6 NOVEMBER 2009 VOL 326 SCIENCE www.sciencemag.org 808 PERSPECTIVES Evolution of Animal Pollination PALEONTOLOGY Jeff Ollerton and Emma Coulthard Animals pollinated specialized seed plants even before ? owering plants evolved. T he evolution of animal pollination in ? owering plants (angiosperms) and the resulting coevolution and diver- si? cation of both angiosperms and major pollinator groups during the late Cretaceous (99.6 to 65.5 million years ago) is one of the classic stories of evolutionary biology ( 1). On page 840 of this issue, however, Ren et al. ( 2) challenge aspects of this story and hint at a much more complex ecological scenario for the evolution of plant-pollina- tor relationships. An important feature of the traditional story is that the non-angiosperm seed plants living during the Mesozoic (251 to 65.5 mil- lion years ago) were mainly wind-pollinated. Although the fossil record of these plants shows evidence of possible animal pollina- tion as early as the late Carboniferous (320 to 300 million years ago) ( 3), this evidence is open to interpretations of the size of pol- len grains (apparently too large to be wind- dispersed), the possibly attractive function of reproductive organs, and patterns of damage by insects. The assumption is that although animal pollination may predate the evolution of the ? owering plants, it was rare and unspe- cialized relative to what was to follow in the late Cretaceous ( 4). Ren et al. now marshal evidence from an impressive range of sources?from compara- tive morphology of fossil insect mouthparts to elemental analysis of the fossils and their surrounding matrix?to propose that a previ- ously overlooked group of Mesozoic scorpi- on? ies was able to feed on a nectarlike ? uid ( 5) produced by a group of now-extinct non- angiosperm seed plants. The authors suggest that the scorpion? ies in turn pollinated the plants. This may be the earliest known exam- ple of coevolution between plants and polli- nators. The evidence that Ren et al. present is compelling, and if they are correct, it will change our understanding of the early ecol- ogy and evolution of pollination by insects. As Darwin ( 6) famously recognized when he speculated about the coevolution of ? ower and tongue length between the Madagascan orchid Angraecum sesquipedale and its (then unknown) moth pollinator, ? owering plants have often evolved tubular structures that hold nectar or protect reproductive organs. The plants can thus discriminate between ? ower visitors, enabling them to specialize on pollinators with appropriately sized mouth- parts ( 7, 8). This match between mouthparts and ? ower depth ( 9) facilitates more accurate pollen placement, meaning that less pollen is wasted, and prevents nonpollinating animals from robbing nectar. It may be a major fac- tor driving the diversi? cation of some angio- sperm genera ( 8) and structuring the patterns of interaction with pollinators in plant com- munities ( 10), but until Ren et al.?s study, it was considered unimportant in non-angio- sperm pollination. The 11 scorpion? y species described by Ren et al. have mouthparts that are both rela- tively long and consistent with ? uid feeding by sucking. The species represent three dif- ferent families, pointing to repeated conver- gent evolution of this feeding strategy, which in turn suggests that substantial quantities of nectarlike ? uid ( 5) were available to these taxa. Ren et al. believe that the source was a group of non-angiosperm seed plants belong- ing to diverse, and mostly extinct, lineages. All these species have repro- ductive organs that are poorly adapted to wind pollination (the previously presumed mode of reproduction for these taxa). Wind pollination requires that pollen-receptive areas are easily accessible to windborne pollen, which is clearly not the case for these plants. Pollen transfer by insects thus seems most likely, and the scorpion? ies are the best candidates so far identi? ed. Some will ? nd these claims controversial, particularly as a key piece of evidence is miss- ing: The authors failed to ? nd any pollen associated with these fossils. This is especially sur- prising for the amber-encased insects, where pollen preserva- tion would be expected ( 11). Absence of evidence is not, however, evidence of absence, and further fossils may provide this information. Biotic pollination was the dominant angiosperm pollina- tion strategy by the late Cre- taceous ( 12). Ren et al.?s research tests old notions that angiosperms evolved in a pre- dominantly wind-pollinated world. Further- more, it challenges us to rethink assumptions that early pollinators were short-tongued gen- eralists that could only exploit open ? owers with easily accessible nectar. Living repre- sentatives of the earliest diverging flower- ing plants have a diverse range of pollination systems ( 13), and many are far from general- ized in their interactions with pollinators. The presence of long-tongued pollinating scorpi- on? ies both before and at the same time as the ? rst angiosperms allows us to imagine that ? owering plants evolved deep ? owers early in their radiation, coexisting with open, general- ist ? owers and gymnosperms as part of a spe- cialist-generalist spectrum similar to modern plant communities ( 14). Can modern assemblages of plants and pollinators be viewed as analogous to their Mesozoic counterparts, at least in terms of functionality, if not phylogenetic identity? Other Mesozoic insects have been suggested to be ? uid-feeding pollinators ( 15, 16), and if this is so, then the scorpion? ies described by C R E D I T : E M M A C O U L T H A R D / U N I V E R S I T Y O F N O R T H A M P T O N Landscape and Biodiversity Research Group, School of Applied Sciences, University of Northampton, Park Cam- pus, Northampton NN2 7AL, UK. E-mail: jeff.ollerton@ northampton.ac.uk Ancient legacy. Mesozoic scorpionflies (Mecoptera) probe the female reproductive organs of Leptostrobus cancer, a member of the extinct order Czekanowskiales. Modern scorpion? ies are much less diverse than their Mesozoic antecedents. Although they only rarely visit ? owers, this habit may be overlooked in modern species and could be a legacy of their distant ancestors. Published by AAAS o n N ov em be r 6 , 2 00 9 w w w .s ci en ce m ag .o rg D ow nl oa de d fro m www.sciencemag.org SCIENCE VOL 326 6 NOVEMBER 2009 809 PERSPECTIVES Ren et al. were probably part of a larger fauna of insects visiting plant reproductive struc- tures. The Mesozoic Era was biotically richer and more complex than previously realized. Studies such as that by Ren et al. provide a tantalizing glimpse of lost worlds of interac- tions between partners that are now extinct or considerably less diverse. References 1. M. Proctor, P. Yeo, A. Lack, The Natural History of Pollina- tion (Timber Press, Portland, OR, 1996). 2. D. Ren et al., Science 326, 840 (2009). 3. W. L. Crepet, Bioscience 29, 102 (1979). 4. W. L. Crepet, Rev. Palaeobot. Palynol. 90, 339 (1996). 5. M. Nepi et al., Ann. Bot. 104, 205 (2009). 6. C. Darwin, On the Various Contrivances by Which British and Foreign Orchids are Fertilised by Insects, and on the Good Effects of Intercrossing (John Murray, London, 1862). 7. J. Ollerton, S. Liede, Ecotropica 9, 107 (2003). 8. J. B. Whittall, S. A. Hodges, Nature 447, 706 (2007). 9. A. Pauw et al., Evolution 63, 268 (2009). 10. M. Stang et al., Ann. Bot. 103, 1459 (2009). 11. S. R. Ramirez et al., Nature 448, 1042 (2007). 12. S. Hu et al., Proc. Natl. Acad. Sci. U.S.A. 105, 240 (2008). 13. L. B. Thien et al., Am. J. Bot. 96, 166 (2009). 14. N. M. Waser, J. Ollerton, Eds., Plant-Pollinator Interac- tions: From Specialization to Generalization (Univ. of Chicago Press, Chicago, 2006). 15. D. Ren, Science 280, 85 (1998). 16. C. C. Labandeira et al., Taxon 56, 663 (2007). 10.1126/science.1181154 Capturing the Complexities of Molecule-Surface Interactions CHEMISTRY Eckart Hasselbrink T he simplest picture of a chemical reac- tion is that two molecules approach, climb a potential energy barrier as bonds get pulled apart in the transition state, and then separate, forming the new products. Molecules move in three dimensions and have internal motions such as vibrations, so a quan- titative model requires molecules to move over a potential energy surface ( 1). Two reports in this issue address the added complexities that result when one of the reactants is a metal sur- face (see the ? gure). On page 829, Shenvi et al. ( 2) describe a method for the quantitative evaluation of one of the open questions in modeling these reactions: How is energy dis- sipated as the molecule approaches the metal surface? On page 832, D?az et al. ( 3) present a pragmatic ? x for the problem of calculat- ing the potential energy surface that describes how molecular hydrogen (H 2 ) reacts with an atomically ? at copper surface. Their approach allows almost every aspect of the experimen- tal ? ndings for this system to be reproduced. For reactions between small molecules in the gas phase, potential energy surfaces cal- culated from ? rst principles (without using inputs from experiments) have become quite successful in predicting product formation ( 4). However, for the more complex case of molecules interacting with a metal sur- face?which is of interest for understand- ing reactions in industrial catalysts?getting the calculations to agree with experiment is much more challenging, because the system to consider is extended and because energy dissipation to electronic excitations may become important ( 5). Modeling a chemical reaction is typically a two-step process. First, the interaction energy of the reactants is calculated for various geo- metric arrangements and then used to cre- ate the potential energy surface. The latter is then the basis for the calculation of the dynam- ics?that is, how the reactants will move and exchange energy until the products are formed. These analyses can be used to calculate reac- tion probabilities for real systems, where the molecules have a distribution of energies. This two-step procedure rests on the thesis work of Robert Oppenheimer: the so-called Born-Oppenheimer approximation (BOA), which allows calculations of electronic energy independent of the motion (such as bond vibrations) of the much heavier nuclei ( 6). This approximation has been very successful for predicting molecular structures and reac- tion dynamics between molecules. However, the interaction with a metal surface is a partic- ularly problematic case for applying the BOA because metals have a continuous electronic excitation spectrum, not discrete energy levels like single atoms and molecules. It has been rigorously shown that the inter- action of a molecule with a metal surface must dissipate energy into substrate electronic exci- tations; processes that dissipate energy in this way are called nonadiabatic ( 7? 9). However, it is still not established how to predict the mag- nitude of this energy, and the error introduced by the BOA, for a particular reaction system. The steadily increasing number of experimen- tal studies that have reported evidence of non- adiabatic behavior emphasizes the need to understand the error created by the BOA [see ( 10) and references therein]. Shenvi et al. present a method that allows a quantitative account of nonadiabaticity and apply it to the particular case of a nitric oxide (NO) molecule scattering off a gold surface. They use data from a sophisticated experiment as a benchmark for their theoretical study. Wodtke and co-workers prepared NO mol- ecules in a highly vibrationally excited state by means of laser techniques and observed that it loses vibrational energy several quanta at a time in the collision ( 11). Their experi- mental data are only consistent with a non- adiabatic coupling mechanism. An adiabatic model for this interaction would predict that the molecules scatter from a metal surface with the vibrational excitation intact, because the vibrational frequency is not in resonance with the vibrations of the surface atoms or the molecule-surface bond. The vibrational motion of the NO mole- cule is connected with an oscillating shift in electronic energy for adding one more elec- tron to the NO molecule. The vibrational motion is hence connected with an oscillat- C R E D I T : P . H U E Y / S C I E N C E Fakult?t f?r Chemie and Center for NanoIntegration, Uni- versit?t Duisburg-Essen, 45117 Essen, Germany. E-mail: eckart.hasselbrink@uni-due.de When molecules meet surfaces. The inter- action of a molecule with a metal surface provides a challenge for the most sophis- ticated molecular structure calculations. Theory is rapidly moving toward a quanti- tative account of how molecules scatter off surfaces, as discussed by Shenvi et al., or react and form bound surface species, as discussed by D?az et al. Two large computational studies describe how theory can better account for the way in which molecules scatter from or react with metal surfaces. Published by AAAS o n N ov em be r 6 , 2 00 9 w w w .s ci en ce m ag .o rg D ow nl oa de d fro m