Patterns of Infection as a Function of Host Genetic Diversity and Host Density in Natural Populations of the Apomictic Crucifer, holboellii Author(s): B. A. Roy Source: Evolution, Vol. 47, No. 1 (Feb., 1993), pp. 111-124 Published by: Society for the Study of Evolution Stable URL: http://www.jstor.org/stable/2410122 . Accessed: 29/01/2014 15:38

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

.

Society for the Study of Evolution is collaborating with JSTOR to digitize, preserve and extend access to Evolution.

http://www.jstor.org

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions Evolution,47(1), 1993, pp. 11 1-124

PATTERNS OF RUST INFECTION AS A FUNCTION OF HOST GENETIC DIVERSITY AND HOST DENSITY IN NATURAL POPULATIONS OF THE APOMICTIC CRUCIFER, ARABIS HOLBOELLII

B. A. Roy' Rancho Santa Ana Botanic Garden, Claremont,CA 91711 USA and Rocky Mountain Biological Laboratory,Crested Butte, CO 81224 USA

Abstract.-It is oftenassumed that geneticdiversity contributes to reduced disease incidence in natural plant populations. However, littleis known about the geneticstructure of natural popu- lationsaffected by disease. Here I presentdata fromthree apomictic (asexual) populationsof Arabis holboelliiinfectedby therusts monoica and P. thlaspeos.An averageof 300 hostindividuals per populationwere genotyped (using seven variable allozyme loci) and scoredfor disease presence. Arabis holboelliipopulations are geneticallydiverse; the numberof clones detectedper population rangedfrom 6 to 27. There was substantialvariation in frequencyof host clones withinand among sites, and significantvariation among clones in susceptibilityto the differentrusts. Contraryto predictionsbased on frequency-dependentselection theorythere was not a consistentpositive relationshipbetween clone frequencyand disease incidence withinany of the populations (Spear- man's r = -0.096, P > 0.5). In addition, clonally diverse populations did not necessarilyhave decreased disease incidence. The population with the lowest overall (both pathogenscombined) disease incidence(7.5 ? 1.9%) had the smallestnumber of clones (6), the lowestspatial variability, and the highestArabis density.By comparison, another population had 22 clones, high spatial variability,low Arabis densityand significantlymore disease overall (16.8 ? 2.7%). Althoughthis studydoes not eliminate the possibilityof frequency-dependentpathogen attack in these popula- tions, the resultssuggest that it is likelyto be weak or intermittent.

Key words.-Arabis, density,epidemiology, frequency-dependence, plant pathogens,Puccinia.

Received December 6, 1991. Accepted June 12, 1992.

Genetically variable clonal populations goingpathogen-mediated frequency-depen- are difficultto explain undertraditional eco- dent selection: logical theory,because the most productive Prediction 1: Within populations, geno- and competitive clone should drive all of types (clones) occurringat low frequency the othersto extinction(Bell, 1982; Sebens should have a fitnessadvantage ("rarityad- and Thorne, 1985; Begon et al., 1986). vantage") (Haldane, 1949; Jaenike et al., However, there are several ways genetic 1980; Hamilton, 1980; Antonovics and variabilityin clonal populations could be Ellstrand,1984; Parker,1989; Lively et al., maintained, including frequency-depen- 1990; Alexander, 1991; Parker,1992). That dent selectionby pathogens,heterogeneous is, genotypeswith rare resistance alleles will environmentsfavoring one clone then an- escape infectionuntil the resistantgenotype other, the continuous generation of new is common enough in the population that clones from sexual progenitors,rapid mu- selectionon the pathogenfavors strains that tation,and coexistenceof similarclones for can overcome the resistanthosts. short times due to neutral stability(Bell, Prediction2: Populations withmore vari- 1982; Sebens and Thorne, 1985). One ofthe ation in resistanceshould have less disease least studiedof these mechanismsis the po- thanmore uniform ones (Adams et al., 1971; tential for pathogens or parasites to exert Harlan, 1976; Bremermann, 1980, 1983; frequency-dependentselection. Barrett,1981; Alexander, 1988). Decreased Two predictionsare typicallymade con- disease incidence is generallyexpected in cerningpopulations thoughtto be under- variable versus uniform populations be- cause the absolute numberof hosts suscep- tible to a given pathogen genotypewill be ' Presentaddress: BotanyDepartment, University of lower, the distance between susceptible California,Davis, CA 95616 USA. plants will be greater, making pathogen 111

?0 1993 The Society forthe Study of Evolution. All rightsreserved.

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions 112 B. A. ROY transmissiondifficult, and pathogenpropa- geneticdifferences in susceptibility(Falcon- gules landing on resistanthosts will be lost er, 1981). Second, the clonal nature of the to the system (Burdon, 1987; Alexander, plants may make it easier to detect fre- 1988). quency-dependenteffects. This is because a The distance between susceptible plants population composed of clones originating can be largebecause eithera) hostresistance fromdifferent parental stocks may be more alleles are diverseand thusthe frequencyof stronglydifferentiated than a sexual popu- individuals in a susceptiblegenotype is low, lation in whichthere may be more genomic or b) because the actual physicalproximity sharing(Tibayrenc and Ayala, 1988). (density)of hosts is low. Thus, host density, like host frequency,is typicallypredicted to MATERIALS AND METHODS be positivelyassociated with disease inci- dence because more and closerhost 'targets' The Hosts are assumed to improve the probabilityof The hosts, Arabis holboellii (Brassica- pathogentransmission (Anderson and May, ceae), are biennial to short-livedperennial 1979; Burdon and Chilvers, 1982; Alex- plants occurring on well-drained soils ander, 1988; Burdon et al., 1989). Never- throughoutNorth America and extending theless,host densityis not always positively north into Greenland (Rollins, 1941). Re- associated with disease incidence (Burdon productionin A. holboelliimay be sexual or and Chilvers, 1982; Burdon, 1987). apomictic (Bocher, 1947, 1951, 1969; Ny- The presentstudy is partof a largereffort gren, 1954). I have verified,by emascula- to determinewhether frequency-dependent tion and crossingexperiments (Roy, 1992; selection by pathogenicrust fungiis struc- unpubl. data) and by electrophoresis of turingnatural populations of the apomictic progenyarrays (Roy and Rieseberg, 1989), (asexual) crucifer,Arabis holboellii.Arabis that the plants in the studypopulations re- holboelliidisplays heritable variation in re- produce by pseudogamous apomixis. In sistanceto rustinfection, and host fitnessis pseudogamous apomicts, pollen is neces- reduced by infection(Roy, 1992). For ex- sary for successfulseed production (prob- ample, in a singleseason, uninfectedplants ably requiredfor endosperm formation) but floweredabout 31% of the time versus0.4% does not fertilizethe ovules (Richards,1986; for those infectedby Puccinia monoica or Asker and Jerling,1992). The geneticcon- neverfor those infected by P. thlaspeos.Sur- sequence of thisform of apomixis is thatall vival of uninfectedplants and of those in- progenyfrom a single motherare identical fectedby P. thlaspeos was relativelyhigh, to each otherand to theirmother (Richards, 65% and 64% respectively,but declined to 1986). Arabis holboelliiplants do not spread 40% or less forthose infectedby the more vegetatively,but theyare neverthelessclon- common rust,P. monoica. The goals of the al (sensu Buss, 1985; Ellstrandand Roose, presentstudy were to determine,in natural 1987), because all the membersof a family populations of Arabis holboellii,(1) the ex- are geneticallyidentical. I have chosen to tentof clonal variability,(2) whetherclones call theseindependent plants individuals in- varied in frequencyand in disease incidence stead of ramets to emphasize that there is withinand among populations,and (3) the no physiologicalconnection between plants. effectsof host densityon disease incidence. I referto groups of individuals with iden- The fact that Arabis holboellii is clonal tical allozyme phenotypesas clones. has importantimplications for this study. First,it should be possible to assess disease The Pathogens incidenceat the clonal level,in situ,because The pathogens, Puccinia monoica (Pk.) each genotypein a given population is rep- Arth. and P. thlaspeos C. Schub. [listed as resented by many replicate, genetically P. holboelliiin Anonymous (1960) and Ar- identical individuals. I am assuming that thur(1962)], cause systemicrust disease on the environmentalinfluences on infection Arabis species (Anonymous, 1960; Farr et probabilityare spread among these repli- al., 1989). Althoughthese rustsare closely cated genotypesand that differencesin in- related to each other (Savile, 1974; Cum- fectionat the clonal level thereforereflect mins and Hiratsuka, 1983), and may rep-

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions PATTERNS OF RUST INFECTION 113 resent a progenitor-derivativespecies pair A. Puccinia monoica Arth. (Cummins and Hiratsuka, 1983), theyhave dissimilarlife cycles (Fig. 1) and the differ- II. Uredia & Uredospores >ll.Tella & Tellospores (2n) ences may have epidemiological conse- (Repeating stage) IV. Basidla & Basidiospores (1n) quences. P. monoica has a "typical" rustlife "__ / GRASS cycle: it alternateshosts and it has fivedif- (#7 (Primary Host) ferentspore states (i.e., it is heteroecious and macrocyclic).For about fourweeks of the year it occurs on the primaryhost, a I. Aecia & Aeclospores (n+n) grass belongingto any of the genera Trise- ARABIS (Alternate host) tum,Koleria, or ; then it must switch to its alternatehost, Arabis or a few other genera in the to complete its 0. Spermogonia & Spermatia (in) (Fertilization) life cycle (Arthur,1962; Farr et al., 1989). P. thlaspeos,however, is microcyclic,hav- ing only threekinds of "spore stages" (sper- matia, teliospores,basidiospores) insteadof B. Puccinia thiaspeos C. Schub. five;it completesits entirelife cycle on Ar- abis. In the field,these rustsare easily distin- guished in mid-May by the color of the spores and spore-bearingstructures on the 0. Spermogonia& Spermatia(in) undersidesof leaves on infectedplants. At thistime P. thlaspeos'teliospores are black, ARABIS whereasP. monoica's aeciospores are bright 111. Telia & Teliospores (2n) yellow. V. Basidia and basidiospores (in) The Sites Three sites at similarelevations (approx- imately 2,700 m), but located in different drainagesystems minimizeintersite FIG. 1. Arabis rustlife cycles. A. Puccinia monoica. (to gene This macrocyclicrust alternates between a grass(of the flow)were selected formonitoring in Gun- generaKoleria, ,or Stipa) and Arabis(or other nison County,Colorado. All threesites were mustard species depending on the rust race). During in sagebrush (Artemisiatridentata) mead- the spermatial stage (stage 0), cellular fusion (plas- ows approximately10,000 m2 in size, and mogamy) occurs between opposite matingtypes, and dikaryotic(n + n) aeciospores (Stage I) are formed bounded by foreston three sides and by a shortlyafter fusion. The aeciospores disperse to the road on the fourthside (see Roy, 1992, for primaryhost, a grass.Not longafter infecting the grass, detailed site descriptions). the pathogen produces urediospores(Stage II), which are capable of infectingother grass individuals. It is Population Sampling duringstage II thatthe pathogen'spopulations expand greatly.Eventually, nuclear fusion occurs (karyogamy), Large scale population samplingwas nec- and diploid teliospores are formedat Stage III. The essary to determine how clonal variation teliosporesgerminate in the nextrains to formbasidia was partitionedwithin and among sitesand and basidiospores (Stage IV). Meiosis occurs in the for calculations of population and clonal basidia, so the basidiosporesare haploid. The basidio- I spores dispersefrom the grasses and cause infectionon disease incidence. In April of 1990 estab- Arabis,completing the cycle. The spermatialand aecial lished at each site eight nonoverlapping, stagesoccur in theearly spring, the uredialstage during randomly chosen, linear transects(nine at mid-latesummer, and thetelial and basidial stagestake Gold Creek) 25 m long by 2 m wide. A place late summer-earlyfall. B. Puccinia thlaspeos.This random sample of 200 individuals per site microcyclicrust has onlythree known spore stagesand is probably derived fromthe macrocyclicspecies, P. were permanentlymarked with aluminum monoica (Cummins and Hiratsuka,1983). P. thlaspeos tags along the eighttransects (25/transect). does not alternatehosts, and is only foundon mustard To minimizethe potential for introducing species such as Arabis. Life cycle terminologyfollows spores between sites and thus changingthe that of Petersen(1974). geneticsof the pathogens present in thepop-

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions 114 B. A. ROY ulations, I always changed clothing,shoes, tinctclonal groups,hereafter referred to as and notebooksbetween sites and thorough- CGs (see Appendix for a list of CGs and ly cleansed my hands and measuringequip- their clonal make-up). Five of these CGs ment with alcohol. werebased on beingidentical at a minimum of PGI and LAP (i.e., identical at 57% or Clone Identification more of the seven variable loci scored). The Clones were identifiedby differencesin othertwo clonal groups ("E" and "G") are allozymebanding patterns. I used allozymes more heterogeneousand representindivid- because it is relativelyeasy to screen large uals that may be sexual (based on higher numbers of individuals with them, and I levels ofhomozygosity and less linkagethan expectedthat if clones differedin resistance, in known apomictic Arabis). then theywould probably also have differ- ent allozyme phenotypesbecause in apo- Disease Incidence mictic taxa all loci act as a single linkage Disease incidence (measured by percent group. Inoculation testson a small number of individuals infected)was determinedfor of clones indicated that this was a reason- each population, and for each clone and able assumption since there was variation clonal group that occurredin all threepop- in resistanceamong differentallozyme phe- ulations. I classified each genotypedindi- notypes,and identicalallozyme phenotypes vidual forpresence or absence of infection. had the same degree of resistance (Roy, Infectedplants are easily identifiedbecause 1992). For electrophoresis,small amounts the systemicrust infections cause distorted of leaf tissue (approximately 1 cm2) were growthin the host and the undersidesof the removed from the marked individuals at leaves are covered withblack (P. thlaspeos) each of the sites on May 8, 1990. The tissue or yellow (P. monoica) spores. Population samples were stored on ice for two days disease incidencewas measuredby counting duringtransportation from the fieldsites to the numberof infectedindividuals encoun- the laboratory,then kept in an ultralow teredin the random sample of 200 individ- freezer(- 80?C) until use. Electrophoretic uals fromeach of the sites. Standard errors procedures followed Roy and Rieseberg of the proportionswere calculated accord- (1989). Of the 9 putativeloci scored,7 were ing to Cochran (1963, p. 64). variable (an additional 15 loci wereresolved I was interestedin learningwhether host in tests,but were not polymorphicor were clones occurred in differentfrequencies too difficultto score). To facilitateaccurate within these populations and whetherfre- scoring of banding patternsand compari- quencyof infection (percent infected) varied sons among gels, markerindividuals with withinand among clones. Because diseased known banding patternswere included in plants were rare (7.5 to 16.8% of the pop- each gel. ulation), the random sample did not pro- Clone identitywas determinedby sorting vide an adequate sample for assessing dis- the allozyme banding patterns of phos- ease incidence within clones. Therefore,I phoglucoisomerase (PGI, 3 loci), triose- also genotyped all of the infected plants phosphateisomerase (TPI, 2 loci), isocitrate within each transect(hereafter referred to dehydrogenase(IDH, 2 loci) and leucine as the complete infectedsample to differ- aminopeptidase (LAP, 2 loci), and identi- entiateit fromthe random sample). Because fyingunique patterns(see Appendix). Clon- the frequencyof infectionwithin the pop- al diversitywas estimatedfor the threesites ulation (percent infectionin the random using standard diversityindices [Shannon- sample) and the distributionof genotypes Weaver and Simpson's Diversitycorrected in theinfected class (completeinfected sam- forfinite sample size (Peet, 1974)]. ple) were necessarilyestimated from differ- Because therewere a largenumber of sim- ent,but overlappingsamples, actual percent ilar, but not identical, clones, I also com- infection per genotype (=clone or clonal bined clones into classes that shared many group, depending on the analysis) was es- alleles to increase sample sizes and statis- timatedby weightingthe totalinfected sam- ticalpower for some analyses.The 41 unique ple of diseased individuals in each genotype clones were readily sorted into seven dis- by their proportion in the population at

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions PATTERNS OF RUST INFECTION 115

TABLE 1. Clonal diversityof Arabis holboelliiat threesites.

Site Cement Creek Taylor River Gold Creek Number of individuals genotyped 455 219 230 Number of clones 6 27 22 Number of unique clones 1 18 13 Clones shared with: Cement Creek - 5 5 Taylor River 5 - 9 Gold Creek 5 9 Number of clones/samplesize 0.01 0.12 0.10 Simpson's DiversityIndex, D 0.40 0.92 0.88 Shannon-Weaver,H' 0.80 2.8 2.4 Population disease incidence 7.5 ? 1.9 10.9 ? 2.2 16.8 ? 2.7

large. Standard errorswere calculated using RESULTS Gaussian error propagation analysis Clone Frequencyand Variationin (Bartsch, 1974, pp. 475-476), because each Disease Incidence proportionin the equation was subject to Arabis holboellii populations are geneti- error. cally diverse;the numberof clones per pop- To determinewhether there was a rela- ulation ranged from 6 to 27 (Table 1 and tionship between clone frequencyand dis- Appendix). Across all three sites a total of ease incidenceI performedtwo differentsta- 41 distinctclones were found (Table 1 and tistical tests: Spearman's rank correlation Appendix). Five clones werecommon to all on the ranksof clone frequencyand disease populations, but each also incidence, and a two-waytest of indepen- population con- tained between 1 and 18 unique clones (Ta- dence to determinewhether disease inci- ble 1). Clonal diversity,as measuredby two dence was independentof clone frequency. standardindices (Table 1), was verysimilar For the independence test I classifiedeach at the Taylor River and Gold Creek sites, clone as rare or common by whetherit fell which were both much more diverse than above or below the average frequencyfor Cement Creek. all clones (across all sites), and by whether Spatial variation in the genetic compo- it was above or below average in termsof sition of the sites can be evaluated visually infection(averaged over all sites); a chi- by mapping the spatial distributionof the square test was used to evaluate signifi- clonal groups (CGs) along the transects,as cance. in Figure 2. Cement Creek is dominated by a single CG (=clone Dl in this case). The Host Density Taylor River site is dominated by CG A on Density of Arabis was measured in ran- the northside of the site but is occupied by domly placed 1 m2 plots (15-48 per site in a combination of CG A, B, and D on the 1990). Density of the grass Koleria nitida south side, perhaps indicatingthe existence Nutt. (the primaryhost forP. monoica) was of an environmentalor historicalgradient. estimated by counting the number of re- Gold Creek is by far the most heteroge- productive stems encounteredin springof neous, with the genotypes spread fairly 1991 in each of the eightrandom transects evenly throughout. at each site (the reproductivestems were Disease incidence (as measured by per- produced during the fall of 1990). Repro- cent of individuals infected)varied signifi- ductiveindividuals of Koleria werecounted cantly among sites, clones, and clonal instead of all plants because it is extremely groups.At the population level, the disease difficultto identifynonflowering grasses; incidences at Cement Creek and Taylor thus, this measure is probably an underes- River were similar (7.5 ? 1.9 and 10.9 ? timate of actual density. 2.2 respectively)and significantlylower than

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8

A. C

CementCreek Taylor River B. D. 2 /3 4 6 7/ 8

1 2 34 56789 014 1 0010| 0001| C"B

CG "A' CG"B" OCG"C"

0CG "E" CG "F" COG'G E CGUNK

E

GoldCreek F. 2~~~~ '~~~~

FIG. 2. Maps showing distributionof clonal groups (CGs) in the random sample of 200 Arabisholboellil fromeach of threesites. (A) Cement Creek CGs (B) map of transectsat Cement Creek, (C) Taylor River CGs, (D) map of transectsat Taylor River, (E) Gold Creek CGs, (F) map of transectsat Gold Creek. The placement of transectsand their relationshipto each other within a site are only approximatelyto scale. The number identifyingeach transectis placed at theend of the transectcorresponding to the top of thegenotype "histogram" shown foreach transect.

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions PATTERNS OF RUST INFECTION 117

TABLE 2. Clone frequency(estimated percentof all individuals sharingidentical electrophoreticphenotypes, see methods for calculation) and their associated disease incidence (percentinfection) at three sites in 1990. Only the clones shared among all threesites are shown. Standarderrors calculated withGaussian erroranalysis.

Site Clone Variable Cement Creek Taylor River Gold Creek

A3 Clone frequency 2.1 ? 1.1 6.1 ? 1.8 13.3 ? 2.4 Disease incidence 2.6 ? 2.3 10.2 ? 6.2 34.9 ? 8.0 B4 Clone frequency 4.2 ? 1.3 4.0 ? 1.5 24.3 ? 3.5 Disease incidence 27.9 ? 10.1 0.0 1.8 ? 1.3 C1 Clone frequency 0.1 4.8 ? 1.5 5.1 ? 1.9 Disease incidence 100.0 17.2 ? 9.1 0.0 DI Clone frequency 84.6 ? 2.5 5.9 ? 1.3 10.78 ? 2.5 Disease incidence 5.5 ? 1.4 49.2 ? 12.9 10.3 ? 4.8 F2 Clone frequency 2.1 ? 0.8 1.0 ? 0.8 5.2 ? 1.8 Disease incidence 52.0 ? 19.2 0.0 12.6 ? 7.5 Population disease incidence 7.5 ? 1.9 10.9 ? 2.2 16.8 ? 2.7 disease incidence at Gold Creek (16.8 + high densitypopulation. Infectionby Puc- 2.7). Contraryto prediction,disease inci- cinia thlaspeos,the autoecious rust(passes dence was lowest at the site with the least fromArabis to Arabis), was highestat the geneticdiversity (Cement Creek,see Tables Cement Creek sitewhere Arabis density was 1, 2, 3, and Fig. 2). Disease incidence was highest,and infectionby P. monoica, the low at the Cement Creek site because the heteroecious rust (passes from Koleria to site was dominated by one common clone, Arabis), was highestwhere Koleria was the D1 (85% of the sample), whichhad low dis- most dense (Gold Creek). ease incidence (5.5%) (Table 2, Fig. 2). There was substantial variation for dis- DIsCUSSION ease incidence among the five clones that occurred at all three sites, rangingfrom 0 Prediction1: Clones Occurringat to 52% (Table 2). At the level of clonal a Low FrequencyShould Have groups,the incidence of infectionalso var- a Fitness Advantage ied stronglywithin and among sites (Table A widespread assumption about geno- 3). The factthat disease incidenceper clone types within populations undergoingfre- (or CG) was not consistentacross sites (Ta- quency-dependentselection is that com- bles 2 and 3) suggeststhat there were dif- mon genotypeswill sufferdisproportionately ferencesin the compositionof the pathogen more than rare ones (Haldane, 1949; Wil- populations presentat the sites, or that the liams, 1975; Maynard Smith, 1978; Ham- same electrophoreticphenotype may have ilton,1980; Livelyet al., 1990). In thisstudy, differentresistance phenotypes among dif- however, disease incidence within a clone ferentpopulations. Contrary to expectation, was not typicallyproportional to thatclone's however, within populations there was no frequencywithin a population(Tables 2 and apparentrelationship between a clone's fre- 3). For example, clone Dl was the most quency and its disease incidence (Spear- common clone at Cement Creek (84.6 + man's rank correlation,r = -0.096, P > 2.5% of the population), but experienced 0.5; chi-square test of independence,x2, 1 very low disease incidence (5.5 ? 1.4%), df= 0.277, P = 0.599). whereas clone F2 was rare, and its disease incidencewas much higher(52.0 ? 19.2%). Host Density One hypothesisto explain this patternis Host densityvaried stronglyamong sites that common clones (e.g., Dl at Cement (Table 4), with Cement Creek having 8 to Creek) are common preciselybecause they 16 timeshigher Arabis density than the oth- have thus far escaped infection, either er two sites. Contraryto expectation,pop- throughresistance or througha fortuitous ulation disease incidence was lowest in this lack of a pathogen population capable of

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions 118 B. A. ROY

TABLE 3. Frequency (estimated percentwithin a population, see methods for calculation) of clonal groups (CGs; CGs are formed by pooling clones with similar electrophoreticbanding patterns,see Appendix and methods) and associated disease incidence (frequencyof infection)at three sites in 1990. Standard errors calculated with Gaussian erroranalysis.

Site Clone Variable Cement Creek Taylor River Gold Creek A CG frequency 2.1 ? 1.1 50.7 ? 3.7 17.2 ? 2.7 Disease incidence 2.6 ? 2.3 9.8 ? 2.5 32.1 ? 6.9 B CG frequency 4.2 ? 1.3 16.1 ? 2.9 45.5 ? 3.9 Disease incidence 27.9 ? 10.1 1.3 ? 1.3 14.1 ? 3.1 C CG frequency 0.1 4.8 ? 1.5 5.1 ? 1.9 Disease incidence 100.0 17.2 ? 9.1 0.0 D CG frequency 84.6 ? 2.5 8.9 ? 1.9 14.8 ? 2.9 Disease incidence 5.5 ? 1.4 32.6 ? 9.4 7.4 ? 3.5 E CG frequency 1.0 ? 0.8 9.3 ? 2.2 8.1 ? 2.3 Disease incidence 0.0 8.9 ? 4.7 5.5 ? 4.0 F CG frequency 2.1 ? 0.8 1.0 ? 0.8 5.2 ? 1.8 Disease incidence 52.0 ? 19.2 0.0 12.6 ? 7.5 Population disease incidence 7.5 ? 1.9 10.9 ? 2.2 16.8 ? 2.7 infectingthem. Theoretically,over time, a Simms, 1992), or have shown that other pathogen strain virulent on the common factorssuch as selectionon characterslinked clone will be selected forand the common to resistancecan controlvariation in resis- clone will become heavily infectedand de- tance (Parker, 1988a), or that morpholog- cline. Across a largenumber of clones, how- ical defensescan be adequate to protectge- ever, the expectation is that, on average, neticallyuniform clonal populations from there should be a positive relationshipbe- disease (Parker, 1988b). There are, how- tween clone frequencyand disease inci- ever,two studysystems involving clonal or- dence. In theseclonally diverse populations ganisms where there is evidence for fre- therewas not a consistentpositive relation- quency-dependentpathogen attack. Lively ship between clone frequencyand disease et al. (1 990) surveyedparasite loads in clon- incidence (Tables 2 and 3, Spearman's r = al and sexual genotypesof fishin the genus -0.096, P > 0.5). Poeciliopsisin threepopulations. In thissys- Studies of selectionby pathogensin nat- tem, incidence of was strongly ural populations are relativelyfew in num- correlatedwith genotypecommonness. In ber. Many ofthese studies have not detected clonal plants, the best direct evidence for frequency-dependence(e.g., Parker, 1989; pathogen-mediated frequency-dependent Alexander, 1991, reviewed in Fritz and selection comes from an example of bio-

TABLE 4. Density (individuals m-2 ? SE) of the hosts Arabis holboelliiand Koleria nitida and percentof A. holboelliiinfected by the two rustspecies, P. monoica and P. thlaspeos,at threesites. Puccinia monoica requires Koleria to complete its life cycle,whereas P. thlaspeosoccurs only on Arabis.

Site Species Variable Cement Creek Taylor River Gold Creek Arabisa Density 16.1 ? 5.4 2.2 ? 0.4 2.1 ? 0.3 Arabis % infectedbby P. thlaspeos 29.07 ? 2.73 15.79 ? 5.95 2.78 ? 2.70 Koleriac Density 0.01 ? 0.001 0.32 ? 0.01 1.21 ? 0.04 Arabis % infectedby P. monoica 70.93 ? 2.73 84.21 ? 5.95 97.22 ? 2.70 Population disease incidence 7.5 ? 1.9 10.9 ? 2.2 16.8 ? 2.7 a Arabis densityis reportedas the mean from 15, 45, and 48 random 1 m2 quadrats per site in 1990, respectively. b % infectionby the differentrusts is calculated fromthe complete sample of infectedplants (see materialsand methods). Individuals that were infectedby both species of rustwere divided evenly betweenthe two rustspecies forthe percentagecalculations. c Koleria densityis reportedas mean fromeight random 25 by I m transectsper site by countingevery reproductive stem encounteredin spring of 1991 (reproductivestems were produced duringthe fallof 1990).

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions PATTERNS OF RUST INFECTION 119 controlwhere a rustfungus (Puccinia chon- a reciprocal transplantexperiment, using drillina) is being used to control an intro- rare and common clones, is currentlyun- duced apomicticweed (Chondrilla). Studies derwayto moreexplicitly test for frequency- in both North America and Australia have dependent selection. shown that disease incidence changes host There are, however, reasons to believe frequency,and thathost frequencydepends thatfrequency-dependent selection may in- on disease incidence (Burdon et al., 1981; deed be weak in this system.One potential Supkoffetal., 198 8). In boththe Poeciliopsis cause for weakened frequency-dependence and Chondrilla studies there were only a are pathogengenotypes that are capable of small number of clones in the host popu- infectingmore than one hostgenotype (Bar- lations. In the study on Poeciliopsis there rett,1988; Parker, 1992). In this situation, were two asexual clones and an unknown even when a host clone becomes rare, dis- numberof sexual genotypes,and forChon- ease incidence on this rare clone may stay drilla there were three clones. Even if fre- high if the pathogen population is main- quency-dependentselection is acting,it may tained by livingon otherhost clones. A con- be much moredifficult to observethe effects currentexperiment under controlled con- in more complex systems,such as the pop- ditions revealed that, although there are ulations described herein. significantdifferences in susceptibilityto Frequency-dependentselection is cyclic Puccinia monoica among clones, all four and dynamic. A resistant host genotype pathogeninoculants tested were capable of should increase in frequencyuntil a patho- infectingall threeclones tested(Roy, 1992). gen race virulenton the resistantclone is The fact that more than one genotype of selected for (or migratesin), then the fre- pathogencan infectmore than one host ge- quency of the clone will decline until it is notypeofArabis holboellii would clearlytend rareenough to "escape" disease, whereupon to weaken frequency-dependence. thenow rarehost genotype will again be able It is also possible that the rust-hostas- to increasein frequency.Because frequency- sociations noted here are not the productof dependent selection is cyclic, and because selectionon resistancealleles. In apomictic in natural populations there may be nu- hosts there is complete linkage disequilib- merous host and pathogen genotypes,fre- rium; thus, it is possible for selection on quency-dependenceis difficultto evaluate other charactersto change resistanceasso- simply by determiningcurrent host fre- ciationsnonadaptively (Burdon, 1985; Bur- quencies and disease incidence as was done don and Miiller,1987; Parker,1988a, 1991). in the currentstudy. In a given population, For example, in wild oats with high levels differenthost genotypesare likelyto be in of linkage disequilibrium due to selfing, differentparts of the frequency-dependent Burdon and Muller (1987) found that dif- cycle. For example, one genotypemay be ferencesin disease resistancewere associ- rarebut stillbe heavilyinfected because the ated with differencesin seed germination. pathogenpopulation has not yetcrashed, or Thus, in wild oats, selectionon germination a common genotypemay be uninfectedbe- would also change resistance.It is possible cause there is not a pathogen genotypein that selectionby pathogenson resistancein the population capable of infectingit. For Arabismay be swamped by selectionon oth- this reason, a descriptivestudy that exam- er traits.Although rust infectiondecreases ines a population at only one point in time fitnessof Arabis (Roy, 1992), it is likelythat can only establish whetheror not disease the overall effectof rustdisease on the pop- varies with host frequency,and cannot de- ulations is small because only a fewpercent termine whether frequency-dependentse- of the plantsin each populationare infected lection is occurring.The consistentlack of at a time. In addition, herbivoryon Arabis correlation between clone frequencyand has the potentialto be a strongerselective disease found in the currentstudy does, forcethan pathogenattack (Roy, 1992, and however,suggest that frequency-dependent unpubl. data) and thereforemay cause cor- selection may be weak or infrequent.Now related,non-adaptive change in rust resis- that it has been established that clones do tance alleles. An experiment in progress varyin frequencywithin these populations, should help estimatethe importanceof her-

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions 120 B. A. ROY bivoryand its interactionwith infection on abis. For infectionby P. monoica to become selectionin Arabis. common in Arabis populations,there must be sufficientdensity of both theprimary and Prediction2: Populations with alternate hosts to facilitate transmission. More Variationin Resistance Frequency-dependentselection could occur Should Have Less Disease at any density,but it is unlikelyfor popu- Than More UniformOnes lation level disease incidence of Puccinia The prevalentview in both evolutionary monoica to be high,regardless of the level biologyand plant pathologyis thathost ge- ofgenetic diversity in thepopulation, unless netic diversityis one of the primaryforces the densityof both host species is high.The controllingdisease incidencein populations relationship between infection and host (Adams et al., 1971; Browning,1974; Har- densityis likelyto be positive forthe other lan, 1976; Bremermann,1980, 1983; Bur- rust,P. thlaspeos,because Arabis is its only don and Shattock,1980; Barrett,1981, 1988; host, and indeed, the densest population of Alexander, 1988; for a review including Arabis(Cement Creek) supportedthe largest negativeevidence see Kranz, 1990). In the population of thisless common rustspecies presentstudy only three populations were (Table 4). characterizeddue to the amount of work involved in determiningmultilocus geno- Clonal Diversity types;thus, it is difficultto generalizeabout The numberof genotypesper population the effectsof host geneticdiversity on pop- uncovered in this study ranged from 6 to ulationlevel disease incidence.Although not 27 (Table 1) and the average number,18.3, conclusive, it is interestingthat the pre- was slightlyhigher than the average of 16.1 dicted relationshipbetween genetic diver- reportedby Ellstrandand Roose (1987) for sity and disease incidence was not borne several species and populations of asexual out in any of the populations, that is, high plants. What is the originof the clonal vari- clonal diversitywithin populations was not ation in these populations? In an earlier associated withdecreased disease incidence. study (Roy and Rieseberg, 1989), electro- For example,the CementCreek population, phoresisof progenyarrays from Arabis hol- which had the lowest overall (both patho- boellii collected at the Cement Creek site gens combined) disease incidence (7.5 + revealed no recombination,suggesting that 1.9%) had the smallest number of clones the clonal variation uncovered there may (six), the least amount of clonal geneticdi- have resulted from polyphyleticorigin of versity(Table 1), the lowest spatial hetero- the apomicts fromsexual parentscombin- geneity(Fig. 2), and the highestdensity (Ta- ing differentalleles. The currentsurvey of ble 4). By comparison, Gold Creek had 22 two otherpopulations uncovered several in- intermingledclones, much lower Arabis dividuals thatmay be sexual (Appendix,CG density,but significantlyhigher disease in- "E"). Their banding patternsare more ho- cidence (1 6.8 ? 2.7%). mozygous than the typical fixedheterozy- gote patternfor apomictic plants, and one The Relationshipof Host Density group,CG "G," appear to be heterozygotes to Disease Incidence combiningdifferent "E" patterns.It would Host densityis clearlyan importantand not be too surprisingfor sexual individuals complex factorin theArabis-Puccinia path- to occur at the Gold Creekand Taylor River osystem.It is unusual forthere to be a neg- sites,as diploids are rarelyapomictic (Bier- ative relationshipbetween host densityand zychudek, 1985; Asker and Jerling,1992) disease as was foundat the population level and I have made diploid chromosome in this study of Arabis (Burdon and Chil- countsfrom Taylor River plants(Roy, 1992 vers, 1982; Burdon et al., 1989). However, and unpubl. data). Other workershave re- the amount of disease caused by the most ported occasional sexual plants of A. hol- common ruston Arabis,Puccinia monoica, boellii on the basis of chromosomenumber is clearlyrelated to densityof the primary and embryology(Bocher, 1951, 1969), and host, Koleria, fromwhich the inoculum is on the basis of putativehybrids with other received,and not simplythe densityof Ar- species (Rollins, 1983). Because pollen of

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions PATTERNS OF RUST INFECTION 121 proven apomicts retains high viability at suggestions of P. Bierzychudek, N. Ell- these sites(Roy, unpubl. data), it is possible strand, J. W. Kirchner, L. Rieseberg, R. that apomictic pollen can fertilizesexual Scogin, M. Stanton,and reviewersfor Evo- plants and give rise to both sexual and ap- lution. This researchwas supportedby an omictic progeny (Nogler, 1975; Asker, NSF Dissertation Improvement Grant 1979). Progenytests of the putative sexual (BSR-900970), a Grant-In-Aidof research plants are necessary to determine their from Sigma Xi, a Lee Snyder Grant from breedingsystems. If any of these more ho- theRocky Mountain Biological Laboratory, mozygous plants are sexual, then thereare and a Fellowshipfrom the Claremont Grad- somewhat fewerapomictic clones than re- uate School. ported in Table 1 (CGs E and G are the LITERATURECITED most likely sexual genotypesand account ADAMS,M. W., A. H. ELLINGBOE, AND E. C. RossMAN. for 9 of the 41 clones). 1971. Biological uniformityand disease epidem- ics. Bioscience 21:1067-1070. CONCLUSIONS ALEXANDER,H. M. 1988. Spatial heterogeneityand disease in naturalplant populations,pp. 144-164. The studypopulations are clearlydissim- In M. J. Jeger(eds.), Spatial Components of Epi- ilar; the primaryhost of one of the patho- demics. Prentice-Hall,N.Y., USA. gens is nearly absent at one site (Cement . 1991. Plant population heterogeneityand Creek) yetquite abundant at another(Gold pathogenand herbivorelevels: A fieldexperiment. Creek), differenthost clones dominate the Oecologia 86(1):125-131. ANDERSON, R. M., AND R. M. MAY. 1979. Population various sites(Fig. 2), and the pathogenpop- biology of infectiousdiseases: Part I. Nature 280: ulations are also apparently variable in 361-367. composition and abundance (Tables 2 and ANONYMOUS. 1960. Index of Plant Diseases in the 3). In some cases, host clones appear to be United States. Handbook No. 165. U.S. Depart- a lack of ment of Agriculture,Washington, DC USA. disease-free,suggesting pathogen ANTrONOvIcS, J., AND N. C. ELLSTRAND. 1984. Ex- genotypes capable of infectingthem. Al- perimentalstudies of the evolutionarysignificance though this study does not eliminate the of sexual reproduction.I. A test of the frequency- possibility of frequency-dependentpatho- dependentselection hypothesis. Evolution 38:103- the lack of correlationbetween 115. gen attack, ARTHUR,J. C. 1962. Manual of the Rusts in the clone frequencyand disease incidencewith- United States and Canada. HafnerPublication Co., in populations does suggestthat it may be N.Y., USA. weak or intermittent.If the environmentis ASKER,G. A. 1979. Progressin apomixis research. sufficientlyheterogeneous in space and time, Hereditas 91:231-240. AsKER, S. E., AND L. JERLING. 1992. Apomixis in due to a combination of, for example, in- Plants. CRC Press, Ann Arbor, MI USA. termittentpathogen attack, historical rea- BARRETT,J. A. 1981. The evolutionaryconsequences sons, and physicalattributes, then different of monoculture,pp. 209-248. In J. A. Bishop and clones could have differentialsuccess often L. M. Cook (eds.), Genetic Consequences of Man- maintain popula- Made Change. Academic Press, London, UK. enough to polymorphic 1988. Frequency-dependentselection in plant- tions (Maynard Smith, 1978; Falconer, fungalinteractions. Philos. Trans. R. Soc. London 198 1). Alternatively,the likelihood of sex- B 319:473-482. ual reproductionin two ofthese populations BARTSCH,H. J. 1974. Handbook of Mathematical suggeststhe possibilitythat clonal diversity Formulas. Academic Press, N.Y., USA. BEGON, M., J. L. HARPER, AND C. R. TOWNSEND. 1986. may be maintained by continuous genera- Ecology: Individuals, Populations and Communi- tion of new clones and not, or only second- ties.Sinauer Associates Inc., Sunderland,MA USA. arilyso, by selection,pathogen-mediated or BELL,G. 1982. The Masterpieceof Nature:The Evo- otherwise. lutionand Geneticsof Sexuality.University of Cal- iforniaPress, Berkeley,CA USA. Patterns in parthe- ACKNOWLEDGMENTS BIERZYCHUDEK, P. 1985. plant nogenesis. Experientia41:1255-1264. I thank L. Rieseberg forthe use of labo- B1OCHER,T. W. 1947. Cytological studies of Arabis ratoryfacilities, J. W. Kirchner for assis- holboellii.Hereditas 32:573. with the erroranalysis, P. Lehr and 1951. Cytologicaland embryologicalstudies tance in the amphi-apomicticArabis holboelliicomplex. A. Mears for housing duringthe fieldsea- Dan. Vid. Biol. Skr. 6(7):1-58. son, and H. Renkin forhelp withfield work. 1969. Furtherstudies in Arabisholboel/ii and The manuscriptwas much improvedby the allied species. Sv. Bot. Tidskr. 48(l):31-44.

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions 122 B. A. ROY

BREMERMANN, H. J. 1980. Sex and polymorphismas LIVELY,C. M., C. CRADDOCK,AND R. C. VRIJENHOEK. strategiesin host-pathogeninteractions. J. Theor. 1990. Red queen hypothesissupported in sexual Biol. 87:671-702. and clonal fish.Nature 344: 864-866. 1983. Theoryof catastrophicdiseases of cul- MAYNARD SMITH, J. 1978. The Evolution of Sex. tivated plants. J. Theor. Biol. 100:255-274. Cambridge UniversityPress, Cambridge,UK. BROWNING, J.A. 1974. Relevance of knowledgeabout NOGLER,G. A. 1975. Genetics of apospory in Ra- natural ecosystemsto development of pest man- nunculusauricormis IV. Embryologyof F3 and F4 agementprograms for agro-ecosystems. Proc. Am. backcrossoffspring. Phytomorphology 25:485-490. Phytopathol.Soc. 1:191-199. NYGREN,A. 1954. Apomixis in theangiosperms. Bot. BURDON, J. J. 1985. Pathogensand the geneticstruc- Rev. 20:577-649. tureof plant populations,pp. 313-325. In J. White PARKER,M. A. 1988a. Disequilibrium betweendis- (eds.), Studies on Plant Demography. Academic ease-resistancevariants and allozyme loci in an an- Press, N.Y., USA. nual legume. Evolution 42:239-247. 1987. Diseases and Plant Population Biology. 1988b. Genetic uniformityand disease re- Cambridge UniversityPress, Cambridge,UK. sistance in a clonal plant. Am. Nat. 132:538-549. BURDON, J. J., AND G. A. CHILVERS. 1982. Host den- . 1989. Disease impact and local geneticdi- sityas a factorin plantdisease ecology.Annu. Rev. versityin the clonal plant Podophyllumpeltatum. Phytopathol.20:143-166. Evolution 43:540-547. BURDON, J. J.,R. H. GROVES, AND J. M. CULLEN. 1981. 1991. Nonadaptive evolution of disease re- The impactof biologicalcontrol on thedistribution sistance in an annual legume. Evolution 45:1209- and abundance of Chondrillajuncea in south-east- 1217. ern Australia. J. Appl. Ecol. 18:957-966. 1992. Disease and plant population genetic BURDON, J. J.,A. M. JAROSZ, AND G. C. KIRBY. 1989. structure,pp. 345-362. In R. S. Fritz and E. L. Patternand patchiness in plant-pathogeninterac- Simms (eds.), Plant Resistance to Herbivores and tions-Causes and consequences. Annu. Rev. Ecol. Pathogens: Ecology, Evolution and Genetics. Uni- Syst. 20:119-136. versityof Chicago Press, Chicago, IL USA. BURDON, J. J., AND W. J. MULLER. 1987. Measuring PEET, R. K. 1974. The measurementof species di- the cost of resistanceto the pathogenPuccinia cor- versity.Annu. Rev. Ecol. Syst. 5:285-307. onata in Avena fatua. J. Appl. Ecol. 24:191-200. PETERSEN, R. H. 1974. The rustfungus life cycle. Bot. BURDON, J. J., AND R. C. SHATTOCK. 1980. Disease Rev. 40:453-513. in plant communities.Appl. Biol. 5:145-219. RICHARDS, A. J. 1986. PlantBreeding Systems. George Buss, L. 1985. The uniqueness of the individual re- Allen and Unwin Ltd., London, UK. visited, pp. 467-505. In J. B. C. Jackson, L. W. ROLLINS,R. C. 1941. A monographicstudy of Arabis Buss, and R. E. Cook (eds.), Population Biology in westernNorth America. Rhodora 43:289-325, and Evolution of Clonal Organisms.Yale Univer- 348-411, 425-481. sityPress, New Haven, CT USA. 1983. Interspecifichybridization and taxon COCHRAN,W. G. 1963. SamplingTechniques. Wiley uniformityin Arabis(Cruciferae). Am. J.Bot. 70(4): and Sons, N.Y., USA. 625-634. CUMMINS,G. B., AND Y. HIRATSUKA. 1983. Illus- Roy, B. A. 1992. Genetic diversity,disease inci- tratedGenera of Rust Fungi. American Phytopath- dence, and pathogen-mediatedselection in apo- ological Society,St. Paul, MN USA. mictic Arabis. PhD Diss. Claremont Graduate ELLSTRAND, N. C., AND M. L. ROOSE. 1987. Patterns School. of genotypicdiversity in clonal plant species. Am. Roy, B. A., AND L. H. RIESEBERG. 1989. Evidence J. Bot. 74:123-13 1. forapomixis in Arabis. J. Hered. 80(6):506-508. FALCONER, D. S. 1981. Introductionto Quantitative SAVILE,D. B. 0. 1974. Some new or poorly known Genetics. Longman Scientificand Technical,Essex, rustsof Brassicaceae. Can. J. Bot. 52:1501-1507. UK. SEBENS,K. P., ANDB. L. THORNE. 1985. Coexistence FARR, D. F., G. F. BILLS, G. P. CHAMURIS, AND A. Y. of clones, clonal diversityand the effectsof distur- ROSSMAN. 1989. Fungi on Plants and Plant Prod- bance, pp. 357-398. In J.B. C. Jackson,L. W. Buss, ucts in the United States. American Phytopatho- and R. E. Cook (eds.), Population Biologyof Clonal logical Society Press, St. Paul, MN USA. Organisms.Yale UniversityPress, New Haven, CT FRITZ, R. S., AND E. L. SIMMs (eds.) 1992. Plant Re- USA. sistance to Herbivores and Pathogens: Ecology, SUPKOFF,D. M., D. B. JOLEY,AND J. J. MAROIS. 1988. Evolution and Genetics. University of Chicago Effectof introduced biological control organisms Press, Chicago, IL USA. on the densityof Chondrillajuncea in California. HALDANE, J. B. S. 1949. Disease and evolution. La J. Appl. Ecol. 25:1089-1095. Ricerca Scientifica19 (suppl.):68-76. TIBAYRENC, M., AND F. J. AYALA. 1988. Isozyme HAMILTON, W. D. 1980. Sex versus non-sex versus variabilityin Trypanosomacruzi, the agentof Cha- parasite. Oikos 35:282-290. gas' disease: Genetical, taxonomical, and epide- HARLAN, J. R. 1976. Diseases as a factorin plant miological significance.Evolution 42:277-292. evolution. Annu. Rev. Phytopathol.14:31-51. WILLIAMS, G. C. 1975. Sex and Evolution. Princeton JAENIKE, J.,E. D. PARKER, AND R. K. SELANDER. 1980. UniversityPress, Princeton,NJ USA. Clonal niche structurein theparthenogenetic earth- worm Octolasion tyrtaeum.Am. Nat. 116(2):196- CorrespondingEditor: M. Parker 205. KRANZ, J. 1990. Fungal diseases in multispeciesplant communities.New Phytol. 116:383-405.

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions PATTERNS OF RUST INFECTION 123

APPENDIX Description of clonal groups (CGs) and clones, includinginformation on theirnumbers at the threesites. Each clone has a two part identifier,a letterfor the clonal group to which it belongs (definedby banding patternat PGI and LAP) and a unique number for individual clones withineach clonal group. The lettersunder each allozyme locus representthe bands present for each clone (the bands include heteromersas well as alleles). Because these banding patternsoften represent unbalanced heterozygotes(due to polyploidy),they are coded to reflectdosage. If the band was strong,then the letteris underlined.UNK means unknown.

Allozyme locus Site Clone PGI 2 and 3 IDH 2 TPI 1 TPI 2 LAP 1 LAP 2 Cement Taylor Gold Clonal group A A 9 abc a ab ac cde ab 0 37 0 A 11 abc a ab c cde ab 0 1 0 A 23 abc a abc abc cde ab 0 0 3 A 10 abc a abc bc cde ab 0 4 0 A 12 abc a abc c cde ab 0 2 0 A4 abc abc a bc cde ab 0 2 0 A 22 abc abc abc a cde UNK 0 3 0 A 25 abc abc abc abc cde ab 0 1 0 A 2 abc abc abc ac cde ab 0 6 0 A 3 abc abc abc abc cde ab 6 16 41 A 26 abc abc abc abc cde ab 0 0 1 Al abc abc abc bc cde ab 0 21 5 A 24 abc abc ab ac cde ab 0 1 0 A 14 abc abc abc ac cde ab 0 2 0 A 15 abc abc abc abc cde ab 0 9 0 A 17 abc abc abc bcc cde ab 0 17 0 A 20 abc c a ac cde ab 0 1 0 Total A 6 123 50 Clonal group B B 9 cde a abc abc cde b 0 0 3 B 11 cde a abc abc cde b 0 0 12 B 4 cde a c bc cde b 51 8 53 B 8 cde abc abc abc cde b 0 1 0 B cde abc abc abc cde b 0 5 30 B 10 cde abc abc abc cde b 0 0 5 B 6 cde abc c bc UNK UNK 0 0 1 B 7 cde abc abc abc cde b 0 14 0 B 3 cde abc abc bc cde b 0 0 1 Total B 51 27 106 Clonal group C C I ace abc abc abc cf ab 3 14 11 Clonal group D D 5 bcde a abc abc bce b 0 4 0 D 4 bcde a ac abc bce b 0 2 8 D 1 bcde abc abc abc bce b 346 23 25 D 3 bcde abc abc abc bce b 0 1 0 Total D 346 30 33 Clonal group E E 7 a a a abc e b 0 6 3 E2 a a a bc e b 0 16 0 E 8 a a a bc e a 0 0 1 E 6 a c c bc e b 2 0 0 E 9 abc abc abc abc e a 0 0 1 E 4 abc abc abc abc ce a 0 0 9 Total E 2 23 14

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions APPENDIX. Continued.

Allozyme locus Site Clone PGI 2 and 3 IDH 2 TPI I TPI 2 LAP 1 LAP 2 Cement Taylor Gold

Clonal group F F 2 abcde a abc bc de ab 47 2 14 Clonal group G G I ac a abc abc cde b 0 0 1 G 2 ac a abc abc e a 0 0 1 Total G 0 0 2 Grant total 455 219 230

This content downloaded from 128.223.93.178 on Wed, 29 Jan 2014 15:38:41 PM All use subject to JSTOR Terms and Conditions