Chloroplast - PNAS · Evolution: Soltis et at againwithTBRbranch swapping, butwith all mostparsimo-nious trees saved. The size ofthe data set precluded internal assessments of

Proc. Natl. Acad. Sci. USAVol. 92, pp. 2647-2651, March 1995Evolution

Chloroplast gene sequence data suggest a single origin of thepredisposition for symbiotic nitrogen fiLxation in angiospermsDOUGLAS E. SoLTIs*, PAMELA S. SoLTIs*, DAVID R. MORGANt, SUSAN M. SWENSENt, BETH C. MULLIN§,JULIE M. DOWDI, AND PETER G. MARTIN II

*Department of Botany, Washington State University, Pullman, WA 99164-4238; tDepartment of Biology, Western Washington University, Bellingham, WA98225; tDepartment of Biology, Indiana University, Bloomington, IN 47405; §Department of Botany, Center for Legume Research, University of Tennessee,Knoxville, TN 37996; and IDepartment of Botany, University of Adelaide, Adelaide, South Australia 5005, Australia

Communicated by Michael T. Clegg University of California, Riverside, CA, November 14, 1994

ABSTRACT Of the approximately 380 families of angio-sperms, representatives of only 10 are known to form symbi-otic associations with nitrogen-fixing bacteria in root nod-ules. The morphologically based classification schemes pro-posed by taxonomists suggest that many of these 10 familiesof plants are only distantly related, engendering the hypoth-esis that the capacity to fix nitrogen evolved independentlyseveral, if not many, times. This has in turn influencedattitudes toward the likelihood of transferring genes respon-sible for symbiotic nitrogen fixation to crop species lackingthis ability. Phylogenetic analysis of DNA sequences for thechloroplast gene rbcL indicates, however, that representativesof all 10 families with nitrogen-fixing symbioses occur to-gether, with several families lacking this association, in asingle clade. This study therefore indicates that only onelineage of closely related taxa achieved the underlying geneticarchitecture necessary for symbiotic nitrogen fixation in rootnodules.

Nitrogen-fixing symbioses in root nodules are known in only10 of the approximately 380 families of angiosperms. Althoughonly a fraction of legumes (Fabaceae) have been studied,nodulation has been found in more than 90% of the plantsexamined in subfamilies Mimosoideae and Papilionoideae andin 30% of Caesalpinioideae (1, 2). In addition to occurring inthe legumes, nitrogen-fixing symbioses involving root nodulesalso occur in some members of Betulaceae, Casuarinaceae,Coriariaceae, Datiscaceae, Elaeagnaceae, Myricaceae, Rham-naceae, Rosaceae, and Ulmaceae (Table 1) (4, 5, 7, 8). Nodulesare induced and inhabited by either of two very differentgenera of bacteria. Species of Rhizobiaceae (Gram-negativemotile rods) nodulate the legumes and Parasponia (subfamilyCeltoideae, Ulmaceae) (9). Actinomycetes of the genusFrankia (Gram-positive, non-endospore-forming, mycelialbacteria) nodulate hosts in the remaining eight families (Table1), plants referred to as actinorhizal (4, 5, 10).The 10 families with nitrogen-fixing symbioses are distrib-

uted among four of Cronquist's (3) six major subgroups(subclasses) of dicotyledons: Magnoliidae, Dilleniidae, Rosi-dae, and Hamamelidae. Many of these families have beenconsidered to be only distantly related (3, 11-14), engenderingthe hypothesis that nitrogen fixation evolved independentlyseveral, if not many, times (7, 8). This view has in turninfluenced attitudes toward the likelihood of transferringgenes responsible for symbiotic nitrogen fixation to cropspecies that do not possess this ability (15, 16). That is, theapparently great phylogenetic distance between some hostplants suggests that the bacterial component can adapt to awide range of genetic backgrounds. In the hope of elucidatingphylogenetic relationships among the 10 families of angio-

Table 1. Angiosperm families that participate in nodularnitrogen-fixing symbioses and the frequency of this association ineach family

Total no. of genera*/generaProkaryote Family having root nodulestRhizobium Fabaceae 640/most

Ulmaceae 18/1Frankia Betulaceae 6/1

Casuarinaceae 4/4Elaeagnaceae 3/3Myricaceae 3/2Rhamnaceae 55/7Rosaceae 100/5Datiscaceae 3/1Coriariaceae 1/1

*From Cronquist (3).tFrom Akkermans and van Dijk (4), Bond (5), and Torrey and Berg(6).

sperms engaged in nitrogen-fixing symbioses, we employedcomparative gene sequencing of rbcL. * *

MATERIALS AND METHODSRecently, phylogenetic trees were presented for angiospermsbased on 499 sequences of the chloroplast gene rbcL, whichencodes the large subunit of ribulose-1,5-bisphosphate car-boxylase/oxygenase (EC 4.1.1.39), the chief carbon-fixingenzyme in photosynthesis (17). Included in that analysis wererbcL sequences representing the three subfamilies of legumes,as well as representatives of eight other families with nitrogen-fixing symbioses: Betulaceae, Casuarinaceae, Coriariaceae,Datiscaceae, Myricaceae, Rhamnaceae, Rosaceae, and Ul-maceae. Contrary to traditional systematic treatments, thephylogenetic trees based on rbcL sequences (17) suggested thatangiosperm families that participate in nodular symbiosesoccur together, interspersed with families that lack thesesymbioses, in one clade corresponding generally to part of thedicot subclass Rosidae (called "Rosid-I"; Fig. 1). However, formost of the families with nitrogen-fixing symbioses the actualgenera involved in root nodule symbiosis were not included inthe analysis (17). Furthermore, the analysis of 499 rbcLsequences (17) resulted in 3900 trees. Because of the largenumber of sequences involved and the numerous most parsi-monious trees obtained, Chase et al. (17) admit that theiranalysis suffers from uncertainties, although it provides ageneral guide for more detailed studies.To elucidate the evolutionary origin of nodular nitrogen-

fixing symbiosis in angiosperms, we therefore conducted abroad phylogenetic analysis of rbcL sequences, including sev-

IDeceased December 15, 1994.**Sequences reported in this paper have been deposited in theGenBank data base (accession nos. V17038, V17039, and U20805).

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

2647

Proc. NatL Acad ScL USA 92 (1995)

FIG. 1. Summary of the major clades identified in the strictconsensus of 3900 equally parsimonious trees based on rbcL sequencesfor 499 taxa (17).

eral sequences for actinorhizal taxa (Table 2) not includedpreviously by Chase et al. (17), to ensure that all angiosperm

families known to participate in nodular symbioses were repre-sented by taxa actually involved in nitrogen-fixing symbioses.Sequences were determined for members of Casuarinaceae(Allocasuarina) and Elaeagnaceae (Elaeagnus -and Shep-herdia), and recently published sequences for Rhamnaceae(Ceanothus) (22), Rosaceae (Purshia and Cercocarpus) (22),and Betulaceae (Alnus) (21) were also included (Table 2). Asa result, all families of angiosperms (except Ulmaceae) knownto host nitrogen-fixing bacteria in root nodules are repre-sented in this analysis by rbcL sequences for genera that trulyare hosts. In Ulmaceae, nitrogen fixation is known to occur

only in Parasponia, a member of subfamily Celtoideae (9).Celtoideae are represented herein by sequences for the relatedgenera Celtis and Trema. Also included were 78 additional rbcLsequences from genera lacking nitrogen-fixing symbioses inboth nitrogen-fixing and non-nitrogen-fixing families of theRosid I clade, chosen using the rbcL trees of Chase et al. (17)as a general guide, but with additional sequences added.

Five genera (Trochodendron, Tetracentron, Platanus, Sabia,and Lambertia) previously shown to be basal to the Rosid Ilineage (17) were used as outgroups to root the phylogenetictrees. These 99 rbcL sequences were subjected to phylogeneticanalysis using PAUP version 3.1.1 (24), which attempts to findthe shortest (most parsimonious) trees. Because of the smallerscope of this project, we were able to conduct a more detailedphylogenetic analysis of the Rosid I dlade than was possible inthe larger analysis of Chase et al. (17). The heuristic searchinvolved 800 replications with random taxon addition, usingtree bisection-reconnection (TBR) branch swapping and sav-ing a single shortest tree from each replicate. The shortest ofthese trees were then used as starting trees for further analyses,

Table 2. Representatives of angiosperm families with nitrogen-fixing symbioses for which rbcLsequences were used in the current study

Species Family Ref.

Species with nodular nitrogen-fixing symbioses and the families they representPisum sativum L. Fabaceae (Papilionoideae) 18Medicago sativa L. Fabaceae (Papilionoideae) 19Albizia julibrissin Durazz. Fabaceae (Mimosoideae) 17Bauhinia sp. Fabaceae (Caesalpinioideae) 17Coriaria myrtifolia L. Coriariaceae 17Datisca cannabina L. Datiscaceae 20Datisca glomerata L. Datiscaceae 20Myrica cerifera L. Myricaceae 17Casuarina litorea L. Casuarinaceae 17Allocasuarina muelleriana Miq. Casuarinaceae GenBank U20805Alnus crispa (Ait.) Pursh Betulaceae 21Purshia tridentata (Pursh) D.C. Rosaceae 22Cercocarpus ledifolius Nutt. Rosaceae 22Ceanothus sanguineus Pursh. Rhamnaceae 22Shepherdia canadensis (L.) Nutt. Elaeagnaceae GenBank V17039Elaeagnus angustifolia L. Elaeagnaceae GenBank V17038

Species lacking nodular nitrogen-fixing symbioses and the families they representOctomeles sumatrana Miq. Datiscaceae 20Tetrameles nudiflora R. Br. Datiscaceae 20Betula nigra L. Betulaceae 17Rhamnus cartharticus L. Rhamnaceae 17Celtis yunnanensis C. K. Schneid. Ulmaceae 17Trema micrantha Blume Ulmaceae 17Geum chiloense Balb. ex Ser. Rosaceae 23Photinia fraseri Dress Rosaceae 23Prunus virginiana L. Rosaceae 23Rubus idaeus L. Rosaceae 22Spiraea vanhouttei (Briot) Zabel Rosaceae 23

Species with and without nodular nitrogen-fixing symbioses were included. References indicate thesource for each sequence. The remaining 72 sequences included in the analysis, from families not knownto possess nodular nitrogen-fixing symbioses, were taken from Chase et al. (17). Legume subfamilies aregiven in parentheses.

2648 Evolution: Soltis et al

Evolution: Soltis et at

again with TBR branch swapping, but with all most parsimo-nious trees saved.The size of the data set precluded internal assessments of

support by using bootstrap (25) and standard decay (26, 27)analyses. To assess support for the "nitrogen-fixing clade"described below, we performed a decay analysis on this single

Proc. NatL Acad Sci USA 92 (1995) 2649

clade. We constructed a constraint tree having a basal poly-tomy and a single branch corresponding to the "nitrogen-fixing clade" and then saved all trees one and two steps longerthan the most parsimonious trees that were incompatible withthis specified topology. Groups of genera that collapse (i.e.,cease to be monophyletic or "decay") in trees two or more

X>ByrsonimaThryallisOchnaHumiriaLicanisChrysobalanusJr onla,rtrxlum

ViolaEuphorbiaPassifloraLepuropetalonEuonymusEucryphiaBaueraCeratopetalumCehal tusPltteca

PhotiniaRubusCercocarpus*Purshla*MorusFicusCannabisHumulusCaltis**Trema**PilesBoehmeriaUlmusCeanothus*Eieaoanus*Shcphierdia*RhamnusBauhinia*AIbIzIa*Pisum*Medicago*StylobaslumSecuridacaPolygaia(LuilbojaCorylusAlnus*BetuisCasuarina*Allocasuarina*TrigonobalanusChrysoapisMyrica*CaryaFaaus

u

CorlariaOctomelesTetramelsDatisca cann.*Datisca glom.*LuffaCucurbitaCucumisBeqonlaSeaumV rassulaotracarpaea

MyriophyllumRibesAstilbeFrancosGreylaCrossosomaKrameriaGuaiacumLythrumClarkia

C isqualisC ualqeaCapparisCaricaTropaeolumAcerCupaniopsisAilanthusGossypiumHamamelisVivianiaCercidiphyllumMpo;oniaGunners _SabiaLambertlaPlatanusTetracentronTrochodendron

CD(U50)

._XC4-C9CD0)0I-4-az

FIG. 2. Strict consensus of 558 shortest trees resulting from phylogenetic analysis of 99 rbcL sequences representing the Rosid I lineage (seeFig. 1) and additional taxa. All taxa engaged in nodular nitrogen-fixing symbioses are indicated by an asterisk. Datisca cann., Datisca cannabina;Datisca glom., Datisca glomerata. In Ulmaceae, nitrogen fixation is known to occur only in Parasponia, a member of subfamily Celtoideae. Celtoideaeare represented herein by sequences of the related genera Celtis and Trema, each of which is marked by two asterisks. Gunnera (Gunneraceae),which hosts nitrogen-fixing cyanobacterial symbionts in leaf glands, rather than root nodules, is noted by an arrow. The "nitrogen-fixing clade"is supported by a decay value of 2, as indicated at the base of this clade (see text for further discussion). Letters (A-D) above lines designate thefour subclades of the nitrogen-fixing clade that actually contain taxa involved in nitrogen-fixing symbioses. These four subclades form amonophyletic group in 74% of the 558 shortest trees.

Proc. Natl. Acad. Sci USA 92 (1995)

steps longer than the shortest trees are considered bettersupported than those that decay at only a single step longer.

RESULTS AND DISCUSSIONThe phylogenetic analysis produced 558 trees of 3552 stepsdistributed on three islands (28). The strict consensus of thesetrees (Fig. 2) indicates that all of the families with nitrogen-fixing symbioses occur together as part of the same large clade(labeled "nitrogen-fixing clade") with several families lackingthese symbioses. This clade is maintained in all trees found thatwere a single step longer than the shortest trees, but it collapsesin the strict consensus of all trees found that were two stepslonger than the shortest trees. Although the nitrogen-fixingclade collapses in trees only two steps longer than the shortesttrees, the maintenance of this clade through the first step of thedecay analysis provides additional support for the monophylyof the nitrogen-fixing clade, especially given the large numberand taxonomic breadth of taxa included in the analysis. Two ofthe six lineages of this clade do not contain any membersknown to host nitrogen-fixing bacteria. The four lineages ofthis clade that contain symbiotic nitrogen-fixing genera forma monophyletic group in 74% of the 558 trees, suggesting thatall of the families with nodular nitrogen fixation may haveindeed shared a common ancestor not shared by other familiesof angiosperms.

Phylogenetic analysis of rbcL sequences suggests a closerelationship among the Rosaceae, the celtoid line (Trema andCeltis) of Ulmaceae, Rhamnaceae, and Elaeagnaceae, all withgenera that host nitrogen-fixing bacteria, and the Moraceae,Cannabaceae, and Urticaceae, which do not. This relationshipis supported not only by rbcL base substitutions but also bystructural variation in the region flanking the 3' end of rbcL.These families are characterized by a 4-bp duplication located12 bp 3' to the terminus of rbcL, although the sequences ofHumulus (Cannabaceae) and Pilea and Boehmeria (Urti-caceae) were not analyzed for this feature (Fig. 3). Thisduplication is absent from all other species of Rosidae exam-ined (26 taxa in 19 families; Fig. 3) (22).The presence in a single clade of all families that engage in

symbiotic nodular nitrogen fixation indicates a much closer

1- -1-i1Rubus TAAtctagcaattacttactgttagttctcttaatt

GeuO TAAtccagcaattacttactcttagttcttttaatt

Elasagnus TAAtcttgtaattacttacagctcgttcttttaatt

Celtis TAAtccagcaattacttctgttccttttcttaatt

Morus

Rhauus

Ceanothus

TAAtccagcaattacttactgttccttttcttaatt

TAAtccagtaattacttactgttcgttctcttaatt

Chryao7l.pis TAAtccagtaattac----cgctcgttctcttaatt

Myrica

Casuarlna

Ceratopetal=us

TAAtccaataattac----ccctcgttcttttaatt

TAAtccagtaattat----cgctcgttctcttaatt

TAAtccagtaattcc----tgttcgttctcttaatt

Bauera TAAtccagtaattcc----tgttcgttctcttaatt

Cephalotus TAAtccagtaattac----tgttcgttctcttaatt

FIG. 3. Duplicated sequence (in brackets) adjacent to 3' end (basesin uppercase) of rbcL present in several families with nitrogen-fixingsymbioses. Chrysolepis, Myrica, and Casuarina, representing familieswith nitrogen-fixing symbioses, lack the duplication. Ceratopetalum,Bauera, and Cephalotus, representing families without nitrogen-fixingsymbioses, also lack the duplication.

relationship among these taxa than is suggested by currentclassification schemes. Thus, molecular systematic data chal-lenge the morphologically based inference that many familiesthat host nitrogen-fixing bacteria are distantly related (3,11-14). These findings, furthermore, suggest a single evolu-tionary origin of the underlying capacity for symbiotic nodularnitrogen fixation. Significantly, Gunneraceae, an angiospermfamily that hosts nitrogen-fixing cyanobacterial symbionts inleaf glands (rather than in root nodules), are not a member ofthis nitrogen-fixing clade (Fig. 2). Gunneraceae, therefore,clearly represent an independent evolution of symbiotic nitro-gen-fixing ability in angiosperms.The clade containing symbiotic nitrogen-fixing taxa also

includes nine families that do not participate in root nodulesymbiosis (Begoniaceae, Cucurbitaceae, Fagaceae, Juglan-daceae, Cannabaceae, Moraceae, Polygalaceae, Surianaceae,Urticaceae). Furthermore, with the exception of subfamilyPapilionoideae and Mimosoideae of Fabaceae, the vast ma-jority of species in families such as Rosaceae, Rhamnaceae,Ulmaceae, and the Caesalpinioideae (Fabaceae) are notknown to nodulate and fix nitrogen (Table 1). For example, ofthe 18 genera in Ulmaceae, only Parasponia of subfamilyCeltoideae has a symbiotic nitrogen-fixing association. There-fore, if the underlying predisposition to engage in symbioticnitrogen fixation stems from a single, common origin, mem-bers of this clade that do not share this symbiosis must have lostthe ability to form such associations. Alternatively, the ances-tor of the nitrogen-fixing clade may have evolved the geneticcomponents that would ultimately permit the evolution ofsymbiotic nodular nitrogen fixation. Following the establish-ment of these conditions, the necessary genetic backgroundwas present to allow parallel, recurrent evolution of symbioticnitrogen fixation in the subsequent diversification of this clade.Both hypotheses call for mutations that established the sym-biotic association with nitrogen-fixing bacteria, or at least thebasis for this symbiosis, in the ancestor of the nitrogen-fixingclade.Within the nitrogen-fixing clade, the legume/rhizobia sym-

bioses and actinorhizal/frankiae symbioses fall into distinctlineages. One of the four subclades that contain members withsymbiotic nitrogen-fixing associations consists solely of le-gumes and related nonsymbiotic relatives. Actinorhizals occur,along with nonsymbiotic taxa, in each of the other three clades.The celtoid line (Trema and Celtis) of Ulmaceae, representingthe Parasponia/rhizobia symbiosis, occurs nested within alarger clade that includes actinorhizal/frankiae symbioses inRosaceae, Rhamnaceae, and Elaeagnaceae. The distinctnessof the legumes from all actinorhizals is further supported byaspects of the symbioses themselves. A major feature thatdistinguishes the legume/rhizobia symbioses from those in-volving actinorhizals and frankiae or Parasponia and rhizobialies in the ontogeny of the nodule. Whether initiated by roothair infection, intercellular penetration, or infection at woundsites, all legume nodules result from cell divisions within theroot cortex and have a stem-like anatomywith peripheralvasculartissue. All actinorhizal and Parasponia nodules, regardless of themode of infection, form by modification of lateral roots andmaintain a central vascular tissue. That organogenesis is directedby the host plant is confirmed by observations that Parasponiarhizobia induce typical legume nodules on compatible legumesand legume rhizobia induce root-like nodules onParasponia (29).Also, some legumes may develop spontaneous nodules in theabsence of a bacterial symbiont, confirming that the develop-mental pathway leading to the formation of nodules is encodedin the plant genome (30).The fact that the Celtoid line of Ulmaceae and the legumes

(families symbiotic with Rhizobiaceae) occur in two separateclades, with the former more closely allied with angiospermfamilies that are nodulated by actinomycetes (Frankia), may berelevant to the origin of root-nodule symbioses. The steps

2650 Evolution: Soltis et at

Proc. Natl. Acad. Sci. USA 92 (1995) 2651

leading to the formation of effective nodules in host plantsinvolve a combination of plant and bacterial signals. Thedetails of this "two-way molecular conversation" (31) varydepending on the host and bacteria involved. It seems unlikelythat a new angiosperm host with the capacity for nodulationwould successfully "converse" with two different bacterialspecies in separate recognition events. Rather, this angio-sperm host more likely acquired the necessary cellular ma-chinery that permitted symbiosis, of which one or the othertype of bacteria could take advantage.

Literature in the area of symbiotic nitrogen fixation providesfew clues as to the nature of the critical steps involved in theformation of the symbiotic association (reviews in refs. 32 and 33).Several lines of evidence actually suggest that the plant genes andproteins involved in nodulation are not unusual (reviewed in ref.2). For example, recent studies suggest that the infection threadsinvolved in nodulation may be related to normal cell functionsthat occur in all plants (34). Furthermore, the nodulation genesin rhizobia can be induced by flavonoids common to manyangiosperms, as demonstrated by Peters et al. (35). In addition,one of the earliest appearing nodule-specific proteins (nodulins),ENOD12, is a cell-wall component also found in stems andflowers (36). Of the late nodulins, leghemoglobin may be presentin all plants (37), and at least six of the enzymes involved innodulation are normal "housekeeping" proteins (38). Evidencefrom legumes also indicates that particular steps essential tosymbiotic establishment in one species may simply be by-passedor achieved differently in other species (39). Single-gene muta-tions can simultaneously prevent symbioses with nitrogen-fixingbacteria and association with vesicular-arbuscular mycorrhizae(40). Nodulation may therefore involve interactions among manydifferent common genes, rather than unique genes ready to beswitched on when appropriate microsymbionts are available (2).The longstanding idea that symbiotic nitrogen-fixing species

are taxonomically diverse (3, 11-14) has engendered the beliefthat nodulation could be extended to other plants by themolecular transfer of genes critical for nodular symbiosis. Forexample, rhizobial-induced galls were recently obtained onrice (41). Although the transfer of root nodulation and nitro-gen fixation to plants such as rice may ultimately be possible,phylogenetic analyses of rbcL sequence data indicate clearlythat of the approximately 250,000 to 300,000 species and 380families of angiosperms, only one lineage of closely relatedtaxa achieved the underlying genetic architecture necessary forroot-nodule symbiosis. The likelihood of only a single origin ofthe predisposition for root-nodule symbioses in angiospermstherefore has implications regarding the crucial evolutionaryevents required in this process. Future efforts to unravel theprocess and evolution of nitrogen-fixing symbioses, and thetransfer of this capacity to nonnodulating species, should firstfocus on related taxa from the nitrogen-fixing clade thatpossess and lack symbiotic nitrogen-fixing ability. Concomi-tantly, nodulating and nonnodulating members of this nitro-gen-fixing clade should be examined to ascertain whetherrecurrent losses or recurrent gains of nitrogen-fixing abilityhave occurred.

The surviving authors dedicate this paper to our friend and col-league, Peter Martin, who died in December 1994. Peter providedinspiration, stimulus, and a wealth of ideas to a generation of botanists.This research was supported in part by National Science FoundationGrant BSR-9007614 to D.E.S., National Science Foundation GrantDEB-9306913 to S.M.S. and L. H. Rieseberg, and U.S. Department ofAgriculture Grant 93-37305-9082 to B.C.M.

1. Allen, 0. N. & Allen, E. K. (1976) Symbiotic Nitrogen Fixation inPlants (Cambridge Univ. Press, Cambridge, U.K.).

2. Sprent, J. I. & Sprent, P. (1990) Nitrogen Fixing Organisms(Chapman & Hall, New York).

3. Cronquist, A. (1981) An Integrated System of Classification ofFlowering Plants (Columbia Univ. Press, New York).

4. Akkermans, A. D. L. & van Dijk, C. (1981) in Nitrogen Fixation,ed. Broughton, W. J. (Clarendon, Oxford), Vol. 1, pp. 57-103.

5. Bond, G. (1983) in Biological Nitrogen Fixation in Forest Ecosys-tems, eds. Gordon, J. C. & Wheeler, C. T. (Nijhoff, Dordrecht,The Netherlands), pp. 55-87.

6. Torrey, J. G. & Berg, R. H. (1988) Am. J. Bot. 75, 864-874.7. Baker, D. D. & Mullin, B. C. (1992) in Biological Nitrogen

Fixation, eds. Stacey, G., Burris, R. H. & Evans, H. J. (Chapman& Hall, New York), pp. 259-291.

8. Mullin, B. C., Swensen, S. M. & Goetting-Minesky, P. (1990) inNitrogen Fixation Achievements and Objectives, eds. Gresshoff,P. M., Roth, L. E., Stacey, G. & Newton, W. E. (Chapman &Hall, New York), pp. 781-787.

9. Trinick, M. J. & Galbraith, J. (1980) New Phytol. 86, 17-26.10. Torrey, J. G. & Tjepkema, J. D. (1979) Bot. Gaz. Suppl. 140, i-ii.11. Thorne, R. F. (1992) Aliso 13, 365-389.12. Thorne, R. F. (1992) Bot. Rev. 58, 225-348.13. Takhtajan, A. (1987) System of Magnoliophyta (U.S.S.R. Acad.

Sci., Leningrad).14. Dahlgren, R. M. T. (1980) Bot. J. Linn. Soc. 80, 91-124.15. Sprent, J. I. & Raven, J. R. (1992) in Biological Nitrogen Fixation,

eds. Stacy, G., Burris, R. H. & Evans, H. J. (Chapman & Hall,New York), pp. 461-496.

16. Mullin, B. C. (1992) in Nodulation and Nitrogen Fixation in Rice,eds. Khush, G. S. & Bennett, J. (Int. Rice Res. Inst., Manila, ThePhilippines), pp. 67-75.

17. Chase, M. W., Soltis, D. E., Olmstead, R. G., Morgan, D., Les,D. H., Mishler, B. D., et al. (1993) Ann. Mo. Bot. Gard. 80,528-580.

18. Zurawski, G., Whitfeld, P. R. & Bottomley, W. (1986) NucleicAcids Res. 14, 3975.

19. Aldrich, J., Cherney, B., Merlin, E. & Palmer, J. (1986) NucleicAcids Res. 14, 9535.

20. Swensen, S. M., Mullin, B. C. & Chase, M. W. (1994) Syst. Bot.19, 157-168.

21. Bousquet, J., Strauss, S. J. & Li, P. (1992) Mol. Biol. Evol. 9,1076-1088.

22. Morgan, D. R., Soltis, D. E. & Robertson, K. R. (1994) Am. J.Bot. 81, 890-903.

23. Morgan, D. R. & Soltis, D. E. (1993) Ann. Mo. Bot. Gard. 80,631-660.

24. Swofford, D. L. (1993) PAUP: Phylogenetic Analysis Using Parsi-mony (Ill. Nat. Hist. Survey, Champaign, IL), Version 3.1.1.

25. Felsenstein, J. (1985) Evolution 39, 783-791.26. Bremer, K. (1988) Evolution 42, 795-803.27. Donoghue, M. J., Olmstead, R. G., Smith, J. F. & Palmer, J. D.

(1992) Ann. Mo. Bot. Gard. 79, 333-345.28. Maddison, D. R. (1991) Syst. Zool. 40, 315-328.29. Becking, J.-H. (1992) in Biological Nitrogen Fixation, eds. Stacey,

G., Evans, H. J. & Burris, R. H. (Chapman & Hall, New York),pp. 497-559.

30. Caetano-Anoll6s, G., Joshi, P. & Gresshoff, P. M. (1992) in PlantBiotechnology and Development, ed. Gresshoff, P. M. (CRC, BocaRaton, FL), pp. 61-70.

31. Fisher, R. F. & Long, S. R. (1992) Nature (London) 357,655-670.32. Young, J. P. W. & Johnston, A. W. B. (1989) Trends Ecol. Evol.

4, 341-349.33. Denarie, J., Debelle, F. & Rosenberg, C. (1992) Annu. Rev.

Microbiol. 46, 497-631.34. VandenBosch, K. A., Bradley, D. J. & Knox, J. P. (1989) EMBO

J. 8, 335-342.35. Peters, G. A., Frost, J. W. & Long, S. R. (1986) Science 223,

977-980.36. Scheres, B., Van de Wiel, C., Zalensky, A., Horvath, B., Spaink,

H., Van Eck, H., Zwarthruis, F., Wolters, A. M., Gloudemans, T.,Van Kammen, A. & Bisseling, T. (1990) Cell 60, 281-294.

37. Landsmann, J., Dennis, E. S., Higgins, T. J. V., Appleby, C. A.,Kortt, A. A. & Peacock, W. J. (1986) Nature (London) 324,166-168.

38. Sanchez, F., Padilla, J. E., Perez, H. & Lara, M. (1991)Annu. Rev.Plant Physiol. Plant Mol. Biol. 42, 507-528.

39. Sprent, J. I. (1989) New Phytol. 111, 129-153.40. Duc, G., Trouvelot, A., Giannazzi-Pearson, V. & Giannazzi, S.

(1989) Plant Sci. 60, 215-222.41. Al-Mallah, M. K., Davey, M. R. & Cocking, E. C. (1989) J. Exp.

Bot. 40, 473-477.

Evolution: Soltis et al