Mutation Research 566 (2004) 249–262

Review

The use of RAPD in ecotoxicology

Hans De Wolfa,∗, Ronny Blusta, Thierry Backeljaua,ba Department of Biology, University of Antwerp (UA), Groenenborgerlaan 171, B-2020 Antwerp, Belgium

b Royal Belgian Institute of Natural Sciences (KBIN), Vautierstraat 29, B-1000 Brussels, Belgium

Received 5 September 2003; received in revised form 13 October 2003; accepted 17 October 2003

Abstract

Toxic compounds may interfere with the genetic constitution of populations, either directly through mutagenic activity, orindirectly via population-mediated processes (i.e. selection, bottleneck). These processes are initiated when toxic compoundsreduce the survival and/or fecundity of exposed organisms, either through the accumulation of unfavorable mutations orwhen they adversely affect the physiology of an organism and/or the environment in which it has to survive. In this review,we describe how the RAPD technique can be applied in an ecotoxicological context, providing information on all directand indirect routes through which toxicants may affect the genetic structure of populations. Based on RAPD band intensity,gain/loss and band numbers, three major types of RAPD fingerprint analyses are discussed, yielding diagnostic, phenetic andgenetic information. Ecotoxicological literature examples demonstrate that, under strictly standardized conditions, the RAPDtechnique can be a useful tool to preliminary assess toxicological population genetic effects, particularly since this techniqueis relatively inexpensive and yields information on a large number of loci without having to obtain sequence data for primerdesign. However, currently only a small fraction of its potential is used in ecotoxicology. Statistical tools and parameters, asused in other RAPD studies, should be applied in ecotoxicological research as well in order to fully exploit the potential ofthis technique. Finally, due to their random nature, RAPD data often must be considered as preliminary until they are furtherdocumented by cloning, sequencing and probing techniques.© 2003 Elsevier B.V. All rights reserved.

Keywords:Random amplified polymorphic DNA; DNA damage; Selection; Bottleneck; Genetic variability

1. Introduction

Toxic compounds adversely affect individual healthwhen they interfere with the normal physiological pro-cesses of an organism. This interference may occur ona direct, toxicant to organism basis (i.e. primary toxiccompounds), or it may occur after the toxic compoundis transformed via physico-chemical interactions withits environment (secondary toxic compounds). As aresult, toxicants may partially shape the demography

∗ Corresponding author. Tel.:+32-3-218-04-78;fax: +32-3-218-04-97.E-mail address:hans.dewolf@ua.ac.be (H. De Wolf).

of populations, potentially affecting their fate, includ-ing the risk of extinction[1]. Obviously, this risk is notonly determined by environmental toxicants, but willalso depend on (1) the populations’ demographic prop-erties [2], (2) the available genetic potential neededto cope with and adapt to this environment[3] and(3) other external conditions[4]. All these factors are,nevertheless, closely related as toxicant exposure af-fects the genetic potential of populations[5,6], whilethe extent of these effects may depend on various ex-ternal factors[7,8].

Toxicant induced population genetic effects mayarise from the direct action of the toxicant at the DNAlevel (mutagenic effects) or may indirectly result

1383-5742/$ – see front matter © 2003 Elsevier B.V. All rights reserved.doi:10.1016/j.mrrev.2003.10.003

250 H. De Wolf et al. / Mutation Research 566 (2004) 249–262

Fig. 1. Flow chart representing the major routes by which toxicants are known to affect the genetic structure of natural populations, eitherincreasing (↑), decreasing (↓) or not affecting the genetic variability (→). The intensity of the arrows that are directed towards the geneticvariability boxes, are indicative for the likelihood of occurrence (diagonal arrows represent increments and decrements).

from population-mediated processes that are relatedto the toxicant exposure (Fig. 1) [5,6,9]. Indeed,germ cell mutations may not only alter the geneticconstitution of a population directly by introducingnew heritable variation[10], but often also reducethe viability of gametes, embryos and neonates, af-fecting the population genetic structure indirectly aswell [11,12]. In contrast somatic mutations are notpassed on to the next generation, except in asex-ual or clonal organisms, but may lead to cell deathor may transform the cell into malignancy[12,13],eventually reducing the fitness of affected individ-uals as well[11,14]. Hence, mutagenic compoundsmay either directly or indirectly affect the geneticpopulation structure, respectively, through heritablegerm cell mutations or through mutagenic-mediatedmortality and/or curtailment of reproduction (Fig. 1).Obviously, non-mutagenic toxicants can only affectthe genetic constitution of a population indirectlyby interfering with the physiology of an organism(physiological effects)[15–17] or by affecting theenvironment in which it has to survive (ecological ef-fects)[18–20], potentially decreasing the individuals’fitness and/or reproductive capacity (Fig. 1).

When these adverse population effects are drasticand affect individuals randomly (i.e. irrespective of thegenotype), they will lead to a bottleneck[21] (Fig. 1).During this process, the genetic variation is expectedto decrease due to the random loss of low frequencyand rare alleles[12]. Genetic drift is especially harm-ful in small populations, as it reduces the availablegenetic potential in each generation by a factor 1/2Ne(Ne = effective population size)[22]. Moreover, dele-terious mutations may become easily fixed in smallpopulations, leading to inbreeding depression, whichmight ultimately result in a mutational meltdown ofthe population[12,23]. In contrast, a bottleneck maytheoretically increase the genetic variability in a pop-ulation as well, as rare alleles may become more com-mon due to inbreeding, while previously unexpressedgenes (e.g. epistatic, epigenic effects) might becomeexpressed, increasing the phenotypic variation[24].

When differences in mortality and/or reproductivecapacity are fitness dependent, they will affect indi-viduals in a non-random fashion (i.e. genotype de-pendent) and are assumed to be driven by naturalselection (Fig. 1). Indeed, if certain genotypes aremore susceptible to mutagenicity and/or toxicant re-

H. De Wolf et al. / Mutation Research 566 (2004) 249–262 251

lated physiological or ecological effects than others,the toxicant may act as a selective agent upon thoseloci that are crucial for the survival and reproductionof an organism[6,24]. Allele frequencies at those locimay then be shifted in a relatively short period oftime, depending on the strength of the selective force[24]. Generally, genetic variability at those loci willdecrease, except when the fitness of heterozygotes issuperior to the fitness of homozygotes for the locus ofinterest (i.e. heterosis). However, if “survivors” withresistant alleles vary at other loci, the overall geneticvariability in an impacted population may not be af-fected at all. In contrast, when toxicant resistance ishighly polygenic, the toxicant induced selection maydecrease the overall genetic variation[24]. Hence, de-pending upon the genetic basis for resistance, selec-tion for toxicant tolerance may range from no changeto very significant genetic changes in exposed popu-lations[24] (Fig. 1).

Population genetic analyses offer powerful tools toassess these effects, as these analyses examine the cur-rent population genetic diversity, infer its recent his-tory and may predict future population directions aswell [6]. However, the predictive value of populationgenetics must be viewed with caution, since the pre-cise future of a specific population may be impossibleto infer when genetic drift is a major determinant. In-deed, in this case one can make general predictionswithin a given probability framework but it will, bydefinition, be impossible to exactly predict how al-lele frequencies will change in any particular popu-lation. Nonetheless, population genetic analyses maystill pick up population genetic phenomena, associatedwith negative demographic effects, that would other-wise remain largely undetected[25].

Initially, protein markers (i.e. allozymes) were usedto infer the population genetic effects of toxicant expo-sure[24,26–30], but currently a wide variety of DNAmarkers/techniques are available to this end. One ofthese techniques involves the random PCR amplifi-cation of anonymous DNA, termed random amplifiedpolymorphic DNA (RAPD;[31]), arbitrarily primedPCR (AP-PCR;[32]) or DAF (DNA amplification fin-gerprinting;[33]). In this paper we will review the useof RAPD in an ecotoxicological context, demonstrat-ing that it can be applied to infer all routes throughwhich toxicants may affect the genetic structure of ex-posed populations (Fig. 1).

2. RAPD

Allozyme electrophoresis has long been the stan-dard population genetic tool in ecotoxicological re-search, since (1) allozyme alleles are co-dominant andare generally supposed to exhibit simple Mendelianinheritance and (2) comparison of homologous lociacross populations and related species are straightfor-ward[34]. Nonetheless, allozymes have some seriousdrawbacks too. Firstly, they only survey genes encod-ing a small number of soluble metabolic enzymes andthus represent a biased genomic sample. Secondly,they require fresh tissue samples and thirdly their ex-pression may be regulated by a series of transcriptionaland post-transcriptional processes. Hence, allozymepatterns may represent true genomic variability, butcan equally well represent an environmentally inducedphenotypic expression of the genome (i.e. gene regu-lation effects).

In contrast to allozyme electrophoresis, RAPD[31]and related DNA fingerprinting methods[32,33] inferthe genetic variability directly at the genome level,provide a less biased genomic sample and are able togenerate a nearly unlimited number of markers, giventhat DNA fragments are flanked by sequences that arecomplementary to a specific primer and located ondifferent template strands that are within 3 kb of eachother[35]. Generally, RAPD reactions are performedwith a single 10 bp primer and amplified fragmentsare visualized by agarose or polyacrylamide gel elec-trophoresis and subsequent staining with ethidiumbro-mide. The resulting DNA profiles may differ amongindividuals, depending on the (1) presence/absence(p/a) of priming sites, (2) priming complementarycompleteness/incompleteness or (3) the distance be-tween priming sites. Hence, RAPD bands are lost orgained when point mutations, inversions, deletions,additions or gross chromosomal rearrangements af-fect the p/a of primer sites, their complementarity toprimers and/or the distance between priming sites.RAPD bands of different molecular weight are in-terpreted as separate loci which are scored on apresent (amplification)—absent (non-amplification)basis [35]. Additionally, band intensity differencesmay be interpreted as well through visual inspectionor by using image analysis software. Intensity differ-ences are suggested to result from product copy num-ber differences, competition between PCR products,

252 H. De Wolf et al. / Mutation Research 566 (2004) 249–262

heterozygosity, co-migration or partial mismatchingof primer sites[36].

Although RAPD has some clear advantages overallozyme electrophoresis, it also possesses someimportant limitations. Reproducibility is one of itsmajor pitfalls and is influenced by PCR conditions[36]. However, even if RAPD reproducibility canbe achieved[37–40], replicates should always beincluded and bands that fail to be reproduced consis-tently should not be considered[41–45]. A secondlimitation arises from the dominant/recessive charac-ter of RAPD bands[35]. As a consequence heterozy-gotes cannot be distinguished from homozygotes ofthe dominant allele[35,46]. Thirdly, the nature of thegenomic change that is scored is not known. Fourthly,RAPD may not screen the genome as randomly asexpected. Given that most RAPD primers have a highGC content, necessary for successful annealing at lowtemperatures, they may tend to screen GC-rich regionswhich are not evenly distributed across the genome[36]. Finally, due to the random nature of amplifica-tion, both nuclear and organelle DNA may be ampli-fied during PCR. For example, Aagaard et al.[42]showed that up to 19% of the bands from the RAPDfingerprints of the Douglas fir (Pseudotsuga men-ziesii) could be attributed to organelle DNA. UnlikeRAPD bands, generated by nuclear DNA, amplifiedorganelle fragments cannot be regarded as Mendelianmarkers, since they are haploid and clonally inherited.Hence, organelle DNA amplification may complicatethe interpretation of RAPD fingerprints.

Next to these specific limitations, there are sev-eral general constraints that apply to RAPD as to anyDNA technique involving amplification and/or elec-trophoretic separations. These include, DNA contami-nation, amplification competition and band homology[36,47]. All of these problems can, however, be over-come by technical means such as the use of blanks ineach PCR run, competitive PCR, southern blot and se-quencing analysis of the amplified products or by im-proving electrophoretic separation and band stainingprocedures[47].

3. RAPD as diagnostic markers

RAPDs are commonly used as markers to discrimi-nate particular taxa by the presence/absence of certain

“diagnostic” RAPD bands[48–50]. They have, how-ever, also been used successfully in an ecotoxicologi-cal context, to diagnose the genotoxicity of a varietyof compounds, including endocrine disruptors[51],benzo{a}pyrene (BaP)[25,52,53], mitomycin-C[13],Cu [54] and UV radiation[55]. This has been donethrough the analysis of band intensities and/or bandgain/loss variation between exposed and non-exposedindividuals. Indeed, the gain/loss or intensity dif-ferences of RAPD bands may be related to DNAdamage, mutations or structural rearrangements in-duced by genotoxic agents, affecting the primer sitesand/or interpriming distances[53]. Although theRAPD assay does not provide information on thenature and extent of these genotoxic induced DNAalterations, it can sometimes be used in a quantitativeway [53]. For instance, Atienzar et al.[25] have usedthe RAPD assay to determine the genotoxic effectsof benzo{a}pyrene in clonalDaphnia magna. TwoRAPD primers revealed different values in RAPDband numbers, size and intensities between exposedand non-exposed individuals[25]. Atienzar et al.[25] used these differences to estimate the percent-age of genomic template stability, calculated as 100− (100a/n), wheren is the number of bands in thecontrol DNA (i.e. non-exposed individuals) and a isthe average number of DNA changes in the exposedindividuals. In this way, Atienzar et al.[25] demon-strated that the DNA template stability parameterpicked up significant effects at BaP exposure levelsat which fitness parameters failed to detect significanteffects.

Next to genotoxicity, a variety of other factorsincluding DNA copy number differences, heterozy-gosity, co-migration or competition between PCRproducts[36] may affect the intensity of RAPD bands,but these issues were not considered by Atienzar et al.[25]. Yet, because of these confounding factors, someauthors argue that only well resolved highly intenseRAPD bands should be used when analyzing RAPDprofiles [41,42,44,45], thus rejecting RAPD bandswhich reveal intensity differences between individu-als. In this respect, Atienzar et al.[52] performed adaphnid-BaP exposure experiment in which RAPDprofile changes were also noted in the non-exposedBaP individuals. The authors argued that these effectsprobably resulted from (1) steady levels of genetic al-terations, (2) changes in metabolic processes, and/or

H. De Wolf et al. / Mutation Research 566 (2004) 249–262 253

(3) variation in gene expression (e.g. methylation),although gene expression related structural effects arelikely to be lost during the denaturation step in thePCR[52]. Atienzar et al.[52], nonetheless, concludedthat DNA damage and mutations are the main factorsthat influence RAPD pattern variation between BaPexposed and non-exposed individuals, provided thata sufficient number of cells are affected by the geno-toxicity [53]. In addition, Jones and Kortenkamp[56]demonstrated that genomic alterations can only bepicked up by the RAPD assay if they affect at least2% of the cells, which makes this assay less sensitivethan for instance the Ames or hprt test[56]. Finally,given that the RAPD assay predominantly screensGC-rich regions which often contain fast mutatingsites[47], observed genotoxicity effects may be up-wardly biased. In view of these limitations RAPDscan be considered as a relatively cheap but “quickand dirty” method to preliminary screen populationsfor genotoxic effects, if replicates and non-exposedindividuals are carefully monitored and analyzed aswell. Moreover, eventual suggestive genotoxicity re-lated RAPD alterations require further investigationusing cloning, sequencing and/or probing techniques,in order to verify and determine the extent and natureof the DNA alterations.

4. RAPD phenetic numerical analysis

The phenetic numerical analysis of RAPD profilesis the most popular way to deal with RAPD in ecotox-icology and has been used to determine the pheneticdiversity in natural populations exposed to a variety ofpollutants including metals[57,58], and radionuclidecontamination[59–61]. The phenetic approach differsfrom the genetic approach (see[5]) in that RAPD pro-files are not considered as genotypes. Hence, diversityis estimated on a phenetic basis only in which no as-sumptions have to be made regarding Hardy–Weinbergequilibrium conditions or the allelic constitution ofthe RAPD markers used. Instead, diversity is esti-mated by comparing the absence/presence of RAPDbands in exposed and non-exposed individuals (i.e.similarity analysis)[62], or by counting their RAPDband numbers and considering band abundancies (i.e.diversity analysis)[63,64]. A variety of methods,taken from community ecology, are available to calcu-

late these RAPD similarity and diversity measures orindices.

RAPD similarity is usually expressed via the simi-larity index proposed by Lynch[62] (S), which is thesame as the index of Czekanowski[65], Dice [66],Sørensen[67], or Nei and Li[68]. It is defined as

S = 2nij

ni + nj

wherenij is the number of bands shared by individ-uals i and j, and ni and nj are the total number ofRAPD bands censored in individualsi and j, respec-tively [62]). S ranges between zero and one, indicat-ing, respectively, that two individuals share no bandsor that they have identical band profiles. If populationsare differentiated it is expected that the mean propor-tion of bands shared by individuals from the samepopulationS̄w, exceeds the mean number of sharedbands from pairs of individuals from distinct popula-tionsSb [69]. Given thatS-values, sharing a commonindividual (e.g.Sij andSik), are non-independent, ap-propriate test statistics should be applied in order tocompareSw andSb correctly. In this respect Danforthand Freeman-Gallant[70] proposed a modified t-testin whichSw, Sb and their standard errors are estimatedfrom subsets of independentS-values, rather than fromthe completeS-value matrix. More recently, Bertorelleet al.[69] described five permutation tests that can beused to compareSw andSb as well. Dissimilarity be-tween populations is defined asD = 1 − Sb, and canbe used to construct a matrix of interphenotypic dis-tances, which in turn can be used for clustering andordination of populations (see below).

In community ecology, diversity measures aregrouped into three categories, viz. species richnessindices (i.e. number of species), species evenness/dominance indices (i.e. proportional species abundan-cies) and species abundance models[71]. In RAPDresearch the Shannon–Wiener (H′) [72,73] and Simp-son (D) index[74] are most commonly used[75–77].Both indices belong to the second category. The Simp-son index is a proportional abundance index, whereasH′ combines both richness and evenness into a singlefigure [71]. D andH′ are, respectively, formulated as

D =k∑

i=1

(ni

N

)2

254 H. De Wolf et al. / Mutation Research 566 (2004) 249–262

or

D =k∑

i=1

[ni(ni − 1)

N(N − 1)

],

and

H ′ = −k∑

i=1

pi ln(pi)

wherek is the number of RAPD bands considered,pi

is the frequency of theith RAPD band,ni is numberof individuals that possess theith RAPD band andNis the number of individuals analyzed. Both indicesdiffer in their sensitivity to pick up effects of com-mon or uncommon observations[78]. By analogywith community ecology, the Simpson index shouldbe more sensitive to common RAPD bands, whileH′ is more sensitive to rare RAPD fragments. Sincediversity indices represent independent data, anal-ysis of interpopulation diversity is not constraintedby non-independency, as is the case in the similar-ity analysis. Although both diversity indices havebeen used on numerous occasions in RAPD relatedresearch[75–77], they have only marginally beenadopted in an ecotoxicological context[61]. Indeed,the frequency and mean number of RAPD bandsare generally reported and tested for their differenceamong exposed and non-exposed populations, whilethe phenetic diversity is generally estimated on a bandsharing based similarity index only.

Theodorakis and Shugart[59] used the numericalphenetic RAPD approach to determine the similarityin mosquito fish (Gambusia affinis) populations col-lected at two radionuclide contaminated and two pris-tine sites. Mosquitofish taken from one of the con-taminated sites originally descended from one of thepristine sites, but had been transplanted to the ra-dionuclide contaminated area, 20 years prior to theRAPD survey. From the RAPD profiles generated by15 RAPD primers, band numbers were counted andband frequencies were determined. Out of 142 am-plified bands, 17 were significantly more frequent inexposed populations and were therefore regarded asdiagnostic for radionuclide contamination. However,there was a substantial level of variation, as the fre-quency of 10 out of these 17 diagnostic bands dif-fered significantly between both radionuclide popu-lations (see Table 2 in[59]). In addition, six RAPD

primers yielded a significant higher average number ofbands at the contaminated sites and were by analogyregarded as contaminant indicative primers. Similar-ity and dissimilarity were calculated using the equa-tion proposed by Lynch[62] (see above) and indicatedan increased dissimilarity in the radionuclide contam-inated areas relative to the reference sites. These re-sults showed that on a relatively short period of time,individuals with a common background, may revealphenetic differentiation when placed under differentialenvironmental conditions. Theodorakis and Shugart[59] were also able to associate fecundity with someof the contaminant indicative RAPD bands, since inboth contaminated areas, they found that mosquitofishpossessing some of these bands had a higher fecunditythan fish which did not. These findings suggest that thedifferences among the contaminated and pristine sitesare due to radionuclide induced selection, although ge-netic drift or selection due to other environmental fac-tors, not necessarily related to radiation exposure, maybe responsible as well[59]. Nevertheless, in a compa-rable study onGambusia holbrooki, new radionuclideindicative RAPD bands were detected, which aftersouthern blot analysis revealed homology with someof the contaminant indicative RAPD bands found inG. affinis[60]. The fact that both species reveal sim-ilar radionuclide indicative RAPD bands at differentradionuclide contaminated localities, provides furtherevidence for a selection induced differentiation.

Hence, phenetic numerical analyses provide usefulinformation on toxicant exposure related populationeffects, while they simultaneously screen populationsfor diagnostic RAPD loci as well. Unfortunately, onlya limited portion of their potential is currently ex-plored in ecotoxicology. Indeed, diversity indices areonly rarely estimated, while similarity based distancematrices are only sporadically constructed[59,61].Yet, such distance matrices are useful, as they form thebasis to uncover phenetic structuring of exposed andnon-exposed populations either via clustering or viaordination techniques. For instance, Theodorakis andShugart[59] used the UPGMA clustering algorithm,while Turuspekov et al.[61] performed a principalcoordinate ordination (i.e. metric multidimensionalscaling) analysis, describing the RAPD relationshipsamong radionuclide contaminated and pristine popula-tions. Although these techniques helped to investigatethe population structure, they could not directly infer

H. De Wolf et al. / Mutation Research 566 (2004) 249–262 255

the diagnostic nature of the RAPD bands analyzedin both studies. In order to obtain such information,both the relationships between RAPD bands and thepopulations they were taken from must be viewedsimultaneously in one plot. This type of ordinationcan be achieved via reciprocal averaging using for in-stance correspondence analysis (CA)[79,80]. CA hasbeen used for a long time in community ecology, or-dinating populations and species in a reduced dimen-sional space on the basis of species presence/absencedata[81–83]. Although powerful and highly applica-ble to RAPD data, this ordination technique has onlymarginally been applied in a RAPD analyses[84],and never in an ecotoxicological context. However,CA can simply be performed starting with a matrix ofRAPD abundancies in which the columns and rows,respectively, represent the populations (1 top) andRAPD bands (1 ton) under investigation. Each rowis subsequently associated with a single “RAPD bandvalue” (a1 to an), while each column is associatedwith a single “population value” (b1 to bp) throughreciprocal averaging. Consequently, theith RAPDband valueai is a weighted average of the populationvalues, with populationj having a weight that is pro-portional to its RAPD bandi occurrence, relative tothe number of timesi is expressed in all populations.Similarly, the jth population valuebj is a weightedaverage of the RAPD band values, with RAPD bandihaving a weight that is proportional to its occurrencein populationj, relative to the total number of RAPDbands that are expressed in that population[85]. As aresult of this reciprocal averaging, RAPD bands andpopulations can be plotted simultaneously along thesame axes, visualising both the RAPD band distribu-tion as well as the resulting interpopulation relation-ships. The eigenvalues (λ) that correspond to the newlyconstructed axes will measure how well RAPD bandand population values are correlated[82]. Alternativeto CA, one could also perform a two-way indicatorspecies analysis (TWINSPAN)[86,87]. This tech-nique is also based on reciprocal averaging and organ-ises the data along an axis as an ordination techniquewould, but then divides the axis into two parts, form-ing a dichotomy[86]. The classification is further re-fined using discriminant function analysis and weight-ing algorithms, placing RAPD bands/populations oneither side of the dichotomy[82]. Each side of thisdichotomy is then further subdivided into smaller di-

chotomies using the same process. Classification ofpopulations and corresponding RAPD bands contin-ues until some predefined criteria set by the user isreached[82]. Compared to CA, TWINSPAN has thedisadvantage of dividing populations which could inpractice be grouped together[82]. However, given thatTWINSPAN has proven to be extremely useful wheninferring the species distribution along environmen-tal gradients of both natural[83] and anthropogenicorigin [88], we expect it to be well-suited for the phe-netic numerical data analysis of RAPD profiles takenfrom individuals along pollution gradients.

5. RAPD genetic numerical analysis

Because RAPDs behave as dominant markers (i.e.present: AA or Aa, absent: aa), the double recessivegenotype fraction (x) is the only fraction which can ac-tually be inferred from the RAPD fingerprints, giventhat di-allelism is assumed. Based on this genotypefrequency, one can readily obtain the recessive allelefrequencyq, as

√x. However, due to the dominant

recessive nature of RAPD alleles, this estimate maybe downwardly biased. Lynch and Milligan[46] sug-gested to reduce this bias by restricting RAPD analy-ses to RAPD bands which are expressed at a frequencybelow 1− 3/N, whereN is the number of individualssampled at a given population, and to include a cor-rection factor, so that the “unbiased” recessive allelefrequencyq̂, can be obtained as

q̂ =√

x̂

[1

1 − (varx̂/8x̂2)

],

with

varx̂ = x̂(1 − x̂)

N

Under the assumption that each RAPD band repre-sents a di-allelic locus whose genotype frequencies arein Hardy–Weinberg equilibrium, one can then easilyobtain the dominant allele frequency (p) as 1− q̂, andthe heterozygotic and dominant homozygotic geno-type frequencies, respectively, as 2pq̂ andp2. Based onthese allele and genotype frequencies, other standardpopulation genetic parameters may be estimated. Thelevel of population differentiation, expressed asFST[89] may be estimated via the formulas proposed by

256 H. De Wolf et al. / Mutation Research 566 (2004) 249–262

Lynch and Milligan[46] or may be inferred from theapproach, proposed by Black and co-workers[90,91]as

FST =[

varp

p̄(1 − p̄)

]

Genetic variability can also be estimated at the nu-cleotide level through the nucleotide diversity (π) [68].Defined as the average number of nucleotide differ-ences per site, between two randomly chosen DNA se-quences in a population[68], π can be obtained from:

π =k∑

i=1

xixjdij ,

wherek is the total number of alleles,xi andxj are thefrequencies of theith andjth allele, respectively, anddij is the number of nucleotide differences or substi-tutions per site between both alleles[68].

Although allele frequency based RAPD analyseshave been often used[77,91–93], they have onlymarginally been applied in ecotoxicology. For in-stance, Theodorakis et al.[94] estimated heterozy-gosity and calculatedFST values to assess the geneticvariability and structure of radionuclide exposed andnon-exposed kangaroo rat (Dipodomys merriami)populations, while Ma et al.[95] estimated nucleotidediversity to infer levels of genetic variability in popu-lations of the musselMytilus galloprovincialisand thebarnacleBalanus glandula, from pristine and contam-inated sites in southern California bay. Theodorakiset al. [94] detected similar heterozygosity values andsmall FST values among the four analyzed kangaroorat populations, indicating little if any genetic popu-lation differentiation. These results where explainedin terms of kangaroo rat migratory patterns, sincethe contaminated areas represented ecological sinks[94]. Ma et al.[95] found consistently lower levels ofnucleotide diversity in bothB. glandulaandM. gal-loprovincialis at polluted sites compared to pristineareas, although the diversity levels differed dependingon the primer choice, and species and or sites thatwere considered. A genetic distance based UPGMAtopology, revealed that within both species, popu-lations from impacted sites were genetically moresimilar to one another than they were to populationsfrom nearby clean sites[95]. Hence, Ma et al.[95]concluded that the adaptive potential ofB. glandula

and M. galloprovincialis had been compromised atthe contaminated bay areas, despite both species arestill occurring at high densities at those polluted sites.

Both literature examples[94,95] illustrate that thenumerical genetic RAPD analysis may be useful toassess pollution related population genetic effects, de-termining the genetic structure and variability and/orassessing how phenomena like dispersal and migra-tion interfere with pollution related influences[94].However, despite its potential, this RAPD approachhas several limitations. Firstly, since RAPD loci aresupposed to be neutral markers, they will only pickup effects of selection, when linked to non-neutralloci under selection, in a process termed genetichitchhiking [6,96]. Secondly, if selection, or otherfactors such as drift, non-random mating, mutation,or migration shape the genetic population structure ofexposed populations, Hardy–Weinberg assumptionsmay be violated so that allele frequencies can nolonger be inferred from the RAPD loci. Consequentlyallele frequency based population parameters and teststatistics (e.g. heterozygosity, nucleotide diversity,F-statistics,FST based gene flow estimates, etc.) canno longer be applied and should be replaced by othertechniques. In this respect, Borowsky[97] proposedan alternative method for the assessment ofπ whichis not based on allele frequencies but instead usesthe proportion of mismatched RAPD bands betweentwo individuals drawn at random from a population(φ) and the number of discriminating sites in theamplification system. Indeed, because of the linearrelationship betweenπ and φ, at least within a bio-logically relevant range of values,π can be inferredfrom φ, using the following equation[97]:

π ≈ 3

4

φe

m,

where m is two times the number of bases of theprimer, when one primer is used, or where m is the sumof the bases when two primers are involved and whereφe is the number of mismatched bands between tworandomly chosen individuals of a population, averagedover all RAPD profiles generated by the primer(s) used[97]. Theπ estimates that are obtained via replicatesare generally consistent within and among populationsof the same species[97]. According to Borowsky[97]this can be expected, given that RAPDs (1) randomlyscreen the genome (but see above) and (2) provide an

H. De Wolf et al. / Mutation Research 566 (2004) 249–262 257

extended sample size from whichφ may be inferred.Similar to the nucleotide diversity estimate[97], ge-netic population differentiation can also be inferredon a non-allele frequency related basis. In this re-spect, Excoffier et al.[98] described how populationgenetic variation can be partitioned into pre-definedhierarchical levels of structuring through an analysisof molecular variance (AMOVA), using differencesbetween molecular sequences. In AMOVA, Euclediandistances are calculated among all pairs of haplotypes,which are represented by 1/n present/absent vectors,wheren represents the number of scored RAPD loci.RAPD bands that are shared by 95% or more of theindividuals should be regarded as monomorphic andshould be excluded from the analysis, since they willnot contribute to the estimation of the distance[21].Because the sum of squares in a conventional ANOVAcan be written as the sum of squared distances be-tween pairs of haplotypes, the squared Euclidian dis-tance matrix can be converted into a nested analysis ofvariance from which the between and within variancecomponents (σ2) can be extracted[45,98,99]. Hence,based on molecular sequence information, these vari-ance components summarise the degree of differen-tiation between population division and can be usedto calculate theφ statistics[98], which are analogousto the allele frequency basedF-statistics proposedby Wright [89]. Although described for co-dominantgenetic markers, the AMOVA test statistic proposedby Excoffier et al.[98] has been used on numerousoccasions for RAPD data[43–45,100–102], includingtoxicological related topics[94,103]. When assuminglinkage equilibrium, AMOVA variance componentscan easily be inferred from RAPD profiles taken fromcontaminated and pristine areas when a three levelhierarchical organization is assumed in which popu-lations (P) are nested within areas of contamination(A), yielding the followingφ statistics:

φAT = σ2A

σ2w + σ2

P + σ2A

φPA = σ2P

σ2w + σ2

P

(1 − φPT) = (1 − φPA)(1 − φAT),

whereσ2A = variance among areas,σ2

w = variancewithin populations,= σ2

P variance among populations,

φAT = ratio of area variance to total variance,φPA =ratio of within area variance to population variance andφPT = ratio of variance among populations to totalvariance. In this particular case,φPT is analogous toFST, and can be used to infer and test levels of geneticpopulation differentiation[43,45,100,101]or may beused to estimate gene flow among populations[102]according to the formula of Wright[89], which canbe rewritten as

Nm≈ 1

4

(1

φPT− 1

)

6. Conclusions

The examples taken from literature illustrate thatRAPDs are not only useful to infer genotoxic relatedpopulation genetic effects by considering band inten-sity differences or gain/loss of RAPD bands (diagnos-tic analysis), but they may also provide informationon the overall genetic diversity (phenetic numericalanalysis) and structure of populations (genetic numer-ical analysis) (Fig. 2). Hence, given that RAPDs arerelatively inexpensive and yield information on a largenumber of loci without having to obtain sequence datafor primer design, they are useful to investigate or pre-liminary assess the different routes in which toxicantexposure may affect the genetic structure of naturalpopulations (Figs. 1 and 2), especially when primersfor more informative, co-dominant microsatellite lociare lacking. Nevertheless, given its high sensitivity toPCR conditions and its random amplification, repro-ducibility remains its major pitfall. Hence, in orderto obtain reliable results, DNA isolation and PCRconditions must be strictly standardized, while repli-cates should be carefully analyzed with respect topresence/absence and intensity differences of RAPDbands. In addition, since pollutants can have multipleeffects on the genetic structure of populations (Fig. 1),most of the obtained RAPD information is bound toremain preliminary until corroborated by a broadermolecular/ecological analysis. Indeed, if the new ormore frequently expressed RAPD bands in exposedpopulations are not sequenced and their function orlinkage to other loci is not determined and/or if theycannot be associated with fitness parameters, fewconclusions can be drawn from the ecotoxicological

258 H. De Wolf et al. / Mutation Research 566 (2004) 249–262

Fig. 2. Flow chart representing the different levels at which RAPD based analyses can be used to infer the different relations shown inFig. 1 (diagonal arrows represent increments and decrements).

related RAPD analysis. The time and resources neededto gain this additional information may perhaps notoutweigh the ease and speed for which the RAPDanalysis was initially developed. Hence, as pointedout above the RAPD analysis is a useful tool to pre-liminary assess the genetic structure of natural popu-latons but will not provide clear cut answers withoutfurther laborious investigation of the newly or morefrequently expressed bands (see also[7]). Finally it isclear that several data exploring tools discussed hereremain to be applied in RAPD related ecotoxicology.The introduction of these statistical tools may, indeed,prove to be of great interest and value, especially withrespect to ecotoxicological related diagnostics.

7. Future perspectives

Several RAPD applications have been developedwhich have proven or at least have the potential tobe useful in ecotoxicological research. One of theseRAPD applications involves the use of bulked segre-gant analysis (BSA) to detect genetic markers linkedto any specific trait of interest[104]. The RAPD pro-

files of two pooled (i.e. bulked) DNA samples of in-dividuals from a segregating population, originatingfrom a single cross are compared[104]. Within eachbulk, the individuals are identical for the trait or geneof interest but are arbitrary for all other genes[104].The presence of a RAPD fragment in one bulk andthe absence in the other provides evidence for a puta-tively linked marker[104]. BSA–RAPD has been usedto detect DNA markers linked to disease resistance[105,106], morphological and biochemical character-istics [107–110]. Obviously, BSA–RAPD has greatpotential with respect to ecotoxicological related di-agnostics, discovering DNA sequences that are linkedto toxicant resistance genes.

Another application has been proposed by Theodor-akis et al. [60], in which diagnostic bands of onespecies are excised and purified from the agarose geland used to synthesize oligonucleotide probes. Subse-quently, these probes can be used in hybridization re-actions involving other species from the same locationalongside matched controls. In the case of Theodorakiset al. [60], probes were labelled with biotin and hy-bridization of the probes was investigated with chemi-luminescent detection. As mentioned inSection 4,

H. De Wolf et al. / Mutation Research 566 (2004) 249–262 259

the probe/hybridization approach of Theodorakis et al.[60] yielded useful information on the underlying evo-lutionary processes (selection versus drift) that shapedthe radionuclide contaminated mosquitofish popula-tions. Obviously, this technique may be highly use-ful in diagnostic research, while it could facilitatecross-laboratory comparisons as well. However, alsohere the identity/functionality of the diagnostic loci ortheir linkage to other loci under selection, remainedunclear without further molecular characterization.

Acknowledgements

The authors would like to thank two anonymousreferees who helped to improve the manuscript.This review was supported by the Fifth Frame-work programme of the European Commission un-der contract number EVK3-CT-2001-00048 (EU-MAR). Additional support was received fromOSTC-projects MO/36/003 and MO/36/008, as wellas from UA-project RAFO1 DEWOH KP02. HDWis a Postdoctoral Fellow of the Fund for ScientificResearch—Flanders (Belgium) (F.W.O.).

References

[1] H.L. Koh, T.G. Hallam, H.L. Lee, Combined effects ofenvironmental and chemical stressors on a model Daphniapopulation, Ecol. Model. 103 (1997) 19–32.

[2] J.D. Stark, J.A.O. Banken, Importance of populationstructure at the time of toxicant exposure, Ecotoxicol.Environ. Safety 42 (1999) 282–287.

[3] R. Burger, M. Lynch, Evolution and extinction in a changingenvironment—a quantitative genetic analysis, Evolution 49(1995) 151–163.

[4] W.D. Allmon, Nutrients, temperature, disturbance, andevolution: a model for the late Cenozoic marine record ofthe western Atlantic, Palaeogeogr. Palaeoclimat. Palaeoecol.166 (2001) 9–26.

[5] C. Becerril, H. Acevedo, M. Ferrero, F. Sanz, A. Castaño,DNA fingerprint comparison of rainbow trout and RTG-2cell line using random amplified polymorphic DNA,Ecotoxicology 10 (2001) 115–124.

[6] N.M. Belfiore, S.L. Anderson, Effects of contaminants ongenetic patterns in aquatic organisms: a review, Mutat. Res.489 (2001) 97–122.

[7] P. Bjerregaard, M.H. Depledge, Cadmium accumulation inLittorina littorea, Mytilus edulisand Carcinus maenastheinfluence of salinity and calcium ion concentrations, Mar.Biol. 119 (1994) 385–395.

[8] B.P. Jackson, P.J. Lasier, W.P. Miller, P.W. Winger, Effectsof calcium, magnesium and sodium on elevating cadmiumtoxicity to Hyalella azteca, Bull. Environ. Contam. Toxicol.64 (2000) 279–286.

[9] C.W. Theodorakis, Integration of genotoxic and populationgenetic endpoints in biomonitoring and risk assessment,Ecotoxcicology 10 (2001) 245–256.

[10] J.L. Staton, N.V. Schizas, G.T. Chandler, B.C. Coull,J.M. Quattro, Ecotoxicology and population genetics:the emergence of “phylogeographic and evolutionaryecotoxicology”, Ecotoxicology 10 (2001) 217–222.

[11] P.D.N. Hebert, M.M. Luiker, Genetic effects of contaminantexposure—towards an assessment of impacts on animalpopulations, Sci. Total Environ. 191 (1996) 23–58.

[12] J.W. Bickham, S. Sandhu, P.D.N. Hebert, L. Chikhi,R. Athwal, Effects of chemical contaminants on geneticdiversity in natural populations: implications for biomonito-ring and ecotoxicology, Mutat. Res. 463 (2000) 33–51.

[13] C. Becerril, M. Ferrero, F. Sanz, A. Castaño, Detection ofmitomycin C-induced genetic damage in fish cells by useof RAPD, Mutagenesis 14 (1999) 449–456.

[14] M.A. Cronin, J.W. Bickham, A population genetic analysisof the potential for a crude oil spill to induce heritablemutations and impact natural populations, Ecotoxicology 7(1998) 259–278.

[15] A.M. Farag, C.J. Boese, D.F. Woodward, H.L. Bergman,Physiological-changes and tissue metal accumulation inrainbow-trout exposed to foodborne and waterborne metals,Environ. Toxicol. Chem. 13 (1994) 2021–2029.

[16] M.H. Depledge, Z. Billinghurst, Ecological significance ofendocrine disruption in marine invertebrates, Mar. Pollut.Bull. 39 (1999) 32–38.

[17] L. Troncoso, R. Galleguillos, A. Larrain, Effects of copperon the fitness of the Chilean scallopArgopecten purpuratus(Mollusca: Bivalvia), Hydrobiologia 420 (2000) 185–189.

[18] R. Lande, Anthropogenic, ecological and genetic factors inextinction and conservation, Res. Popul. Ecol. 40 (1998)259–269.

[19] G. Beasley, P. Kneale, Reviewing the impact of metals andPAH’s on macro invertebrates in urban watercourses, Prog.Phys. Geogr. 26 (2002) 236–270.

[20] C.A. Edwards, Assessing the effects of environmentalpollutants on soil organisms, communities, processes andecosystems, Eur. J. Soil Biol. 38 (2002) 225–231.

[21] D.L. Hartl, A.G. Clark, Principles of Population Genetics,second ed., Sinauer Associates, Inc., Suderland, USA, 1998.

[22] S. Wright, Evolution in Mendelian populations, Genetics 16(1931) 97–159.

[23] A.T. Bernardes, Mutation load and the extinction of largepopulations, Physica A 230 (1996) 156–173.

[24] R.B. Gillespie, S.I. Guttman, Chemical-induced changes inthe genetic structure of populations: effects on allozymes,in: V.E. Forbes (Ed.), Genetics and Ecotoxicology, Taylor& Francis, Philadelphia, 1999, pp. 55–77.

[25] F. A Atienzar, M. Conradi, A.J. Evenden, A.N. Jha,M.H. Depledge, Qualitative assessment of genotoxicityusing random amplified polymorphic DNA: comparison

260 H. De Wolf et al. / Mutation Research 566 (2004) 249–262

of genomic template stability with key fitness parametersin Daphnia magnaexposed to benzo[a]pyrene, Environ.Toxicol. Chem. 18 (1999) 2275–2282.

[26] B. Lavie, E. Nevo, Heavy metal selection of phosphoglucoseisomerase allozymes in marine gastropods, Mar. Biol. 71(1982) 17–22.

[27] B. Lavie, E. Nevo, Genetic selection of homozygoteallozyme genotypes in marine gastropods exposed tocadmium, Sci. Total Environ. 57 (1986) 91–98.

[28] B. Nevo, R. Ben-Shlomo, B. Lavie, Mercury selection ofallozymes in marine organisms: predictions and verificationin nature, Proc. Natl. Acad. Sci. U.S.A. 81 (1984) 1258–1259.

[29] B. Nevo, R. Noy, B. Lavie, A. Beiles, S. Muchtar, Geneticdiversity and resistance to marine pollution, Biol. J. Linn.Soc. 29 (1986) 139–144.

[30] L. Tranvik, M. Sjörgen, G. Bengtsson, Allozymepolymorphism and protein profile inOrchesella bifasciata(Collembola): indicative of extended metal pollution?Biochem. Syst. Ecol. 22 (1994) 13–23.

[31] J.G.K. Williams, A.R. Kubelik, K.J. Livak, J.A. Rafalski,S.V. Tingey, DNA polymorphisms amplified by arbitraryprimers are useful as genetic markers, Nucleic Acids Res.18 (1990) 6531–6535.

[32] J. Welsh, M. McClelland, Fingerprinting genomes usingPCR with arbitrary primers, Nucleic Acids Res. 18 (1990)7213–7218.

[33] G. Caetano-Anolles, B.J. Bassam, P.M. Grensshoff, DNAamplification fingerprinting using very short arbitraryoligonucleotide primers, Biotechnology 9 (1991) 553–557.

[34] B.A. Schaal, W.J. Leverich, S.H. Rogstad, A comparison ofmethods for assessing genetic variation in plant conservationbiology, in: D.A. Falk, K.E. Holsinger (Eds.), Genetics andConservation of Rare Plants, Oxford University Press, NewYork, 1991, pp. 123–134.

[35] P. Fritsch, L.H. Rieseberg. The use of random amplifiedpolymorphic DNA (RAPD) in conservation genetics, in: T.B.Smither, R.K. Wayne (Eds.), Molecular Genetic Approachesin Conservation, Oxford University Press, 1996, pp. 55–73.

[36] S.A. Harris, RAPDs in systematics—a useful methodology?in: P.M. Hollingsworth, R.M. Bateman, R.J. Gornal (Eds.),Molecular Systematics and Plant Evolution, The SystematicsAssociation Special Volume Series 57, Taylor & Francis,London, UK, 1999, pp. 211–228.

[37] T. Benter, S. Papadopoulos, M. Pape, M. Manns, H.Poliwoda, Optimization and reproducibility of randomamplified polymorphic DNA in human, Anal. Biochem. 230(1995) 92–100.

[38] A. Gravesen, T. Jacobsen, P.L. Møller, F. Hansen, A.G.Larsen, S. Knøchel, Genotyping ofListeria monocytogenes:comparison of RAPD, ITS, and PFGE, Int. J. FoodMicrobiol. 57 (2000) 43–51.

[39] J. Fraga, L. Rojas, I. Sariego, C.A. Sarria, Optimizationof random amplified polymorphic DNA techniques for itsuse in genetic studies ofTrichomonas vaginalisisolates,Infection, Genet. Evolut. 2 (2002) 73–75.

[40] Y. Blixt, R. Knutsson, E. Borch, P. Rådström, Interlaboratoryrandom amplified polymorphic DNA typing ofYersinia

enterocoliticaandY. enterocolitica-like bacteria, Int. J. FoodMicrobiol. 83 (2003) 15–26.

[41] B. Kump, B. Javornik, Evaluation of genetic variabilityamong common buckwheat (Fagopyrum esculentumMoench) populations by RAPD markers, Plant Sci. 114(1996) 149–158.

[42] J.E. Aagaard, K.V. Krutovskii, S.H. Strauss, RAPD markersof mitochondrial origin exhibit lower population diversityand higher differentiation than RAPDs of nuclear origin inDouglas fir, Mol. Ecol. 7 (1998) 801–812.

[43] J.L. Bouzat, The population genetic structure of the greaterrhea (Rhea americana) in an agricultural landscape, Biol.Conserv. 99 (2001) 277–284.

[44] X.-J. Ge, Y. Yu, N.-X. Zhao, H.-S. Chen, W.-Q. Qi,Genetic variation in the endangered inner Mongolia endemicshrub Tetraena mongolicaMaxim. (Zygophyllaceae), Biol.Conserv. 11 (2003) 427–434.

[45] F. Maltagliati, P. Domenici, C.F. Fosh, P. Cossu, M.Casu, A. Castelli, Small-scale morphological and geneticdifferentiation in the Mediterranean killifishAphaniusfasciatus(Cyprinodontidae) from a coastal brackish-waterpond and an adjacent pool in northern Sardinia, Oceanol.Acta 26 (2003) 111–119.

[46] M. Lynch, B.G. Milligan, Analysis of population geneticstructure with RAPD markers, Mol. Ecol. 3 (1994) 91–99.

[47] T. Backeljau, L. De Bruyn, H. De Wolf, K. Jordaens, S.Van Dongen, R. Verhagen, B. Winnepenninckx, Randomamplified polymorphic DNA (RAPD) and parsimonymethods, Cladistics 11 (1995) 119–130.

[48] F.P. Nicese, J.I. Hormaza, G.H. McGranahan, Molecularchracterisation and genetic ralatedness among walnut(Juglans regia L.) genotypes based on RAPD markers,Euphytica 101 (1998) 199–206.

[49] D.R. Lacerda, J.P. Lemos, M.D.P. Acedo, M.B. Lovato,Molecular differentiation of two vicariant neotropicaltree species,Plathymenia foliolosa and P. reticulata(Mimosoidea), inferred using RAPD markers, Plant Syst. E235 (2002) 67–77.

[50] H.K. Barman, A. Barat, B.M. Yadav, S. Banerjee, P.K.Meher, P.V.G.K. Reddy, R.K. Jana, Genetic variationbetween four species of Indian major carps as revealed byrandom amplified polymorphic DNA assay, Aquaculture 217(2003) 115–123.

[51] F.A. Atienzar, Z. Billinghurst, M.H. Depledge, 4-n-Nonylphenol and 17-� estradiol may induce common DNAeffects in developing barnacle larvae, Environ. Pollut. 120(2002) 735–738.

[52] F.A. Atienzar, A.J. Evenden, A.N. Jha, M.H. Depledge, Useof the random amplified polymorphic DNA (RAPD) assayfor the detection of DNA damage and mutations: possibleimplications of confounding factors, Biomarkers 7 (2002)94–101.

[53] F.A. Atienzar, P. Venier, A.N. Jha, M.H. Depledge,Evaluation of the random amplified polymorphic DNA(RAPD) assay for the detection of DNA damage andmutations, Mutat. Res. 521 (2002) 151–163.

[54] F.A. Atienzar, V.V. Cheung, A.N. Jha, M.H. Depledge,Fitness parameters and DNA effects are sensitive indicators

H. De Wolf et al. / Mutation Research 566 (2004) 249–262 261

of copper induced toxicity inDaphnia magna, Toxicol. Sci.59 (2001) 241–250.

[55] F.A. Atienzar, B. Cordi, M.E. Donkin, A.J. Evenden, A.N.Jha, M.H. Depledge, Comparison of ultraviolet-inducedgenotoxicity detected by random amplified polymorphicDNA with chlorophyll fluorescence and growth in a marinemacroalgae,Palmaria palmate, Aquat. Toxicol. 50 (2000)1–12.

[56] C. Jones, A. Kortenkamp, RAPD library fingerprintingof bacterial and human DNA: applications in mutationdetection, Teratogen. Carcinogen. Mutagen. 20 (2000) 49–63.

[57] C. Conte, I. Mutti, P. Puglisi, A. Ferrarini, G. Regina, E.Maestri, N. Marmiroli, DNA fingerprinting analysis by aPCR based method for monitoring the genotoxic effects ofheavy metals pollution, Chemosphere 37 (1998) 2739–2749.

[58] K. Ross, N. Cooper, J.R. Bidwell, J. Elder, Genetic diversityand metal tolerance of two marine species: a comparisonbetween populations from contaminated and reference sites,Mar. Pollut. Bull. 44 (2002) 671–679.

[59] C.W. Theodorakis, L.R. Shugart, Genetic ecotoxicology. II.Population genetic structure in mosquitofish exposed in situto radionuclides, Ecotoxicology 6 (1997) 335–354.

[60] C.W. Theodorakis, J.W. Bickham, T. Elbl, L.R. Shugart,R.K. Chesser, Genetics of radionuclide-contaminatedmosquitofish populations and homology betweenGambusiaaffinis andG. holbrooki, Environ. Toxicol. Chem. 17 (1998)1992–1998.

[61] Y. Turuspekov, R.P. Adams, C.M. Kearney, Genetic diversityin three perennial grasses from Semipalatinsk nuclear testingregion of Kazakhstan after long-term radiation exposure,Biochem. Syst. Ecol. 30 (2002) 809–817.

[62] M. Lynch, The similarity index and DNA fingerprinting,Mol. Biol. E 7 (1990) 478–484.

[63] E.A. James, G.R. Ashburner, Intraspecific variation inAsteliaaustraliana(Liliaceae) and implications for the conservationof this Australian species, Biol. Conserv. 82 (1997) 253–261.

[64] A.P. Maldonado San Martin, L. Adamec, J. Suda, T.H.M.Mes, H. Štorchová, Genetic variation within the endangeredspeciesAldrovanda vesiculosa(Droseraceae) as revealed byRAPD analysis, Aquat. Bot. 75 (2003) 159–172.

[65] J. Czekanowski, Zarys method statystycznyck (Diegrundzuge der statistisken methoden), Warsaw, 1913.

[66] L.R. Dice, Measures of the amount of ecological associationbetween species, Ecology 26 (1945) 297–302.

[67] T. Sørensen, A method for establishing groups of equalamplitude in plant sociology based on similarity of speciescontent and its applications to analysis of the vegetation onDanish commons, Biol. Skr. 5 (1948) 1–34.

[68] M. Nei, W.H. Li, Mathematical model for studying geneticvariation in terms of restriction endonucleases, Proc. Natl.Acad. Sci. U.S.A. 74 (1979) 5267–5273.

[69] G. Bertorelle, L. Bucchini, A. Pilastro, C. Matessi, DNAfingerprinting data and the analysis of population geneticstructure by comparing band-sharing patterns, Mol. Ecol. 8(1999) 1851–1866.

[70] B.N. Danforth, C.R. Freeman-Gallant, DNA fingerprintingdata and the problem of non-independence among pairwisecomparisons, Mol. Ecol. 5 (1996) 221–227.

[71] A.E. Magurran, Ecological Diversity and Its Measurement,Chapman and Hall, London, 1988.

[72] N. Wiener, Cybernetics: or Control and Communication inthe Animal and Machine, Wiley, New York, 1948.

[73] C.E. Shannon, W. Weaver, The Mathematical Theory ofCommunication, University of Illinois Press, Urbana, IL,1949.

[74] E.H. Simpson, Measurement of diversity, Nature 163 (1949)688.

[75] M.A. Cardoso, J. Provan, W. Powell, P.C.G. Ferreiras, D.E.De Oliveira, High genetic differentiation among remnantpopulations of the endangeredCaesalpinia echinataLam.(Leguminosae—Caesalpinioideae), Mol. Ecol. 7 (1998) 601–608.

[76] T.M. Gabrielsen, C. Brochmann, Sex after all: high levels ofdiversity detected in the arctic clonal plantSaxifraga cernuausing RAPD markers, Mol. Ecol. 7 (1998) 1701–1708.

[77] M. Gaudeul, P. Taberlet, I. Till-Bottraud, Genetic diversityin an endangered alpine plant,Eryngium alpinum L.(Apiaceae), inferred from amplified fragment lengthpolymorphism markers, Mol. Ecol. 9 (2000) 1625–1637.

[78] H. Nagendra, Opposite trends in response for the Shannonand Simpson indices of landscape diversity, Appl. Geogr.22 (2002) 175–186.

[79] H.O. Hirschfeld, A connection between correlation andcontingency, Proc. Cambridge Philos. Soc. 31 (1935) 520–524.

[80] R.A. Fisher, The precision of discriminant functions, Anal.Eugen. 10 (1940) 422–429.

[81] I. Shennan, A.J. Long, M.M. Rutherford, F.M. Green, J.B.Innes, J.M. Lloyd, Y. Zong, K.J. Walker, Tidal marshstratigraphy, sea level change and large earthquakes, I: a5000-year record in Washington, USA, Quat. Sci. Rev. 15(1996) 1023–1059.

[82] V. Thomas, P. Treitz, D. Jelinski, J. Miller, P. Lafleur, J.H.McCaughey, Image classification of a northern peatlandcomplex using spectral and community data, Remote Sens.Environ. 84 (2002) 83–99.

[83] K.H. Shaltout, M.G. Sheded, H.F. El-Kady, Y.M. Al-Sodany,Phytosociology and size structure ofNitraria retusa alongthe Egyptian Red Sea coast, J. Arid Environ. 53 (2003)331–345.

[84] L. Amsellem, J.L. Noyer, T. Le Bourgeois, M. Hossaert-McKey, Comparison of genetic diversity of the invasiveweedRubus alceifoliusPoir (Rosaceae) in its native rangeand in areas of introduction, using amplified fragment lengthpolymorphism (AFLP) markers, Mol. Ecol. 9 (2000) 443–455.

[85] B.F.J. Manly, Multivariate Statistical Methods: A Primer,Chapman and Hall, London, 1994.

[86] M.O. Hill, TWINSPAN—A FORTRAN Programme forArranging Multivariate Data in an Ordered Two-WayTable by Classification of the Individuals and Attributes,Cornell University, Ithaca, NY, 1979.

262 H. De Wolf et al. / Mutation Research 566 (2004) 249–262

[87] H.G. Gauch Jr., R.H. Whittacker, Hierarchical classificationof community data, J. Ecol. 69 (1981) 537–557.

[88] F. Mezquita, R. Hernandez, J. Rueda, Ecology anddistribution of ostracods in a polluted Mediterranean river,Palaeogeogr. Palaeoclim. Palaeoecol. 148 (1999) 87–103.

[89] S. Wright, Variability Within and Among NaturalPopulations, vol. 4, The University of Chicago Press,Chicago, USA, 1978.

[90] B. Black, RAPDFST 4.0.1—A FORTRAN Program toEstimate F(ST) and Effective Migration Rates AmongSubpopulations Using RAPD–PCR Files, Colorado StateUniversity, 1997.

[91] B.L. Apostol, W.C. Black, P. Reiter, B.R. Miller, Populationgenetics with RAPD–PCR markers: the breeding structure ofAedes aegyptiin Puerto Rico, Heredity 76 (1996) 325–334.

[92] H. De Wolf, T. Backeljau, R. Verhagen, Congruence betweenallozyme and RAPD data in assessing macrogeographicalgenetic variation in the periwinkleLittorina striata(Mollusca, Gastropoda), Heredity 81 (1998) 486–492.

[93] R. Bártfai, S. Egedi, G.H. Yue, B. Kovács, B. Urbányi,G. Tamás, L. Horváth, L. Orbán, Genetic analysis of twocommon carp broodstocks by RAPD and microsatellitemarkers, Aquaculture 219 (2003) 157–167.

[94] C.W. Theodorakis, J.W. Bickham, T. Lamb, Integration ofgenotoxicity and population genetic analyses in kangaroorats (Dipodomys merriami) exposed to radionuclidecontamination at the Nevada test site, USA, Environ.Toxicol. Chem. 20 (2001) 317–326.

[95] X.L. Ma, D.L. Cowles, R.L. Carter, Effect of pollution ongenetic diversity in the bay musselMytilus galloprovincialisand the acorn barnacleBalanus glandula, Mar. Environ. Res.50 (2000) 559–563.

[96] G.Y. Yan, D.D. Chadee, D.W. Severson, Evidence for genetichitchhiking effect associated with insecticide resistance inAedes aegypti, Genetics 148 (1998) 793–800.

[97] R.L. Borowsky, Estimating nucleotide diversity from randomamplified polymorphic DNA and amplified fragment lengthpolymorphism data, Mol. Phylogen. Evol. 18 (2001) 143–148.

[98] L. Excoffier, P.E. Smouse, J.M. Quatro, Analysis ofmolecular variance inferred from metric distances amongDNA haplotypes: application to human mitochondrial DNArestriction data, Genetics 131 (1992) 479–491.

[99] J.S. Porter, J.S. Ryland, G.R. Cravalho, Micro- and macro-geographic genetic structure in bryozoans with differentlarval strategies, J. Exp. Mar. Biol. Ecol. 272 (2002) 119–130.

[100] C.D. Todd, A.M. Walker, K. Wolff, S.J. Northcott, A.F.Walker, M.G. Ritchie, R. Hoskins, R.J. Abbott, N. Hazon,Genetic differentiation of populations of the copepod sealouse Lepeophtheirus salmonis(Krøyer) ectoparasite onwild and farmed salmonids around the coasts of Scotland:evidence from RAPD markers, J. Exp. Mar. Biol. Ecol. 210(1997) 251–274.

[101] B.X. Huang, R. Peakall, P.J. Hanna, Analysis of geneticstructure of blacklip abalone (Haliotis rubra) populationsusing RAPD, minisatellite and microsatellite markers, Mar.Biol. 136 (2000) 207–216.

[102] C. Jäggi, T. Wirth, B. Baur, Genetic variability insubpopulations of the asp viper (Vipera aspis) in the SwissJura mountains: implications for a conservation strategy,Biol. Conserv. 94 (2000) 69–77.

[103] A. Mengoni, C. Gonnelli, F. Galardi, R. Gabbrielli, M.Bazzicalupo, Genetic diversity and heavy metal tolerancein populations ofSilence paradoxaL. (Caryophyllaceae): arandom amplified polymorphic DNA analysis, Mol. Ecol. 9(2000) 1319–1324.

[104] R.W. Michelmore, I. Paran, R.V. Kesseli, Identificationof markers linked to disease-resistance genes by bulkedsegregant analysis: a rapid method to detect markers inspecific genomic regions by using segregating populations,Proc. Natl. Acad. Sci. U.S.A. 88 (1991) 9828–9832.

[105] S.J. Chun, Y.H. Lee, Genetic analysis of a mutation onappressorium formation inMagnaporthe grisea, FEMSMicrobiol. Lett. 173 (1999) 133–137.

[106] Y.PaltiJ.E. Parsons, G.H. Thorgaard, Identification ofcandidate DNA markers associated with IHN virus resistancein backcrosses of rainbow (Oncorynchus mykiss) andcutthroat trout (O. clarki), Aquaculture 173 (1999) 81–94.

[107] K.J. Dehmer, W. Friedt, Development of molecular markersfor high oleic acid content in sunflower (Helianthus annuusL.), Ind. Crops Prod. 7 (1998) 311–315.

[108] M.G. Wirthensohn, G. Collins, G.P. Jones, M. Sedgley,Variability in waxiness ofEucalyptus gunniifoliage forfloriculture, Sci. Hortic. 82 (1999) 279–288.

[109] L. Ramirez, V. Muez, M. Alfonso, A. Garcia Barrenechea,L. Alfonso, A.G. Pisabarro, Use of molecular markersto differentiate between commercial strains of the buttonmushroomAgaricus bisporus, FEMS Microbiol. Lett. 198(2001) 45–48.

[110] J.-Y. Wu, H.-K. Wu, M.-C. Chung, Co-dominant RAPDmarkers closely linked with two morphological genes in rice(Oryza sativaL.), Bot. Bull. Acad. Sin. 43 (2002) 171–180.