DNA and morphology: Inference of plant phylogeny

TREE vol. 5, no. 4, April 1990

helpful in identifying critical factors for population management. Tony Starfield (University of Minneapolis) described another population simulation model and showed how a different approach to modelling is required when many parameters and their ranges are unknown. Unlike engineering applications, models should not be used to seek optimal solutions, but instead to identify and minimize risks. A policy of ‘minimum regret’ may be a good one for conservation biology.

A second theme was ‘altered states’, a term introduced by William Conway (New York Zoological So- ciety) in his opening remarks. The world is changing rapidly, especially for wild species in tropical areas, and the conservation community will soon face a variety of new problems arising from the need to manage many species, even those in reserves

DNA and Morphology: Plant Phylogeny Kenneth J. Sytsma

In recent gears, there has been an increas- ing number of reports of discrepancies between analyses of DNA and morphology in the estimation of phylogeny. In plants, the discrepancies can be attributed to procedural problems (apparent discrepancies) or to biological attributes of the organisms (real discrepancies). The problems can arise from within both morphological and molecular aspects of the study. A better understanding of both hinds of problem permits a more thorough synthesis of DNA and morphology in the inference of plant phylogeny, and can result in the further refinement or resolution of a morphologically based phylogeny by molecular evidence, and of a molecularly based phylogeny 6y morphological evidence.

In the past decade, DNA-based inference of plant phylogeny has developed into a multi-faceted systematic endeavor. Although this approach is recent, it has already had a significant impact on many levels of plant systematics’-3. The great interest in, and major movement toward,

104

Kenneth Sytsma isat the Botany Dept, University of Wisconsin, Madison, WI 53706, USA.

and protected areas. Long-term sur- vival of many species will depend upon the development of principles very different from those presently adopted in wildlife management - principles that may conflict with our more traditional concepts of wildness and freedom. As intensive population management techniques start to incorporate both captive and re- introduced populations, the distinc- tion between captive population management and management in reserves and protected areas breaks down. There are aesthetic and moral issues to be resolved in this area that are even more demanding than the biological ones.

Acknowledgements The meeting was sponsored by the

National Zoo, American Association of Zo- ological Parks and Aquariums, Institute of Museum Studies and Chicago Zoological

Society. GMM was also supported by a grant from the Wellcome Trust.

References 1 Gilpin, M.E. and Soul& M.E. (1986) in Conservation Biology: The Science of Scarcity and Diversity (Soul& M.E., ed.), pp. 19-34, Sinauer 2 Soul& M.E. (1987) Viable fopuiations for Conservation, Cambridge University Press 3 Rails. K. and Ballou, J.D. (1986) Zoo Biol. i,‘81-238 4 Kleiman, D.G. et a/. (1986) in Pimates: The Road to Self-sustaining Populations (Benirschke, K., ed.), pp. 959-979, Springer-Verlag 5 Avise. J.C. (1989) Trends Ecol. Eva/. 4, 279-28; 6 Lynch, M. (1988) Mol. Bioi. Evol. 5, 584-599 7 Haig, S., Ballou, J.D. and Derrickson, S. Conserv. Biol. (in press) 8 White, T.J., Arnheim, N. and French, H.A. (1989) Trends Genet. 5, 185-189 9 Lande, R. (1988) Science 247, 1455-1460

Inference of

the use of DNA in phylogenetic studies of plants raises the question, ‘Why is molecular evidence so appealing in plants?’ The appeal comes from several beliefs, justified or not, concerning DNA and morphology: I I) that molecular analysis provides numerous and independent characters whereas morphological analysis provides fewer characters, often of questionable homology; (21 that morphology, unlike most DNA regions, is prone to considerable convergence; (31 that the genetic basis of convergence in molecules is better known; and 14) that molecular analysis is free of the subjective kinds of character analysis and outgroup selection that might mislead morphological analysis. As Patterson4 aptly queried, ‘have molecules superseded morphology as guides to the history of life, or are the approaches sides of the same coin, with same problems and limitations? Do molecules and morphology give the same picture of the history of life, or two more or less distorted views of the same picture, or two quite different pic- tures?’ These questions have been

recently examined for animal+j, but the situation in plants is unresolved.

The primary goal in phylogenetic studies is to infer the single histori- cal genealogy - i.e. the true phylogeny - of a group of organisms. It might thus be expected ( II that systematic studies of any set of geneti- cally determined characters within a group of organisms should be congruent with other such studies based on different suites of characters, (2) that congruency provides strong evidence that the true phylogeny has been inferred, and 131 that conflict among the results may indicate low resolution power of the data sets, invalid and inappropriate


Jnderiying assumptions, or pro- ,:edural problems in some or all of :he analyses5. Although these ex- oectations often hold, the situation ‘or plants appears to be even more ,:omplex.

Most studies of plant groups that 7ave been examined by both DNA <snd traditional approaches indicate l:lose congruence in many of the -esulting phylogenies. This con- ;;ruence is reassuring and indicates ,:hat both approaches are important n inferring plant phylogeny. In this

I-eview, however, I examine DNA- based inferences of plant phy- iogeny that apparently conflict with morphologically based inferences. These examples can be placed in !;even categories, each character- tzed by features generating the apparent or real lack of congruency between the DNA and morphological data sets; the seven categories tall loosely into two groups (Box I).

The first group consists of categories that deal with procedural problems in data analysis or in- terpretation (i.e. the data might in lact provide the true phylogeny if correctly analysed or interpreted). The second group consists of categories in which inherent biological or evolutionary features of the plants, either in DNA or morphology, yield the discrepancies (i.e. the available data are such that I he true phylogeny is not possible IO obtain without other knowledge). These categories and the groups tvithin which they fall are admittedly artificial, in that many examples of morphological and molecular incon- rristency result from combinations of both these problems.

Equating overall morphological similarity with phylogenetic relationships

Reliance on overall morphological similarity (the phenetic approach) as an estimate of phylogenetic relationships, rather than on shared derived character states I the ciadistic approach I, can gener- ;:lte misleading estimates of phylogenetic relationships (Box 2). The phenetic approach fusually using clustering or principal components inalyses) often explicitly assumes the operation of some type of clock- like rate of change in morphology, whereas the cladistic approach does not. A more subtle aspect of this problem is that much of the extant taxonomic and evolutionary

literature is viewed as reflecting cladistic relationships, but may in fact be based on the traditional ‘phenetic/evolutionary’ approach6, with varying degrees of reliance upon morphological divergence as the main criterion for establishing relationships.

Although molecular data sets can be, and often are, treated in a phenetic manner and are thus open to the same criticism, the phenetic and cladistic approaches when used on the same molecular data sets have often shown remarkable con- gruence7. Additionally, overall molecular divergence data are often analysed with ‘phenetic’ algorithms (the Fitch and Margoliash algorithm for example) that do not assume

equal rates of divergence and thus are not strictly phenetic. One spec- tacular example will serve to illus- trate the problem of relying on the traditional ‘pheneticlevolutionary’ approach rather than on a cladistic approach, although many other studies undoubtedly will fall into this category as rigorous cladistic techniques are increasingly applied to both morphological and DNA data sets.

During the past 30 years, six different schemes of phylogenetic relationships have been proposed for the Asteraceae (Compositae), with no consensus regarding the number and relationships of monophyletic tribes’,s. One of the reasons suggestedP,8 as contributing to the

105


a

Mutisieae - Gochnalimae

Cardueae I I I

Mutlsieae - Muf~smae Mutisieae - Mut~s~mae SubfamIly Mutweae - Nassauvimae

Mut~seae - Nassauvimae Clchorioldeae A Cardueae

l”YWSlO” Vemonieae

j

VWllCXW.%3

Liabeae Llabeae

Cichoreae

sublamily Clchorioldeae

all tribes of subfamily Asteroideae b I

I------ “chor’eae I I all tribes of subfamily Aslero~deae

Fig. I. Cladograms of selected tribes and subtribes within the Asteraceae: lal based on morphological characters and one molecular character’ (redrawn with permission oftheauthorl; and fbl based on restriction-site mutations and one inversion in the chloroplast genome’ K (redrawn with permission ofthe authbrsl. The shaded bars in la) represent unresolved nodes

confusion was the lack of application of cladistic principles to morphological characters for phylogenetic reconstruction within the family, at least until recently. A recent morphology-based cladogram for the Asteraceae9 (Fig. la) is start- lingly similar to a cladogram based on chloroplast DNA (c~DNA)‘,~ (Fig. lb). The cpDNA evidence is based on both comparative restriction-site analysis and the distribution of a major inversion found only in the Asteraceae, excluding the subtribe Mutisieae-Barnadesiinae. The morphological cladistic analysis unfor- tunately also uses the presence or absence of this inversion, thus mak- ing the two approaches not com- pletely independent. Nevertheless, the DNA and morphological cladograms have produced some important complementary results: (I) the small and monophyletic subtribe Mutisieae-Barnadesiinae is placed as the most basal lineage in the

family, (2) the tribe Mutisieae and subfamily Cichorioideae are paraphyletic, and (3) the subfamily As- teroideae is monophyletic.

The basal status within the As- teraceae of a number of different tribes has been suggested over the years, but these suggestions lacked rigorous outgroup analysis of all characters. The application of strict cladistic analysis on these morphological characters, rather than using ad hoc assumptions about primitive and advanced character states, allowed for a more objective approach. The close congruence of the DNA and morphological cladograms suggests that a better estimate of the true phylogeny for the Aster- aceae is now becoming available. These studies indicate that the fail- ure of applying cladistic methods to morphological characters can be an important reason for the apparent discrepancy of morphology and DNA. This example underscores the

SpSCiSS subsection

C. dudleyana 1 I

+- C. modesta

C. lingulata

Lautiflorae

Clarkia epilobioides 1 Micranthae

C. fostfata

C. lewisii

C. cylindrica I

Sympherica

Ueterogaura heterandra

C. biloba

Fig. 2. Cladogram of Clarkia and Heferogaura IOnagraceael, based on cpDNA restriction-site analysis and gene duplications and subsequent loss of activity of 6Pgdh. Restriction-site changesare indicated by solid vertical bars. The two gene duplications (+ 1 and subsequent losses (-I are indicated by open bars fC. cytoplasmic; P, plastid17.“**

need for more detailed cladistic analysis of groups with both morphological and molecular characters.

Equating crossing relationships with phylogenetic relationships

The ability of two plant species to cross as a function of their genetic relationship has long been accepted in the biosystematic literature. It has been argued, and demonstrated, however, that the lack of ability of two species to cross does not necessarily reflect lesser genetic relationshipto. The ability to cross can in fact be considered a shared primitive character at some taxonomic level, and a barrier to crossing can be considered, in some instances, to be a shared derived character between two closely related species”. Equating crossing relationships with phylogenetic relationships can thus potentially yield misleading estimates of the true phylogeny. Two examples, in Clarkia (Onagraceae) and in Eleusine (Poaceae), demonstrate this kind of inconsistency between traditional and molecular data sets.

Clarkia rostrata has been placed within the subsection Sympherica, based on both the overwhelming phenetic similarity to, and the ability to produce fertile hybrids with, the other two species of the subsection (C. lewisii and C. cylindrical (see Refs 7 and 12). Both isozyme gene duplication dataI and cpDNA restriction-site analysis7 ‘2 indicate, however, that C. rostrata is in fact more closely related to C. epilobioides of subsection Micranthae than it is to the other two species with which it can cross (Fig. 21. The ability to cross within the paraphyletic subsection Sympherica is thus best considered as a shared primitive state, and the lack of crossability between the closely related C. rostrata and C. epilobioides


I Chlamydomonas

is best considered as a derived state.

incongruities between very de- :ailed cytological/crossing studies ,and DNA results have been demonstrated in domesticated and wild :etraploid finger millets (Eleusine ‘zoracana subspp. coracana and kicana) and a putative diploid orogenitor (E. Mica)‘“. Eleusine Mica was first considered to be one of the diploid progenitors for the ~:etraploid finger millets on the basis of morphology and biogeography. It was later concluded, however, that C. indica could not have contributed any of the genomes of finger millet, due to the almost complete lack of chromosome pairing in synthetic hybrids between subspp. coracana or africana and E. indica. The c:pDNAs of the three, however, are identical and differ substantially horn other diploid species, suggesting that E. indica served as the maternal progenitor for the tetraploidsr4. Chromosome pairing behavior was argued as not accu- rately reflecting phylogenetic af- finities in these species.

inappropriate molecule(s) for taxonomic level or question addressed

The choice of molecule or DNA sequence to be examined is an important first step in any evolutionary Ior systematic study. A priori knowledge of rates of change, amounts of variation at different taxonomic levels, considerations of homology, cmd degree of expected homoplasy, is essential before and after data tire collected. As more and more studies involving a given molecule or sequence are done, the value ;Ind limitations of each molecule or c#equence at different taxonomic levels become more evident. Thus, for example, cpDNA restriction-site analysis in angiosperms has been occasionally useful at the species level, extremely useful at the gen- eric and familial level, but seldom used at the ordinal level or above’. Similarly, analysis of restriction-site 2nd fragment-length variation in the non-coding spacer of nuclear ribosomal DNA (rDNA) is possible within and among populations and among congeneric species’5,‘6, but it is analysis in the coding portions of rDNA that is useful only at higher l~2vels2,‘7.

Perhaps more of a problem in es-

timations of the true phylogeny is that certain portions of a given molecule or nucleic acid sequence evolve at quite different rates. One portion of the sequence will exhibit too few shared mutations between species, while other portions will have multiple substitutions and thus show ‘evolutionary noise’ between species. The use of 5s rRNA sequence data to reconstruct relationships in green plants demon- strates these problems.

Phenetic analysis using an evolutionary distance modified for tran- sition and transversion biases resulted in the phenogram presented in Fig. 3a (Ref. 18). This phenogram has some unusual relationships relative to currently accepted ideas based on traditional kinds of data. The two most unusual include the placement of the Bryophyta (mosses, hornworts and liverworts) within the lower vascular plants rather than basal to all vascular plants, and wheat with tomato rather than with duckweed, the other monocot.

Cladistic analysis of the same data set generated numerous equally parsimonious cladograms, one of which is presented in Fig. 3b (Ref. 19). This cladogram shows unusual placements of the Bryophyta (now as the sister group to the gymnosperms and flowering plants), wheat, as well as the green alga Chlamydomonas. The latter oc- cupies the most basal position in the phenogram, but is nested within the angiosperms in the cladogram. The wide departure of all most parsimonious cladograms from expected relationships and from the phenogram was attributed in part to a high level of homoplasy (43%) - ‘evolutionary noise’ - in some portions of the 5s rRNA data set.

The separation of ‘noisy’ characters from those providing meaningful estimates of the true phylogeny is no easy task. It requires a considerable amount of a priori knowledge about the structure and evolution of 5s rRNA, weighting of characters or character states, elim- ination of hypervariable sites, and/ or some amount of (necessary?) circular reasoning involving secondary analysis of the data by a priori defining of already accepted groups.

It is noteworthy that phenetic cluster analysis resulted in less dis-

Stoneworf Liverwort Liverwort MOSS Homwofl Whiskfern Clubmoea Horaetail Fern CY=d Ginkgo Dawn redwood Duckweed

L Wheat Tomato Sunflower Spinach FlaX

7 Lupine

I I - Homwort ‘1,

41 Dawn redwood Ginkgo Cycad Spinach Dwarf bean Broad bean Lupine Sunflower

I ,-’ Flax

b Wheat Tomato Duckweed Chlamydomonas

Fig. 3. la) Phenogram of green plants based on 5s rRNAs” lredrawn with permission of the authors). (b) Cladogram of green plants based on 5s rRNAs” (redrawn with permission of the authorsl.

torted relationships fin reference to generally accepted relationships) than did cladistic analysis of the same data. Perhaps at great evolutionary divergences (e.g. all green plants) and with data sets exhibiting considerable levels of convergence (e.g. 5s rRNA), distances rather than cladistic procedures provide better estimates of true relationships. In simulation studies involving amino acid sequence data at great evolutionary distances, a genetic distance analysis that did not assume a molecular clock resulted in better phylogenetic approximations of the known true phylogeny than did the more ‘cladistic’ ancestral sequence method20. It has been demonstrated that when large differences in rates of evolutionary change exist between characters, cladistic methods

107


Hetevogaura heterandra

I ?.mm

Fig. 4. Flowers and fruits of Clarkia dudleyana and Heterogaura heterandra of the Onagraceae: A, inflorescence: B, fruit; C, style; D. stamen: E. staminode.

that attempt to minimize the total number of character state changes (parsimony) do not work as well as other methods such as character compatibility or likelihood estimation*‘.

Unequal rates of morphological evolution As described earlier, reliance on

phenetic similarity rather than on shared derived character states, whether morphological or molecular, can give rise to instances of apparent conflict between the two types of data set. However, in cases where wholesale changes have occurred rapidly in one lineage, it is possible that both phenetic and cladistic analyses of morphological data would have difficulty in aligning this lineage correctly with sister lineages. A number of recent molecular studies have demonstrated such discrepancies.

One dramatic increase in rate of morphological divergence giving rise to a discrepancy between traditional and molecular techniques is seen in Heterogaura (Ona- graceael. The monotypic genus is closely related to Clarkia based on floral morphology, stigma sur- face, seed coat structure, anther anatomy and flavonoids, but differs markedly from Clarkia in having only four fertile anthers (four are sterile), an unlobed stigma, and a round nut- like indehiscent fruit with one or two seeds (Fig. 4). The floral and fruit differences between the two genera are so distinctive that they have

been maintained as separate genera since 1866. Chloroplast and nuclear rDNA analyses, however, indicated that Heterogaura is actually derived from within Clarkia subsect. Lautiflorae, with C. dudleyana as its sister species (Fig. 217,22. The extreme floral and fruit reduction seen in Heterogaura relative to Clarkia masked the close phylogenetic relationship of Heterogaura to an advanced subsection within Clarkia.

Cladistic analysis of morphological characters would have difficulty in aligning Heterogaura properly not only because of its lack or great modification of some characters important in defining monophyletic groups within Clarkia (e.g. capsular and anther characters), but also because there is no a priori reason not to accept Heterogaura as the sister group for all of Clarkia. Synapomorphic (shared derived) morphological characters uniting Heterogaura with these Clarkia species might exist, but they will be found a posteriori and as a direct result of the compelling isozyme and cpDNA data. To date, however, no morphological character has yet been identified that unites Hetero- gaura solely with Clarkia subsect. Lautifiorae, let alone C. dudleyana.

Rapid morphological change in other monotypic genera that has obscured close phylogenetic relationships to species within more species-rich genera has also been documented in the Saxifragaceae2’.

The genera Bensoniella and Coni- mite/la, like Heterogaura, have un- dergone unusually rapid floral and fruit evolution relative to their closest relatives in the progenitor genus Mite/la. Evidence from cpDNA has also indicated similar kinds of rapid but asymmetric rates of morphological evolution within monophyletic lineages of the Onci- diinae (Orchidaceae), again demon- strating a surprising degree of plasticity in floral morphology24. The ‘mule-ear’ succulent vegetative con- dition seen in the genus Tricho- centrum and one series within the genus Oncidium has been viewed as arising by parallelism, because the two groups are radically different in floral morphology. Tricho- centrum, however, has a chloroplast genome that is closest to this ‘mule- ear’ series within the larger On- cidium. Thus, it is the vegetative features previously considered to be convergent that provide the meaningful phylogenetic relationships in the group, and not the ‘conservative’ floral features previously emphasized in cladistic studies of orchids.

Convergence of morphological and molecular characters

Rapid and plastic changes in (supposedly! 1 conservative (and thus taxonomically important) floral or fruit characters is only one inherent problem in estimating phylogenetic relationships with morphological characters. Morphological convergence or parallelism is another problem that can result in misleading phylogenetic relationships. Two recent studies, in the subtribe Oncidiinae (Orchidaceael and in Polystichum (Dryopteridaceae), ex- emplify this problem.

A cpDNA phylogeny of the subtribe Oncidiinae based on comparative cpDNA restriction-site mapping reveals that strong regional evolution in this neotropical group of orchids has produced quite similar floral morphologies in parallel among relatively distantly related species’,25. For example, several sections of Oncidium in Brazil are phylogenetically more closely related, based on DNA, to other genera in the same region than they are to other sections of Oncidium with similar floral morphologies, both gross and detailed, but occurring in


other geographical areas. The similarity of floral morphologies seen in unrelated sections of Oncidium is a striking case of floral parallelism that had previously misled taxono- mists into assuming that these sections, although geographically widespread, were closely related on the basis of floral synapomorphies. A cpDNA phylogeny for PO/Y- stichum and related fern genera indicates that frond venation, tra- ditionally important in these genera, is fraught with convergences, but that rhizome characters provide results congruent with c~DNA~~.

Although convergence in molecular characters occurs, in some instances to a considerable extent (as noted above in 5s sequence data for all plants), the molecular data sets are more amenable to dealing with convergences than are their morphological counterparts. This difference is due to our knowledge of the genetic basis for convergences in molecules as opposed to in morphology. For example, the relative probabilities of different types of restriction-site convergence are well known2’, allowing for further discrimination among equally short cladograms12. Also, the bias towards transitions rather than transversions in most DNA sequence data is documented28, permitting additional analysis on sequence data with high levels of convergences*.

Hybridization and/or introgression A number of cases of discrep-

ancies between morphology and DNA have been attributed to hybridization and/or introgression between two species. These gene movements can lead to situations in which maternally inherited genomes (chloroplast and mitochon- drion in most cases) of one species are more closely related to those of a distinct species than to those of other conspecifics. Thus, contra- dictory cpDNA and nuclear phylogenies obtained in two accessions of Brassica napus led to the hypoth- esis that introgressive hybridization has been involved in their recent evolution29. In this case the second species providing the cytoplasm including the chloroplast genomel

$:ould not be identified. Likewise, t>ne of I 3 examined cultivated lines ‘3f the garden pea Pisum sativum

did not exhibit the chloroplast gen- Polyploidy ome of northern populations of P. Polyploids can show several humile, the putative wild progeni- kinds of discordance between mor- tor30. Rather, this one accession had phology and molecular results. One a chloroplast genome similar to P. kind of discrepancy arises when elatius and southern populations of reciprocal crosses between two P. humile, indicating that secondary diploid species generate the hybridization may have occurred in allopolyploid. Aegilops triuncialis the domestication of the garden thus appears diphyletic with two pea. quite different maternal chloroplast

In a recent cpDNA and rDNA genomes, due to independent and phylogenetic analysis of Populus reciprocal crosses between the pro- lSalicaceae13’, the chloroplast gen- genitor species A. caudata and A. omes of all varieties of the Euro- umbellulata, each acting as the pean black poplar (P. nigral were chloroplast donor in turn34. distinctly different from those of the A second kind of discrepancy American cottonwoods I/? deltoides arises when a suite of closely re- and P. fremontii), all placed in sec- lated diploid species can generate tion Aigeiros. However, the black the allopolyploid. The tetraploid poplar chloroplast genome was Andean Solanum tuberosum subsp. most similar to that of the European andigena complex displays sev- white poplar (P. alba) of section eral chloroplast genomes due to Populus. The nuclear genome of the crossings of numerous cultivated black poplars, however, was distinct diploid species35. A third kind of from both the cottonwoods and the discrepancy arises when allo- or white poplar, suggesting that the autopolyploids originate indepen- origin of the black poplar included dently numerous times and in dif- hybridization with the white poplar ferent places, such as in the as the maternal parent. It is note- allopolyploid moss Plagiomnium worthy that the black and white medium36 and in autopolyploid poplars can be forced to cross, but populations of Heuchera micrantha only with the latter as the maternal lSaxifragaceae137. parent.

Surprisingly, molecular data con- Conclusion tradict a ‘classic’ example of For most studies in which both introgression based on morphologi- DNA and morphology have been cal evidence - the recent origin of examined, the majority of the re- weedy races of Helianthus bolan- sults are encouragingly congruent. deri by the introgression of genes Many of these studies, however, from H. annuus into a serpentine show some instances of discrep- race of H. bolanderi. Isozyme, rDNA ancy. These discrepancies in most and cpDNA data strongly suggest cases can be attributed to certain that the weedy race of H. bolanderi problems with techniques and pro- is relatively ancient in origin and cedures, or to biological phenom- did not arise by introgression32. It ena that provide alternative should be noted that phylogenetic perceptions on how evolution has sorting of matriarchical lineages progressed. In many of the ex- during speciation, independent of amples cited here, the discovery hybridization, can in theory also of the discrepancies has actually give rise to discrepancies between initiated studies to address new species boundaries and chloroplast questions and avenues of research. or mitochondrial genealogies33. For example, the robustness with When the time since separation of which various molecular data place two species is relatively short, the the genus Heterogaura within probability is highest that some Clarka now provides a model sys- chloroplast or mitochondrial lin- tern to re-examine the rate and type eages from one species will be more of, and genetic basis for, morpho- closely related to lineages from an- logical evolution in plants. other species than they are to other Where is plant phylogenetics lineages from the same species (i.e. heading? Undoubtedly, the new the first species appears either molecular evidence is forcing a paraphyletic or polyphyleticl. No serious reassessment of relation- examples of this phenomenon are ships based on the traditional sys- yet known in plants. tematic approaches. The examples

109


reviewed here, in which problems can arise from strict dependency on either molecular or morphological characters, indicate that we should ‘cease the debate about “molecules versus morphology” and begin to cooperate on the basis of molecules and morphology.‘4 An interesting by-product of the emergence of molecular studies is their impact on morphological studies. Because of both the presence and the absence of congruence of phylogenetic results in the two kinds of study, there has been a resurgence of interest in more critical cladistic analysis of morphological characters, and a movement towards studies that in- volve both morphology and molecules. It is just these kinds of multidisciplinary study that ulti- mately will have the most impact and benefit for the issue of molecules and morphology.

The utilization of DNA approaches will become greater and even more powerful as rapid se- quencing via polymerase chain re- action mediated amplification of specific DNA segments becomes routine in phylogenetic studies38. The phylogenetic potential of im- mense molecular data sets will be realized, however, only if the theor- etical bases for phylogenetic analysis of sequence information and the development of sophisticated com- puter programs to handle the data can keep pace with the actual data collection. The utilization of DNA approaches to clarify relationships that are difficult to ascertain with morphology will also increase, especially in plant groups that exhibit remarkable divergence related to

adaptive radiation. Understanding the adaptive radiation of such groups by the phylogenetic use of the very morphological characters that are important in the radiation is tantamount to circular reasoning. Molecular cladograms of these evol- utionarily interesting and important groups will thus be overlaid with morphological characters in at- tempts to understand better the di- rections and rates of morphological evolution and ecological radiation.

A final aspect of the ‘molecule and morphology’ dilemma that will have to be addressed is how to deal with both molecular and morphological characters in the inference of plant phylogeny5. Four possibilities exist and merit examination, although in practice only the first three have been considered. (I 1 Combine the two sets of characters with equal weighting for each character. In this way the morphological characters might provide strength to branches that are weak based on molecular characters, and vice versa. A variant on this approach, but perhaps difficult to justify and implement, is to provide equal weighting for each set of characters, thus not discriminat- ing against the (usually) fewer morphological characters. (2) Analyse the two sets of characters indepen- dently and construct a consensus cladogram that depicts only relationships recognized by both approaches. (3 I Use only the molecular characters to generate a cladogram and secondarily overlay the morphological characters onto the molecular cladogram. (41 Use only the morphological characters to generate a cladogram and secondarily overlay the molecular characters onto the morphologicat cladogram.

These issues related to handling both (or many) kinds of data set in inferences of plant phylogeny must be dealt with if we are going to cooperate successfully on the basis of molecules and morphology.

References I Palmer, I.D., jansen, R.K., Michaels, H.I., Chase. M.W. and Manhart, 1.R. ((9881 Ann. MO. Bat. Card. 75, I180-1206 2 Zimmer, E.A., Hamby. R.K., Arnold, M.L., LeBlanc, D.A. and Thiriot. E.C. (19891 in The Hierarchy of Life IFernholm. B., Bremer, K. and liirnvall. H., edsl. pp. 205-2 14, Elsevier Science Publishers 3 Doyle, 1.1. II9871 Int. Organ. Plant Biosyst. News/. 8. 3-7 4 Patterson, C. I I9871 Molecules and

Morphology in Evolution: Conflict or Compromise, Cambridge University Press 5 Hillis, D.M. (1987) Annu. Rev. Ecol. Syst. 18, 23-42 6 Donoghue, M.I. and Cantino. P.D. f 19881 Bat. Rev. 54, 107-128 7 Sytsma, K.I. and Cottlieb, L.D. ( 19861 Evolution 40, 1248-1261 8 lansen, R.K. and Palmer, I.D. (19881 Am. / Bot. 75, 753-766 9 Bremer, K. II9871 C/ad&tics 3. 210-253 IO Dewet, I.M.I. and Harlan, I.R. II9721 Taxon 2 I, 67-70 II Rosen, D.E. (19791 Boll. Am. Mus. Nat. Hist. 162, 267-376 I2 Sytsma, K.I. and Smith, I.F. ( I9881 Ann. MO. Bot. Card. 75, I2 17-l 237 I3 Odrzykoski, 1.1. and Gottlieb, L.D. ( 19841 Syst. Bot. 9. 479-489 14 Hilu. K.W. 119881 Genetics 118, 163-167 I5 Sytsma, K.I. and Schaal, B.A. (19851 Evolution 39, 594-608 16 Schaal, B.A. and Learn, G.H., jr (19881 Ann. MO. Bat. Gard. 75, I207- I2 I6 I7 Appels, R. and Honeycutt, R.L. (19861 in DNA Systematics, Vol. II: Plants (Dutta, SK, ed.1, pp. 81-135. CRC Press I8 Hori, H.. Lim, B-L. and Osawa, S. (I9851 Proc. Nat/ Acad. Sci. USA 82,820-823 I9 Bremer, K., Humphries, C.1.. Mishler. B.D and Churchill, S.P. (19871 Taxon 36, 339-349 20 Schwartz, R. and Dayhoff, M. ( 19781 Science 199, 395-403 21 Felsenstein. I. t 1981 I Biol. 1. Linn. Sot 16, 183-196 22 Sytsma, K.I. and Gottlieb. L.D. ( 19861 Proc. Nat1 Acad. Sci. USA 83. 5554-5557 23 Soltis, D.E., Soltis, P.S. and Bothel, K.D Syst. Bot. lin press) 24 Chase, M.W. and Palmer, I.D. (19891 Am. /. t?ot. 76, I 720- I730 25 Chase, M.W. and Palmer, I.D. t 19881 Am. / Bot. 75, 163-l 64 26 Yatskievych. G., Stein. D.B. and Gastony. G.I. ( 19881 Proc. Nat/ Acad. Sci. USA 85, 2589-2593 27 Templeton, A.R. I 1983) in Statistical Analysis of DNA Sequence Data (Weir, B.S. ed.), pp. 151-179, Marcel Dekker 28 Lake, I.A. 11987) Mol. Bio/. Evol. 4, 167-191 29 Palmer, I.D.. Shields, C.R., Cohen, D B. and Orton, T.I. 11983) Theor. App/ Cenet 65, 181-189 30 Palmer, I.D., lorgensen. R.A. and Thompson, W.F. ( I9851 Genetics 109. 195-213 31 Smith, R.L. and Sytsma, K.I. (19881 Am. /. Bot. 75, 206 32 Rieseberg, L.H.. Soltis, D.E. and Palmer, I.D. II9881 Evolution 42, 227-238 33 Neigel, I.E. and Avise. I.C. II9861 in Evolutionary Processes and Theory I Karlin, S. and Nevo. E., edsl. pp. 5 15-534, Academic Press 34 Ogihara, Y. and Tsunewaki, K. ( 19881 Theor. Appl. Genet. 76, 32 l-332 35 Hosaka, K. and Hanneman, R.E.. jr 119881 Theor. Appl. Genet. 76. 333-340 36 Wyatt, R , Odrzykoski, Il.. Stoneburner, A., Bass, H.W. and Calau, G.A (19881 Proc Nat/ Acad. Sci. USA 85. 560 I-5604 37 Soltis, D.E., Soltis, P.S. and Ness, B.D I I9891 Evolution 43, 650-656 38 Wilson, A.C.. Zimmer. E.A, Prager, E M and Kocher, T.D. (19891 in The Hierarchy of Life Ifernholm, B , Bremer, K. and lornvall, H., edsl, pp. 407-419. Elsevier Science Publishers

DNA and morphology: Inference of plant phylogeny

Documents

Transcript of DNA and morphology: Inference of plant phylogeny