Differentiation in populations of Hordeum spontaneum Koch ... · and the organism’s ability to accurately assess envi-ronmental variation (Levins, 1968); and (2) on biolog-ical

*Corresponding author. E-mail: [email protected]†Current address: Department of Nature Conservation,Stellenbosch University, P. Bag X01, Matieland 7602, SouthAfrica

Biological Journal of the Linnean Society, 2002, 75, 301–312. With 4 figures

INTRODUCTION

Studies of evolution in a heterogeneous environmentusually deal with different aspects of a reaction norm, a set of phenotypes produced by a genotype

in response to different environments (Schmalgausen,1949). The genetics, ecology and evolution of reactionnorms, and its main attribute, phenotypic plasticity,have been intensively studied (Bradshaw, 1965;Schlichting, 1986; Sultan, 1987; Thompson, 1991;Scheiner, 1993). Phenotypic plasticity, defined as thedegree and direction of departure of the reaction normfrom a parallel to the environmental axis flat line(Lewontin, 1974; Pigliucci et al., 1995), depends on: (1)the environmental range encountered by a genotype

© 2002 The Linnean Society of London, Biological Journal of the Linnean Society, 2002, 75, 301–312 301

Differentiation in populations of Hordeum spontaneumKoch along a gradient of environmental productivityand predictability: plasticity in response to water andnutrient stress

SERGEI VOLIS1,2*, SAMUEL MENDLINGER1 and DAVID WARD2†

1The Institutes for Applied Research, Ben-Gurion University of the Negev, POB 653, Beer Sheva 84105,Israel2Mitrani Department for Desert Ecology, Blaustein Institute for Desert Research, Ben-GurionUniversity of the Negev, Sede Boqer 84990, Israel

Received 4 June 2001; accepted for publication 2 November 2001

Plants from four populations of Hordeum spontaneum originating in distinct environments of Israel were comparedfor stress induced phenotypic plasticity. The environments ranged along a gradient of increasing rainfall amountand predictability from low (desert) to moderate (semisteppe batha) to high (Mediterranean grassland and moun-tain, the latter also experiencing frost stress). The plants were exposed to a set of four treatments: no stress(optimum water and nutrients), water, nutrient and both water and nutrient stress. Plants from the four popula-tions (or ecotypes) exhibited different patterns of plasticity in response to the different stresses (water and nutri-ents) and in different trait categories (reproductive, fitness and resource allocation). The importance of plasticityin response to water stress appears to decrease, and to nutrient stress appears to increase along the increasingrainfall gradient. The mountain ecotype, growing in an area with high potential productivity (amount of rainfall)but experiencing periodic frosts, was the most plastic among ecotypes in resource allocation under both water andnutrient stress, but exhibited low plasticity in other trait categories. In contrast, the desert ecotype had low plas-ticity in resource allocation under water stress and the lowest plasticity among the four ecotypes in all trait cate-gories in response to nutrient stress. The ecotype originating in Mediterranean grassland, a predictable and mostfavourable environment, was highly plastic in fitness and allocation traits in response to low nutrient levels whichis likely to occur due to competition in productive environment. We discuss the observed differences in ecotype plas-ticity as part of their environmentally induced adaptive ‘strategies’. We found no support for the hypothesis thatplants originating in environments with greater variation and unpredictability are more plastic. © 2002 TheLinnean Society of London, Biological Journal of the Linnean Society 2002, 75, 301–312.

ADDITIONAL KEYWORDS: canonical discriminant analysis – ecotype – local adaptation – reaction norm –stress – plant strategy – plasticity.

and the organism’s ability to accurately assess envi-ronmental variation (Levins, 1968); and (2) on biolog-ical constraints preventing or imposing too high a costto plasticity (Bradshaw, 1965). Reaction norm experi-ments test the genotype plasticity in response to asingle factor, a combination of factors or across naturalenvironments. The use of either experimentallymanipulated or natural environmental conditions has both advantages and drawbacks. The advantage of asingle or a few factor reaction norm experiment comesfrom its analytical power and ability to classify envi-ronmental factors by their importance for genotypesurvival. However, any given natural environment isa combination of biotic and abiotic effects and selectsfor genotypes adapted not to a single or several factorsbut to the whole set of factors, including their directand indirect interactions. Therefore, the adaptive sig-nificance of plasticity should ultimately be tested byexamining a genotype in a series of natural environ-ments including its indigenous one. However, for thequestion of whether plants from different populationenvironments differ in amount and direction of plasticity in specific traits as well as over all traits,experimentally manipulated conditions are moreappropriate.

In this paper, we compare phenotypic plasticity infour populations of Hordeum spontaneum in responseto experimentally induced environmental stresses. A stressed environment may reveal functional trade-offs and evolutionary trends that otherwise are con-cealed under favourable conditions (Pigliucci et al.,1995). We used two stress-inducing factors, water and nutrients, as well as their combination against an ‘optimal’ control. The assessment of phenotypicplasticity was carried out in two ways: (1) in responseto a single factor (either water or nutrients while thesecond factor was optimal), i.e. in ordered treatments;and (2) to a set of unordered treatments, i.e. with-out a priori ranking of the conditions. As Pigliucci et al. (1995) states, plastic response to unordered vs.ordered environments may have more biological sense,as the former may better reflect the complexity ofabiotic constrains on phenotypic evolution than thelatter.

A study of populations of the same species distrib-uted along clear environmental gradients, e.g. amountof annual rainfall, is potentially a powerful tool toidentify environmentally induced effects of naturalselection and resulting plant strategies. Wild barley is native to a wide range of environments in Israel,including extremes such as deserts and mountains.Recently, we reported of locally adapted ecotypes thatdiffer in life history and phenotypic traits (Volis, 2001). The present study investigates the role of plasticity as part of plant strategies in response towater and nutrient stress and their combined effect.

MATERIAL AND METHODS

STUDIED SPECIES AND CHOICE OF POPULATIONS

Wild barley, Hordeum spontaneum Koch, is a winterannual, predominately selfing grass (Harlan & Zohary,1966). In Israel, despite its mainly Mediterranean andIrano-Turanian distribution in steppe-like formations,wild barley penetrates into desert (<200mm annualrainfall) and mountain (up to 1600m elevation) envi-ronments where it has stable populations. The popu-lations selected here represent four environments in order of increasing rainfall: desert (Sede Boqer,hereafter SB), batha (Beit Guvrin, BG), grassland(Ammiad, AM) and mountain (Mount Hermon, MH)with 90, 400, 600 and 1600mm of rainfall per annum,respectively. In addition, in desert and mountain localities plants are exposed to contrasting stresses(drought and frosts).

EXPERIMENTAL DESIGN

Fifteen mother plants from each population (referredhere as ‘ecotype’) were planted under uniform condi-tions in a greenhouse and their offspring used in thisexperiment. The offspring of each mother plant can be considered genetically identical as wild barley is predominantly autogamous (98% or more; Nevo et al.,1979) and can be considered as a single genotype.Between 10 and 15 seeds from each mother plant were sown in Petri dishes. After emergence, they were transplanted into plastic trays and at the two-leaf stage (at 2 weeks) four randomly chosenseedlings/mother plant were transferred to 10-litrepots containing terra rossa soil with a single plant/pot.Each of the four plants per genotype per ecotype wassubjected to one of four treatments, comprising a totalof 240 plants.

The experiment was carried out in the EcologicalGrowth Facility at the Institute for Desert Researchat the Sede Boker Campus, in the Negev Desert ofIsrael. The facility allowed close simulation of naturalgrowing conditions by staying open at all times withthe exception of nights, during rainfall, or when thetemperature fell below 5°C. The terra rossa soil wasleached prior to the experiment and its organic nitro-gen content was only 1.02 ± 0.17mg·kg-1, which can beconsidered to be nitrogen deficient.

The four treatments were:

(1) High level of water and nutrients (HH): amount ofwater equivalent to 500mm of rainfall during thegrowing season applied as 1.5L of water once perweek (which is more or less in accord with thenatural pattern of precipitation in the Mediter-ranean region), 10g of slow-release fertilizer at the

302 S. VOLIS ET AL.

© 2002 The Linnean Society of London, Biological Journal of the Linnean Society, 2002, 75, 301–312

beginning of experiment and 100mg of 20 :20 :20(NPK) weekly.

(2) High level of water and low level of nutrients (HL):the same water treatment as in HH but with nonutrients added.

(3) Low level of water and high level of nutrients (LH):the same nutrient treatment as in HH but amountof water equivalent to 150mm of rainfall duringthe growing season applied as 100mL of wateronce per week plus 1.5L once per month.

(4) Low level of water and nutrients (LL): the samewater treatment as in LH and the same nutrienttreatment as in HL.

After the appearance of reproductive tillers, the firstthree reproductive tillers of each plant were taggedwith coloured tags representing the first, second andthird tiller in order of awn appearance and the fol-lowing traits were measured: days to awn appearance,days to anthesis, tiller height, number of nodes pertiller, flag leaf length and width, penultimate leaflength and width, awn length, spike length, numberof spikelets per spike. At senescence, we measured thenumber of spikes per plant, yield (total number ofspikelets produced) per plant, reproductive biomass,

spikelet weight, root biomass, ratio of root to vege-tative shoot biomass, reproductive effort (ratio ofbiomass of fertile spikelets to the total biomass), percentage of fertile spikelets to the total number offertile and aborted spikelets. The characters were cat-egorized into the four classes: phenology, reproductive,fitness and resource allocation (Table 1).

DATA ANALYSIS

Two types of statistical analyses were used: analysis ofvariance and canonical discriminant analysis. Arepeated measures analysis of variance with onegrouping factor (populations) and one within-groupfactor with unordered levels (treatments) was per-formed to assess univariate environmental, geneticand interaction effects. A genotype was not replicated.This analysis was the most appropriate to test theoverall treatment effects (phenotypic plasticity), popu-lation effects (genetic variation for population traitmeans across treatments) and treatment ¥ popu-lation interactions (genetic variation for plasticity). Inthis analysis, phenotypic plasticity was estimated inresponse to all four unordered treatments (HH, HL,LH and LL) without a priori ranking of the treatments.

PLASTICITY IN HORDEUM SPONTANEUM POPULATIONS 303


Table 1. Repeated measures analysis of variance for the response to four unordered treatmentsin four populations of H. spontaneum. The treatment effect estimates extent of phenotypic plas-ticity and treatment–population interactions indicate the interpopulation genetic variation forplasticity

Source of variation

Trait Type of trait Population Treatment P*T

Days to awn appearance Phenology 17.6*** 31*** 1.7nsDays to anthesis Phenology 19.2*** 27*** 1.4nsTiller height Reproductive 34.4*** 99*** 4.0***No. of nodes Reproductive 11.6*** 1ns 0.6nsFlag leaf length Reproductive 6.2** 40*** 2.5*Penultimate leaf length Reproductive 4.9** 52*** 3.8***Flag leaf width Reproductive 9.7*** 43*** 2.9**Penultimate leaf width Reproductive 5.6** 21*** 1.0nsAwn length Reproductive 1.6ns 6** 1.8nsSpike length Reproductive 22.5*** 95*** 3.2**Spikelets per spike Reproductive 9.5*** 97*** 3.7***No. of spikes Reproductive 3.5* 110*** 1.1nsYield (no. of spikelets) Fitness 14.0*** 204*** 4.2***Reproductive biomass (g) Fitness 1.4ns 201*** 4.1***Spikelet weight (mg) Fitness 26.3*** 16*** 3.4**Root biomass Allocation 23.4*** 60*** 4.1***Root/shoot biomass Allocation 22.4*** 5** 2.5**Reproductive effort Allocation 3.4* 45*** 2.6**Percentage of fertile spikelets Allocation 4.6** 10*** 2.8**

* P < 0.05; ** P < 0.01; *** P < 0.001.

However, in this model, we could not estimate the population responses to a single factor (either water ornutrients). This can be done only by considering thegenotypes within each population as replicates in athree-way analysis of variance. The population originand treatments (water and nutrients) were consideredas fixed effects. The data (except ratios) required notransformation to satisfy the assumptions of the analysis (homogeneity of variance and normality). Theratios were arcsin ÷̄ ¯ transformed.

A canonical discriminant analysis was used to esti-mate the overall plastic response of ecotypes (popula-tions) to water, nutrient and water ¥ nutrient stress.We used discriminant analysis because in this tech-nique new variables are created to maximize the vari-ation between groups relative to the variation withingroups. A key assumption of discriminant analysis is homogeneity of within-group variance–covariancematrices (Williams, 1983). However, even in caseswhen this requirement is not met, the results of discriminant analysis can be used as descriptive andexploratory, without estimation of the significance of group differences (McGarigal et al., 2000). A fewstudies of phenotypic plasticity have successfully usedcanonical discriminant analysis in this manner (Blais & Lechowicz, 1989; Gurevitch, 1992; Zhang &Lechowicz, 1994; Pigliucci et al., 1995; Zhang, 1995).As adaptation to different stresses by plants of differ-ent origins may be achieved by differential trait andtrait complex plasticity, the estimation of overall plasticity in this study was carried out separately forthe three trait categories: reproductive, fitness andallocation (phenology was found to possess no inter-population genetic variation for plasticity). This wascarried out by: (1) calculating the Mahalanobis dis-tances between population centroids at two differentlevels of the same factor; and (2) visual representationof the distances between population centroids in acanonical variable plot where the axes are the first twocanonical variables, and the points are the canonicalvariable scores. The second method allows comparisonof not only the amount of ecotype plasticity but alsothe direction of plasticity in different populations. Inaddition, the canonical scores of populations and treat-ments were calculated separately and plotted againstthe first and second canonical axes to determine thediscrimination in the two main effects.

The relative contribution of different traits to seg-regation of ecotypes under each of four experimentalconditions was tested by stepwise discriminant analy-sis with automatic forward stepping. In this analysis,the variables for which the group means are the mostdifferent are identified and entered in a model step bystep. The limit for F-to-enter, which corresponds to theF for one-way analysis of covariance where the covari-ates are the variables already included, was 4.0. The

stepping stopped when no variable had an F-to-enterabove this limit. All the data analyses were done usingSYSTAT version 7.0 (Systat, 1997).

RESULTS

REACTION NORMS AND PLASTICITY OF TRAITS

A repeated measures analysis of variance showed highphenotypic plasticity in all the traits (except numberof nodes) in response to the four unordered treatments(Table 1). The population means across the four treat-ments were significantly different (i.e. there was sig-nificant genetic variation among populations) in alltraits except awn length and reproductive biomass. Asignificant population * treatment interaction, whichindicates interpopulation genetic variation for plastic-ity, was found in 13 out of 19 traits: all fitness and allocation traits, 6 of 10 reproductive traits, but nophenological trait. No interpopulation variation forplasticity was detected in the phenological traits daysto awn appearance and days to anthesis.

Three-way analysis of variance detected higherplasticity in all traits (except number of nodes) inresponse to water stress when compared to nutrientstress (Table 2). Only the number of nodes was unaf-fected by water stress, although 9 of 19 traits were notaffected by nutrient stress (G-test, 8.6; P < 0.01). Thecombined effect of water and nutrient stress had a sig-nificant impact only on number of spikes, yield andabortion rate. There was significant interpopulationgenetic variation for plasticity in a response to waterstress in three reproductive traits (penultimate leaflength, flag leaf width and spikelets per spike), twofitness traits (yield and reproductive biomass) and allthree allocation traits. However, in response to nutri-ent stress, no significant population * treatment inter-action was found in any allocation traits, althoughsignificant interactions were found in all three fitnesstraits and in four reproductive traits (tiller height,number of spikes, flag and penultimate leaf length).

The population (ecotype) across-treatment reactionnorms (Fig. 1) show the following patterns of plastic-ity in the four trait categories:

(1) Phenology. Plasticity is low and does not differamong ecotypes. Water stress causes a delay in the onset of reproduction in all populations but a combination of water and nutrient stresses causeddelay in MH plants only;

(2) Reproductive traits. The differences in plasticityamong ecotypes were not very pronounced. Theleast affected were the desert (Sede Boqer) plantsfor leaf length and mountain (MH) plants fornumber of spikes;

(3) Fitness. Ecotypes differed in amount of plasticity

304 S. VOLIS ET AL.


but the patterns of plasticity in fitness traits weresimilar in most traits, with a few exceptions. Theplants from the two southern populations (SedeBoqer and Beit Guvrin) (but not Ammiad andMount Hermon) responded to high water lownutrients (HL) by decreasing yield and reproduc-tive biomass, and only AM plants among the fourecotypes increased spikelet weight in response tolow water low nutrients (LL);

(4) Resource allocation. The ecotypes were distinctlydifferent in both amount and pattern of plasticity.AM plants, in contrast to other three ecotypes,increased reproductive effort in response to highwater low nutrients (HL) but did not respond to low water high nutrients (LH). MH plantsincreased root biomass under low water low nutri-ents (LL), which is the opposite direction to thatexhibited by other three ecotypes. MH and AMplants showed higher plasticity and distinct fromSB and BG plants pattern in ratio root/shootbiomass. In percentage of fertile spikelets (abor-tion rate), SB plants showed very low if any plas-ticity under all three stress treatments (HL, LHand LL), BG and MH plants responded under LHand LL treatments, and AM plants respondedunder HL and LL treatments.

PATTERNS OF OVERALL ECOTYPE PLASTICITY

Water stressIn response to water stress (LH vs. HH treatment) thefour ecotypes’ plasticity in reproductive traits weresimilar in both amount (distance between centroids)and pattern (direction of change from HH to LH)(Figs 2 and 3). The SB and BG plants were the mostplastic in fitness traits; however, the plastic responseof the SB ecotype was different from the other threeecotypes. In resource allocation traits, MH plantsexhibited the highest plasticity and the allocationpattern differed in MH, AM vs. SB, BG ecotypes.

Nutrient stressIn both reproductive and fitness trait classes, BG andAM plants had distinctly higher plasticity than SBand MH plants. For resource allocation, the rankingof ecotype plasticity was: SB < BG < AM < MH.

Water ¥ nutrient stressThe amount and patterns of ecotype plasticities weresimilar to those under conditions of water stress alone(Figs 2 and 3).

The differences in plant plasticity as induced by thetwo stresses and their combined effects across all four



Table 2. Analysis of variance for the response to the two fixed treatments, water and nutrients in four populations of H. spontaneum. The effects of water and nutrients estimate extent of phenotypic plasticity and treatment ¥ populationinteractions indicate the interpopulation genetic variation for plasticity

Source of variation

Trait Type of trait Population Water Nutirents W*N P*W P*N

Days to awn appearance Phenology 60*** 39*** 20*** 3.5ns 2.2ns 0.5nsDays to anthesis Phenology 64*** 34*** 18*** 2.5ns 1.9ns 0.5nsTiller height Reproductive 72*** 234*** 15*** 0.1ns 1.5ns 4.8**No. of nodes Reproductive 28*** 0.1ns 0.7ns 2.4ns 0.6ns 0.1nsFlag leaf length Reproductive 12.6*** 87*** 0.9ns 2.3ns 2.4ns 3.7*Penultimate leaf length Reproductive 9.3*** 116*** 3.9* 1.3ns 2.9* 5.2**Flag leaf width Reproductive 22*** 97*** 0.1ns 2.4ns 3.3* 1.7nsPenultimate leaf width Reproductive 11*** 63*** 0.1ns 0.5ns 1.9ns 1.2nsAwn length Reproductive 3.4* 13*** 0.5ns 0.1ns 1.9ns 0.9nsSpike length Reproductive 61*** 156*** 9.9** 0.1ns 2.2ns 2.0nsSpikelets per spike Reproductive 22*** 207*** 15*** 1.7ns 3.4* 2.0nsNo. of spikes Reproductive 2.2ns 392*** 4.4* 8.4** 1.1ns 3.0*Yield (no. of spikelets) Fitness 20*** 529*** 1.9ns 5.1* 4.9** 3.0*Reproductive biomass (g) Fitness 2.0ns 589*** 0.1ns 0.6ns 5.6** 3.5*Spikelet weight (mg) Fitness 79*** 23*** 6.8** 2.6ns 1.5ns 5.6**Root biomass Allocation 32*** 195*** 13*** 0.3ns 9.8*** 0.8nsRoot/shoot biomass Allocation 20*** 13*** 4.4* 0.2ns 6.4*** 0.2nsReproductive effort Allocation 2.8* 121*** 0.1ns 0.4ns 4.2** 0.3nsPercentage of fertile spikelets Allocation 7.1*** 19*** 0.9ns 10.1** 5.3** 0.2ns

* P < 0.05; ** P < 0.01; *** P < 0.001.

306 S. VOLIS ET AL.


Fig

ure

1.R

eact

ion

nor

ms

of f

our

popu

lati

ons

of w

ild

barl

ey t

o a

set

of t

reat

men

ts,

incl

udi

ng

opti

mal

(H

H)

and

thre

e co

mbi

nat

ion

s of

nu

trie

nt

and

wat

er s

tres

s(H

L,

LH

an

d L

L).

populations (Fig. 4) show a consistent pattern inreproductive, fitness and allocation traits, and over alltraits. Namely: (i) water stress causes a strong plasticresponse in wild barley plants; (ii) plant response tothe combined water ¥ nutrient stress is very similarto the effect of water stress alone; and (iii) nutrientstress when applied alone causes smaller and distinctplastic response as compared to water stress.

DISCRIMINATION OF ECOTYPES ACROSS TREATMENTS

The four ecotypes were discriminated across the fourexperimental environmental conditions in a mannerthat was consistent among reproductive, fitness andallocation traits, as well as in all traits together

(Fig. 4). The position of SB, BG and AM ecotypes wasalmost a linear, with SB and AM at opposite extremesand BG in between. The MH ecotype was distant fromthe other three populations. This trend was the mostapparent on a plot where group scores were estimatedusing all 19 traits (Fig. 4).

The high number of traits contributing to discrimi-nation of ecotypes under favourable conditions (seven)decreased under conditions of stress (four for highwater low nutrients and low water low nutrients andtwo for low water high nutrients) (Table 3). Most contributing traits were the same under favourableand stressed conditions except for root biomass andspikelets per spike where ecotypes were discriminatedonly under HL and awn length and yield where ecotypes were discriminated only under LL. The traitsdiscriminating ecotypes under different conditions arethose showing high interpopulation genetic variation,but with a wide range of plasticity from non-plastic(days to awn appearance) to relatively highly plastic(tiller hight and spikelet weight).

DISCUSSION

A set of unordered experimental stress treatmentsinduced a significant plastic response in 18 of 19analysed traits in barley plants. The traits (all excepttwo) exhibited genetic variation among populationsand showed a different interpopulation pattern ofplasticity across treatments in most reproductive andall fitness and allocation traits (but not in phenologi-cal traits). These two sources of genetic variationamong four populations can not be ascribed to randomgenetic processes because the ecotypes were found tobe locally adapted (Volis, 2001). As the main abioticdeterminants of the four population environments are known, their role in creating ecotype-specific trait complexes and trait plasticity can be identified.The ecotypes, originating in desert (SB), batha (BG),grassland (AM) and mountain (MH) represent a gradient of increasing environmental productivity and predictability (increasing amount and decreasing variation in interannual rainfall). However, in themountains, the main stress factor is winter frost,which may override other effects in certain aspects ofplant biology.

The ecotypes exhibited a strong and significanttrend for delay in the onset of reproduction andincrease in root/shoot biomass with increasing pro-ductivity of their environments, which was treatmentindependent. The same was true for root biomass, butonly under high water high nutrients and high waterlow nutrients treatments.

A strong pattern of consistent differences was foundamong SB, BG and AM plants when they were con-sidered alone (in all these cases, MH plants had inter-



Figure 2. Mahalanobis distances between ecotype meansin optimal (HH) and stressed (either LH, HL or LL) envi-ronments, calculated over three trait categories.

mediate values). There was an increase in the numberof nodes, tiller height, spike length and spikeletweight, and a decrease in the number of spikelets per spike and yield with increasing environmentalproductivity-predictability. When plotted in a multi-variate space defined by the two first canonical axes,the SB, BG and AM ecotypes were arranged accordingto their position along the gradient of environmentalproductivity-predictability; the MH ecotype was apartand most distant from the SB ecotype.

The ecotypes did not differ in plasticity in both phenological and about half the reproductive traits,but differed significantly in all fitness and allocationtraits. The overall population plasticities were similarunder water stress for reproductive and fitness traits,but not for resource allocation. The direction of overallresponse in allocation was different for AM and SBplants, with BG intermediate. AM plants did notchange their reproductive effort, but invested moreinto roots than to shoots and increased abortion rate.SB plants, in contrast, decreased reproductive effort

but did not change root/shoot ratio. The response ofBG was almost identical to that of SB in reproductiveeffort and of AM in root/shoot biomass.

Under nutrient stress, AM and BG plants hadhigher overall plasticity than SB and MH plants forreproductive and fitness traits, and were more plasticthan SB plants in resource allocation. The nutrientstress induced in AM plants increased spikelet weight,reproductive biomass and reproductive effort anddecreased abortion rate. However, the response of BGplants was different; there was a decrease in yield,reproductive biomass and reproductive effort undernutrient stress.

MH plants were the most plastic in resource alloca-tion under both water and nutrient stress, but exhib-ited low plasticity in other trait classes. The MHecotype responded to water stress by decreasing, andto nutrient stress by increasing, the absolute rootbiomass and increased root/shoot ratio in both cases.

Despite different patterns of ecotype plasticity inparticular traits, there was no difference between

308 S. VOLIS ET AL.


Figure 3. Amount of plasticity (distance between class centroids) and direction of overall responses (from high to the lowlevel of the treatment) in reproductive, fitness and resource allocation traits of plants of different origin to water, nutri-ent and water * nutrient stresses.

Figure 4. Treatment and population centroids in a discriminant plot defined by the first two canonical axes. Discrimi-nation of treatments and populations (distance between centroids) and pattern of plant responses to different treatments(directions) are shown for reproductive (A), fitness (B), allocation (C) and all traits together (D).

ecotypes in patterns of plasticity over all traits and forall trait classes, viz. reproductive, fitness and allo-cation. In a methodologically similar study, Zhang &Lechowicz (1994) compared plastic responses of Ara-bidopsis thaliana populations to nutrient stress. Theamount of overall plasticity differed among popula-tions but the pattern of plasticity remained similar.They concluded that a shared pattern of plastic adjust-ments reflects coordinate changes in the magnitude oflinked traits, thereby maintaining the stability of theirfunctional relationship. Our results suggest that thisis true only for highly interrelated traits (i.e. lengthand width parameters of leaves). The pattern ofoverall plasticity (over all traits or for trait classes)conceals rather than reveals the mechanism of plasticresponse that may vary across environments. Whenpopulation environments are similar, differences onlyin amount of plasticity are observed (Schlichting &Levin, 1986, 1990). However, when environmentsdiffer sharply different patterns of plasticity evolve(Macdonald & Chinnappa, 1989).

The response of plants to water and nutrient stressin all ecotypes in this study followed a sequence inyield component plasticity, described by Harper et al.(1970), namely spike number, spikelets per spike andthen spikelet weight. However, the resulting yield wasalso a function of abortion rate which was nonplasticin SB and plastic in the other three ecotypes. Produc-

tivity and predictability of environment appear todetermine whether abortion is involved in response to stress or not. In deserts, where amount and dis-tribution of rainy events is highly unpredictable andprobability of stress is very high, abortion of alreadydeveloped reproductive organs can be too costly.Therefore, a less risky and more conservative resourceallocation strategy with respect to abortion rateappears to have evolved in annuals from environmentswith both low productivity and low predictability.

The different plastic responses of the four ecotypesto water and nutrient stress may also reflect the rolesof these stresses as limiting factors in different environments. Water stress caused a greater plasticresponse in SB and BG than in AM and MH ecotypes,whereas for nutrient stress, plasticity was higher inthe BG and AM ecotypes than SB and MH. Oftenwater is the main limiting factor in deserts, whilenutrients (for instance, nitrogen) are less limiting. Wefound (unpubl. observ.) that, on average, the soil nitro-gen content of SB was higher than in the more pro-ductive (with respect to available water) BG locality.This may be because, in the Negev desert, most of veg-etation is localized in wadis (ephemeral river valleys)that accumulate water and nutrients from runoff andwhere most animal droppings occur. However, nutri-ents may be a limiting factor in more productive envi-ronments because competition may cause their quick

310 S. VOLIS ET AL.


Table 3. Contribution of variables to the function than best discriminates among the four ecotypes under four experi-mental environmental conditions. All F-values are significant with P < 0.001.

Variable No. of variables in a model Wilk’s lambda Approximate F-value

HHDays to awn appearance 1 0.38 27.1Spikelet weight 2 0.17 22.9Root/shoot biomass 3 0.09 21.8Flag leaf width 4 0.06 19.8Tiller height 5 0.03 21.0Reproductive effort 6 0.02 21.5No. of nodes 7 0.01 19.9

HLSpikelet weight 1 0.35 34.5Root biomass 2 0.2 22.4Tiller height 3 0.13 18.5Spikelets per spike 4 0.08 17.6

LHTiller height 1 0.43 21.4Days to awn appearance 2 0.25 15.7

LLDays to awn appearance 1 0.35 20.4Tiller height 2 0.19 16.4Awn length 3 0.14 12.6Yield 4 0.11 10.9

depletion. As a result, efficient and rapid uptake ofavailable nutrients may be part of plant strategy inproductive areas (Grime et al., 1986).

In contrast to the widely held view that the onset of reproduction in desert, and sometimes in mesic,annuals is stimulated by drought (Went, 1948, 1949;Rathcke & Lacey, 1985; Lacey, 1986), our resultssuggest the opposite. Water stress significantlydelayed onset of reproduction, a pattern found forother annuals from arid environments (Mott &McComb, 1975; Fox, 1990; van Rooyen et al., 1991;Steyn et al., 1996). There was no difference in delaying response among plants of different origin,suggesting a similar mechanism for switch from vegetative to reproductive stage and similar pheno-typic plasticity for this transition for the four ecotypes.

Our findings are not consistent with the hypothesisthat plants originating in environments with greatervariation and unpredictability are more plastic(Schlichting, 1986; Bell et al., 1993). Contrary to pre-diction, the desert ecotype showed the lowest plastic-ity in all trait categories in response to nutrient stressand did not have the highest plasticity under waterstress. No ecotype was superior in degree of plasticityunder both water and nutrient stresses and in all traitcategories. The importance of plasticity in response towater and nutrient stress appears to decrease andincrease, respectively, along the increasing rainfallgradient for the three ecotypes that do not experiencefrost stress (SB, BG and AM).

In summary, the SB desert ecotype appears to be themost specialized with reduced plasticity for resourceallocation. It starts reproduction earlier than the otherecotypes, matures earlier, has the highest shoot/rootbiomass and produces the largest number of, but thesmallest seeds. In contrast, the AM ecotype, originat-ing in Mediterranean grassland that is a favourableand predictable environment, is highly plastic forfitness and allocation traits in response to reducednutrient availability which may occur due to competi-tion in the productive environment. High morphologi-cal plasticity as part of a foraging mechanism waspredicted by Grime et al. (1986) for plants of produc-tive habitats, corresponding to the ‘competitive strat-egy’ of the triangular ordination model (Grime, 1974).The AM ecotype exhibited increased root/shoot ratio in response to water, nutrient and water * nutrientstress, a pattern not observed in either SB or BGplants. The AM plants were the tallest, producing thelargest but fewest seeds. The BG ecotype from thesemisteppe batha, an environment intermediatebetween SB and AM in productivity and predictabil-ity, was very plastic in response to both water andnutrient stress. This may indicate that either water ornutrients can be limiting during plant growth due toplant competition and irregular supply of water and

nutrients. The BG plants are intermediate betweenSB and AM for initiation of reproduction, plant size,yield (total number of spikelets) and spikelet weight.The mountain MH ecotype, which was found in recip-rocal tests (Volis, 2001) to be a locally adapted spe-cialist, exhibited a highly plastic response to waterstress in root allocation. A similar case was describedin Agrostis stolonifera (Bradshaw & Hardwick, 1989)where distinct dwarf ecotypes occur in response towind and a plastic pattern of root growth occurred inresponse to nutrient stress. The explanation of Bradshaw & Hardwick was that the same plant canevolve both specialized response(s) to distinct perma-nent stress(s), as well as a plastic phenotype to copewith specially or temporally variable stresses. The MHplants start reproduction later, have larger roots anda higher root/shoot ratio than other ecotypes. With-standing winter frosts requires physiological adap-tation, as seeds germinate and become seedlings oryoung plants before the frosts start. Intensive develop-ment of the root system may serve both purposes,namely to be a part of a frost-tolerant mechanism andto evolve competitive superiority for nutrients andwater.

ACKNOWLEDGEMENTS

We thank Kamshat Keldibekov, Maxim Madorsky andEdward Berezovsky for assistance in the nethouse andRuti Soto for laboratory work. A grant from the IsraelAcademy of Sciences (86293101) and a local grant ofthe Mitrani Department for Desert Ecology supportedthis study. This is publication number 336 of theMitrani Department for Desert Ecology.

REFERENCES

Bell G, Lechowicz MJ, Appenzeller A, Chandler M,DeBlois E, Jackson L, Mackenzie B, Preziosi R, Schallenberg M, Tinker N. 1993. The spatial structure of the physical environment. Oecologia 96: 114–121.

Blais PA, Lechowicz MJ. 1989. Variation among populationsof Xanthium strumarium (Compositae) from natural andruderal habitats. American Journal of Botany 76: 901–908.

Bradshaw AD. 1965. Evolutionary significance of phenotypicplasticity in plants. Advance in Genetics 13: 115–155.

Bradshaw AD, Hardwick K. 1989. Evolution and stress –genotypic and phenotypic components. Biological Journal ofthe Linnean Society 37: 137–155.

Fox GA. 1990. Components of flowering time variation in adesert annual. Evolution 44: 1404–1423.

Grime JP. 1974. Vegetation classification by reference tostrategies. Nature 250: 26–31.

Grime JP, Crick JC, Rincon E. 1986. The ecological significance of plasticity. In: Jennings DH, Trewavas AJ, eds.



Plasticity in plants. Cambridge: Company of Biologists,5–29.

Gurevitch J. 1992. Sources of variation in leaf shape amongtwo populations of Achillea lanulosa. Genetics 130: 385–394.

Harlan RJ, Zohary D. 1966. Distribution of wild wheats andbarley. Science 153: 1074–1080.

Harper JL, Lovell PH, Moore KG. 1970. The shape andsizes of seeds. Annual Review of Ecology and Systematics 1:327–356.

Lacey EP. 1986. Onset of reproduction in plants: size- versusage-dependency. Trends in Ecology and Evolution 1: 72–75.

Levins R. 1968. Evolution in fluctuating environments.Princeton: Princeton University Press.

Lewontin RC. 1974. The analysis of variance and the analy-sis of causes. American Journal of Human Genetics 26:400–411.

Macdonald SE, Chinnappa CC. 1989. Population differen-tiation for phenotypic plasticity in the Stellaria longipescomplex. American Journal of Botany 76: 1627–1637.

McGarigal K, Cushman S, Stafford S. 2000. Multivariatestatistics for wildlife and ecology research. New York:Springer.

Mott JJ, McComb AJ. 1975. Effects of moisture stress on thegrowth and reproduction of three annual species from anarid region of Western Australia. Journal of Ecology 63:825–834.

Nevo E, Zohary D, Brown AHD, Haber M. 1979. Geneticdiversity and environmental associations of wild barley,Hordeum spontaneum, in Israel. Evolution 33: 815–833.

Pigliucci M, Schlichting CD, Whitton J. 1995. Reactionnorms of Arabidopsis. II. Response to stress and unorderedenvironmental variation. Functional Ecology 9: 537–547.

Rathcke B, Lacey EP. 1985. Phenological patterns of terres-trial plants. Annual Review of Ecology and Systematics 16:179–214.

van Rooyen MW, Grobbelaar N, Theron GK, van RooyenN. 1991. The ephemerals of Namaqualand: effects of pho-toperiod, temperature and moisture stress on developmentand flowering of three species. Journal of Arid Environments19: 133–145.

Scheiner SM. 1993. Genetics and evolution of phenotypic

plasticity. Annual Review of Ecology and Systematics 24:35–78.

Schlichting CD. 1986. The evolution of phenotypic plasticityin plants. Annual Review of Ecology and Systematics 17:667–693.

Schlichting CD, Levin DA. 1986. Phenotypic plasticity: anevolving plant character. Biological Journal of LinneanSociety 29: 37–47.

Schlichting CD, Levin DA. 1990. Phenotypic plasticity inPhlox. III. Variation among natural populations of P. drum-mondii. Journal of Evolutionary Biology 3: 411–428.

Schmalhausen II. 1949. Factors of evolution. The theory of stabilizing selection. Chicago: Chicago University Press.

Steyn HM, Van Rooyen N, Van Rooyen MW, Theron GK.1996. The phenology of Namaqualand ephemeral species.The effect of water stress. Journal of Arid Environments 33:49–62.

Sultan SE. 1987. Evolutionary implications of phenotypicplasticity in plants. Evolutionary Biology 21: 127–178.

Systat. 1997. SYSTAT 7.0 for Windows. Chicago: SPSS Inc.Thompson JD. 1991. Phenotypic plasticity as a component of

evolutionary change. Trends in Ecology and Evolution 6:246–249.

Volis S. 2001. Adaptation in heterogeneous environments: evidence of different plant strategies in populations ofHordeum spontaneum. Ph.D. Thesis, Ben Burion Universityof the Negev.

Went FW. 1948. Ecology of desert plants. I. Observations of germination in the Joshua Tree National Monument, California. Ecology 29: 242–253.

Went FW. 1949. Ecology of desert plants. II. The effect of rainand temperature on germination and growth. Ecology 30:1–13.

Williams BK. 1983. Some observations on the use of dis-criminant analysis in ecology. Ecology 69: 1275–1285.

Zhang J. 1995. Differences in phenotypic plasticity betweenplants from dimorphic seeds of Cakile edentula. Oecologia102: 353–360.

Zhang J, Lechowicz MJ. 1994. Correlation between time offlowering and phenotypic plasticity in Arabidopsis thaliana(Brassicaceae). American Journal of Botany 81: 1336–1342.

312 S. VOLIS ET AL.


Documents

Differentiation in populations of Hordeum spontaneum Koch ... · and the organism’s ability to accurately assess envi-ronmental variation (Levins, 1968); and (2) on biolog-ical