in angiosperm taxonomy

16 Pages·2012·2.08 MB·English

ActaBot. Need.39(4), December 1992,p. 369-384
Review
Nuclear DNA markers in angiosperm taxonomy
K. Bachmann
HugodeVriesLaboratory,University ofAmsterdam, Kruislaan318,NL-I098 SMAmsterdam,
TheNetherlands
CONTENTS
Chloroplast vs. nuclearDNA 369
Approaches to thenucleargenome 370
Restriction fragmentlength polymorphisms (RFLPs) 371
RibosomalRNA genesas anexample 372
RFLPanalysis withrandomgenomicandcDNA probes 374
Mini-satellitevariation:VNTR fingerprints 375
Polymorphisms inPCR amplification products: RAPDs 376
ConclusionsandOutlook 378
Key-words: angiosperms, molecularphylogeny, RAPD, reticulate
evolution, RFLP,VNTR.
CHLOROPLAST VS. NUCLEAR DNA
New methods for the analysis ofpolymorphisms in DNA sequences have a profound
influence on systematic botany. Several recent books dealwith molecular methods in
evolutionandtaxonomy(Clegg &O'Brien 1990;Doolittle1990;Hillis& Moritz 1990;Li
& Graur 1991),someofthemspecificallywithplants(Crawford 1990;Soltisetal. 1992).A
general revisionoftheevolutionary relationships ofplants withmolecularmethodsseems
tobewellunderway(Palmeretal. 1988).Chloroplast DNA(cpDNA) hasturnedouttobe
themostconvenientcommonmoleculeforthe reconstructionofplant phylogeny. Inthis
review, the use ofmarkersin the nuclear genomewill be examined.Obtaining reliable
phylogenetic information from nuclear DNA tends to be considerably less straight-
forward than working wth cpDNA. This pertains both to the collection and to the
interpretation ofthe data. However, the featuresofcpDNA that make it such a useful
molecule for phylogenetic analysis also limitits applicability. A brief and simplified
comparison ofthese featureswith thoseofnuclearDNA can help point out whereand
why nuclearmarkersbecomeimportant.
ChloroplastscontainclosedcircularDNA moleculesabout 150000 basepairs(150kb)
inlength.Thesmallestnucleargenomesinangiosperms areabout500-foldbigger,andthe
variationinhaploid nucleargenomesizes(‘C values’)amongtheangiosperms isconsider-
ablefromArabidopsis thaliana(Brassicaceae) withabout 105kb in thefivechromosomes
ofthehaploid genometoFritillariaassyriaca (Liliaceae) with1-23x 108kb in 12chromo-
somes (Bennett et al. 1982; Price 1988). There can be considerable differencesin the
genomesizes ofclosely relatedcongeneric species and thereis anincreasing numberof
reports on intraspecific variationin genomesize (Price 1991). A variationin C values
from 1-84pg to 7-14pg (1 pg=9-5 x 105kb) among populations of Collinsia verna
369
370 K. BACHMANN
(Scrophulariaceae) seems to be thewidest intraspecific range on record(Greenlee et al.
1984).Therangeofabout2:1betweenthe largest andthesmallestcpDNA inlandplants
(Palmer 1985)issmallincomparison andeasyto explain. Theavailability ofthecomplete
sequences of several cpDNAs facilitates considerably the evolutionary comparison
(Ohyamacta/. 1986;Shinozakietal. 1986).
Allchloroplast genomescontainmoreorlessthesamesetofgenes.Eventheorderofthe
genes on the chloroplast map changes relatively slowly. A segmentcontaining, among
others, thegenes forthe ribosomalRNAs typically ispresenttwice in thecpDNA inthe
formofaninvertedrepeat.Thiscp-rDNAshouldnotbeconfusedwiththenuclearrDNA,
which I shall discuss below, or withmitochondrialrDNA, which up to now has played
only aminorroleinplant evolutionary studies(Grabau 1985).Thetwo repeatcopies in
the cpDNA separate a large single-copy region from a small single-copy region. The
problem ofmultiple copies ofidenticalor similar sequencesthereforeplays virtually no
role in cpDNA evolution while it is ever present in comparisons of nuclear genomes
(Doyle 1991, 1992).DifferencesamongcpDNAs are mostlybasesubstitutionsatstrictly
homologous (orthologous) sites. Major changes in genome architecture such as inser-
tions, deletions, inversionsand transpositions involvingseveral genes,arerelatively rare.
Whentheyoccur,they canbecrucialmarkersfor majormonophyletic groups(Downie &
Palmer 1992).
Typically, the inheritanceofchloroplasts is uniparental. TheevolutionofcpDNA is
essentially thatofasexual haploid clones. Eachmutationinitiatesa monophyletic line,
andcladisticanalysis isstrictly applicable atalltaxonomiclevels. Acladogram basedon
cpDNA should parallel that ofthe plants containing the chloroplasts. Usually it does
(e.g. Crowhurst et al. 1990). However, recombination in the nuclear genomes within
species and reticulateevolution (hybridization) between species are essential aspects
oforganismal evolution, and discrepancies between thephylogenies ofchloroplast and
nuclear DNAs can reveal crucial evolutionary events. Moreover, chloroplast DNA is
essentially a neutral independent marker of phylogeny, while the vast majority of
evolutionary change relevant to the living plant is encoded in the nuclear DNA. Ifwe
want to go beyond phylogenetic reconstruction and analyse the interactions between
organismic andmolecularevolution,wehaveto study thenucleargenome.
One of the advantages of cpDNA for phylogenetic analysis is the relatively slow
sequenceevolution. Thetaxonomiclevel at which cpDNA sequencesare optimally in-
formativeis the differentiationof genera. With an increasing amount ofnoise (homo-
plasy), cpDNA can be compared at higher taxonomic levels (Palmer 1985). At lower
taxonomic levels, especially for studies on population genetics andspeciation, cpDNA
usually provides littleinformativevariation.Here, ofcourse, recombinationandreticu-
lateevolution play an essential role, and nuclear DNA markers reveal their fullvalue
(Schaal et al. 1991). However, there is informativesequence variation for all levels
of phylogenetic analysis in the nuclear genome, and we may state that, aside from
fossils documenting extinctlineages, an event inphylogenetic history that has not left a
recognizable trace inthe nuclearDNA is notdocumentedanywhere.
APPROACHES TO THE NUCLEAR GENOME
Therecan be considerablevariationin genomesizes amongclosely related species, and
thisshowsthattheevolutionofnucleargenomesislessconstrainedandfasterthanthatof
theorganisms harbouring thesegenomes.Theselectiveamplification ofsmallparts ofthe
DNA MARKERS AND ANGIOSPERMTAXONOMY 371
genomecan create large fractions of repetitive DNA that may be specific for young
evolutionary lineages (McIntyre etal. 1988).Asignificant fractionofthe DNA foundin
onespecies maybeabsentinaclosely relatedone(Ganal & Hemleben 1986b; Schmidtet
al. 1991; Unfriedet al. 1991). SuchDNA sequencescannot playa roleincoding forthe
phenotype, eventhough theirbulkmaywell havea phenotypic effect viacell-cycle time
and cell size (Bennett 1985; Price 1991). Thecoding sequencesshared by all plants are
embeddedinavastand variable majorityofnon-coding DNA,much ofitconsisting of
very many repeats ofa few basic sequences. Efforts to derive oneoverall measure ofa
genetic similarityamong organisms fromthecommoncoding sequencesin nuclearDNA
become technically difficultwhengenomeswithlargeandvariableamounts ofDNAare
compared. The most general method, DNA-DNA hybridization (Bendich & Bolton
1967),is nowonly ofhistoricalinterestinangiosperm taxonomy.
The nearest we can come to an overall measure of genetic relatedness is a random
sampling of variation atmany strictly homologous sites all across the genomeas it is
attempted in iso-enzyme studies (Gottlieb 1977). A typical problem with iso-enzyme
studies is the limited numberof loci that are available for analysis. A more extensive
randomsampling ofgenetic variationallacross the genomecanbeachieved withDNA
fingerprinting, RFLP and RAPD methods. I shall discuss these methods and their
limitationsbelow.
Alternatively, ofcourse, thereare manygenes commonto all plant genomes,and the
nucleotidesequenceofeachonecanbe usedforphylogenetic reconstruction. Thisyields
the phylogeny ofthat onegene, and (as in the caseofcpDNA phylogenies) it will be
necessary to check ifthis reflects the phylogeny ofthe organisms (Pamilo & Nei 1988;
Doyle 1992). The key concept here is that of concordance(Avise & Ball 1990) of gene
genealogies (Hudson 1990). A general consensus seems toemerge that the best possible
molecularapproach to angiosperm phylogeny is the precise sequence comparison of
severalgenetically unlinkedandfunctionally diversegenes,wherethechloroplast genome
may be consideredone linkage group(Martin & Dowd 1991). There is, however, no
consensus yet about the nuclear genes that should be includedin a general taxonomic
survey.Theeffortinvolvedincharacterizing oneplantthisway excludesthisapproach for
thetimebeing frompopulation sampling. Implicitly thiscreatesa typological approach
(i.e., the individualsequence from one specimen represents a species, genus or even
family).This seems to bean acceptable practice forthephylogeny ofhigher categories,
whereintraspecific sequencevariationhastobetreatedas noiseunlessitcanberecorded
statistically as allele frequencies. The continuous improvement in the methods for
isolating strictly homologous genes from genomesand for determining their sequences
willsoonmakephylogenetic analysis viagenetrees generally available.
Othermethodswillhavetobeusedatthespecies level.Here,arepresentative sampling
ofindividuals and of variation throughout the genome requires simpler methods for
determining polymorphisms. Theavailability ofallkindsofdirectandindirectmethods
forthe comparison ofDNA sequencevariationmakes it possible to choose the proper
combinationoftechniques foreachprojectandeachorganism.Inthefollowingsections, I
shalltry to illustratesometypical approaches.
RESTRICTION FRAGMENT LENGTH POLYMORPHISMS (RFLPs)
Manybacterialrestrictionendonucleasesarenowcommerciallyavailable.Eachrecognizes
ashort nucleotidesequence, its specific restriction site, atwhich itcuts double-stranded
372 K. BACHMANN
DNA. Restriction sites 4 or6 bp long are typical. Such sites occur statistically in any
longer stretch of DNA unless it consists ofrepetitive simple sequences. A restriction
enzyme thereforewilldigest DNA so that homologous sequencesare cutinto identical
restrictionfragments. A nucleotidesubstitutionin a restriction site will prevent recog-
nitionandcutting, andwill replace tworestrictionfragmentswithoneoftheircombined
length. It will create a restriction fragment length polymorphism (RFLP). Restriction
fragments can besortedaccording to length by gelelectrophoresis. All fragments ofone
length then form a band in the gel. Banding patterns on gels based on the restriction
fragments fromapiece ofDNA areroutinely used foridentificationandcomparison of
DNA sequences.Ifapiece ofDNA is digested withseveral differentrestriction enzymes
singly and together, therelative positions of the various cutting sites canbe determined
andtheir distancesin basepairs canbe shownona physical mapofthe DNA (Fig. 4of
Lanaudetal. 1992).Sucha physical mapis an extractofthe actual nucleotidesequence
andcanbepredicted ifthesequenceisknown.
Restriction fragments ofidentical length need not be homologous. The morebands
thereareinapattern,thelesseasywillitbetodeterminetheirhomology.Thisisextremein
the digest ofanentirenucleargenomecontaining thousandsoffragments whichforma
smearallalongthegel. Inthiscase,theDNA fragments canbeblottedfromthegelontoa
membranefilter(Southern 1975) andindividualbands homologous to a(radioactively)
labelledclonedDNA-probecanbedetectedselectively by DNA/DNA hybridization. The
probe may be cloned from the species itself (a ‘homologous’ probe) or from another
species witha sufficiently similarsequencesothatthe probe findsits homologue by base
pairing(a ‘heterologous’ probe).
Forthechloroplast genomesofseveralspecies fromdiverseangiosperm familiesthere
existcomplete setsofprobes covering the entiremolecule(e.g. the dicot, lettuce: Jansen
& Palmer 1987; and the monocot, Oncidium excavatum: Chase & Palmer 1989). The
conservativeevolution ofcpDNA facilitatesthe cross-hybridization with heterologous
probes, and phylogenetic comparisons ofcpDNA can be based on complete physical
maps. Much ofthe variationamongthe mapsfromrelated species is dueto thegain or
loss of a restriction site due to a point mutation.Such mutationscan be arranged as
character-statechanges inastrictcladistic analysis. Inthisway, RFLPsare astatistically
characterizedrandom sample (based on the random occurrence ofrestriction sites) of
sequencevariationallacross thecpDNA. RFLPscanalsoarisebytheinsertion, deletion,
or inversionofDNA. Suchevents canbe recognized and differentiatedfromrestriction
sitemutationsincpDNA analyses. RFLPanalysis ofnuclearDNA is neveras complete
andusually notas precise. Itinvolveseitherthedetailedcomparison ofa gene(essentially
asimplified sequencecomparison) orthe searchforRFLPsatarandomsetoflociacross
thegenome.Thefirstis exemplified by RFLPanalysis oftheribosomalRNA genes.The
latterdepends ontheavailability ofasetofappropriate clonedprobes for variousregions
ofthegenome.
RIBOSOMAL RNA GENES AS AN EXAMPLE
The nuclear genes coding for ribosomal RNAs (rDNA) occur hundredsof times as
tandem(one afterthe other) repeats atthe nucleolarorganizer regions on one or more
chromosomesofa haploid set. This and the abundanceoftheir RNA products in the
cellhas facilitatedtheir isolationand they have been used for a variety of molecular
studiesincludingtheearliestattempts atmolecularphylogeny ofangiosperms (Vodkin &
DNA MARKERS AND ANGIOSPERM TAXONOMY 373
Fig. 1.Structureandterminologyofthenuclearribosomal genes(rDNA).Eachoftherepeatunits(itoi)consists
ofatranscriptionunitbetweenatranscriptioninitiation site(T1S,i)andatranscriptionterminationsite(TTS;I)
and anon-transcribed spacer (NTS).Black bars mark theregionscodingforribosomal RNAs, 18S,5.8S, and
25S.Theseareprecededbyanexternal transcribed spacer(ETS)and separatedbythetwo internal transcribed
spacers,ITS1and ITS2.Theentire regionbetweenthe endofone25Scodingsequence and thestartofan18S
codingsequence istheintergenicspacer(IGS).
Katterman1971).EachrDNA geneistranscribedintoone continuousprimary transcript
(precursor rRNA), fromwhich theribosomal18S, 5-8S,and 25S RNAs arecut,whereby
sequencescorresponding toan‘externaltranscribedspacer’ (ETS)andtwoshort‘internal
transcribed spacers’ (ITS1 andITS2) arediscarded(Fig. 1). Theribosomal5S RNAs are
coded by a separate,unlinkedtandemarray(Gerlach &Dyer 1980;Hemleben& Werts
1988; Dvorak etal. 1989; Lapitan etal. 1991).TherDNA transcription unitsare separ-
ated by ‘non-transcribedspacers’ (NTSs, intergenic spacers: IGSs). In general, all the
numerouscopiesofthe 18Sand25SrRNA genesina genomehavethesamesequence,and
thesequencesofthesetwo RNAsevolveslowly (Zimmeretal. 1989;Martin&Dowd 1991;
Hamby & Zimmer 1992). Thisconservationofthe coding sequencescontrasts with the
nearly unrestrainedevolutionofthenon-transcribedspacerregion (Jorgensen &Cluster
1988).DNAprobes fromtheIGSofCucurbitapepo(courgette) donoteven hybridizewith
DNA fromother generaofthe Cucurbitaceaesuch as Cucumis(cucumber and melon;
Torres etal. 1989; Torres& Hemleben, 1991).Each intergenic spacertypically contains
tandemrepeatsofsimplesequences,80-325bp inlengthindifferentspecies. Thesesimple-
sequencesub-repeats areusuallyconservedwithina genusoramongclosely relatedgenera
(Ganal& Hemleben1986a).Length heterogeneityamongtherDNArepeatunitswithina
species can arise when the numberofsimple-sequence repeats in the IGSregion varies
(Yakura etal. 1984; Rogers & Bendich 1987; Hemlebenet al. 1988; Reddy etal. 1990).
Enzymes thatdonotcutwithinthesimple-sequence repeatsoftheIGS transformspacer
length heterogeneity into RFLPs (e.g. Bellarosa et al. 1990; Reddy et al. 1990). This,
ratherthanmutationsinrestrictionsites, is thetypical originofRFLPs inthe rDNA. As
(heterologous) probes fortherDNAcoding sequenceshavebeenavailablefor sometime,
there are quite a few taxonomicand evolutionary studies using RFLP analysis ofthe
rDNA.
Thevalueofanypolymorphism forphylogenetic analysis dependsonourknowledge of
theway inwhichthispolymorphism evolves. Evenwherewehaveageneral understanding
fortheevolutionofatypeofpolymorphism (suchas forbasepair replacements), wehave
tocheckthespecificparametersempirically ineach case. Thereare severalpossible mech-
anismsleading tochanges intherepeatnumberoftandemlyrepetitiveDNA,but thereare
also mechanismsthateliminatevariationamongtherepeatsinonelinkedcluster(Appels
& Dvorak 1982; Dover 1986; Sytsma & Schaal 1990). Ingeneral, recombinationtends
to promote sequence homogeneity within populations, and the magnitude of
fragment length differences among species neednot be a measure of thephylogenetic
distance.
374 K. BACHMANN
RFLPsbasedonsimple-sequence repeatsinspacersare beingusedforthedetermination
of (phylo)genetic relatedness, mostly among populations of one species and among
congeneric species. Thereare considerabledifferencesamong the levels ofvariability in
differentcases. Twenty lengthvariants were detectedamongtherDNA repeatsofasingle
plant ofViciafaba (Yakura etal. 1984)andupto fiveperindividualin Clematisfremontii
(Learn & Schaal 1987). Nointraspecific spacer length variation, but some variationin
restriction sites was observedinanextensive survey ofRudbeckiamissouriensis(King &
Schaal 1989). Inmany species restrictionsite variationoccurs only amongraces (Doyle
et al. 1984) or atthe interspecific level (Sytsma & Schaal, 1985; Doyle & Beachy 1985;
Rieseberg et al. 1988). Qualitative and quantitative differentiationamong populations
and subspecies was found in wild barley (Saghai-Maroof el al. 1984), Phloxdivaricata
(Schaaletal. 1987)andClematisfremontii(Learn& Schaal 1987). Sytsma & Schaal(1985)
haveanalysed theLisianthiusskinnericomplex inPanama.Lisianthiusisoneoftheappar-
ently rareexamplesoflengthvariationintheITSratherthantheintergenicspacer(Sytsma
& Schaal 1990). Bellarosa etal. (1990) have used RFLPs in the rDNA to determinethe
relationship amongmediterraneanoaks (Quercus spp.), D’Ovidio et al. (1990a)that of
poplars (Populus spp.),Reddy etal. (1990)havecompared 23populations from12species
ofSecale, andZentgrafetal. (1992) haveused diagnostic RFLPs inthe rDNA clusterto
distinguish theAsianandAfrican species groupswithinCucumis.
Thereis nouniformmethodfortreating thedatainthevariousstudiesusing RFLPs in
the nuclear rDNA. While Reddy et al. (1990) employ an essentially phenetic approach
basedonfrequencies ofspacer-length variants, theclassificationby Bellarosaetal.(1990)
isbasedonthesharedpresenceoflengthvariantsandis thereforeessentially cladistic.The
recent analysisofnuclearrRNA variationinKrigia (Asteraceae) by Kim& Mabry(1991)
is aninstructiveexample fortheevaluationofrDNA RFLPdata.
RFLP ANALYSIS WITH RANDOM GENOMIC AND cDNA
PROBES
RFLPsallacross thegenomecanbeusedto determinetherelatednessandthephylogeny
ofgenomes.Thecrucialfactoris theavailabilityofclonedDNA-probes forthedetection
of homologous sequences by DNA/DNA hybridization on blots ofrestriction digests.
RFLPs inthe nuclear DNA can beusedas co-dominantmarkersingenetic crosses, and
their segregation and linkage relationships can be determinedby Mendeliananalysis.
Theycanbeusedtoproduce anRFLPmapofthegenome.Incontrast tothephysical map
ofa segment ofa genomein which restriction sitesare mapped in base-pair distances,
distanceson an RFLP mapofthe genomeare in recombinationunits. Thegenetic loci
detectedby RFLPs can be incorporated in the general genetic mapofthe species and
greatly facilitate Mendeliananalysis as they show a simple monogenic co-dominant
inheritance pattern (Helentjaris el al. 1985; Landry et al. 1987; Paterson et al. 1988).
NuclearRFLPs forma directlink betweenthegeneticanalysis ofthegenomeandphylo-
genetic analysis. Ideally theirmappositions areknown, and theircontribution to a truly
representative sampling of the genome can be determined(Kesseli et al. 1991). The
sequenceresponsible foranRFLP andthe relevantchromosomalregion are experimen-
tally accessible via the cloned probe, and a direct link can be laid between molecular
evolutionandthe geneticbasis ofmorphological evolution.
Unfortunately,theexpenses intimeandmoneyforthefulluseofnuclearRFLPanalysis
can become limiting. This primarily concerns the production of a set of informative
DNA MARKERS AND ANGIOSPERMTAXONOMY 375
probes, the preparation and the characterizationofa clone‘library’. Theseprobes are
eithercDNA, i.e. DNA that hasbeensynthesized in vitroby reverse transcriptionfrom
mRNA (Helentjaris et al. 1985; Bernatzky & Tanksley 1986), or pieces of DNA cut
directlyfrom thenuclear genome(Kochert et al. 1991;Neuhausen 1992). cDNA clones
containonly thecoding sequencesofexpressed genes,whilerandomgenomic clonesalso
containregulatory and non-coding repetitive sequences, which require a moreintensive
screening for useful probes (Xu & Sleper 1991) butprovide access to a wider rangeof
variation.
Dueto theconsiderabletechnicalinvestment, nuclearRFLPanalysis hasbeenapplied
mainlytocropspecies. Thesestudies,however, frequentlyinvolveinterestingphylogenetic
problems, mostly on the level of species withina genus. The levels of intraspecific or
interspecific variationfornuclearRFLPs areratherunpredictable inplants fromvarious
generaor families(Shattuck-Eidens et al. 1990). Neuhausen(1992) found that RFLPs
withinspecies ofmelons(Cucumis)mostlyweredueto mutationsinrestrictionsites,while
RFLPs between species frequently involved chromosome restructuring. Song et al.
(1988a,b, 1990) have constructed cladograms of Brassica species and cultivars from
nuclearRFLPs,whichagreewellwithotherdataandprovidenewinsightsintheevolution
ofthis group. Their investigation raises the difficult question of the incorporation of
polyploid hybrid taxaamongdiploids in a cladisticanalysis. Brummeretal.(1991) con-
structed separatecladograms for diploid and tetraploid cultivars ofalfalfa(Medicago),
but also faced the problem ofreticulateevolution via hybridization at one ploidy level
(‘homoploid reticulateevolution’). Kochert etal. (1991) compared 8 cultivars and 14
species of peanuts (Arachis) and found that they could recognize specific restriction
fragments of the diploid parental species in the banding pattern of the tetraploids.
Two cottonwoodspecies (Populus fremontiiand P.angustifolia) along the Weber River
drainage in Utah couldbe identifiedby interspecific RFLPs and the parentageofindi-
vidualplants ofahybrid swarm occurringintheiroverlap zonecouldbe determined.The
hybrids were eitherFI plants orbackcrosses to the P. angustifolia parent (Keim etal.
1989).Thetopology oftherelationship amongfourspecies ofActinidiabasedonnuclear
RFLPs withcDNA and randomgenomic clonesagreeswiththatbasedon RFLPs in the
cpDNA (Crowhurst etal. 1990).
MINI-SATELLITE VARIATION: VNTR FINGERPRINTS
Various relatively short DNA sequencesseem to cause frequent irregularities in DNA
replication and/orrecombinationwhereverthey occur (Levinson & Gutman1987). Such
sequencesare foundasmoderately repetitivetandemrepeats(‘mini-satellites’) ofmoreor
less identicalunits thatare up toabout60bp in length (Jeffreys etal. 1985).Many such
mini-satellites seem to occur at various positions throughout all eukaryote genomes.
Martienssen& Baulcombe (1989) illustrateamini-satellitein the wheat genomethat
consists of 21 repeats of 15-26bp in length each containing a 14bp ‘core sequence’
(averaged overall 21 repeats, thereare 121of 14bp identicalwiththecore consensus).
Thenumberofrepeatsinaparticularmini-satellitecanvaryamongtheindividualsofa
population (variable number tandem repeats: VNTRs). This variation is detected as
restriction fragmentlength variationwhenDNAis digested withrestriction enzymesthat
donotcutwithinthe repeat.As VNTRs basedon onebasic sequenceregularly occur at
various places in the genome, a probe detecting the repeat sequence usually reveals
severalor many restrictionfragments perindividual.Someofthesemaybealleles from
loci heterozygous for repeat numbers. Due to the high rate of length variation, the
376 K. BACHMANN
corresponding pattern ofbands often is specific for anindividualand is calleda ‘DNA
fingerprint’.
Severalprobes containing coresequencesofVNTRs havebeenusedforfingerprinting
invariousorganisms. Thefirst demonstrationsoftheexistenceofmini-satellitesinplants
madeuseofahumanmini-satelliteprobe (Dallas 1988;Rogstad etal. 1988a) orofaprobe
derived from bacteriophage M13(Rogstad et al. 1988b, 1991a,b; Ryskov et al. 1988;
Nybom et al. 1989;Zimmermanet al. 1989). Relatively simple and reliabletechniques
fordetectingfingerprints inplantsmakeuseofsynthetic oligonucleotides thatcanbe four-
baserepeatssuch as (GATA)4or(GACA)5(Weisingetal. 1989, 1990, 1991).Suchprobes
are easilyavailableandlikelytorevealRFLPs atmanylocithroughout thegenome(Tautz
& Renz 1984;Tautz 1989;Condit& Hubble 1991).Thegreatadvantage ofthismethodis
setagainst two limitationsto its usein phylogenetic analysis.
Themost severe limitationis thehighrate ofchange ofVNTR lengths. VanHouten et
al. (1991), using (GATA)4 as a probe, foundthat very few plants in a population of
Microserispygmaea(Asteraceae) had identicalfingerprintpatternsand thatthe patterns
couldchange unpredictablybetween aplantand itsoffspring fromselling. Inthe closely
relatedspecies, M.elegans, fingerprint patternswere considerably morestableandcould
be used to determinethegenetic relationship of genotypesisolatedfrom variouspopu-
lations(Van Heusden& Bachmann, 1992a). Inacomparison ofthreespeciesofPlantago,
Wolffetal.(in press)foundnointrapopulation variationwiththeM13 probeintheselling
species P. majorand high levels ofmini-satellitevariationinthe outcrossing species P.
lanceolata. Population differentiationformini-satelliteswas foundin Acer negundo by
Nybom &Rogstad (1990) andinAsiminatriloba(Annonaceae; Rogstad etal. 1991a). The
interpretation of fingerprint patterns confronts us with the second limitationof the
method.Itisimpossible to determinethehomology (or theallelism)ofthebandswithout
performing breeding experiments. Thecomparison is usually limitedto the presenceor
absenceofabandata certainposition in thegel, andit isassumed thatthis comes from
identical allelesofone locus in allthe plants. The typical datamatrix will list, foreach
plant, the absence orpresence ofallbands foundin the entire sample. Thesedata can
eitherbe treatedas cladistic charactersor, using anindexofband sharing, ina distance
matrix(Lynch 1988, 1990).
Themostuseful applicationofmini-satelliteprofiles inplant evolution seems tobe for
thedeterminationofclonalidentityandevolutioninasexually reproducing plantspecies
such as apomictic blackberries (Rubus, Nybom & Schaal 1990; Nybom & Hall 1991)or
clones of quaking aspen (.Populus tremuloides, Rogstad et al. 1991b). Fingerprinting
of apomictic dandelions(Taraxacum) has shown that plants of the ‘microspecies’ T.
hollandicum from The Netherlands, Czechoslovakia and France are indeed members
ofone clone, but has thrown doubton the clonal nature of some other Taraxacum
microspecies (Van Heusdenetal. 1991).
Theterm‘DNAfingerprinting’ is notexclusively applied to banding patternsbased on
mini-satellitelength variation, butmaysignify any individual-or strain-specific banding
pattern on a gel based on length polymorphisms of DNA fragments (Gebhardt et al.
1989). This may include banding patterns based not on restriction fragments but on
amplification products(D’Ovidioetal. 1990b;Welsh& McClelland 1990).
POLYMORPHISMS IN PCR AMPLIFICATION PRODUCTS, RAPDs
Theintroductionofthepolymerase chainreaction(PCR) forthein-vitroamplification of
specific DNA segmentsout ofa mixtureofmany others(Saiki et al. 1985) has greatly
DNA MARKERS AND ANGIOSPERM TAXONOMY 377
simplified many molecular protocols. PCR has made comparative sequencing of
homologous stretches of DNA accessible to systematists, because the purification of
homologous sequencesfromnew material,whichusedtobeforbiddingly complex, isnow
asimple routine.Themethod dependson sufficientknowledge ofthedesiredsequenceto
synthesize stretchesofabout30bp lengthflanking thepiece tobeamplifiedwithopposite
complementarity. These hybridize to the homologous sequences on denatured, single-
strand (template) DNA and serve as starting points (primers) for the synthesis of the
complementary strand including the site ofthe opposite primer. A heat-stableDNA
polymerase is usedthat survives thetemperaturesused for denaturing DNA. Repeated
denaturing, primerattachmentandcomplementary strandsynthesis eventually produces
anenormous amountof double-strandedDNA corresponding precisely to the stretch
between the opposing primers. Sequencing the amplification product will thenreveal
polymorphisms atthislocusamongthevariousspecimens fromwhichtemplateDNA has
beenisolated(Barbier etal. 1991).
PCR canalsobeusedforarandomsamplingofsequencevariationalloverthegenome.
Forthis, in place oftwo long (c. 30bp) primers flanking aspecific sequence,oneshorter
(10bp) primer with an arbitrary sequence is used. Amplification will depend on the
existencesomewhereinthegenomeoftwosites thatarecomplementary to theprimerand
lieon opposite strands ata distancesothattheintervening sequencecanbefilledinby in
vitro DNA synthesis. The statistical chance for such a constellation is minute, but
genomesare largeandthereis a virtually unlimitedvarietyofpossible 10bp primers. In
fact, quiteregularlyseveralamplificationproductsofdifferentlengthsareobtainedfroma
plantgenomewithanarbitraryprimeranddetectedasbandsonagelafterelectrophoresis
(Williams el al. 1990; Quiros et al. 1991; Van Heusden & Bachmann 1992a). These
bandingpatternsareusually polymorphicinpopulations (randomamplifiedpolymorphic
DNAs: RAPDs; Williamsetal. 1990), and thevariouspatterns obtainedwithdifferent
primers forma detailedDNA ‘fingerprint’ for genomerecognition (Welsh & McClelland
1990; Klein-Lankhorstet al. 1991). Mendeliananalysis shows that the polymorphisms
usually consist ofthe absence or presence of an amplification product from a site on
the genome,possibly reflecting base substitutions in the primer attachmentsites. The
dominantinheritanceofRAPDs makes them less convenientgenetic markersthan the
co-dominantly inheritedRFLPs, but RAPD amplification products can be mapped
(Williams et al. 1990; Quiros et al. 1991). They can also be used as presence/absence
charactersforphylogenetic analysis.
Populations of theCalifornian autogamousannual, Microseriselegans (Asteraceae),
are geographically isolatedfromeachotherandconsist essentially ofonelocal ‘biotype’
each.Themorphologicalandgeneticdifferencesamongthesebiotypeshavebeenanalysed
ininbredstrainsderivedfromvariouspopulations. Thevirtualabsenceofheterozygosity
inthesestrainssimplifiestheircomparison withdominantRAPDmarkers(Van Heusden
& Bachmann 1992a). These revealed monophyletic groups of strains, of which one
represented populations that were not geographically nearest neighbours. Apparently,
populations can arise after long-distance fruit dispersal to unoccupied sites withinthe
range ofthe species and then resemble theirsource populations more than theirneigh-
bours. Theclose genetic relationship among thesepopulations could not beconcluded
frommorphologicalcharacters, butonceitwasknown,itcouldbe detectedinthefirsttwo
factors ofaprincipalcomponentsanalysis ofthemorphological variation(Bachmann &
van Heusden 1992). Similaranalyses of the closely related species, M. bigelovii (Van
Heusden&Bachmann 1992b)and M.pygmaea(Van Heusden& Bachmann 1992c) have
378 K. BACHMANN
revealedsubtledifferencesamongthe species bothin the degree ofrecombination(gene
flow) amongpopulations and in the overall variationwithin the species. The method
seems to be very useful for the analysis ofintraspecific variation, atleast in inbreeding
species. However,VanHeusden& Bachmann(1992c) haveshownthatthereare RAPDs
forvarioustaxonomiclevelsofanalysis. Inacomparison ofthetwo mostdivergentstrains
ofeachofthreespecies ofMicroseriswith22 primers, 143bands were sharedby atleast
two strains.Ofthese,6, 12,and 14amplification products were sharedby thetwo strains
ofeach ofthe three species and are thereforecandidatesfor species identificationwith
DNA.Forty-nine amplification products werepresentin all six strainsand are possibly
informativeathigher taxonomiclevels.
RAPD polymorphisms are typical molecularmarkers. They are individually oflittle
value, but can become immensely valuablewhen they give informationabout many
uniquelocirandomly distributedacross thegenome.Themethodisrelativelynewandnot
yetfullytested. Infact,theoreticalpredictions basedonthe assumption that theprimers
recognize and reveal allcomplementary sites inthe genomeare frequently not met. For
instance, combinationsoftwo primersusually donotproduce thebandsofthetwo single-
primerpatternsplusnewones(Klein-Lankhorst etal. 1991;Williamsetal.inpress), anda
combinationoftwo template DNAs doesnot necessarily produce thecombined bands
amplified fromthe single templates(Williamset al. in press). Moreover,ifamplification
products are isolated froma gel, labelledandhybridized to ablot ofthis gel, morebands
are labelledthanare seenin theoriginal ethidiumbromidestainedgel (Williams etal. in
press; K.Buchmann, unpublished data).Thissuggeststhatthereis somesortofselection
amongvarious potential sites foramplification during the amplification process. More-
over, it pays to rememberthat the bands revealed by RAPD primers (and unspecific
fingerprint probes) may come fromany genomeincluding thoseof insects and molds
infecting theplants under study. Withinlimits and with relevant controls, though, the
methodis idealfora quickand easy genetic survey. As theamplification products canbe
isolatedandcloned,themethodcanbe convertedintoa rigorous RFLPtest.
CONCLUSIONS AND OUTLOOK
Thegenetic variationinnuclear-DNAmarkers covers theentirerangefrompolymorphisms
amongrepeated sequenceswithinasingle plantto evolutionary stability overlong geolo-
gical periods. The discussion ofthe ribosomalRNA genes has shown that all levels of
variability can occur within differentparts of a single gene.Various genomic fractions
revealed by the differentmethods have their typical levels of variability. The examples
citedabovehaveshownmuchvariationwithintheexpected rangeofeachfraction, andthe
level at whicha particular marker is informativein a specific plant groupneeds to be
empirically determined.
Most ofthe recent studies in plant evolution employing nuclear-DNA markers use
them atthe level ofspecies and populations. There they complement cpDNA databy
revealing variationat all levels of genetic relationship, by their biparental inheritance
and bytheirgenetic linkage withthe genes involvedin morphological andphysiological
evolution.Withnuclearmarkers, theparental genomesin diploid andpolyploid hybrids
canberecognized (Arnold etal. 1990;Kochertetal. 1991; Zhang& Dvorak 1991).When
polymorphisms in the nuclearDNA ofa diploid parental species are foundback inthe
allotetraploid derivative, they suggesta multiple origin ofthe tetraploid (Soltis & Soltis
1991).