
11. A. M. P. McDonnell, K. O. Buesseler, Limnol. Oceanogr. 55,
12. S. J. Giovannoni et al., Science 309, 1242–1245 (2005).
13. J. J. Morris, R. E. Lenski, E. R. Zinser, MBio 3, e00036-12
14. R. E. Lenski, Int. Microbiol. 1, 265–270 (1998).
15. B. K. Swan et al., Science 333, 1296–1300 (2011).
16. S. J. Giovannoni, J. Cameron Thrash, B. Temperton, ISME J. 8,
17. B. M. Satinsky et al., ISME J. 11, 1852–1864 (2017).
18. B. M. Satinsky et al., Proc. Natl. Acad. Sci. U.S.A. 111,
19. B. L. Zielinski et al., PLOS ONE 11, e0160929 (2016).
20. V. Martin-Jézéquel, M. Hildebrand, M. A. Brzezinski, J. Phycol.
36, 821–840 (2000).
21. F. Azam, Science 280, 694–696 (1998).
22. J. I. Goes et al., Prog. Oceanogr. 120, 29–40 (2013).
23. N. Loick-Wilde et al., Limnol. Oceanogr. 61, 460–481 (2015).
24. S. C. Weber et al., Limnol. Oceanogr. 62, 618–631 (2017).
25. T. S. Kostadinov, D. A. Siegel, S. Maritorena, Biogeosciences 7,
26. M. van der Does, L. F. Korte, C. I. Munday, G. J. A. Brummer,
J. B. W. Stuut, Atmos. Chem. Phys. 16, 13697–13710 (2016).
27. R. R. Hood, V. J. Coles, J. Geophys. Res. 109, C06006 (2004).
28. W. J. Pasciak, J. Gavis, Limnol. Oceanogr. 19, 881–888 (1974).
29. A. A. L. Waite, P. A. Thompson, P. J. Harrison, Limnol.
Oceanogr. 37, 468–477 (1992).
30. A. Waite, A. Fisher, P. A. Thompson, P. J. Harrison, Mar. Ecol.
Prog. Ser. 157, 97–108 (1997).
31. J. R. Taylor, R. Stocker, Science 338, 675–679 (2012).
32. T. A. Villareal et al., Nature 397, 423–425 (1999).
33. H. Agogué, M. Brink, J. Dinasquet, G. J. Herndl, Nature 456,
34. J. M. Smith, F. P. Chavez, C. A. Francis, PLOS ONE 9, e108173 (2014).
35. V. J. Coles, R. R. Hood, M. Pascual, D. G. Capone,
J. Geophys. Res. 109, C06007 (2004).
36. C. Yoshikawa, V. J. V. J. Coles, R. R. R. Hood,
D. G. D. G. Capone, N. Yoshida, J. Geophys. Res. Oceans 118,
37. M. R. Stukel, V. J. Coles, M. T. Brooks, R. R. Hood,
Biogeochemistry 10, 13931–13976 (2013).
38. S. M. Gifford, S. Sharma, J. M. Rinta-Kanto, M. A. Moran,
ISME J. 5, 461–472 (2011).
39. V. J. Coles et al., J. Geophys. Res. Oceans 118, 6894–6913 (2013).
40. A. Subramaniam et al., Proc. Natl. Acad. Sci. U.S.A. 105,
41. V. J. Coles, R. R. Hood, in Aquatic Microbial Ecology and
Biogeochemistry: A Dual Perspective, P. M. Glibert, T. M. Kana,
Eds. (Springer International Publishing, 2016), pp. 45–63.
42. M. Doherty et al., Front. Microbiol. 8, 882 (2017).
43. C. Burke, P. Steinberg, D. Rusch, S. Kjelleberg, T. Thomas,
Proc. Natl. Acad. Sci. U.S.A. 108, 14288–14293 (2011).
44. P. F. Sale, Environ. Biol. Fishes 3, 85–102 (1978).
45. D. R. Nemergut et al., Microbiol. Mol. Biol. Rev. 77, 342–356 (2013).
46. F. L. Hellweger, E. van Sebille, N. D. Fredrick, Science 345,
Sequences from the June 2010 study are available from the
National Center for Biotechnology Information under accession
numbers SRP037334 (metagenomes), SRP037995 (nonselective
metatranscriptomes), and SRP039544 (polyadenylate-selected
metatranscriptomes). The model computer code is available
Atlantic Meridional Transect Consortium nutrient data (NER/0/5/
2001/00680) were provided by the British Oceanographic Data
Centre and supported by the Natural Environment Research
Council. Particle concentration fields were provided by
T. Kostadinov. This research was funded by The Gordon and
Betty Moore Foundation through the “River Ocean Continuum of
the Amazon” project (ROCA; grants GBMF2293 and GBMF2928).
Ship time was provided by the National Science Foundation (grant
OCE-0933975). V.C. carried out the modeling effort in
collaboration with R.H., M.S., M.A.M., M.B., and A.B. “Omics” were
measured and analyzed by M.A.M., J.P., B.S., B.Z., and B.C.,
all of whom contributed to the conceptual model development.
P. Y. led the ANACONDAS (“Amazon iNfluence on the Atlantic:
CarbOn export from Nitrogen fixation by DiAtom Symbioses”) /ROCA
teams. This is UMCES contribution 5435.
Materials and Methods
Figs. S1 to S5
Tables S1 to S4
3 May 2017; accepted 19 October 2017
1154 1 DECEMBER 2017 • VOL 358 ISSUE 6367 sciencemag.org SCIENCE
RESEARCH | RESEARCH ARTICLES