Impact of Melon Accessions Resistant to Aphids on the Demographic Potential of Silverleaf Whitefly

PLANT RESISTANCE

Impact of Melon Accessions Resistant to Aphids on the DemographicPotential of Silverleaf Whitefly

NICOLAS SAUVION,1 VALERIE MAURIELLO, BENJAMIN RENARD, AND NATHALIE BOISSOT

Unite de Recherche en Productions Vegetales, Centre Antilles-Guyane, Institut National de la Recherche Agronomique,Petit-Bourg, Guadeloupe, 97170, France

J. Econ. Entomol. 98(2): 557Ð567 (2005)

ABSTRACT Bemisia tabaci (Gennadius) and Aphis gossypiiGlover are devastating melon, Cucumismelo L., pests. The geographic areas where they occur overlap, and the same chemicals are used tocontrol both of them. Therefore, to reduce pesticide use, it would be necessary to breed melon linesthat simultaneously express a resistance to both insects. Female survival; the time when reproductionstarts, peaks, and ends; the number of female offspring at the reproductive peak; and total reproduction(S) were determined under semicontrolled conditions for B. tabaci kept in clip-cages on a susceptiblemelon genotype Vedrantais, and 12 potential resistant accessions, particularly genotypes expressingthe Vat gene controlling resistance to A. gossypii. By using the Lewontin triangular reproductivefunction and Bootstrapping, the intrinsic rate of increase (r) and its variance were calculated.Statistical analysis showed that the parameter S was as relevant as r for discriminating between themelon accessions. Three genotypes were potential genitors of resistance to the whiteßy: PI 161375,PI414723, andPI532841.Thosepossessing theVatgenewereeither resistant(PI161375andPI414723)or susceptible (Margot, IsoVat R, and AR 5). This demonstrated the ineffectiveness of Vat against B.tabaci. In this article, we propose a strategy to breed lines that express resistance to aphids andwhiteßies on the short term.

KEY WORDS Cucumis melo, Bemisia tabaci, host plant resistance, population dynamics, Va gene

SWEET POTATO WHITEFLY, Bemisia tabaci (Gennadius),and the cotton (melon) aphid, Aphis gossypii Glover,are devastating melon pests,CucumismeloL., found inmajor melon production areas (United States, Mexico,Central America and South America, southern France,and Spain), where they cause major economic losseseither as herbivores (Slosser et al. 1989, Gerling andMayer 1996) or as virus vectors (Lecoq et al. 1998).The losses due toB. tabacihave increased over the past20 yr because of the appearance of new biotypes andthe emergence of their associated viruses (Jones2003).

Integrated pest management programs to controlthese two pests consist of a combination of chemicals(in particular, imidacloprid and pymetrozin) and cul-tivation practices to maintain insect populations be-low an economic threshold level (McCreight 1998).But the development of resistance against the chem-icals (Devonshire 1989, Prabhaker et al. 1998) makesit necessary to Þnd alternative strategies such as theuse of resistant breeding lines. In melon, resistantcultivars have shown evidence of their effectiveness incontrolling A. gossypii in France, where 80% of the

registered varieties are resistant to this aphid. Never-theless, these varieties are susceptible to B. tabaci andbecause the same chemicals are used to control pierc-ing-sucking insects, the extension of the whiteßy castsdoubts on the advantage of using these commercialvarieties as a means of limiting pesticide use. It wouldtherefore be advantageous for the breeding lines tosimultaneously express resistance to the aphid and thewhiteßy to limit the use of chemicals. Presently, nostudy has been done on this subject.

Research into melon genotypes resistant toB. tabaciis recent and has revealed material that could be usefulfor breeding because screening was essentially basedon tolerance (i.e., loss of biomass) in the Þeld(Moreno et al. 1993, Simmons and McCreight 1996).Riley and Palumbo (1995a, b) estimated losses in theÞeld on the basis of the density of whiteßy popula-tions. They evoked the possibility of the glabrous char-acter of leaves as a resistance factor, but it has not yetbeen applied to breeding. In Þeld trials, Boissot et al.(2003) observed partial resistance to B. tabaci (i.e.,reduction of the density of adults and larval insects onleaves) in PI 161375, PI 414723, PI 164723, and 90625,but the study did not allow an assertion that thesegenotypes were suitable candidates to retain for agenetic analysis of resistance. Under free-choice andno-choice conditions, Soria et al. (1999)) quantiÞed

1 Currentaddress:UMRBGPI INRAÐCIRADÐENSAM,EquipeEpi-demiologie-virus/vecteurs, CIRAD TA 41/K, Campus Internationalde Baillarguet, Montpellier Cedex 5, 34398, France.

0022-0493/05/0557Ð0567$04.00/0 � 2005 Entomological Society of America

the level of resistance by considering the life historyparameters: net reproduction, adult longevity, adultsurvival, and length of developmental time. Two ge-notypes turned out to be potential donors of resistanceto B. tabaci: agrestis Ô87Õ and TGR-1551.

For the past 30 yr, research into melon genotypesresistant to A. gossypii has essentially focused on twoaccessions: LJ 90634 (later called PI 414723) fromIndia (Kishaba et al. 1971, McCreight 1992) and PI161375 from Korea (Pitrat and Lecoq 1980). Anti-xenosis and antibiosis expressed in the two sources ofgermplasm were controlled by one dominant geneknown as Vat (Pitrat and Lecoq 1982, Perin et al.2002). It was cloned and belongs to the NBS-LRRresistant gene family (Dogimont et al. 2003). Othermelon accessions were identiÞed as potential sourcesof resistance to A. gossypii. At INRA Avignon, resis-tance was identiÞed in �5% of the 500 tested acces-sions (Pitrat 1997). Nevertheless, few of them werethe object of detailed research into the mechanismsinvolved (i.e., detailed behavioral analysis of theaphidÐplant interaction) in resistance expression, ex-cept for a genotype from Zimbabwe, TGR 1551 (Soriaet al. 2000, 2003), and a genotype from India, 90625(Boissot et al. 2000), which seemed all the more in-teresting because they expressed resistances differentfrom those identiÞed in PI 161375 and P1 414723.

At least four genotypes characterized as being re-sistant toA. gossypii also express resistance toB. tabaciundercontrolledconditionsor inÞeld trials: PI 161375,P1 414723, 90625, and TGR-1551. These results led usto wonder whether it was the same or independentfactors that controlled resistance to both pests. Withthe genotypes PI 161375 and PI 414723, it was possibleto answer this question: A. gossypii resistance (Vatgene) has been introgressed into melon breeding linesfrom both the Korean and Indian germplasm sourcesto create nearly isogenic lines Margot and AR5, re-spectively (Chen et al. 1997, Kingler et al. 1998). Thus,adequate plant material for speciÞcally studying the

mode of action of theVat gene on B. tabaciwas Þnallyavailable.

The aim of our research was to quantify and qualifyresistance to B. tabaci expressed by melon genotypesunder semicontrolled conditions to 1) conÞrm resis-tance expressed in the Þeld (McCreight 1995, Boissotet al. 2003), 2) study the Vat gene effects on thewhiteßy; and Þnally, 3) determine the most interestinggenotypes to be used for a genetic analysis of resis-tance to the whiteßy and more in-depth studies of themechanisms involved in the plantÐinsect interaction.Six life history components were used as indicators ofresistance: female survival; the time when reproduc-tion starts, peaks, and ends; and the number of femaleoffspring at the reproductive peak and total repro-duction. Using the Lewontin triangular reproductivefunction (Lewontin 1965), we estimated the intrinsicrate of natural increase and its variance for each ge-notype to discriminate between them.

Materials and Methods

Insects.Whiteßies were collected from colonies inGuadeloupe (French West Indies) and reared on cab-bage, Brassica oleracea L. ÔCopenhagueÕ. The colonieswere placed in an air-conditioned room maintained at25Ð27�C, 60Ð90% RH, and a photoperiod of 12:12(L:D) h. Whiteßies were identiÞed by analysis ofesterases and random ampliÞcation of polymorphicDNA polymerase chain reaction patterns such as bio-type B ofB. tabaci (De Barro and Driver 1997, Sauvionet al. 1999).Plants. All the accessions (Table 1) used in this

study showed partial resistance toB. tabaci in the Þeld(McCreight 1993, 1994, 1995; Boissot et al. 2000, 2003),and, as much as possible, resistance to A. gossypii,except for Vedrantais and Bola de Oro, which wereused as controls. The seeds used were taken from thecollection of INRA Avignon. The identity and the

Table 1. Description of the thirteen genotypes of melon tested for their resistance to the whitefly, B. tabaci

Varieties Type OriginResistance

A. gossypii B. tabaci

Vedrantais cantalupensis France Susceptible Susceptible [1][2]Bola de Oro inodorus Spain - Susceptible [3]PI 161375 chinensis Korea Vat gene [4] -PI 414723 momordica India Vat gene [4] -Margot near-isogenic

Vedrantaiscantalupensis France Vat gene, introgressed from

PI 161375 [5]-

IsoVat R near-isogenicCharentais T

cantalupensis France Vat gene, introgressed fromMargot [5]

-

AR5 quasi-isogenic of PMR5 reticulatus USA Vat gene, introgressed fromPI 414723 [6]

-

90625 acidulus India Tolerant to leaf curling [7];antixenosis [1]

Partially resistant [1][2]

Meloncillo chito Colombia - Partially resistant [2]Ogon 9 makuwa Japan - Partially resistant [2]PI 164723 acidulus India - Partially resistant [2]PI 179248 - Turkey - Partially resistant [2]PI 532841 momordica India - Tolerant [8]

[1] Biotests, Boissot et al. 2000; [2] Field trials, Boissot et al. 2003; [3] Lopez-Sese et al. 1996; [4] Pitrat and Lecoq 1982; [5] Pitrat comm.pers.; [6] McCreight et al. 1984; [7] Bohn et al. 1996; [8] McCreight 1995.

558 JOURNAL OF ECONOMIC ENTOMOLOGY Vol. 98, no. 2

origin of the varieties were speciÞed whenever pos-sible (Stepansky et al. 1999).

Three series of assays were made to test the 13genotypes: Þve, four, and seven genotypes were testedduring the Þrst, second, and third series, respectively(Fig. 1). The Vedrantais control was tested in everyseries and the genotype AR 5 was tested twice (series1 and 3). The assays took place in an open greenhouseat Petit-Bourg (Guadeloupe) from April to July 2000.All the plants were sown, individually staked, andrandomly distributed in the greenhouse.Experiment.To describe reproduction over a single

generation of B. tabaci on different melon genotypes,we used the same approach as Romanow et al. (1991),based on the population model developed by Le-wontin (1965). The bioassay consisted of following thedevelopment of a female for 15 d after it hatched andthen recording the emergence of its descendants. Onthe evening before the day of the infestation (day 0),all of the adults were removed from the cages. On themorningof the infestationday(day1),newlyemerged

adults were individually placed in glass tubes. Femaleswere then selected under a stereomicroscope, brießyanesthetizedwithCO2, anddropped intoan individualleaf cage. Females were then assumed to be unmated.Six cages per plant were placed on young leaves (oneor two cages per leaf) and six plants per genotype (sixto seven well-spread leaves) were tested. A daily re-port was kept until day 15, indicating the status offemales: attached to the leaf, moving in the cage,missing, or dead. For 3 d after the infestation, white-ßies found dead or that had disappeared were re-placed by newly emerged adults of the same cohort asthe originals. Premature mortality was considered as amanipulation bias (stressed or wounded insects).Survival. The data obtained from day 0 to day 15

made it possible to establish the survival curve of theadult whiteßies for each tested genotype. Survivalrates (percentage) were then compared at a givenpoint in time by a standard % comparison test (FisherÕsexact test; genotype versus Vedrantais) (Dagnelie1998).

Fig. 1. Graphs of survival of femaleB. tabacion 13 melon accessions tested under semi-controlled conditions during threeseries of assays (AÐC). Number of whiteßies tested for each accession is given in parentheses; arrows indicate the mean valueswhich differ signiÞcantly (FisherÕs exact test; P � 0.01) from the Vedrantais control.

April 2005 SAUVION ET AL.: DEMOGRAPHIC POTENTIAL OF SILVERLEAF WHITEFLY 559

Emergence Curve Modeling. On day 15, all thecages containing live females were removed and theirpositions on the leaves were marked with a pencil.These leaves were then left uncovered and undis-turbed. Empty puparia were counted and removedevery day. Assuming that each empty puparium cor-responded to a viable adult (male or female), weestimated the number of progeny of the female foreach day, and calculated the daily mean number offemale offspring for each genotype, considering a 1:1sex ratio per genotype (Romanow et al. 1991).

We obtained an emergence curve for the popula-tion, V(x) with:

V� x� � l� x�.m� x�

where l(x) designates the probability at birth of beingalive at age x, andm(x) is the mean number of femaleoffspring produced in a unit of time by a female agedx. This curve was modeled with a triangular function,called L(x), Þtted by minimizing the residual sum ofsquares, and including A, T,W, and Y parameters (thebeginning, the peak and the end of the reproductiontime, respectively, and the number of female offspringat the reproductive peak):

L� x� � �0 if x � A

Y� x � A�

T � Aif A � x � T

Y�W � x�

W � Tif T � x �W

0 if x �W

Estimation of r and Life History Components. Tocompare the dynamics of whiteßies on the differentaccessions, the intrinsic rate of natural increase (r)was calculated for each of them. This basic estimateincorporates all salient factors of an insectÕs life history(Dent 1997). Birch (1948) deÞned it as the rate ofincrease per head (i.e., individual female) under spec-iÞed physical conditions, in an unlimited environmentwhere the effects of increasing density do not need tobeconsidered.Witha stableagedistribution, thevalueof r can be calculated from the equation:

�0

��

e�rxl� x�m� x�dx � 1

Lewontin (1965) proposed calculating r as the so-lution of this equation where l(x).m(x) is modeled byL(x): 0

��e�rx L(x)dx 1. The solution of such anequation was approximated with NewtonÕs method.The total number of progeny was thus estimated as 2.0

�� L(x)dx. The other life history components wereestimated as follows: the Þrst and last 1% of totalreproduction were Þrst removed from the data set(see Romanow et al. 1991 for details). The data set wasthen divided into two parts: the maximum value ofV(k) was included as 1) the last value of the Þrst dataset (before the peak), and 2) as the Þrst value of thesecond data set (after the peak). A linear regressionwas then used on each of these data sets to estimateA,T,W, and Y parameters. T and Ywere estimated as the

intercept between two regression lines.A andWwerethe intercept of the Þrst and second lines with thex-axis.Mean Value and Standard Error Calculation byBootstrapping. We used a computer-intensive tech-nique, bootstrapping, to estimate sampling distribu-tion, mean value and standard error of r and the as-sociated parameters. Meyer et al. (1986) described themethod initially presented by Efron (1982) in detail.The procedure is based on 1) recombining the originaldata, 2) calculating pseudo-values of the parametersfor each recombination of the original data, and 3)estimating the mean value and standard error of theparameters from the resulting frequency distributionof pseudo-values. To initiate a bootstrap replicate, weplaced the original data set of N insects (i.e., numberof replicates for each genotype) into a pool fromwhich N values were randomly selected with replace-ment (i.e., each sampled insect was returned to thedata pool before another insect was sampled). Wethen computed the intrinsic rate of increase for thisrandom recombination of the original data and des-ignated it r*i, the r value estimated for the ith Bootstrapreplicate (or ith pseudo-value of r). Random samplingand calculation of r* were repeatedm times (m 1000in our study). Finally, we estimated the mean value(rB) and the standard error of r from the resultingfrequency distribution of pseudo-values as follows:

rB �1

1000�i1

1000

r*i

SE�rB� � �Sr*2where Sr*

2 variance of the 1000 bootstrap r* values,{r*1, r*2, . . . , r*1000}

For each estimated bootstrap r replicate, we alsoobtained a pseudo-value of the parameters A, T, W, Y,and S.We were therefore able to estimate the boot-strap mean values of these parameters (and their stan-dard errors) that we referred to asAB, TB,WB, YB, andSB, respectively.Confidence Interval Calculation.ConÞdence inter-

vals surrounding bootstrap estimates of the parame-ters (PB rB, AB, TB, WB, YB, or SB) were calculatedby the method established by Meyer et al. (1986). Thismethod took advantage of the information containedin the frequency distributions of the 1000 values of P(P*) that were generated during each bootstrapcalculation. Brießy, skewness and kurtosis of thesedistributions were computed, and the 2.5 and 97.5percentile values (P*

2.5% and P*97.5%) were deter-

mined to include the central 95% of the Bootstrapdistribution. The interval from P*

2.5% to P*97.5% was a

candidate for a 95% conÞdence interval on P. Whenthe sampling distributions were skewed (i.e., the con-Þdence interval was biased), we adopted a bias-cor-rected percentile method for the bootstrap (Efron1982) to compute the following asymmetrical 95%conÞdence interval:

95.5% CI � �PB,adj� �PB�P*2.5%�; PB,adj� �P*97.5% �PB��


with PB,adj deÞned as a bias-adjusted bootstrap esti-mate of P: PB,adj 2Pall � PB, with Pall estimate ofP for the entire population under consideration.Comparison of Genotypes. The classiÞcation of ge-

notypes was carried out for each of the three series oftests on the basis of the rmean value (rB) and also onthe basis of the Smean (SB). We calculated the prob-ability of a genotype being resistant in comparison tothe susceptible control (Vedrantais) according to thefollowing logic: given 1000 values of the intrinsic rateof increase for the reference genotype (rref) and forthe tested genotype (rtest), we obtained 106 values ofrref� rtest(totalnumberofcases).Onthebasisof thesevalues, the probability that the tested genotype wasresistant in comparison to the reference, P(rref� rtest 0), was given by the formula

P�rref � rtest� �number of favorable cases

total number of cases,

which can be written algebraically as follows:

P�rref � rtest� �Card�rref � rtest � 0�

106

We used the same logic to discriminate between ge-notypes on the basis of the parameter S, to calculateto probability P(Sref Stest).

Modeling, parameter estimations, standard error,and conÞdence interval calculations, as well as theprocedure for the genotype comparisons, were allcarried out using a program written with Mathematica4.1 software (Wolfram Research, Inc., Champaign,IL).

Results

Survival of Female Adults. We observed the sur-vivalof femaleson the13genotypes for15d. In theÞrstseries (Fig. 1A), survival was not affected on the Ve-drantais control until the 10th day (5% mortality). Itincreased and reached 40% at day 15. The four geno-types tested in this series presented survival curvesthat were very similar to the control. A higher pre-mature mortality rate was observed as of the sixth day.However, the difference with regard to the controlwas signiÞcant (Fisher test, P � 0.01) only for PI161375 at the ninth and 10th day, and PI 164723 at the10th day. These differences, in comparison with thecontrol, remained until the 15th day but were notsigniÞcant.

In the second series (Fig. 1B), the female mortalityrate on the control genotype increased more rapidlythan in series 1: nearly half of the females died be-tween the seventh and 11th day. However, the mor-tality rate at the end of the experiment was very similarfor both series. Females on Meloncillo and Ogon 9 hadmortality rates identical to those tested on Vedrantais.On 90625, a higher mortality rate was observed as ofthe seventh day. The difference with the control wassigniÞcant on the ninth day (Fig. 1B, arrow), and themortality rate remained 20% higher on 90625 than onVedrantais until day 15 when it reached 68%.

In the third series, the survival of the females onVedrantais was little affected. It gradually decreasedfrom the seventh day, reaching 63% by day 15. Almostidentical results were observed with genotypes PI414723, AR 5, PI 532841, and PI 179248. For the Þrstone, a premature mortality rate was observed (i.e., asigniÞcant difference compared with the control onthe Þfth, sixth, and seventh days), but this tendencydid not continue. The mortality rate on Bola de Orowas always lower than on the Vedrantais control (20%difference at day 15; survival rate of Bola de Oro versusVedrantais, 82 versus 63%). On the contrary, a highermortality rate was observed on IsoVat R as of the Þfthday (signiÞcant difference at days 7 and 15), reaching71% at the end of experiment.Classification of Genotypes and Probability of Re-sistance. For each genotype tested for its resistance toB. tabaci, a mean triangle (Fig. 2) was established frombootstrap mean values estimated for the beginning(AB), peak (TB), and end of reproduction (WB), andthe number of female offspring per female adult pro-duced at peak emergence (YB). The total reproduc-tion per female (SB) (Fig. 3) and the intrinsic rate ofincrease also were estimated (rB; Fig. 4) for eachgenotype tested. A statistical discrimination of geno-types was made on the basis of these last two param-eters. This was not possible with the other parametersbecause of their skewed sampling distributions. Theterm “tendency” was used when we observed markeddifferences between genotypes (i.e., when the conÞ-dence intervals did not overlap) for these parameters.

The Þtting by the Lewontin triangular reproductivefunction was very effective, regardless of the genotypeconsidered (R2 0.9). The computed values of S (SB)closely agreed with observed values (Fig. 3, arrows)for most of the genotypes: the overestimation wasbetween 0.5 and 3%, except for the Þrst Vedrantaisseries (10%).

In the Þrst series (Fig. 2A), on the Vedrantais con-trol, the Þrst emergences occurred around the 20thday after the females were deposited (AB 20.4). Theaverage number of emergences per female reached apeak on the 28th day (TB 28.3) of �11 individuals(YB 10.6). The last emergences were recorded onthe 36th day (WB 36.3). Generally, on Vedrantais,a female gave birth to 84 viable individuals (SB 83.9;Fig. 3A), and the value of the rB was 0.161 (Fig. 4A).Margot was a better host for whiteßies than the con-trol: the intrinsic rate of increase was signiÞcantlyhigher by 12.2% (Fig. 4A). This can be explained by ahigher YB (13.5 versus 10.6; Margot versus Vedrantais;Fig. 2A), and accordingly a higher number of emer-gences (SB 109.8 versus 83.9, a difference of 30.9%;Fig. 3A). The results for genotypes PI 164723 and AR5 were very close to those of Vedrantais forAB, TB,WB,YB, and SB (Figs. 2A and 3A), and rB values were notsigniÞcantly different in comparison with the control(0.167 for PI 164723 and 0.169 for AR 5; Fig. 4A). Theseresults reveal a very low probability of resistance (P0.20 and 0.16, respectively) for these two genotypes.However, data obtained with AR 5 were characterizedby a higher variability than those obtained on other


genotypes tested in this series, as demonstrated by theconÞdence intervals of the parameters SB (Fig. 3A)and rB (Fig. 4A). Finally, during the Þrst series, onlythe genotype PI 161375 expressed resistance to B.tabaci (P 1.00). The mean value of the intrinsic rateof natural increase was signiÞcantly lower (16.3%; Fig.4A) on PI 161375 than on Vedrantais. This differencecould be mainly explained by the parameter YB: theaverage number of insects emerging on the day of thepeak was only six as opposed to 11 for the control (Fig.2A), and the total number of progeny engendered bya female on this genotype was very low (SB 46.5versus 83.9).

In the second series, whiteßies placed on Vedrantaisand Ogon 9 behaved in an almost similar way: emer-gence began on the 16th day (Fig. 2B). A delay ofalmost 2 d from the Þrst emergence was observed for90625. The value of the parameter YB seemed to begreater for Ogon 9 and 90625 in comparison withVedrantais (11.0 and 9.7 versus 8.9), but these differ-ences did not seem signiÞcant (i.e., the conÞdence

interval did not overlap). An extension of the emer-gence time was observed on Vedrantais in comparisonwith Ogon 9 and especially for 90625 (WB 33.6versus 38.0; 90625 versus Vedrantais), resulting in anonsigniÞcant decrease of the total number of de-scendants per female (SB 75.7 versus 94.9; 90625versus Vedrantais; Fig. 3B). Finally, r values of thethree genotypes were not signiÞcantly different (Fig.4B). Meloncillo expressed partial resistance to B.tabaci: females had �30% less progeny in comparisonwith females tested on Vedrantais. This difference hadsome impact on the intrinsic rate of increase but re-mained relatively low (0.176 versus 0.186, P 0.93).

In the third series, the results were more clear-cutthan those in the Þrst ones. Three groups of genotypescould be distinguished. For the Þrst group, Vedrantaisand IsoVat R, emergence began around the 15th day,ended on the 37th day (Fig. 2C), and reached a peakon the 24th day with an average of seven emergences.Whiteßies had approximately the same number ofprogeny on Vedrantais and IsoVatR (SB 78.9 versus

Fig. 2. Lewontin reproductive function for B. tabaci on 13 melon accessions tested under semicontrolled conditionsduring three series of assays (AÐC). The vertical and horizontal bars of error represent the 95% conÞdence intervals estimatedby bootstrapping.


65.6; Fig. 3C). The small gap observed between theemergence curves of these two genotypes (Fig. 2C)was enough to infer a relatively high probability ofresistance for IsoVat R (P 0.90; Fig. 4C). However,the difference between the rB values was very small(0.181 versus 0.173, a difference of almost 5%).

The second group was represented by susceptibleto very susceptible genotypes: PI 179248, AR 5, andespecially Bola de Oro, for which the rB values weresigniÞcantly higher than those of the Vedrantais con-trol (0.199, 0.200, and 0.212, respectively, versus 0.181;differences of 9.7, 10.3, and 17.1%, respectively).These differences were essentially due to the precoc-ity and the height of the emergence peak: in compar-ison with Vedrantais, the peak occurred 3 to 4 d earlieron the very susceptible genotypes (TB 21.0, 20.3, and21.5 versus 24.6), and an average of three to six morewhiteßies occurred at that time on those three geno-types. The differences between the control and PI179248 and AR 5 were less signiÞcant with regard tothe total number of descendants by female becausethe emergence peak was more precocious and fol-lowed by a more precocious end of the emergence.However, high susceptibility of Bola de Oro wasclearly revealed as the result of a high value of theparameter S (SB 117 versus 78.9, 48% difference; Fig.3C).

For the third group, two genotypes showed signif-icant differences in their r values in comparison with

the control: PI 414723 and PI 532841 (rB 0.159 and0.156, respectively, versus 0.181, a difference of 12.3and 14.1%; P 0.99 and 1.00, respectively; Fig. 4C).These differences could be essentially explained by alower number of insects emerging at the peak (YB 4.5 and 4.3 versus 7.2; Fig. 2C) and a shorter emer-gence time (WB 35.3 and 32.3 versus 37.5), resultingin an average total number of progeny per female thatwas signiÞcantly less (44 and 53%, respectively) thanon Vedrantais (SB 44.2 and 36.8 versus 78.9; Fig. 3C).

Discussion

We chose r as the criterion of discrimination be-tween the different melon genotypes. It is complex inits mode of calculation and its biological signiÞcancebut has the advantage of integrating all of the impor-tant parameters of the insect population dynamics(Dent 1997). Furthermore, the inßuence of each ofthese parameters on the value of r has been describedpreviously (Lewontin 1965). It would be possible fortwo genotypes to have the same value of S and differby the point in time that emergence begins. In thiscase, the parameter S would not make it possible todistinguish between genotypes, whereas r, which isvery sensitive to this gap, would be indicative of theeffect of the genotype on larval developmental time.In the current study, the parameters S and r almostalways evolved in the same way and resulted in the

Fig. 3. ClassiÞcation of the melon genotypes on the basis of the bootstrap mean value of S (SB) obtained for every seriesof tests (AÐC). Number of female whiteßies taken into account for mean calculation is given in italics. The horizontal barsof error represent the 95% conÞdence intervals estimated by bootstrapping. Arrows indicate the value of S observed (i.e.,the mean number of offspring per female). * or ** indicate the probability (P 90% or P 99%) for a genotype to be moreresistant or more susceptible compared with the Vedrantais control.


same classiÞcation of genotypes with the exception ofAR 5 and PI 164723 in series 1. This was possiblebecause the parameter A (the beginning of emer-gence) in our study varied little from one genotype tothe other.

By staying within the speciÞc framework of ourbiological model, the parameter S seemed as relevantas the intrinsic rate of natural increase for discrimi-nating between melon genotypes. The use of this pa-rameter actually offers several advantages with regardto that of r.

1. The data necessary for its estimation involves fewerexperimental constraints. In particular, it does notrequire daily counting, making it possible to testmore plants, as would be required by e.g., geneticanalysis.

2. The biological signiÞcance is simple to perceivefrom an intuitive point of view.

3. Discrimination between genotypes or plants can bemade on the basis of the methods of classic infer-ential statistics, without requiring the application ofcomplex and time-consuming procedures of calcu-lation such as bootstrapping.

4. The parameter S is a robust criterion (in the sta-tistical sense of the term) because we were able todemonstrate good reproducibility of the results onidentical genotypes (Vedrantais or AR 5) duringthe various series of tests. In particular, the param-eter S is little inßuenced by temperature variations,

which is the contrary in the case of r and that ofother parameters such as larval developmental timeand emergence time. It therefore has the advantageof allowing comparisons between various series oftests with different time and place variables. Thevalue of S that we observed on Bola de Oro wasalmost totally identical to the value obtained on thesame genotype (117 versus 109) by Soria et al.(1999) in Spain. Based on that control, PI 414723and PI 532841 seemed to be as resistant as TGR-1551.

On the basis of robust criteria of quantiÞcation ofthe resistance, our study is indeed the Þrst that allowsto assert that three genotypes can be considered po-tential genitors of resistance to B. tabaci: PI 161375, PI414723, and PI 532841. The parameter r on these ge-notypes was decreased by 16.3, 12.3, and 14.1%, re-spectively, compared with Vedrantais. In theory, thismeans that over a 30-d period, the population increaseon control plants would have potentially been 90-fold (r 0.161), whereas on PI 161375, the corre-sponding population increase over the same time pe-riod of time would only have been 45 times greater(r 0.135). This decline could not be explained by ahigher mortality rate of the adults and/or by an ex-tension of larval developmental time (i.e., values of theparameter A andWwere not signiÞcantly different incomparison with values for Vedrantais). This declinewas correlated to a considerable decrease in fecundity

Fig. 4. ClassiÞcation of the melon genotypes on the basis of the bootstrap mean value of r (rB) obtained for every seriesof tests (AÐC). Number of female whiteßies taken into account for mean calculation is given in italics. The horizontal barsof error represent the 95% conÞdence intervals estimated by Bootstrapping. Arrows indicate the value of r observed (i.e.,the mean number of female offspring per female). * or ** indicate the probability (P 0.90 or P 0.99) for a genotype tobe more resistant or more susceptible compared with the Vedrantais control.


(�45Ð55% less progeny). More appropriate tests willmake it possible to determine whether the mecha-nisms of resistance could be explained by larval mor-tality and/or an effect on oviposition, due, for exam-ple, to a physical barrier as was observed with TGR-1551 (Soria et al. 1999).

Three genotypes did not conÞrm the results ob-tained during previous tests in the Þeld: 90625, PI164723, and PI 179248 (McCreight 1995, Boissot et al.2003). Our study showed that the antibiosis expressedin 90625 and PI 164723 was on the same level as the onepartially expressed on the Vedrantais control (i.e.,compared with the susceptible genotype Bola deOro). This contradiction between the results in theÞeld and our tests with females tested under non-choice conditions indicates that the resistance is an-tixenosis. This hypothesis was supported for 90625 bythe low rate of females on the leaves 6 h after infes-tation on 90625 (Boissot et al. 2000) and the mortalityof the female adults observed in our study. This resultshould be conÞrmed to determine whether 90625could be a genitor of resistance to B. tabaci since thiswould be of great interest to breeders because it alsoexpresses a phenotype of resistance to A. gossypii thatis different from the one expressed by genotypes withthe Vat gene (Boissot et al. 2000). McCreight (1995)observed that PI 179248 was very tolerant in the Þeld,whereas the females in our study gave birth to nu-merous progeny on this genotype. The resistance ob-served by McCreight would therefore be tolerantsensu stricto. It is of little interest in the context of thesustainable control of whiteßy populations becauseinsects manage to remain and to develop on genotypesexpressing this type of resistance.

In this study, we tested Þve genotypes (PI 161375,Margot, IsoVatR, PI 414723, and AR 5) that possess theVat gene that provide resistance to A. gosyypii. AR 5and IsoVat R were as susceptible to whiteßies as Ve-drantais. Margot was even more susceptible, demon-strating the ineffectiveness of Vat against B. tabaci.The genetic factors of resistance to B. tabaci, presentin PI 161375 and PI 414723, were lost when Vat wasintrogressed from these two genotypes into Margot orAR5 (genetic background of cantalupensis and retic-ulates). In the case of Margot, it seems that breedershave even introduced genes of susceptibility ex-pressed by the susceptible genitor. This result dem-onstrates that there is a very strong speciÞcity ofthe plant piercing-sucking insect interaction and in-dicates that the only gene known in melon for resis-tance to aphids cannot be used as part of a strategy tocontrol B. tabaci.Although disappointing, this conclu-sion is not surprising. Indeed, differential expressionsof resistance in melon were already discriminating,not between different families of Hemiptera (Aley-rodidae versus Aphididae) but between different spe-cies. In other words, the Vat gene is effective againstA. gossypii but ineffective against the green peachaphid, Myzus persicae (Sulzer) (Lecoq et al. 1979,Romanow et al. 1986), and, in the same way, Cucumismeeusi Jeffrey expresses resistance to the greenhouse

whiteßy, Trialeurodes vaporariorum (Westwood), butnot to B. tabaci (Soria et al. 1996).

To limit pesticides used to control B. tabaci and A.gossypii, a possible strategy would be to use breedinglines that simultaneously express resistance to theaphid and the whiteßy. Such cultivars do not exist, butit seems like it would be possible to create them in thenear future. It would Þrst be necessary to carry out agenetic analysis of resistance to the whiteßy to facil-itate the introgression of this resistance into new cul-tivars. Recombinant inbred line populations that werecreated from the cross between Vedrantais � PI161375 and Vedrantais � PI 414723, to map the Vatgene (Perin et al. 2002), will make it possible to iden-tify quantitative trait loci (QTLs) that contribute tothe resistance to B. tabaci. In a second step, it wouldbe necessary to choose the populations that also pos-sess QTLs of resistance to B. tabaci from among thenumerous populations used for introgression of Vatinto new cultivars. This would make it possible torapidly undertake the process of selecting cultivarsresistant to the two Hemiptera.

Because the targeted insect is of the piercing-suck-ing type, it is difÞcult to dissociate resistance to theinsect sensu stricto and resistance to the transmissionof the viruses that it carries. This particularly appliesto A. gossypii, which transmits numerous viruses toCucurbitaceae, notably cucumber mosaic virus(Lecoq et al. 1998). But at present, resistance to thetransmission conferred by the Vat gene effectivelycompletes the partial resistance to this virus in theÞeld, from an epidemiological view. However, thesituation is more worrisome with regard to virusestransmitted by B. tabaci on melon because of theirrecent and rapid expansion in all of the regions wheremelon is grown. We can mention, in particular, cu-cumber vein yellowing virus, an ipomovirus that isfound in Israel, the Middle East, and southern Spain;cucumber yellow stunting disorder virus, a crinivirusfound in Spain, the Middle East, and Texas; and waterchlorotic stunt virus, a begomovirus found in Sudan(Lecoq et al. 1998). It is already possible for melonbreeders to attempt to limit the damage done byCYSDV by using the resistant genotypes TGR 1551and PI 414723 (Soria et al. 1999). In this context, itwould be extremely interesting to Þnd out whetherother genotypes resistant to the vector, in particular,PI 161375 or PI 414723, also would be able to limit virusepidemics transmitted to melon by B. tabaci.

Acknowledgments

We are grateful for the support provided by colleagues atthe URPV. We also thank Albert Huc for many hours oftechnical assistance, and Mustapha Bousalem, Gerard Feb-vay, and Michel Pitrat for valuable comments on previousversions. The manuscript beneÞted greatly from review byGail Wagman. This research was supported in part by theSecretariat Francais a lÕOutre-Mer. This paper is dedicated tothe memory of Alain Kermarrec (deceased 21 April 2003).


References Cited

Birch, L. C. 1948. The intrinsic rate of natural increase of aninsect population. J. Anin. Ecol. 17: 15Ð26.

Bohn, G. W., A. N. Kishab, and J. D. McCreight. 1996. Asurvey of tolerance toAphis gossypiiGlover in part of theworld collection ofCucumismeloL. InCucurbits towards2000: Proceedings of the VIth Eucarpia Meeting on Cu-curbit Genetics and Breeding. Malaga, Spain.

Boissot, N., D. Lafortune, C. Pavis, and N. Sauvion. 2003.Field resistance to Bemisia tabaci in Cucumis melo.Hort-Science 38: 77Ð80.

Boissot, N., C. Pavis, R. Guillaume, D. Lafortune, and N.Sauvion. 2000. Insect resistance in Cucumis melo acces-sion 90625. Acta Hortic. 510: 297Ð304.

Chen, J. Q., B.Martin, Y. Rahbe, andA. Fereres. 1997. Earlyintracellular punctures by two aphid species on near-isogenic melon lines with and without the virus aphidtransmission (Vat) resistance gene. Eur. J. Plant Pathol.103: 521Ð536.

Dagnelie, P. 1998. Statistique theorique et appliquee. Tome2: inference statistique a une et a deux dimensions, DeBoeck & Larcier s.a., Paris, France.

De Barro, P. J., and F. Driver. 1997. Use of RAPD PCR todistinguish the B biotype from other biotypes of Bemisiatabaci (Gennadius) (Hemiptera: Aleyrodidae). Aust. J.Entomol. 36: 149Ð152.

Dent, D. R. 1997. Quantifying insect populations: estimatesand parameters, pp. 57Ð110. In D. R. Dent and M. P.Walton [eds.], Methods in Ecological and AgriculturalEntomology. CAB International, Wallingford, UnitedKingdom.

Devonshire, A. L. 1989. Resistance of aphids to insecticides,pp. 123Ð140. In A. K. Minks and P. Harrewijn [eds.],Aphids, their biology, natural enemies and control.Elsevier, Amsterdam, The Netherlands.

Dogimont, C., A. Bendahmane, J. Pauquet, E. Burget, S.Desloire, L. Hagen, M. Caboche, and M. Pitrat. 2003.Map-based cloning of the Vat gene confers resistance toboth aphid colonization and virus transmission. In Pro-ceedings of the 11th International Congress on MolecularPlant-Microbe Interactions, July 2003. St Petersburg, Rus-sia.

Efron, B. 1982. The jackknife, the bootstrap and other re-sampling plans. Society for Industrial and Applied Math-ematics, J.W. Arrowsmith, Philadelphia.

Gerling,D., andR. T.Mayer. 1996. Bemisia 1995: taxonomy,biology, damage, control and management. InterceptLtd., Andover, United Kingdom.

Jones, D. R. 2003. Plant viruses transmitted by whiteßies.Eur. J. Plant Pathol. 109: 195Ð219.

Kingler, J., G. Powell, G. A. Thompson, and R. Isaacs. 1998.Phloem speciÞc aphid resistance in Cucumis melo lineAR5: effects on feeding behaviour and performance ofAphis gossypii. Entomol. Exp. Appl. 86: 79Ð88.

Kishaba, A. N., G. W. Bohnamp, and H. H. Toba. 1971.Resistance to Aphis gossypii in Muskmelon. J. Econ. En-tomol. 64: 935Ð937.

Lecoq, H., S. Cohen, M. Pitrat, and G. Labonne. 1979. Re-sistance to cucumber mosaic virus transmission by aphidsin Cucumis melo. Phytopathology 69: 1223Ð1225.

Lecoq, H., G. Wisler, and M. Pitrat. 1998. Cucurbit viruses:the classics and the emerging. In Proceedings of theCucurbitaceae 98. ASHS & U.S. Dep. Agric.-ARS, PaciÞcGrove, CA.

Lewontin, R. C. 1965. Selection for colonizing ability, pp.79Ð94. In H. G. Baker and G. L. Stebbins [eds.], Thegenetics of colonizing Species. Academic, New York.

Lopez-Sese, A. I., C. Soria, and M. L. Gomez-Guillamon.1996. Resistance to Bemisia tabaci in Cucumis melo. InCucurbits towards2000:Proceedingsof theVIthEucarpiaMeeting on Cucurbits Genetics and Breeding. Malaga,Spain.

McCreight, J. 1992. Preliminary screening of melons forsweetpotato whiteßy resistance. Cucurbit Genetic Co-operative Report 15: 59Ð61.

McCreight, J. 1993. Screening of melons for sweetpotatowhiteßy resistance: 1992. Cucurbit Genetic CooperativeReport 16: 49Ð52.

McCreight, J. 1994. Screening of melons for sweetpotatowhiteßy resistance: 1993. Cucurbit Genetic CooperativeReport 17: 83Ð85.

McCreight, J. 1995. Screening of melons for silverleafwhiteßy resistance: 1994. Cucurbit Genetic CooperativeReport 18: 45Ð47.

McCreight, J. D. 1998. Breeding melon resistance to lettuceinfectious yellows virus. In Proceedings of the Cucurbi-taceae 98. ASHS & U.S. Dep. Agric.-ARS, PaciÞc Grove,CA.

McCreight, J. D., A. N. Kishaba, and G. W. Bohn. 1984. ARHaleÕs Best Jumbo, AR 5, and AR Topmark: melon aphid-resistant muskmelon breeding lines. HortScience 19:309Ð310.

Meyer, J. S., C.G. Ingersoll, L.L.McDonald, andM. S.Boyce.1986. Estimating uncertainty in population growth rates:jacknife vs. boostrap techniques. Ecology 67: 1156Ð1166.

Moreno, V., J. L. Gomez-Aguilera, C. Guereau De Arellano,and L. A. Roig. 1993. Preliminary screening of cucurbitsspecies for Bemisia tabaci Genn. whiteßy resistance. Cu-curbit Genetic Cooperative Report 16: 87Ð89.

Perin,C., L. S.Hagen,V. deConto,N.Katzir, Y.Danin-Poleg,V. Portnoy, S. Baudracco-Arnas, J. Chadoeuf, C. Dogi-mont, and M. Pitrat. 2002. A reference map of Cucumismelo based on two recombinant inbred line populations.Theor. Appl. Genet. 104: 1017Ð1034.

Pitrat, M. 1997. Breeding plants for virus and vector resis-tance. Comptes Rendus de lÕAcademie dÕAgriculture deFrance. 83: 145Ð152.

Pitrat, M., and H. Lecoq. 1980. Inheritance of resistance tocucumber mosaic virus transmission by Aphis gossypii inCucumis melo. Phytopathology 70: 958Ð961.

Pitrat, M., and H. Lecoq. 1982. Relations genetiques entreles resistances par non-acceptation et par antibiose dumelon a Aphis gossypii. Recherche de liaisons avecdÕautres genes. Agronomie 2: 503Ð508.

Prabhaker, N., N. C. Toscano, and T. J. Henneberry. 1998.Evaluation of insecticide rotations and mixtures as resis-tance management strategies forBemisia argentifolii(Ho-moptera: Aleyrodidae). J. Econ. Entomol. 91: 820Ð826.

Riley,D.G., and J.C. Palumbo. 1995a. Action thresholds forBemisia argentifolii (Homoptera: Aleyrodidae) in Canta-loupe. J. Econ. Entomol. 88: 1733Ð1738.

Riley, D. G., and J. C. Palumbo. 1995b. Interaction of sil-verleaf whiteßy (Homoptera: Aleyrodidae) with canta-loupe yield. J. Econ. Entomol. 88: 1726Ð1732.

Romanow, L. R., O.M.B. de Ponti, and C. Mollema. 1991.Resistance in tomato to the greenhouse whiteßy: analysisof population dynamics. Entomol. Exp. Appl. 60: 247Ð259.

Romanow, L. R., J. W. Moyer, and G. G. Kennedy. 1986.Alteration of efÞciencies of acquisition and inoculation ofwatermelon mosaic virus 2 by plants resistant to the virusand to an aphid vector. Phytopathology 76: 1276Ð1281.

Sauvion, N., C. Pavis, A. Huc, M. Rousseau, and N. Boissot.1999. Caracterisation de Bemisia tabaci biotype B(Hemiptera: Aleyrodidae) en Guadeloupe. Annales de laSociete Entomologique de France 35: 46Ð53.


Simmons, A. M., and J. D. McCreight. 1996. Evaluation ofmelon for resistance to Bemisia argentifolii (Homoptera:Aleyrodidae). J. Econ. Entomol. 6: 1663Ð1668.

Slosser, J. E., W. E. Pinchak, and D. R. Rummel. 1989. Areview of know and potential factors affecting the pop-ulation dynamics of the cotton aphid. Southwest. Ento-mol. 14: 302Ð313.

Soria, C., J. A. Diaz, E. Moriones, and M. L. Gomez-Guil-lamon. 2000. Resistance to Aphis gossypii and to virustransmission by this aphid in melon. Acta Hortic. 510:305Ð312.

Soria, C., A. I. Lopez-Sese, and M. L. Gomez-Guillamon.1999. Resistance of Cucumis melo against Bemisia tabaci(Homoptera: Aleyrodidae). Environ. Entomol. 28: 831Ð835.

Soria, C., E. Moriones, A. Fereres, E. Garzo, and M. L.Gomez-Guillamon. 2003. New source of resistance tomosaic virus transmission by Aphis gossypii in melon.Euphytica 133: 313Ð318.

Soria, C., A. I. Lopez-Sese, and M. L. Gomez-Guillamon.1996. Resistance mechanisms of Cucumis melo var agres-tis against Trialeurodes vaporariorum and their use tocontrol a closterovirus that causes a yellowing disease ofmelon. Plant Pathol. 45: 761Ð766.

Stepansky, A., I. Kovalski, and R. Perl Treves. 1999. In-traspeciÞc classiÞcation of melons (Cucumis melo L.) inview of their phenotypic and molecular variation. PlantSyst. Evol. 217: 313Ð332.

Received 30 July 2004; accepted 31 October 2004.


Impact of Melon Accessions Resistant to Aphids on the Demographic Potential of Silverleaf Whitefly

Documents

Transcript of Impact of Melon Accessions Resistant to Aphids on the Demographic Potential of Silverleaf Whitefly