ECOPHYSIOLOGY

E. Kathryn Barto Æ Don Cipollini

Testing the optimal defense theory and the growth-differentiation balancehypothesis in Arabidopsis thaliana

Received: 17 December 2004 / Accepted: 6 July 2005� Springer-Verlag 2005

Abstract Two prominent theories proposed to explainpatterns of chemical defense expression in plants are theoptimal defense theory (ODT) and the growth-differen-tiation balance hypothesis (GDBH). The ODT predictsthat plant parts with high fitness value will be highlydefended, and the GDBH predicts that slow growingplant parts will have more resources available for defenseand thus will have higher defense levels than fastergrowing tissues. We examined growth rate, fitness value,and defense protein levels in leaves of a wild and labecotype of Arabidopsis thaliana to address whether pat-terns of defense protein expression in this plant conformto predictions of either the ODT or the GDBH. We di-vided leaves of A. thaliana into six leaf classes based onthree developmental stages: vegetative, bolting, andflowering; with two leaf ages at each stage: young andold. We assessed the fitness value of leaves by deter-mining the impact of the removal of each leaf class ontotal seed production and germination rates. AlthoughA. thaliana was highly tolerant to defoliation, youngleaves were more valuable than old in general, and youngleaves on bolting plants were the most valuable leaf classin particular. Young leaves on vegetative plants grewfastest in both ecotypes, while old leaves on bolting andflowering plants grew slowest. Finally, defense levelswere assessed in each leaf class by quantifying the con-stitutive and inducible expression of four defense-relatedproteins. Expression of guaiacol peroxidase and chitinaseactivity conformed largely to GDBH predictions.Expression of trypsin inhibitor and polyphenoloxidaseactivity varied by leaf class and treatment, but con-formed to neither GDBH nor ODT predictions.

Introduction

Plants employ a variety of mechanisms to protectthemselves from attack by herbivores and microbialpathogens, including the production of constitutive andinducible chemical defenses. Two important pathways ofchemical defense production in plants are salicylic acid-mediated systemic acquired resistance (SAR) induced bybiotrophic pathogens, and jasmonic acid-mediated in-duced resistance (IR) induced by necrotrophic patho-gens and insects (McDowell and Dangl 2000). Importantdefense proteins associated with SAR include peroxid-ases and chitinases (Samac and Shah 1991; Dowd et al.1997; Franke et al. 2002), while trypsin inhibitors andpolyphenoloxidases are associated with IR (McManuset al. 1983; Koiwa et al. 1997). In Arabidopsis thaliana,expression of these defense proteins benefits resistance(Traw et al. 2003; Cipollini et al. 2004), but is appar-ently costly in the absence of attack (Cipollini 2002). Incontrast to the glucosinolates, important defensive sec-ondary metabolites of mustards (Cipollini et al. 2004),patterns in the expression of these defense proteins inlight of plant defense theory have been little studied.

Constitutive expression of such defenses in plants isneither constant throughout a plant’s lifetime, nor aredefenses expressed evenly throughout a plant (Zangerland Rutledge 1996). The degree of induction of chemicaldefenses also varies with plant age and plant part (Ci-pollini and Redman 1999; Van Dam et al. 2001). Twoprominent hypotheses to explain spatial and temporalvariation in plant defense expression are the optimaldefense theory (ODT) (Rhoades 1979), and the growth-differentiation balance hypothesis (GDBH) (Herms andMattson 1992). These two theories are representative ofthe range of theories raised to explain variation in plantdefense (Stamp 2003) and tests of them provide insightinto predictions made by other theories.

The ODT postulates that defenses are allocated toplant parts based on the fitness value of the plant partand its risk of herbivory, and assumes that defenses are

Communicated by Richard Lindroth

E. K. Barto Æ D. Cipollini (&)Department of Biological Sciences, Wright State University,3640 Colonel Glenn Highway, Dayton, OH 45435, USAE-mail: don.cipollini@wright.eduTel.: +1-937-7753805Fax: +1-937-7753320

Oecologia (2005)DOI 10.1007/s00442-005-0207-0

costly (Rhoades 1979). ODT predicts that plant partsthat are either highly likely to be attacked or that havehigh fitness value should have high constitutive levels ofdefense and low inducibility. Plant parts that are lesslikely to be attacked and/or are less valuable shouldhave low constitutive levels of defense, and higherinducibility.

Zangerl and Rutledge (1996) found that expression ofxanthotoxin in different tissues of wild parsnip (Pastin-aca sativa L.) conformed to ODT predictions based onrisk of insect herbivory, but fitness value of plant partswas not considered. Ohnmeiss and Baldwin (2000) testedthe degree to which nicotine levels in tobacco conformedto ODT predictions, but examined defense levels inrelation to fitness value of tissues, which is arguablymore important than risk of attack. Removing leavesfrom bolting stage plants caused the greatest fitnessdetriment, and these leaves also had higher constitutiveand damage-induced nicotine concentrations than leavesof vegetative or flowering plants. This supports ODTpredictions of high constitutive defense levels in valuableleaves, but high inducibility of valuable leaves does notstrictly support ODT predictions, as defined by Zangerland Rutledge (1996).

The GDBH assumes that a balance must be main-tained between resources used for growth and for dif-ferentiation (which includes chemical defenseproduction) (Herms and Mattson 1992; Wilkens et al.1996a). A key premise of this hypothesis is that defenseis costly. Although a number of studies have foundconcentrations of secondary metabolites to be highest inimmature leaves (Herms and Mattson 1992; Lerdauet al. 1994; Riipi et al. 2002), GDBH predicts that thiscarries an opportunity cost. Plants that divert resourcesto defense of young leaves are predicted to grow slowerthan they otherwise might (Herms 1999). Herms andMattson (1992, 1994) postulated that one mechanism bywhich rapidly growing leaves and seedlings overcomeresource-based constraints on chemical defense is bydrawing on previously acquired stored reserves. Becausethis is not possible in rapidly growing annual plants,GDBH can be inferred to predict that chemical defensewill be resource-limited in immature leaves of suchplants.

Experimental support for the GDBH has varied, butit has infrequently been examined in herbaceous plants(Stamp 2003). In their tests of GDBH, Wilkens et al.(1996a, b) found that levels of two phenolics and tri-chome densities in tomato failed to display reducedlevels in rapidly growing immature leaves, although thispattern is not inconsistent with GDBH for plants withsufficient storage to meet the demands of both growthand defense in such leaves (Herms and Mattson 1992,1994).

We simultaneously tested predictions of both theODT and GDBH by examining the constitutive andinduced expression of four defense proteins in youngand old leaves of A. thaliana, for which we also as-sessed fitness value and relative growth rate. We used

both a common laboratory strain of A. thaliana and awild ecotype, and examined plant responses to fourdefense elicitors. To our knowledge, the ODT and theGDBH have never been tested simultaneously in thesame plant, let alone in the model plant A. thaliana.We demonstrate that specific leaves of A. thaliana varyin fitness value and in growth rate. In turn, we showthat peroxidase and chitinase expression in differentleaf classes largely followed predictions of the GDBHas it pertains to annual plants, and that little supportfor the ODT was found for any defense protein mea-sured here.

Methods

Plant material and growth

We used A. thaliana (mouse ear cress) as a model systemin this study. Although the winter annual life cycle ofA. thaliana permits escape from many herbivores, wildpopulations of A. thaliana are attacked by many ofthe same insects and pathogens that attack other mus-tards (Koch and Slusarenko 1990; Biever et al. 1992).A. thaliana also expresses substantial levels of defenseproteins and secondary compounds (e.g. glucosinolates)(Traw et al. 2003; Cipollini et al. 2004). The plant hor-mones salicylic acid (SA) and jasmonic acid (JA) can beused to induce SAR and IR, respectively, with nodamage to the plant (Cipollini 2002; Traw et al. 2003;Cipollini et al. 2004). Microbial inducers of these path-ways are also available (Dong et al. 1999; Thommaet al. 1998).

The major lab ecotype of A. thaliana, Columbia(Col), and a wild ecotype, Oak Openings (Oak), wereused in this experiment. Col seed was originally obtainedfrom the Arabidopsis Biological Resource Center. AllOak seed used was obtained from one plant grown in thegreenhouse from seed collected in the wild near Toledo,Ohio. Seeds were sterilized in a 20%-bleach, 0.1%-so-dium dodecyl sulfate solution for 5 min before plantingin each experiment. Col seeds were planted directly inpots in the greenhouse as in Cipollini (2002), and thin-ned after 2 weeks to one plant per pot. Oak seeds werestarted on agar plates made with half strength Mu-rashige and Skoog basal salt mixture and stratified at4�C for four days to stimulate germination. Plates werethen moved to an incubator at 22�C with continuouslight for two weeks, and then transplanted to pots in thegreenhouse. All plants were grown in ProMix ‘BX’(Premier Horticulture INC, Red Hill PA) in 150 ml potsat temperatures of 22�C±5�C. Plants were grown in agreenhouse under ambient light with supplementalfluorescent light on a 13-h photoperiod. Plants werewatered with distilled water and were not fertilized.After 45 days, Oak plants were placed in a cold room at4�C for 2 weeks to stimulate bolting, then returned tothe greenhouse. Col plants have no seed stratification orvernalization requirement.

Assignment of leaf classes

The A. thaliana life cycle was divided into three devel-opmental stages. We defined the vegetative stage as thetime from germination until flower buds first appeared.The bolting stage lasted from this point until the firstflower opened, and the flowering stage continued untilthe plant set seed and senesced. Treatments of vegetativeplants occurred when the plants had six true leaves.Bolting stage treatments were applied when the primarybolting stem was 3 cm tall, and flowering stage treat-ments were applied when three flowers had opened.Using this scheme, bolting and flowering plants werecloser in actual age than vegetative and bolting plants.Each developmental stage was further considered tohave two leaf ages. At each developmental stage, thenumerical half of the rosette leaves nearest the top ofthe rosette were considered the young leaves and thenumerical half nearest the base were considered the oldleaves. This characterization of plant stages and leafages yielded six rosette leaf classes: young leaves onvegetative plants (VY), old leaves on vegetative plants(VO), young leaves on bolting plants (BY), old leaves onbolting plants (BO), young leaves on flowering plants(FY), and old leaves on flowering plants (FO).

Experiment 1: value

The fitness value of each leaf class was assessed byremoving it and measuring effects on total seed pro-duction and seed germination. At the appropriatedevelopmental stage, each leaf class was entirely re-moved by cutting at the base of the petiole of each leafwith scissors. Areas or masses of each leaf class removedduring this experiment were not quantified. However, ina follow-up experiment done on the Col ecotype usingthis design, we determined that the total areas of youngand old leaves were not significantly different withinstages [Age (Stage): F3, 54=1.67, P=0.183, Table 3),which paralleled effects on leaf mass. Thus, similaramounts of young and old tissues within a stage wereremoved. The plant was then left to complete its lifecycle, and siliques were collected as they matured.Thirteen replicate plants were used per leaf class. Sili-ques were air dried at room temperature for 2 weeks inan air-conditioned laboratory, and separated from theseeds. All seeds from each plant were weighed to thenearest 0.1 mg. Total seed mass correlates strongly withtotal seed number in A. thaliana (D. Cipollini, unpub-lished data). Seed germination percentages were mea-sured by stratifying about 30 seeds from each plant onmoistened filter paper in sterile petri dishes for 4 days at4�C. After cold stratification, dishes were incubated at22�C with continuous light and percent germination ineach dish was determined after 5 and 13 days. Total seedmass was multiplied by germination percentage at eachdate to give viable seed mass. Two trials of this experi-ment were run for each ecotype.

Viable seed masses were analyzed using a two-waynested ANOVA with plant stage (V,B,F), and leaf age(Y,O) nested within stage as factors. Col and Oak eco-types were analyzed separately. Control plants were notincluded because they could not be grouped by plantstage and leaf age, but are presented for comparison.Type III sums of squares were used. To allow compari-sons with undamaged control plants, Tukey comparisonsshown on Fig. 2 are based on a one-way ANOVA withtreatment (VY, VO, BY, BO, FY, FO, control) as themain effect. Data were transformed as necessary to meetthe assumptions of ANOVA. SAS was used for all sta-tistical analyses (Version 8.02, SAS Institute, Cary, NC).

Experiment 2: relative growth rate

The relative growth rate of a representative leaf fromeach leaf class was assessed by measuring changes in itslength for 5 days. One young and one old leaf at eachdevelopmental stage were measured to the nearest mmon thirteen untreated control plants. Although untreatedcontrols were measured, leaf lengths were recorded from2 days before through 2 days after the time wheninducing treatments would have been applied to treatedplants. A second order polynomial best-fit line was fittedto those five points and the slope of the line, which isequal to absolute growth rate on the day of treatment,was calculated. Absolute growth rate was then dividedby the leaf length just before the day of treatment to giverelative growth rate of the leaf on the day of treatment.

Relative growth rates of leaves from each leaf classwere analyzed as for total viable seed mass. Tukeycomparisons were done for six leaf classes (VY, VO, BY,BO, FY, and FO).

Experiment 3: defense levels

Treatments

Levels of four defense proteins were assessed in each leafclass in response to four elicitors of induced defenses.We used two pathogens, the bacterium Pseudomonassyringae (Ps) and the fungus Botrytis cinerea (Bc), whichinduce SA-mediated defenses and JA-mediated defenses,respectively, in A. thaliana (Dong et al. 1999; Thommaet al. 1998). In addition, we used two plant hormones,SA and JA, which induce SAR and IR when appliedexogenously, respectively.

Pseudomonas syringae pv tomato (strain DC3000) wascultured on King’s B agar. Prior to inoculation, P. sy-ringae was grown in King’s B liquid media for 3 h at23�C. Fifty microliters of a suspension of 4·108 cells/mlwas pressure infiltrated into one young and one old leafof each plant using a 1-ml syringe without the needle asin Summermatter et al. (1995). Infected leaves weremarked with a nontoxic pen and were not sampled foranalysis.

Botrytis cinerea (strain IMI169558) was cultured oncorn meal agar (5% corn meal, 1.5% agar, pH 5.5) for10 days with no light in a Precision 818 Plant Growthincubator. Spores were collected by flooding the platewith 0.1% Tween 20 and scraping the surface with asterile glass rod. Scrapings were filtered through What-man 1 paper (pore size 11 lm) to remove fungal hyphaeand isolate spores. Fifty microliters of a suspension of3·105 spores/ml were infiltrated into one young and oneold leaf on each plant at the appropriate developmentalstage.

A 0.5-mM SA solution or a 0.45-mM JA solution inwater was sprayed on hormone treated plants atappropriate times as in Cipollini (2002). Untreatedcontrol plants were not manipulated in any way.

After treatment, all plants (including controls) wereplaced under plastic domes in the greenhouse for 48 hbefore sampling. Because Col plants were small, allleaves on the plant, except pathogen inoculated leaves,were collected and divided into young and old leaves foranalysis. Oak plants at the vegetative stage were alsosmall and were sampled the same way. Oak plants at thebolting and flowering stages were larger, so approxi-mately 300 mg each of young and old leaves were ran-domly collected at these stages. Leaves were stored at�20�C until analysis. There were 5–13 replicate plantsfor each leaf class in each of the five elicitor treatments(Control, PS, BC, SA, JA).

Protein assays

Soluble proteins were extracted by grinding leaf samplesin 0.05 M sodium phosphate buffer (pH 6.8). Sampleswere then centrifuged at 12,000 rpm for 12 min to pelletthe leaf material. The supernatant was transferred tonew tubes and extracts were stored at �20�C untilanalysis. Samples were kept at 4�C during the entireextraction.

Activities of n-acetylglucosaminidase (exochitinase),guaiacol peroxidase, polyphenol oxidase, and trypsininhibitor were assayed using spectrophotometric andradial diffusion techniques as in Traw et al. (2003). En-zyme activities in leaf extracts were expressed per mgextract protein, which was determined according toBradford (1976) with Bovine Serum Albumin as astandard. In a follow-up experiment using the Col eco-type, we found that POD activities expressed per mgprotein exhibited the same response to plant stages andleaf age within stages as POD activities expressed perunit leaf mass (Table 3), and indeed these two measuresof POD were highly phenotypically correlated (r=0.938, P<0.0001, n=60).

Enzyme activity levels were each analyzed separatelyusing a three-way nested ANOVA with plant stage, leafage nested within stage, treatment (control, Ps, Bc, SA,JA), and their appropriate interactions as factors. Coland Oak ecotypes were analyzed separately. Type IIIsums of squares were used and data were transformed as

necessary. Tukey tests were done on all possible com-binations of effects.

Results

Experiment 1: value

There were marginally significant reductions in totalviable seed mass due to removal of specific leaf classes inone of two trials with Col and in one of two trials withOak (Fig. 1a, 1b). In the first Col trial, removing youngleaves led to a 25% reduction in total viable seed mass at13 days post germination over removing old leavesacross developmental stages, with the difference beingmarginally significant (Age (Stage): F3,72=2.36,P=0.078, Fig. 1a). Removing young leaves on boltingplants (BY) tended to have the greatest fitness impact,and led to a nearly 40% decrease in total viable seedmass over undamaged controls. Removing old leaves atany stage had no effect. In the first Oak trial, viable seedmass of plants defoliated at the vegetative stage was44% lower at 5 days post germination than plantsdefoliated at the bolting stage across leaf ages, and 74%lower than plants defoliated at the flowering stage, withthe differences being marginally significant (Stage: F2,

35=2.69, P=0.082, Figure 1b). There was a tendencyfor reduced viable seed production when young leaveson bolting plants were removed, as for Col, and alsowhen old leaves on vegetative plants were removed. By13 days post germination, these differences disappeared.Although no factors proved to be significant in thesecond trial of this experiment with each ecotype, non-significant patterns in the means largely paralleled thoseof the first trial (data not shown).

Experiment 2: relative growth rate

In the Col ecotype, young leaves grew approximately sixtimes faster than old leaves (Age (Stage): F3,72=67.58,P<0.0001), Fig. 1c). Leaves on vegetative plants grewabout twice as fast as on bolting plants and seven timesfaster than on flowering plants (Stage: F2, 72=72.4,P<0.0001, Fig. 1c). Young leaves on vegetative plantsgrew about twice as fast as young leaves on boltingplants, and ten times as fast as young leaves on floweringplants. Old leaves grew at the same low rate at alldevelopmental stages.

In the Oak ecotype, young leaves grew approximatelythree times faster than old leaves (Age (Stage):F3,67=52.95, P<0.0001, Fig. 1d). Leaves on vegetativeplants grew five times faster than leaves on boltingplants and seven and a half times faster than leaves onflowering plants (Stage: F2,67=166.72, P<0.0001,Fig. 1d). Young leaves on vegetative plants grew aboutthree times as fast as young leaves on bolting plants orold leaves on vegetative plants, and eight and a half

times as fast as young leaves on flowering plants. Oldleaves on bolting and flowering plants had essentiallystopped growing.

Experiment 3: defense

Chitinase (CHI)

Oak plants had higher average CHI activities than Colplants, but patterns were similar between the two eco-types. Plants at the bolting and flowering stages hadhigher CHI activities than plants at the vegetative stage(Col�66%, Oak�223%) (Tables 1, 2, Fig. 2). Oldleaves had higher activity than young leaves in bothecotypes (Col�54%, Oak�117%). Young leaves gener-ally did not respond to elicitor treatments.

Patterns of CHI expression in plants treated withSA and Ps (each SAR inducing) were strikingly similar,as were patterns in plants treated with JA and Bc (eachIR inducing). Bolting Col plants that were not treatedand flowering plants treated with Bc had higher CHIlevels than vegetative plants treated with Bc, JA, or

untreated plants. No other combinations were signifi-cantly different. Young and old leaves on vegetativeOak plants had the lowest CHI levels, followed byyoung leaves on bolting and flowering plants, whoseCHI levels were twice as high as leaves on vegetativeplants. Old leaves on bolting and flowering plants hadthe highest levels, four times higher than leaves onvegetative plants. The difference between young andold leaves was much more pronounced in bolting andflowering plants than in vegetative plants. Both SA andPs induced higher CHI levels in old leaves on floweringplants than in old leaves on bolting plants. Both JAand Bc induced the opposite pattern, with higher CHIlevels in old leaves on bolting plants than in old leaveson flowering plants.

Peroxidase (POD)

Old leaves had approximately five times higher PODlevels than young leaves in both ecotypes (Tables 1, 2,Fig. 3). Bolting and flowering Oak plants had approxi-mately nine times higher levels than vegetative plants. As

Fig. 1 Fitness value and growthrate of young and old leaves ofvegetative (V), bolting (B), andflowering (F) plants. a Totalviable seed mass produced afterremoval of each leaf class in theColumbia ecotype afterallowing 13 days forgermination. b Total viable seedmass produced after removal ofeach leaf class in the Oakecotype after allowing 5 daysfor germination. c Relativegrowth rates of each leaf classin the Columbia ecotype. dRelative growth rate of eachleaf class in the Oak ecotype.Bars represent the mean(+1SE). Bars with differentletters are significantly different(a=0.05)

Table 1 Results of three-way nested ANOVA examining the effects of developmental stage, leaf age, elicitor treatment, and theirinteraction on defense protein activities in the Columbia ecotype of A. thaliana

CHI POD PPO TI

Source df F P F P F P F P

Stage 2 9.22 0.0001 1.23 0.2946 4.13 0.0173 6.03 0.0028Age (stage) 3 9.08 <0.0001 2.66 0.0488 0.86 0.4639 2.83 0.0391Treatment 4 1.35 0.2528 0.80 0.5262 0.66 0.6170 1.56 0.1860Stage·treatment 8 2.39 0.0169 0.79 0.6118 1.51 0.1559 1.96 0.0522Age (stage)·treatment 12 0.90 0.5454 0.15 0.9996 2.20 0.0123 0.84 0.6136

Error-df: 231 Error-df: 231 Error-df: 231 Error-df: 231

CHI chitinase, POD peroxidase, PPO polyphenol oxidase, TI trypsin inhibitor

with CHI, young and old leaves on vegetative Oakplants had the lowest POD levels, followed by youngleaves on bolting and flowering plants, which were threetimes higher. Old leaves on bolting and flowering plantshad the highest levels, approximately 15 times higherthan leaves on vegetative plants. As for CHI, patterns ofresponse to treatments with SA and Ps, and to JA andBc, were strikingly similar. POD levels were higher in oldleaves of flowering plants treated with SA or Ps, whichstimulate SAR, than in old leaves of bolting plantstreated with SA or Ps. In turn, POD levels were lower inold leaves of flowering plants treated with JA or Bc,which stimulate IR, than in old leaves of bolting plantstreated with JA or Bc.

Polyphenol oxidase (PPO)

The highest levels of PPO occurred in vegetative plantsof both ecotypes (Tables 1, 2, Fig. 4). In the Col eco-type vegetative plants had PPO levels 1.3 times higherthan in bolting plants. Levels in flowering plants werenot significantly different from vegetative or boltingplants. In the Oak ecotype, vegetative plants had PPOlevels approximately 9% higher than bolting andflowering plants. This pattern is opposite to that seenfor CHI and POD levels, where vegetative plants hadthe lowest activity levels. Young leaves of Col plants atall stages treated with Ps had PPO levels 1.7 times ashigh as young leaves at all stages treated with Bc. No

Fig. 2 Chitinase levels in young and old leaves of vegetative (V),bolting (B), and flowering (F) plants after treatment with salicylicacid (SA), Pseudomonas syringae (Ps), jasmonic acid (JA), orBotrytis cinerea (Bc). a Columbia ecotype. Bars represent mean of5–12 plants (+1SE). b Oak ecotype. Bars represent mean of 6–13plants (+1SE)

Table 2 Results of three-way nested ANOVA examining the effects of developmental stage, leaf age, elicitor treatment, and theirinteraction on defense protein activities in the Oak openings ecotype of A. thaliana

CHI POD PPO TI

Source df F p F p F p F p

Stage 2 162.04 <0.0001 142.80 <0.0001 8.73 0.0003 14.54 <0.0001Age (stage) 3 33.50 <0.0001 38.79 <0.0001 7.29 0.0001 9.21 <0.0001Treatment 4 1.61 0.1731 0.71 0.5848 1.11 0.3538 5.04 0.0006Stage·treatment 8 3.42 0.0009 1.37 0.2102 1.55 0.1401 2.66 0.0081Age (stage)·treatment 12 1.68 0.0703 2.05 0.0207 1.52 0.1167 1.86 0.0393

Error-df: 266 Error-df: 267 Error-df: 264 Error-df: 267

CHI chitinase, POD peroxidase, PPO polyphenol oxidase, TI trypsin inhibitor

Fig. 3 Peroxidase levels in young and old leaves of vegetative (V),bolting (B), and flowering (F) plants after treatment with salicylicacid (SA), Pseudomonas syringae (Ps), jasmonic acid (JA), orBotrytis cinerea (Bc). a Columbia ecotype. Bars represent mean of5–12 plants (+1SE). b Oak ecotype. Bars represent mean of 6–13plants (+1SE)

other combinations were significantly different. Oldleaves of Oak plants had levels 1.8 times as high asyoung leaves. Young leaves on vegetative Oak plantshad lower PPO levels than all other combinations byabout half.

Trypsin inhibitor (TI)

In the Col ecotype, leaves of vegetative plants had 1.5times higher TI levels than bolting and flowering plants(Table 1, Fig. 5a). In the Oak ecotype bolting andflowering plants had levels 1.2 times higher than vege-tative plants (Table 2, Fig. 5b). Old leaves had higherlevels than young leaves in both ecotypes. Old leaves ofCol had approximately 1.3 times higher levels thanyoung leaves, while old leaves of Oak had approximately1.1 times higher levels than young leaves. Flowering Colplants treated with JA had the highest TI levels in the Colecotype, followed by vegetative plants that were nottreated. Untreated flowering plants had the lowest levels,followed by flowering plants treated with SA. In the OakOpenings ecotype vegetative plants treated with Bc hadthe lowest TI levels, followed by vegetative plants treatedwith SA, then vegetative plants treated with Ps. Youngleaves on vegetative Oak plants had lower TI levels thanall other combinations by almost two times. Control Oak

plants and those treated with JA had 1.2 times higher TIlevels than plants treated with Bc. Plants treated with Psor SA were not significantly different from any othertreatment. Young leaves of vegetative Oak plants hadlower activity than old leaves of vegetative plants for alltreatments except Bc, where activity was higher in youngleaves. Activity was higher in old leaves of bolting plantsthan young leaves of bolting plants treated with Bc, incontrast to all other treatments, where young leaves ofbolting plants had higher TI levels.

Discussion

We related constitutive and induced levels of four de-fense proteins to fitness value and relative growth rate ofleaves of A. thaliana in order to simultaneously testpredictions of the ODT and GDBH.

Fitness value of leaves

In the first trial with Col, removal of young leaves led to a25% decrease overall in total seed mass over undamagedcontrols or plants whose old leaves had been removed.Seeds from these plants also germinated more slowlythan seeds from controls or plants whose old leaves had

Fig. 4 Polyphenol oxidase levels in young and old leaves ofvegetative (V), bolting (B), and flowering (F) plants after treatmentwith salicylic acid (SA), Pseudomonas syringae (Ps), jasmonic acid(JA), or Botrytis cinerea (Bc). a Columbia ecotype. Bars representmean of 5–12 plants (+1SE). b Oak openings ecotype. Barsrepresent mean of 5–13 plants (+1SE)

Fig. 5 Trypsin inhibitor levels in young and old leaves of vegetative(V), bolting (B), and flowering (F) plants after treatment withsalicylic acid (SA), Pseudomonas syringae (Ps), jasmonic acid (JA),or Botrytis cinerea (Bc). a Columbia ecotype. Bars represent meanof 5–12 plants (+1SE). b Oak Openings ecotype. Bars representmean of 5–13 plants (+1SE)

been removed. In addition to reduced total seed pro-duction, delayed seed germination could lead to reducedfitness for the plant if it germinates after its competitors(Rice and Dyer 2001). Young leaves may be more valu-able to fitness than old leaves because of their greatercurrent and future photosynthetic potential for the plant.In a follow-up experiment using the Col ecotype, leafsoluble protein concentrations (mg soluble protein/gfresh leaf mass) were lower in bolting and floweringplants than in vegetative plants (Stage: F2, 54=6.43,P=0.003) and were lower in old leaves than in youngleaves within stages [(Age) Stage]: F3, 54=6.09, P=0.001,Table 3). Declines in total protein concentration withleaf age are often associated with loss of Rubisco(Hopkins 1999), suggesting that old leaves of A. thaliana(particularly those on bolting and flowering plants) havelower photosynthetic value than young leaves.

Young rosette leaves on bolting plants appear to be aparticularly important contributor to flowering and seedfilling in A. thaliana. Loss of these leaves would lead tofewer current resources for reproduction through de-creased photosynthetic capacity and loss of stored re-sources in leaves. Removing old leaves of vegetativeplants were also detrimental to viable seed mass in Oakplants (driven largely by delayed germination of seeds),but not Col plants. This difference could be due to thewinter annual habit of the Oak ecotype that requiresplants to spend the greatest amount of time in the fieldas vegetative plants.

Based on these results, we predicted that if defensesconform to ODT, young leaves in general, and youngleaves on bolting plants in particular, should have thehighest levels of defense because they had the highestfitness value. Old leaves, especially those on floweringplants, should have the lowest defense levels in bothecotypes because they were of little fitness value. Oneexception would be old leaves of vegetative plants on theOak ecotype.

Leaf growth rate

Young leaves grew fast in general, and young leaves onvegetative plants had a higher RGR than any other leafclass for both ecotypes. Old leaves on bolting and

flowering plants had particularly low RGRs for bothecotypes. Therefore, we predicted that if defenses con-form to GDBH as it applies to annual plants, defenselevels should be lowest in young leaves of vegetativeplants, and highest in old leaves of bolting and floweringplants for both ecotypes. Intermediate defense levels areexpected in young leaves of bolting plants for both ec-otypes.

Defense protein activities

In general, young leaves had lower defense levels thanold leaves whenever a significant difference was found.In no instance did we find the highest constitutive orinducible levels of defenses in young leaves of boltingplants, the most valuable leaf class identified for Col,and a highly valuable leaf class for Oak. These patternscontradict ODT predictions of high defense levels invaluable plant parts, but support GDBH predictions oflow defense levels in fast growing plant parts and highlevels in slow growing parts. The effect of developmentalstage on defense levels also varied with the defense. CHIand POD, which are associated with SAR, were at theirlowest levels in rapidly growing leaves of little immediatefitness value on vegetative plants, and at their highestlevels in old leaves of little value on bolting and flow-ering plants. This conforms to GDBH predictions be-cause rapidly growing plant parts had low defense levels,and vice-versa, but contradicts ODT predictions of lowdefense levels in leaves of little value. However, PPO andTI, which are associated with IR, showed the highestdefense levels in vegetative plants with rapidly growingleaves of little immediate fitness value. This patternsupports neither the GDBH nor the ODT. Although theODT has received some recent support (Stamp 2003), wefound no evidence that it correctly predicts the expres-sion of the defense proteins measured here. That said,the expression of these defense proteins has been littleexamined in light of either ODT or GDBH, or any othertheory (but see Cipollini et al. 2003).

When examining induction of different leaf classesseparately by defense protein, some other interestingpatterns emerge. Activities of CHI and POD followedthe same patterns in response to similar inducing treat-

Table 3 Leaf area, leaf mass, soluble protein concentration of leaves, and two measures of peroxidase (POD) activity in young and oldleaves of the Col ecotype at three developmental stages

Stage Leaf age Area (cm2) Mass (g) Protein(mg/g leaf FW)

POD(D 450/min/g leaf FW)

POD(D 450/min/mgextract protein)

V y 0.589 (0.046) 0.00549 (0.00053) 3.31 (0.300) 3.04 (0.917) 1.70 (0.582)o 0.526 (0.046) 0.00539 (0.00048) 2.70 (0.549) 15.9 (4.65) 17.2 (8.54)

B y 2.17 (0.226) 0.0242 (0.00271) 2.31 (0.414) 11.4 (3.09) 8.81 (2.40)o 2.60 (0.191) 0.0273 (0.00209) 1.38 (0.242) 45.5 (9.54) 73.4 (20.3)

F y 1.57 (0.161) 0.0174 (0.00213) 2.70 (0.406) 19.8 (7.14) 12.7 (3.73)o 2.02 (0.247) 0.0206 (0.00260) 1.29 (0.199) 37.3 (6.48) 57.9 (13.4)

Means (±1SEM), n=10

ments. For example, SA and Ps induced higher levels ofCHI and POD in old leaves on flowering plants than oldleaves on bolting plants. In contrast, JA and Bc inducedCHI and POD on old leaves on bolting plants morestrongly than old leaves on flowering plants. TI activityfollowed no consistent pattern in response to treatments,while PPO was not very inducible at all. Even consid-ering differential inducibility, expression of the SAR-associated defenses generally conformed to GDBHpredictions, while IR-associated defenses conformed toneither GDBH nor ODT predictions. No data exists onthe risk of attack by different enemies of different leafclasses in A. thaliana, but our results hint to the possi-bility that different leaf classes are geared to respondmore strongly to some elicitors than to others. Olderleaves, while of little immediate fitness value, may be atgreater risk of attack by certain classes of enemies owingto their position relative to the soil surface, and viceversa.

In addition to the resource driven explanations ofthe GDBH, an alternative explanation for high levelsof some defenses in older plant parts could be con-tinuous accumulation. If defenses were produced con-tinuously and were stable once produced then oldleaves on flowering plants should have the highestlevels, and young leaves on vegetative plants shouldhave the lowest levels (as was seen for CHI and POD).However, the half-life of POD in ginseng is approxi-mately 4 days (Troitskaya et al. 1999), and only about2.5 days in peanut (van Huystee and Zheng 1995).Other defensive proteins likely have high turnoverrates as well, and/or are not made continuously. TIlevels in Brassica napus peaked 3 days after induction(Cipollini and Bergelson 2000), 4 days after inductionin Nicotiana attenuata (Van Dam et al. 2001), and6 days after induction in castor bean (Bruce and West1989), before declining. Therefore, the defense proteinsthat reached high levels in old leaves here are likelynot passively accumulating with time, but are insteadregulated differentially in old versus young leaves bythe plant. The defense proteins measured here (par-ticularly POD) may also play essential developmentalroles in addition to defense (Campa 1991), which couldpartly explain some of the developmental patterns weobserved.

Both the ODT and GDBH could correctly predictpatterns of expression for defensive compounds thatare translocated around the plant (Cronin and Hay1996). Ohnmeiss and Baldwin (2000) found highernicotine levels in the valuable parts of tobacco (youngleaves on bolting plants) in support of ODT predic-tions. Nicotine is synthesized in the roots of tobaccoplants, and then transported to other parts of the plant.GDBH predictions are met because nicotine is syn-thesized in slow growing roots, and ODT predictionsare met because valuable plant parts are protected.However, all defenses are not translocatable; proteinsin particular are not very mobile (Hopkins 1999). Aninvestigation tracking the synthesis of, as well as final

location of many chemical defenses could help to re-solve this question.

Although we found little support for it among thedefense proteins that we measured, ODT could makecorrect predictions if valuable plant parts are protectedby defensive mechanisms not assayed here. Youngleaves may be protected by chemical defenses such asglucosinolates, which are constitutively present, induc-ible, and are associated with insect resistance in A.thaliana (Cipollini et al. 2004). Glucosinolates reachhigher levels in young than in old leaves of A. thaliana,and even higher levels in flowers, siliques and seeds(Brown et al. 2003), a pattern that supports predictionsof ODT. Young leaves may also receive some level ofprotection by being located at the center of the rosettesurrounded by older leaves that do have high levels ofdefensive proteins, thus receiving some within-plant‘‘associational resistance’’ (Tahvanainen and Root1972).

Conclusions

We have shown that constitutive and inducible levels ofseveral defensive proteins varied with ontogeny in A.thaliana, as did the fitness value and growth rate ofleaves. Expression of CHI and POD, defenses typicallyassociated with SAR, followed patterns predicted by theGDBH. Expression of PPO and TI, defenses typicallyassociated with IR, was influenced by the developmentalstage of the plant and the age of the leaves sampled, butfollowed patterns predicted by neither ODT nor theGDBH. Although a wild ecotype of A. thaliana hadhigher defense levels in some cases, defense patterns indifferent leaf classes and their response to elicitors werelargely represented in the major experimental lab strain,Columbia.

Acknowledgements This research was financially supported, by theDepartment of Biological Sciences, Wright State University, andUDSA NRI-CGP Grant 2001-35302-10998 to D.C. We thankTonia Korves for advice and materials for our experiments with P.syringae. Comments by Dan Herms and two anonymous reviewerssubstantially improved this manuscript.

References

Biever KD, Chauvin RL, Reed GL, Wilson RC (1992) Seasonaloccurrence and abundance of lepidopterous pests on collards inthe Northwestern United States. J Entomol Sci 27(1):5–18

Bradford MM (1976) A rapid and sensitive method for the quan-titation of microgram quantities of protein utilizing the prin-ciple of protein-dye binding. Anal Biochem 72:248–254

Brown PD, Tokuhisa JG, Reichelt M, Gershenzon J (2003) Vari-ation in glucosinolate accumulation among different organs anddevelopmental stages of Arabidopsis thaliana. Phytochemistry62:471–481

Bruce RJ, West CA (1989) Elicitation of lignin biosynthesis andisoperoxidase activity by pectic fragments in suspension cul-tures of castor bean. Plant Physiol 91:889–897

Campa (1991) Biological roles of plant peroxidases: known orpotential function. Peroxidases in chemistry and biology. CRCPress, Boca Raton, pp 22–50

Cipollini DF (2002) Does competition magnify the fitness costs ofinduced responses in Arabidopsis thaliana? A manipulative ap-proach. Oecologia 131:514–520

Cipollini DF, Bergelson J (2000) Environmental and developmen-tal regulation of trypsin inhibitor activity in Brassica napus. JChem Ecol 26(6):1411–1423

Cipollini DF, Redman AM (1999) Age-dependent effects of jas-monic acid treatment and wind exposure on foliar oxidaseactivity and insect resistance in tomato. J Chem Ecol 25(2):271–281

Cipollini DF, Busch JW, Stowe KA, Simms EL, Bergelson J (2003)Genetic variation and relationships of constitutive and herbi-vore-induced glucosinolates, trypsin inhibitors, and herbivoreresistance in Brassica rapa. J Chem Ecol 29:285–302

Cipollini DF, Enright S, Traw B, Bergelson J (2004) Salicylic acidinhibits jasmonic acid-induced resistance of Arabidopsis thali-ana to Spodoptera exigua. Mol Ecol 13:1643–1653

Cronin G, Hay ME (1996) Within-plant variation in seaweed pal-atability and chemical defenses: optimal defense theory versusthe growth-differentiation balance hypothesis. Oecologia105:361–368

Dong H, Delaney TP, Bauer DW, Beer SV (1999) Harpin inducesdisease resistance in Arabidopsis through the systemic acquiredresistance pathway mediated by salicylic acid and the NIM1gene. Plant J 20(2):207–215

Dowd PF, Duvick JP, Rood T (1997) Comparative toxicity of al-lelochemicals and their enzymatic oxidation products to maizefungal pathogens, emphasizing Fusarium graminearum. NatTox 5:180–185

Franke R, Hemm MR, Denault JW, Ruegger MO, HumphreysJM, Chapple C (2002) Changes in secondary metabolism anddeposition of an unusual lignin in the ref8 mutant of Arabid-opsis. Plant J 30:47–59

Herms DA (1999) Physiological and abiotic determinants of com-petitive ability and herbivore resistance. Phyton 39:53–64

Herms DA, Mattson WT (1992) The dilemma of plants: to grow orto defend. Quart Rev Biol 67:283–335

Herms DA, Mattson WT (1994) Plant growth and defense. TrendsEcol Evol 9:488

Hopkins WG (1999) Introduction to Plant Physiology, 2nd edn.Wiley, New York. pp 221, 261

van Huystee RB, Zheng X (1995) Half-life of peroxidase in activelygrowing and arrested peanut cells. CR Acad Sci Paris, Sciencesde la vie/ Life Sciences 318:655–658

Koch E, Slusarenko A (1990) Arabidopsis is susceptible to infectionby a downy mildew fungus. Plant Cell 2:437–445

Koiwa H, Subramanian L, Shade RE, Zhu K, Nielsen SS, Mur-dock LL, Bressan RA, Hasegawa PM (1997) Insecticidalactivities of soybean cysteine proteinase inhibitors are corre-lated with their papain-binding affinities as phage-displayedproteins. Plant Physiol 114:1115

Lerdau M, Letvak M, Monson R (1994) Plant chemical defense:monoterpenes and the growth-differentiation balance hypothe-sis. Tree 9:58–61

McDowell JM, Dangl JL (2000) Signal transduction in the plantimmune response. Trends Biochem Sci 25:79–82

McManus J, Lilley TH, Haslam E (1983) Plant polyphenols andtheir association with proteins. In: Hedin PA (eds) plant resis-tance to insects. American Chemical Society, Washington DC,pp 123–137

Ohnmeiss TE, Baldwin IT (2000) Optimal defense theory predictsthe ontogeny of an induced nicotine defense. Ecology81(7):1765–1783

Rhoades DF (1979) Evolution of plant chemical defense againstherbivores. In: Rosenthal GA, Janzen DH (eds) Herbivores:their interaction with secondary plant metabolites. AcademicPress, New York and London, pp 4–53

Rice KJ, Dyer AR (2001) Seed aging, delayed germination andreduced competitive ability in Bromus tectorum. Plant Ecol155:237–243

Riipi M, Ossipov V, Lempa K, Haukioja E, Koricheva J, OssipovaS, Pihlaja K (2002) Seasonal changes in birch leaf chemistry: arethere trade-offs between leaf growth and accumulation ofphenolics? Oecologia 130:380–390

Samac DA, Shah DM (1991) Developmental and pathogen-in-duced activation of the Arabidopsis acidic chitinase promoter.Plant Cell 3:1063–1072

Stamp N (2003) Out of the quagmire of plant defense hypotheses.Quart Rev Biol 78:23–55

Summermatter K, Sticher L, Metraux JP (1995) Systemic responsesin Arabidopsis thaliana infected and challenged with Pseudo-monas syringae pv syringae. Plant Physiol 108:1379–1385

Tahvanainen JO, Root RB (1972) The influence of vegetationaldiversity on the population ecology of a specialized herbivore,Phyllotreta crucifera (Coleoptera: Chrysomelidae). Oecologia10:321–346

Thomma BPHJ, Eggermont K, Penninckx IAMA, Mauch-Mani B,Vogelsang R, Cammue BPA, Broekaert WF (1998) Separatejasmonate-dependent and salicylate-dependent pathways inArabidopsis are essential for resistance to distinct microbialpathogens. Proc Natl Acad Sci USA 95:15107–15111

Traw MB, Kim J, Enright S, Cipollini DF, Bergelson J (2003)Negative cross-talk between salicylate- and jasmonate-mediatedpathways in the Wassilewskija ecotype of Arabidopsis thaliana.Mol Ecol 12:1125–1135

Troitskaya LA, Komov VP, Kirillova NV (1999) Peroxidaseturnover in ginseng strains under standard conditions andtemperature stress. J Plant Phys 155:281–284

Van Dam NM, Horn M, Mares M, Baldwin IT (2001) Ontogenyconstrains systemic protease inhibitor response in Nicotianaattenuata. J Chem Ecol 27(3):547–568

Wilkens RT, Shea GO, Halbreich S, Stamp N (1996a) Resourceavailability and the trichome defenses of tomato plants. Oeco-logia 106:181–191

Wilkens RT, Spoerke JM, Stamp NE (1996b) Differential re-sponses of growth and two soluble phenolics of tomato to re-source availability. Ecology 77(1):247–258

Zangerl AR, Rutledge CE (1996) The probability of attack andpatterns of constitutive and induced defense: a test of optimaldefense theory. Am Nat 147(4):599–608