A / .

particular groups, such as fishes, arthro-pods, and squid, is presumably a reflectionof their composition. No mineralizationoccurred in the polychaete Nereis (1.48%phosphate by weight) under the same ex-perimental conditions.

Phosphatization of Santana Formationfishes has been estimated to occur in life (7)or, on the basis of a comparison with decayrates of modern fish tissues, within 5 hoursof death (29). In our experiments, howev-er, the process was initiated within 2 weeks,but the degree of mineralization of muscletissue increased up to 4 weeks or evenlonger. The rate of decay, and presumablymineralization, would have been furtherslowed at lower temperatures.

This study, together with the evidenceof the fossil record, emphasizes that preciseconditions must prevail in order for miner-alization of soft tissue to take place. But,once the morphology is stabilized by initialmineralization, the potential for preserva-tion of the soft tissue through subsequentdiagenesis is enormously enhanced, al-though the material remains vulnerable tocompaction unless subsequently enclosed ina concretion. There is a critical balancebetween decay to release phosphate andprecipitation to preserve the soft tissue be-fore much morphological detail is lost (1,7). This balance occurs in closed conditionswhere an initial drop in pH triggers phos-phatization. The role of microbial films infossilization may extend beyond the protec-tion of carcasses and inhibition of decom-position (30) to the establishment of theconditions necessary for rapid mineraliza-tion (4).

REFERENCES AND NOTES

1. P. A. Allison, Paleobiology 14, 139 (1988).2. __ and D. E. G. Briggs, in The Processes of

Fossilization, S. K. Donovan, Ed. (Belhaven, Lon-don, 1991), pp. 120-140; in Taphonomy, Releas-ing the Data Locked in the Fossil Record, P. A.Allison and D. E. G. Briggs, Eds. (Plenum, NewYork, 1991), pp. 25-70. Preservation of muscletissue in amber is an exception [A. Henwood,Palaios 7, 203 (1992)].

3. The silicification of soft tissues in the Eocenelignite of Geiseltal is an exception [E. Voigt, Cour.Forsch.-lnst. Senckenberg 107, 325 (1988)].

4. D. M. Martill, Palaeontology 31, 1 (1988).5. , Geol. Today 5, 201 (1989).6. , Nature 346, 171 (1990).7. H.-P. Schultze, Revista Geol. Chile 16, 183 (1989).8. D. M. Martill and D. M. Unwin, Nature 340, 138

(1989).9. J. Mehi, Archaeopteryx 8, 77 (1990).

10. P. A. Allison, Lethaia 21, 403 (1988).11. P. R. Wilby and D. M. Martill, Hist. Biol. 6, 25

(1992).12. R. A. Berner, Science 159, 195 (1968).13. P. A. Allison [thesis, University of Bristol (1987)]

recorded fragments (not exceeding 100 gm) offibrous material (muscle or connective tissue col-lagen) mineralized in an unknown phosphate afterdecay of fish muscle for 8 days in marine mud andwater enriched with calcium phosphate.

14. Live specimens of Crangon crangon (n = 139;size range, 0.09 to 1.77 g) and Palaemon sp. (n =97; size range, 0.10 to 1.95 g) were obtained by

trawling (Plymouth Sound) or hand-netting (PooleHarbour; Starcross Harbour) in southern England.They were held in an aquarium (30 to 1 00C, withinASW of salinity 35 to 37 per mil) until required, andthen killed by anoxia in an anaerobic cabinet,dried with tissue, and weighed.

15. The Tay Estuary was chosen because this site ischaracterized by high rates of degradation oforganic matter as a result of aerobic processesand anaerobic sulfate reduction [R. J. Parkes andW. J. Buckingham, in Proceedings of the 4thInternational Symposium on Microbial Ecology, F.Megasar and M. Gantar, Eds. (Slovene Society forMicrobiology, Ljubljana, Yugoslavia, 1991), pp.617-624; R. J. Parkes, G. R. Gibson, I. Mueller-Harvey, W. J. Buckingham, R. A. Herbert, J. Gen.Microbiol. 135,175 (1989)].

16. D. E. G. Briggs and A. J. Kear, Paleobiology 19,107 (1993).

17. Sampling, which terminated the experiments, wascarried out at 3 days and 1, 2, 3, 4, 6, 8, 10, 15,20, and 25 weeks. The pH and 02 content of theASW and the wet and dry weights of the decayingcarcass were measured. The state of decay anddegree of mineralization were recorded, and min-eralized tissue was examined under a scanningelectron microscope (gold coated with an accel-erating voltage of 4 to 8 kV, or sometimes 12 kV),and analyzed with the use of x-ray diffraction andelectron microprobe techniques.

18. By bubbling 02-free N2 through it overnight whileagitating with a magnetic stirrer.

19. The conditions for experiments A, B, C, and Dcorrespond to those for experiments 1 a, b, c, andd, respectively, in (16).

20. B. A. Cragg, S. J. Bale, R. J. Parkes, Lett. Appl.Microbiol. 15,125 (1992).

21. Similar to bacterially induced precipitates in ex-periments by C. Buczynski and H. S. Chafetz, J.Sediment. Petrol. 61, 226 (1991).

22. C. R. Glenn and M. A. Arthur, Mar. Geol. 80, 231(1988); D. T. Heggie et al., in Phosphorite Re-search and Development, A. J. G. Notholt and 1.Jarvis, Eds. (Special Publication 52, Geological

Society, London, 1990), pp. 87-117; J. Lucas andL. Prevot, C. R. Acad. Sci. Ser. 11 292, 1203(1981); A. Hirschler, J. Lucas, J.-C. Hubert,Geomicrobiol. J. 8, 47 (1990).

23. J. G. Maisey, in Santana Fossils, an IllustratedAtlas, J. G. Maisey, Ed. (T. F. H. Publications,Neptune City, NJ, 1991), pp. 57-88.

24. P. R. Wilby and D. M. Martill (personal communi-cation) have pointed out that, because the analy-ses by Schultze (7) were based on mechanicallyextracted rather than acid-prepared material, anyphosphate peaks on the x-ray diffractometerwould have been swamped by interstitial calciumcarbonate. They observed that the preservation ofmuscle tissue in calcite is not supported by pe-trography or elemental mapping by energy-dis-persive x-ray analysis.

25. Mineralization occurred whether anoxia wasmaintained by anaerocult A or C; the formerresults in a starting level of 18% C02 of gasvolume, the latter only 8 to 9%.

26. Y. Nathan and E. Sass, Chem. Geol. 34,103 (1981).27. A contrary view is expressed in (10).28. H. Willems and M. Wuttke, Neues Jahrb. Geol.

Palaeontol. Abh. 174, 213 (1987).29. D. M. Martill and L. Harper, Palaeontology33, 423

(1990). The experimental conditions (tissue sam-ples of a freshwater fish in an open vessel ofseawater at room temperature) are unlikely tohave been similar to those in the Santana Sea.

30. J.-C. Gall, Lethaia 23, 21 (1990).31. J. L. Nation, Stain Technol. 58, 347 (1983).32. We are grateful to R. J. Parkes for advice and

discussion. T. Kemp, S. Lane, and D. Robinsonassisted with the analysis of samples. S. Powellassisted with photography. F. Frettsome and R.Swinfen of the Plymouth Marine Laboratory coor-dinated the supply of specimens. M. Collins, J. G.Maisey, D. M. Martill, R. Raiswell, K. M. Towe, andP. R. Wilby commented on the manuscript. Thiswork was funded by Natural Environment Re-search Council grant GR3/7235.

18 September 1992; accepted 17 December 1992

Population Structure and the Evolution of Virulencein Nematode Parasites of Fig Wasps

Edward Allen HerreIt is often assumed that parasitic and disease-producing organisms tend to evolve benignrelationships with their hosts over time. In contrast, theoretical arguments suggest thatincreased opportunities for parasite transmission will promote the evolution of increasedvirulence. The natural history of species-specific nematodes that parasitize fig-pollinatingwasps permits the testing of these predictions in natural populations. For 11 species ofPanamanian fig wasps, those species characterized by population structures that result inincreased opportunities for parasite transmission harbor more virulent species of nem-atodes. In addition, differences in population structure are also associated with differencesin other intra- and interspecific phenomena, including sex ratios among the fig waspspecies, the degree of tension in the wasp-fig mutualism, and lethal combat among themales of parasitic wasps.

Parasitic and disease-producing organismsinfluence nearly all aspects of biologicalorganization. These organisms have beenimplicated either directly or indirectly in (i)limiting the geographical range of host or-ganisms (community composition); (ii) reg-

Smithsonian Tropical Research Institute, Apartado2072, Balboa, Republic of Panama or SmithsonianTropical Research Institute, Unit 0948, APO AA34002-0948.

ulating host population densities and dy-namics; (iii) mediating the outcome ofcompetition or predation between theirhosts and other organisms; (iv) maintainingpolymorphisms in blood proteins, and pos-sibly in maintaining general genetic poly-morphisms; (v) underlying sexual selection;and (vi) serving as a selective force leadingto the predominance of sexual reproduction(1-6). The proposed mechanisms by which

SCIENCE * VOL. 259 * 5 MARCH 19931442

on

Janu

ary

13, 2

009

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

parasites might affect all of these ecologicaland evolutionary phenomena depend onthe parasite's virulence (reduction in thelifetime reproductive success of the hostowing to the parasite), begging the questionof what factors ultimately influence theevolution of virulence itself.

It is frequently assumed that, with time,parasites tend to evolve benign relation-ships with their hosts. However, the useful-ness and general validity of this view havebeen questioned (7). Moreover, recentanalyses have suggested instead that para-site virulence should depend on transmis-sion, so that greater opportunities for trans-mission would permit or even promote theevolution of greater virulence (7-13). Con-versely, systems characterized by restrictedopportunities for transmission [for example,systems in which most or all parasite trans-mission is vertical (from parent to off-spring), as opposed to horizontal (fromamong unrelated individuals)] are thoughtto promote neutral or even mutualistic in-teractions among would-be parasites andtheir hosts (14). Whereas evidence sup-ports some aspects of these models (10, 14),many of these proposed relationships havebeen supported on theoretical groundsonly. Unfortunately, limitations imposedby the natural history of the study orga-nisms (for example, difficulties in measuringlifetime reproductive success of individualhosts as a function of parasitism and diffi-culties in measuring opportunities for para-site transmission) often hinder direct test-ing of the predictions made by these modelsin natural populations.

The natural history of fig-pollinatingwasps and the species-specific nematodesthat parasitize them (15-18) permits a di-rect measurement of these crucial parame-ters. In the fig wasps, a number of gravid,pollen-bearing foundress wasps enter a re-ceptive syconium (the enclosed inflores-cence that eventually ripens to become thefig fruit) nearly simultaneously, pollinatethe flowers, lay eggs, and die. The bodies ofthe foundress wasps remain within the figfruit and may be counted to determine thenumber of wasps potentially pollinating andlaying eggs in any given fig. As the fruitripens and seeds develop, the wasp offspringmature (each feeding on the contents ofone seed), eclose, and mate inside the fig.The winged females gather pollen, leavethe syconium, and disperse to begin thecycle anew. When only one foundress waspenters a fig fruit, a count of the offspringgives a direct measure of the wasp's lifetimereproductive success (15).

Just as distinct species of fig-pollinatingwasps are generally associated with distinctspecies of host figs, morphological work onthe Panamanian species has establishedthat distinct species of nematodes are asso-

Table 1. Species of fig, wasp, and nematode studied. The number of fruit crops sampled in orderto estimate Prop, the proportion of single foundress broods in each species, is indicated by n; thenumber of crops sampled per species to estimate the virulence, V (the lifetime reproductivesuccess of nematode-infected wasps relative to uninfected wasps), is given by N; I gives the totalnumber of broods sampled from individual infected wasps; and U gives the total number of broodssampled from individual uninfected wasps. More virulent nematode species are associated withgreater opportunities for transmission.

Pollinator Nematode SingleFig species wasp species species foundress Virulence estimates

(Ficus) (Pegoscapus or (Parasito- broodsTetrapus) diplogaster) n Prop. N I U V

columbrinae P. orozcoi colubrinema 21 0.99 3 26 74 1.01perforata P. standleyi perfornema 22 0.99 6 60 80 0.97paraensis P. sp. paranema 39 0.92 4 56 44 1.00pertusa P. silvestrii pertanema 16 0.90 2 33 9 0.99obtusifolia P. hoffmeyer obtusinema 48 0.83 7 170 89 0.94bullenei P. sp. bullenema 24 0.82 2 12 33 0.99citrifolia P. aesutus citrinema 65 0.78 6 92 129 0.93yoponensis T. ecuadoranus yoponema 15 0.58 1 4 14 0.92nymphaefolia P. amabilis nymphanema 33 0.55 8 183 47 0.90nr. trigonata P. sp. trinema 36 0.31 5 45 103 0.84popenoei P. sp. popenema 61 0.24 8 61 80 0.89

ciated with distinct species of host fig wasps(15-18). The nematodes' life cycle is thusintimately connected with that of the figwasps. In nematode-infested figs, immature,dispersal-phase nematodes crawl onto newlyemerged female fig wasps and are carried bythem to the next fig. At some point, thenematodes enter the body cavity of the waspand begin to consume it. Later, about six orseven adult nematodes emerge from thebody of the dead wasp (personal counts, 16to 18), mate, and lay eggs within the samefig in which the host wasp has laid her eggs.The nematodes' eggs hatch synchronouslywith the emergence of the next generationof fig wasps and begin their cycle anew. Inalmost ripe fig fruits that have been pollinat-ed by only one foundress wasp, the presenceof immature nematodes can be used to de-termine whether or not that individual waspwas infected. Therefore, the number ofwaspoffspring associated with nematode-infectedsingle foundress wasps can be compared tothe number of offspring associated with un-infected single foundresses in order to esti-mate the nematodes' effects on the lifetimereproductive success of their host fig wasps(that is, in order to measure the virulence ofthe parasite).

Studies initiated over 10 years ago (about120 to 140 fig wasp and nematode genera-tions) in the vicinity of the Panama Canalshow that the 11 different species of hostwasp considered here present a continuum ofpopulation structures (in this case, distribu-tions of numbers of foundresses per fig fruit)(19, 20) (Table 1). The opportunities fornematode transmission among the individu-als of any given wasp species depend, inturn, on the population structure of the hostfig wasp. For example, if all the wasp broodsin all of the figs in a population are founded

SCIENCE * VOL. 259 * 5 MARCH 1993

I

a

oh

.3

IIS

1.1,

1.0..0.9 0 0

1.8z 01.0 0.8 0.6 0.4 0.2

Proporion of single foundress broods

Fig. 1. Relation between virulence, measuredas the lifetime reproductive success of nema-tode-infected relative to uninfected female figwasps, and the proportion of multifoundressbroods encountered in 11 Panamanian spe-cies. Increased incidence of multifoundressbroods provides increased opportunities fornematode transmission. Those species of waspcharacterized by increased opportunities fortransmission of their species-specific parasiticnematodes contain more virulent species ofnematodes. All statistical tests indicate a signif-icant relation [P < 0.001 (22)].

by only a single foundress wasp, then allparasite transmission is vertical, from parentwasp to offspring. The offspring of the hostwasp that the nematodes are directly para-sitizing present the only opportunity forpropagation of the nematodes' offspring. Anematode's fitness is therefore completelycoupled to the reproductive success of thehost wasp it is directly parasitizing. In such asystem, any reduction in the wasp's repro-ductive success as a result of the nematodewill ultimately result in the elimination ofthe nematode from the wasp population, ina manner analogous to the elimination of adeleterious dominant allele.

However, if some broods are founded bymore than one foundress wasp, then the

1443

wyvwwwwwwwwwww. pg;nm;;RgnnRmmmmmmo;gmnnong clam

on

Janu

ary

13, 2

009

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

opportunity exists for a nematode's offspringto propagate by parasitizing the offspring ofwasps other than its direct host. As multi-ple foundress broods increase in frequency,a nematode's fitness, on average, becomesprogressively decoupled from the reproduc-tive success of the host wasp it is directlyparasitizing because parasite transmissionincreasingly becomes horizontal, from par-ent wasp to the offspring of another wasp.In such a system, nematodes could afford toreduce the reproductive success of theirdirect hosts, particularly if reduced nema-tode propagation through their host's off-spring were balanced by an equal or greateramount of horizontal transmission (21).

Therefore, within fig wasp species, thelifetime reproductive success of individualfemale wasps can be related to the presenceor absence of nematode infections, therebypermitting the estimation of nematode vir-ulence. Across species, the characteristical-ly different population structures exhibitedby the host wasp species present differentopportunities for the transmission of thenematode species that parasitize them.Consequently, the estimates of virulencefor the different wasp-nematode species as-sociations can be compared in the contextof the different opportunities for nematodetransmission observed among the differenthost fig wasp species.

For each fig-wasp-nematode species asso-ciation, I sampled almost ripe fruit fromdifferent crops and inspected the fruit forthe bodies of the foundresses. In thosebroods with only one foundress, I checkedfor the presence of dispersal-phase nema-todes and allowed the brood to completedevelopment. The wasp offspring were thencounted. For each fruit crop sampled, themean number of offspring per infectedfoundress was divided by the mean numberof offspring per uninfected foundress inorder to provide an estimate of the relativereproductive success of infected versus un-infected single foundresses (virulence).Within each species, the weighted mean ofsuch estimates provides the final estimate ofnematode virulence (Table 1). I then com-pared the rankings of virulence estimatesfor each wasp-nematode species associationwith the rankings of the incidence of singlefoundress broods (those broods in whichonly vertical transmission is possible).

The nematode species with the greatestestimated virulence are associated with hostwasp species that are characterized by pop-ulation structures providing the most fre-quent opportunities for horizontal transmis-sion of their parasites (Table 1 and Fig. 1)'(22). This result indicates that host popu-lation structure and the resulting opportu-nities for parasite transmission can influ-ence the ecological and evolutionary out-come of species interactions in natural pop-

1444

ulations. Moreover, even though thesepairs of species all share the same basicnatural history, estimates of the nematodes'influence on the lifetime reproductive suc-cess of the wasps define a continuum ofrelationships ranging from commensal toparasitic. Furthermore, the result supportsthe proposition that an increase in hostdensity or crowding will be linked to in-creased opportunities for parasite transmis-sion, which in turn will promote the evo-lution of increased virulence of the parasiticor disease-producing organisms infectingthose host populations (7-14).

The preceding analysis addresses ongo-ing relationships in contemporary time. In,addition, several lines of evidence suggest alonger term evolutionary context for thesefindings. The Panamanian nematode spe-cies are all members of the genus Parasito-diplogaster, established by Poinar from nem-atodes collected from the African fig wasp(Elisabethiella stuckenbergi) (16) (Table 1).The parasitic life cycle of this genus ofnematodes represents a significant evolu-tionary shift in an otherwise phoretic group(Diplogasteridae) (16-18). Taken together,the species-specificity, biogeography, andphylogenetic affinities of this group of nem-atodes (16-18) and of their host wasps andhost figs, as well as the DNA sequencedivergence (22) and fossil record amongboth the fig and the wasp species (18,23-25), indicate an ancient association be-tween the nematodes, the wasps, and thefigs that began as a benign, commensalrelationship (18). This strongly suggeststhat, rather than evolving to become morebenign, the nematodes have, if anything,evolved to become more virulent, and thatthe eventual outcome of any particularspecies interaction is largely dependent onthe population structure characteristic ofthe host wasp species.

Independent of nematode virulence, dif-ferences in wasp population structure alsoaffect other important aspects of wasp andfig biology. For example, because of theeffects of local mate competition and in-breeding, wasp species characterized bylower numbers of foundresses produce, onaverage, more female-biased broods (19,20). Because only female wasps performpollination services for the host fig, broodsex ratios are an indicator of the degree oftension (conflict of interest) within themutualistic association between the waspand the fig, such that decreasingly female-biased sex ratios represent a shift away fromthe fig's reproductive interest (15). Further-more, parasitic species of fig wasps arecharacterized by much less subdivided pop-ulation structures than any of the pollinat-ing wasp species (15, 26). These parasiticspecies are often characterized by lethalcombat among the males along with the

SCIENCE * VOL. 259 * 5 MARCH 1993

concomitant morphological adaptations.Therefore, species characterized by themost subdivided population structures havethe greatest female biases in sex ratio, theleast tension in their mutualistic interac-tions, the least tendency for fighting amongmales, and the most benign parasites. Thisobservation points to the diversity of waysin which population structure can influencethe outcome of evolution.

REFERENCES AND NOTES

1 C. A. Toft, in Parasite Host Associations: Coexist-ence or Conflict, C. A. Toft, A. Aeschlimann, L.Bolis, Eds. (Oxford Univ. Press, New York, 1991),pp. 319-343.

2. J. C. Holmes, in Population Biology of InfectiousDiseases, R. M. Anderson and R. M. May, Eds.(Dahlem Workshop Reports, Life Sciences Work-shop Report 25, Springer-Verlag, New York,1982), pp. 37-57.

3. P. W. Price et al., Ann. Rev. Ecol. Syst. 17, 487(1986).

4. J. B. S. Haldane, in On Being the Right Size, J.Maynard Smith, Ed. (Oxford Univ. Press, NewYork, 1985), p. 183.

5. W. D. Hamilton and M. Zuk, Science 218, 384(1982).

6. J. Seger and W. D. Hamilton, in The Evolution ofSex, R. E. Michod and B. R. Levin, Eds. (Sinauer,Sunderland, MA, 1988), pp. 282-290.

7. C. A. Toft and A. Aeschliman, in (1), pp. 1-12.8. S. Levin and D. Pimentel, Am. Nat. 117, 308

(1981).9. R. M. Anderson and R. M. May, Parasitology 85,

411 (1982).10. R. M. May and R. M. Anderson, in Coevolution, D.

J. Futuyma and M. Slatkin, Eds. (Sinauer, Sunder-land, MA, 1982), p. 186.

11. P. W. Ewald, Ann. Rev. Ecol. Syst. 14, 465 (1983).12. A. C. Allison, in Dahlem Workshop Reports, Life

Sciences Research Report 25, Population Biologyof Infectious Diseases, R. M. Anderson and R. M.May, Eds. (Springer-Verlag, New York, 1982), p.245.

13. A. P. Dobson and A. Merenlender, in (1), p. 83.14. J. J. Bull, I. J. Molineux, W. R. Rice, Evolution 45,

875 (1991).15. E. A. Herre, Experientia 45, 637 (1989).16. G. 0. Poinar, Jr., Proc. K. Ned. Akad. Wet. 82,375

(1979).17. _ , The Natural History of Nematodes (Pren-

tice-Hall, Englewood Cliffs, NJ, 1983).18. __ and E. A. Herre, Rev. Nematol. 14, 361

(1991).19. E. A. Herre, Science 228, 896 (1985).20. _ , Nature 329, 627 (1987).21. In wasp species that commonly have single

foundress broods, the nematodes infecting any par-ticular wasp are likely to be highly related to eachother, in contrast to the situation in fig wasp speciesthat commonly have multifoundress broods, whichallows mixing of unrelated nematodes. Consequent-ly, in the latter systems there is more intense selec-tion on nematodes to increase their reproductionrelative to other nematodes infecting the same indi-vidual host. One form a response to such selectionmight take is for the nematodes to eat more of theirparticular host sooner, thereby increasing their ownreproductive output relative to other nematodes in-fecting the same individual host wasp or on otherhosts inside the same fig. The end result would bediminished reproductive success of the host figwasp. Simulation models using population parame-ters obtained from the host wasp species studiedsupport the proposition that increased virulence canevolve, and the results are consistent with the levelsof virulence actually observed. Furthermore, themodels suggest that it is the competition amongdifferent parasite genotypes within individual hoststhat is the critical factor promoting the evolution of

on

Janu

ary

13, 2

009

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

higher virulence [see also H. J. Bremmermann andJ. Pickering, J. Theor. Biol. 100, 411 (1983)].

22. The phylogenetic relationships among all of thespecies of nematodes and their host fig waspsare the subject of ongoing molecular studies andhave not yet been completely resolved. However,the sequence divergence of the chloroplast rbcLgene suggests an ancient split among the differ-ent host fig subgenera. Mitochondrial sequencedivergence among the corresponding pollinatorwasp taxa reflect this split and further suggeststhat relatively greater levels of virulence haveevolved independently several times [(23, 24); E.A. Herre et al., unpublished data]. Following theimplication that virulence evolves quickly withrespect to speciation, and treating species pairsas independent, the Spearman rank correlation(Rs) between weighted means of relative repro-ductive success of infected and uninfected moth-ers and proportion of single foundress broods inthe host wasp species is 0.91 (P < 0.001, n = 1 1)(Table 1 and Fig. 1); with unweighted means, Rs= 0.90 (P < 0.001, n = 11). With the relativenumber of female offspring, Rs = 0.88 (P < 0.001,n = 1 1). Neither the size of the host wasp nor thesize of the nematode species has a significantrelation to virulence.

23. C. C. Berg, Experientia 45, 605 (1989).24. J. T. Wiebes, Neth. J. Zool. 32, 395 (1982).

25. M. E. Collinson, in Evolution, Systematics, andFossil History of the Hamamelidae, P. R. Craneand S. Blackmore, Eds. (Clarendon Press, Ox-ford, 1989), vol. 2, pp. 319-340.

26. W. D. Hamilton, in Sexual Selection and Reproduc-tive Competition in Insects, M. S. Blum and N. A.Blum, Eds. (Academic Press, New York, 1979), p.167.

27. thank J. J. Van Alphen, J. G. Sevenster, G. 0.Poinar, Jr., J. Aho, P. W. Ewald, S. J. Wright, E. G.Leigh, Jr., D. M. Windsor, R. B. Srygley, A. Read, C.A. Toft, E. F. Stockwell, M. J. Ryan, D. W. Schemske,L. E. Gilbert, G. S. Gilbert, M. J. Wade, M. Feldman,and E. Kalko for useful discussions during the de-velopment of this work. E. Jaen, A. Balderson, L.Papemik, A. Gomez, M. Lopez, and R. Mihalikhelped with fieldwork. J. T. Wiebes and W. Ramirezidentified the wasp species. T. Collins, L. Weigt, M.Caries, C. A. Machado, S. S. McCafferty, G. Zuraw-ski, W. Van Houten, A. Graffen, and E. Berminghamhelped with DNA sequencing and analysis. R. V.Cochard was instrumental in developing the simu-lation model. Thisworkwas supported bytwo Smith-sonian Post-Doctoral Fellowships, two SmithsonianScholarly Studies Grants, The Smithsonian TropicalResearch Institute Molecular Program, and a NATOPost-Doctoral Fellowship.

10 August 1992; accepted 30 November 1992

Structure-Based Discovery of Inhibitors ofThymidylate Synthase

Brian K. Shoichet,* Robert M. Stroud,t Daniel V. Santi,Irwin D. Kuntz,t Kathy M. Perry*

A molecular docking computer program (DOCK) was used to screen the Fine ChemicalDirectory, a database of commercially available compounds, for molecules that are com-plementary to thymidylate synthase (TS), a chemotherapeutic target. Besides retrievingthe substrate and several known inhibitors, DOCK proposed putative inhibitors previouslyunknown to bind to the enzyme. Three of these compounds inhibited Lactobacillus caseiTS at submillimolar concentrations. One of these inhibitors, sulisobenzone, crystallizedwith TS in two configurations that differed from the DOCK-favored geometry: a counterionwas bound in the substrate site, which resulted in a 6 to 9 angstrom displacement of theinhibitor. The structure of the complexes suggested another binding region in the activesite that could be exploited. This region was probed with molecules sterically similar tosulisobenzone, which led to the identification of a family of phenolphthalein analogs thatinhibit TS in the 1 to 30 micromolar range. These inhibitors do not resemble the substratesof the enzyme. A crystal structure of phenolphthalein with TS shows that it binds in thetarget site in a configuration that resembles the one suggested by DOCK.

Thymidylate synthase (TS) is a target fordrugs against proliferative diseases and can-cer because it is required for de novo syn-thesis of deoxythymidine monophosphate(dTMP) and, hence, for DNA production.Efforts have focused on the design of inhib-itors similar to the substrate, deoxyuridinemonophosphate (dUMP), or the cofactor,5, 10-methylenetetrahydrofolate (CH2-H4folate). Administered as the premetaboliteB. K. Shoichet, D. V. Santi, I. D. Kuntz, Department ofPharmaceutical Chemistry, University of California,San Francisco, CA 94143.R. M. Stroud and K. M. Perry, Departments of Phar-maceutical Chemistry and Biochemistry and Biophys-ics, University of California, San Francisco, CA 94143.

*B.K.S. and K.M.P. are equally responsible for thework described in this paper.tTo whom correspondence should be addressed.

5-Fluorouracil, 5-Fluorouridylate is a mech-anism-based inhibitor ofTS (1) that is usedin chemotherapy, whereas 10-propargyl-5,8dideazafolate (CB37 17) is a cofactor mimic(2). Although CB3717 has an inhibitionconstant, Ki, of 40 nM (3) for TS, it showsliver and kidney toxicity (4), which em-phasizes the need for more efficaciousagents. The x-ray structure of TS (5, 6)has been used in iterative cycles of crys-tallographic analysis, synthesis, and inhi-bition assays to design high-affinity hetero-cyclic inhibitors that resemble the cofac-tor (7). Anti-TS drugs dissimilar to thesubstrate and cofactor remain attractivebecause they are less likely to have the sideeffects that are produced by the nucleotideand folate mimics.

We sought novel inhibitors ofTS with acomputational screen targeted at the struc-ture of L. casei TS determined at 2.3 Aresolution (8). We used the programDOCK (9-1 1) to explore the active site ofTS with molecules from the Fine ChemicalDirectory (FCD) (12). For each FCD mol-ecule, DOCK examined an average of 104orientations for steric fit to TS and scoredeach on the basis of the energy of theelectrostatic interaction with the enzyme.Of the 55,313 molecules searched, the 600with the highest scores were saved. Weapplied a correction for ligand solvation,which was based on a modified Born equa-tion (13), to sort the selected molecules.

The TS substrate, dUMP, and severalknown nucleotide and non-nucleotide in-hibitors of the enzyme received goodscores (Table 1). The program retrievedsome molecules previously unknown tobind to TS. On the basis of DOCK scoreand lack of similarity to dUMP, we se-lected 25 of these molecules as putativeinhibitors. Three of the compounds, in-cluding sulisobenzone (SB), inhibited TSwith values of the inhibition constant IC50(50% reduction) in the high micromolarrange (Table 1).

The TS-SB complex was crystallizedfrom both phosphate and tris buffers, andthe structure of each was determined to 2.5A resolution (14) (Fig. 1). In neither struc-ture was SB found in the binding modefavored by DOCK. We anticipated that thesulfonate group of SB would bind in the TSsubsite that is occupied by a phosphatemoiety in all of the other structures that wehave solved (15). Instead, both of thesestructures showed a buffer-derived anion,either phosphate or sulfate from the(NH4)2SO4 precipitant, bound at this site(Fig. 1B). In both structures, SB is dis-placed away from the nucleotide bindingregion to different but overlapping regionsof the active site. In the structure crystal-lized from the phosphate buffer, the sul-fonate moiety of SB forms hydrogen bondswith the sulfhydryl of the conserved activesite nucleophile, C198, and with the back-bone NH of D221, a conserved residue thatis involved in cofactor binding (6, 16). Themethyl of SB makes van der Waals contactswith L224, which also contributes to thecofactor binding site. In the structure thatcrystallized from the tris buffer, the SB isrotated 1530 and translated 5.1 A withrespect to the configuration that crystallizedfrom the phosphate buffer. The sulfonate isaccessible to solvent in the vicinity of theTS carboxyl terminus (Fig. iB). The sub-stituted ring of SB is in van der Waalscontact with W85 at the mouth of theactive site. Taken together, the two SBconfigurations partially overlap the sub-strate and cofactor subsites of TS, as deter-

SCIENCE * VOL. 259 * 5 MARCH 1993

1111

1445

on

Janu

ary

13, 2

009

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om