Animal Conservation (2004) 7, 139–146 C© 2004 The Zoological Society of London. Printed in the United Kingdom DOI:10.1017/S1367943004001325

The effects of inbreeding on mortality during a morbillivirusoutbreak in the Mediterranean striped dolphin (Stenellacoeruleoalba)

Elena Valsecchi1, William Amos2, Juan Antonio Raga3, Michela Podesta4 and William Sherwin1

1 School of Biological Science, University of New South Wales, Sydney NSW2052, Australia2 Zoology Department, Cambridge University, Cambridge, CB1 2EJ, UK3 Departamento de Zoologıa and Instituto Cavanilles de Biodiversidad y Biologia Evolutiva, Universitat de Valencia, Spain4 Museum of Natural History of Milan, Milan, Italy

(Received 12 March 2003; accepted 29 September 2003)

AbstractBetween 1990 and 1992, Mediterranean striped dolphins (Stenella coeruleoalba) suffered high mortality due toa morbillivirus epidemic. Ten highly variable microsatellite markers were used to assess the population structureof a sample of these stranded animals and to assess the genetic consequences of the epizootic on present stocks.We found little evidence of population structure within the Mediterranean, but distinct separation betweenthis and the North Sea (Atlantic) population, the latter also showing greater genetic diversity. Using a geneticmeasure of inbreeding, we found that dolphins dying early in the outbreak were significantly more inbred thanthose dying later. Within 10 years of the end of the epidemic, the level of inbreeding among stranded dolphinshad returned to its pre-outbreak levels. However, on average all stranded animals showed elevated levels ofinbreeding, suggesting that animals dying from disease may venture towards the shore more than those dyingof old age. Our results imply an important role for inbreeding in the dynamics of disease spread and that, inmarine mammal research, caution should be exercised when inferring demographic parameters from strandedspecimens.

INTRODUCTION

The striped dolphin (Stenella coeruleoalba) is a cosmo-politan cetacean found in tropical and temperate pelagicwaters. In the central-western Mediterranean it is themost abundant dolphin species (Aguilar & Raga, 1993).Mitochondrial DNA polymorphisms suggest that geneflow between the Mediterranean stock and the EuropeanAtlantic is restricted (Garcia-Martinez et al., 1999).

Over the last 15 years, several species of marinemammal have suffered die-offs attributable to morbilli-virus epidemics (e.g. Osterhaus & Vedder, 1988; Harderet al., 1990, Domingo, Van Bressem & Kennedy, 2001).In 1990–1991; one such outbreak is estimated to havekilled thousands of striped dolphins in the western/centralMediterranean (Raga & Aguilar, 1991; Bortolotto, Casini& Stanzani, 1992; Cebrian, 1995). Although the totalmortality of this epizootic is unknown, populations werenot driven to low numbers, as evidenced by a count ofat least 117 800 animals post-epidemic in the WesternMediterranean (Forcada et al., 1994). Males and femalesappeared to be equally affected (Aguilar & Raga, 1993).

During infection, lesions appear in the nervous system,the lymph nodes and lungs, allowing transmission viaaerosol (Domingo et al., 1990, Van Bressem et al., 1991;

All correspondence to: William Amos. Tel/Fax: ++44(0) 1223336616; E-mail: W.Amos@zoo.cam.ac.uk

Duignan et al., 1992). The striped dolphin is highlygregarious, habitually swimming in groups of 12–40individuals but occasionally congregating in hundreds andthis is thought to have contributed to the rapid spread ofthe infection throughout the Mediterranean basin.

Environmental factors may also have contributed.Aguilar & Borrell (1994) found unusually high poly-chlorinated biphenyl (PCB) levels in the tissues of thediseased dolphins. During the epizootic, most infectedindividuals presented with liver lesions (Domingo et al.,1992). Such lesions tend to compromise the liver’s abilityto deal with PCBs, allowing a build up to levels that mayact to depress the immune system (e.g. Loose et al., 1977).It is currently not known whether the liver lesions facilitatea greater accumulation of PCBs or whether they are aconsequence of high PCB levels.

A third factor that may have played a role is inbreedingand other forms of genetic susceptibility. Recent studieshave shown that heterozygosity and other estimates ofparental similarity often correlate with diverse measuresof fitness, including birth weight (Coulson et al., 1998),juvenile survival (Coltman, Bowen & Wright, 1998),parasite load (Coltman et al., 1999), reproductive success(Slate et al., 2000; Amos et al., 2001) and potential torecover from disease (Acevedo-Whitehouse et al., 2003).Consequently, it is interesting to see whether individualsdying during the outbreak were enriched for individualswith high scores for genetic estimates of parental

140 E. VALSECCHI ET AL.

GV

LSNT

SI

NI

SA

AO

Fig. 1. Sampling locations: Atlantic Ocean (AO), Gulf of Valencia(GV), Ligurian Sea (LS), North Tyrrhenian (NT), South Ionian Sea(SI), North Ionian Sea (NI) and South Adriatic (SA).

similarity. This can be tested by genotyping the samples ofindividuals and then comparing levels of heterozygosityamong animals that died during the epizootic with thoseof animals that died from causes other than morbillivirusafter the outbreak had passed.

We have used 10 microsatellite loci to analyse 104striped dolphins sampled over the last 10 years from oneAtlantic and six Mediterranean regions to test whetherinbred individuals show greater susceptibility to thedisease.

MATERIALS AND METHODS

Sample set

We analysed 104 striped dolphins representing oneAtlantic (North Sea, n = 6) and six Mediterranean (n = 98)locations (Fig. 1). All samples were collected fromstranded individuals. The Mediterranean can be dividedinto several distinct regions referred to as ‘interior seas’.We obtained samples from the Gulf of Valencia (GV, n =28), Ligurian Sea (LS, n = 30), North Tyrrhenian (NT,n = 20), North Ionian (NI, n = 5), Southern Ionian (SI,n = 8) and the Southern Adriatic (SA, n = 7). Sampleswere separated into during-epidemic or ‘D’ samples, col-lected from infected individuals stranded between August1990 and September 1991 (in late 1991 and early 1992the epidemic reached, and was confined to, Greek waters:see Cebrian, 1995) and post-epidemic or ‘P’ samples,collected after August 1994. Broadly speaking, collectionof the D samples followed the spread of the epidemic,beginning in the GV (n = 13) and LS (n = 17) regionsin late 1990, continuing through the NT (n = 20) andNI (n = 5) regions and ending in the SI (n = 4) andSA (n = 7) regions. A small number of samples collectedbetween the end of the die-off and mid-1994 wereexcluded so as to avoid confusion due to the tail end ofthe epidemic.

Our sample set included 59 males, 44 females and oneindividual of unknown sex. The sex ratio in the Medi-

terranean locations was roughly even (i.e. M:F = 1:1),with the exception of the LS region where males were over-represented (M:F = 1:0.45). All six North Sea sampleswere males. For 94 (90.4%) samples, data on animal sizewere available, ranging between 80 and 230 cm. Aguilar(1991) has estimated that weaning occurs at around length165 cm. Using this criterion, 15 individuals for which sizewas known died before weaning, of which five died duringand 10 after the epidemic.

Molecular screening

All samples were screened for the following 10 micro-satellite loci: EV37Mn, EV94Mn (Valsecchi & Amos,1996); 199/200, 417/418 (Schlotterer, Amos & Tautz,1991); MK5, MK6, MK9 (Krutzen et al., 2001); D08,D18 (Shinohara, Domingo-Roura & Takenaka, 1997) andKW12 (Hoelzel, Dahlheim & Stern, 1998). Each poly-merase chain reaction (PCR) contained: 5 ng of genomicDNA, 1 µl 10x Taq Buffer (Promega), 0.5 µl MgCl2(25 mM), 0.4 µl of each primer (10 µM), 0.1 µl dNTPs(10 µM), 3 µg bovine serum albumin (New EnglandBioLabs), 0.08 µl Taq DNA Polymerase (Promega). EachPCR was brought up to 10 µl with sdH2O. PCR profilesconsisted of the following steps: a 3 min denaturation(93 ◦C) phase followed by 35 cycles each consisting ofa denaturation (92 ◦C, 30 s), an annealing (49–60 ◦C, 40 s)and an extension (72 ◦C, 40 s) phase; the programmeterminated with a final extension step of 5 min at 72 ◦C.Microsatellites were visualised on an ABI-377 automatedsequencer. Loci were tested for the presence of null allelesby checking for heterozygote deficit using the programmeNULLTEST (Allen et al., 1995).

Analysis of population subdivision

Genetic differentiation between striped dolphins from theseven different regions was measured using both FST(Wright, 1921) and the unbiased estimation of Slatkin’sRST (Slatkin, 1995) using RST CALC 2.2 by Goodman(1997), which allows for differences in sample size. Forboth approaches, significance of deviation from zero wasdetermined by permutation tests (10 000 permutations).Sequential Bonferroni adjustment for multiple test (Rice,1989) was performed to minimise biases due to multiplecomparisons being made. Population subdivision withinthe Mediterranean was also examined using an assignmenttest (Paetkau et al., 1995, 1997) using the program DOH

(www.biology.ualberta.ca/jbrzusto/Doh.html).

Measuring inbreeding

In the absence of detailed pedigrees, the relatednessof an individual’s parents can be estimated geneticallyby measuring the similarity between maternally andpaternally inherited alleles. It is simplest to score alleles ata locus as same or different, i.e. to calculate the proportionof heterozygous loci. To avoid the problem that, when notall individuals are scored for all loci, those scored for lessvariable loci will appear to be more homozygous, the score

Dolphin mortality and inbreeding 141

at each locus can be weighted by the heterozygosity at thatlocus to give ‘standardised heterozygosity’ (SH: Coltmanet al., 1998). Elsewhere, a measure called the mean d 2 hasbeen proposed for microsatellite data, which exploits thefact that microsatellite alleles that differ greatly in lengthare usually less related to each other than two allelesof similar length (Coulson et al., 1998). However, boththeoretical (Hedrick, Fredrickson & Ellegren, 2001) andempirical (Coltman & Slate, 2003; Slate & Pemberton,2002) studies of mean d 2 suggest that it tends only to beinformative under exceptional circumstances, for examplewhen two dissimilar populations meet and hybridise oras single locus effects (Hansson et al., 2001). A thirdmethod exploits the approach of Queller & Goodnight(1989) to estimate the relatedness between parents basedon heterozygosity but weighted by the frequencies ofthe alleles in each genotype. This measure, known as‘internal’ relatedness (IR: Amos et al., 2001), is calculatedas follows: (

2H − ∑fi

)

(2N − ∑

fi

)

where H is the number of loci that are homozygous, Nis the number of loci and fi is the frequency of the ithallele contained in the genotype. IR values are distributedapproximately normally about zero for individuals bornto ‘unrelated’ parents. Although there is a lot of scatter,on average individuals born to dissimilar parents will givelower IR values compared with individuals born to morerelated parents. IR is highly correlated with SH, in ourstudy r = 0.96, and the two approaches obviously showconsiderable agreement. However, where they have beencompared directly, IR tends to reveal slightly strongereffects than SH (Amos et al., 2001; Acevedo-Whitehouseet al., 2003; W. Amos, unpub. results), in line withtheoretical expectations. Consequently, IR is the measureof choice in this study.

IR values were calculated using region-specific allelefrequency distributions. For regions where sample sizesare small, say fewer than 10 individuals or 20 alleles, thisapproach may itself introduce biases. However, we used asimulation approach to show that the use of population-specific frequencies tends to reduce the estimate of IR,making inferences about high values associated with dis-ease occurrence conservative. The only way to circumventthe problem entirely would be to have access to largersample sizes than are actually available.

RESULTS

Microsatellite polymorphism

The number of alleles detected at the 10 loci we surveyedranged between 12 and 26 (mean 18.7: Table 1). Meanobserved (Ho) and expected (He) heterozygosities were0.71 and 0.80, respectively. Null alleles were identifiedat one locus, KW12 (mean null frequency in our sixpopulation samples = 0.253 ± 0.047 standard error ofthe mean (s.e.m.)), and this was excluded from furtheranalysis.

Table 1. Characteristics of the ten microsatellite markers used inthis study

No. ofindividuals No. of Allele range

Locus typed alleles (bp) Ho He Ref.

199/200 97 15 106–134 0.77 0.82 1417/418 99 15 161–195 0.63 0.69 1EV37Mn 102 21 181–237 0.71 0.77 2EV94Mn 103 22 216–262 0.84 0.86 2KW12 98 14 164–190 0.46 0.81 3MK5 99 26 199–249 0.84 0.88 4MK6 104 22 135–191 0.78 0.85 4MK9 97 12 156–182 0.79 0.74 4D08 102 24 78–134 0.79 0.91 5D18 104 16 73–107 0.48 0.63 5

Means 100.5 18.7 0.71 0.80

References: 1, Schlotterer et al. (1991); 2, Valsecchi & Amos(1996); 3, Hoelzel et al. (1998); 4, Krutzen et al. (2001); 5,Shinohara et al. (1997).Ho, observed heterozygosity; He, expected heterozygosity; bp,base-pair.

Table 2. Average gene diversity over loci (Ho) measured in the sevenstudy locations

Sampling locations n Ho (± SD)

AO, Atlantic Ocean 6 0.91 (± 0.51)GV, Gulf of Valencia 28 0.77 (± 0.40)LS, Ligurian Sea 30 0.71 (± 0.38)NT, Northern Tyrrhenian Sea 20 0.75 (± 0.40)NI, Northern Ionian Sea 5 0.62 (± 0.36)SI, Southern Ionian Sea 8 0.58 (± 0.33)SA, Southern Adriatic Sea 7 0.75 (± 0.42)

SD, standard deviation.

Analysis of population subdivision

Our sample of Atlantic animals have somewhat greaterlevels of allelic diversity than Mediterranean samples(see Table 2: mean Ho within Mediterranean = 0.697 ±0.032; Ho in North Atlantic = 0.91). This difference is notsignificant using a two sample t-test (t = 1.29, n1 = 98,n2 = 6, P = 0.2), although the fact that the Atlantic samplehas the highest heterozygosity of all seven populationssuggests that this may be due to the very small samplesize of Atlantic animals. A higher heterozygosity wouldbe expected, given the more than likely larger size ofthe Atlantic (North Sea) population coupled with theapparently low levels of gene flow between this and theMediterranean (see below).

No differentiation was apparent between the D and Psamples from any of the locations where such comparisonwas possible (i.e. GV and LS). Therefore, all individualssampled in the same region were pooled, regardless of yearof sampling. Estimated values for FST and unbiased RSTfor all pairwise population combinations are summarisedin Table 3. The pooled Mediterranean samples showsignificant differentiation from the Atlantic for bothFST (P = 0.003) and unbiased RST (P = 0.041). Within

142 E. VALSECCHI ET AL.

Table 3. Pairwise estimates of indexes of genetic differentiation between striped dolphins from six Mediterranean locations

Sampling locations GV LS NT NI SI SA

GV, Gulf of Valencia 0.007 − 0.0003 0.057 − 0.029 − 0.0001LS, Ligurian Sea 0.006 − 0.003 0.013 − 0.031 − 0.026NT, Northern Tyrrhenian Sea 0.005 0.006 0.024 − 0.033 − 0.012NI, Northern Ionian Sea 0.021 0.007 0.016 − 0.001 − 0.027SI, Southern Ionian Sea − 0.008 − 0.006 − 0.019 0.018 − 0.033SA, Southern Adriatic Sea 0.015 0.021 0.012 0.017 0.012

FST and Rho (unbiased RST) pairwise values are shown below and above the diagonal, respectively.

Table 4. Results of the assignment test performed across Mediter-ranean regions

the Mediterranean basin, neither measure indicates asignificant differentiation, although in the assignmenttests, 37.8% of individuals are assigned to the populationfrom which they originated (Table 4). Despite being lessthan half, this proportion is none the less significantlyhigher than the number (22.8%) expected by chance(χ2 = 9.9, P < 0.005). If this does indicate some slightdifferentiation, formal confirmation must await largersample sizes. IR did not differ significantly between thetwo populations (GV, LS; t = 0.91, d.f. = 28) for whichenough post-epidemic samples were available. Similarlyno difference in IR values was found before and afterthe epidemic in GV (t = 0.29, d.f. = 16) and LS (t = 0.57,d.f. = 15). In contrast, significant differences in IR do existbetween regions during the epidemic, probably due moreto the geographical progression of the disease rather thanto effective regional differences (see Discussion, below).

Relationship between IR and mortality

Overall, the IR values found in stranded dolphins (meanIR = 0.036) are somewhat greater than would be expectedof a random sample of outbred individuals (t = 1.9,n = 104, P < 0.05, one-tailed test for mean > 0). Wejustify the one-tailed test because, in a population wheredifferentiation is slight at most, outbreeding depressionis unlikely. Having said this, the difference is marginal.Since we did not have a control sample of healthy dolphinswith which to compare the stranded dolphins, we turnedto equivalent microsatellite data sets based on samples ofapparently healthy animals in other natural populationsof sea mammals. In all species where data were availableand indeed including many other species that were not

marine mammals, mean IR values are extremely closeto the expected mean of zero (e.g. long-finned pilotwhales, n = 658, mean IR = 0.007; grey seals, n = 1481,mean IR =− 0.0058: W. Amos, pers. obs.). Furthermore,when data are partitioned into smaller samples with allelefrequencies calculated for each group separately, meanIR values tend to be lower because rare alleles can neverhave estimated frequencies less than the inverse of thesample size of alleles. To test this expectation, we usedthe observed allele frequencies of the Stenella repeatedlyto generate sets of 128 randomised genotypes by MonteCarlo simulation. Each set of genotypes was then dividedinto N equally sized sets, where N = 1, 2, 4, 8 or 16,and the IR was calculated on the basis of group-specificallele frequencies. For each value of N, 100 replicates weremade. Mean IR declined monotonically from − 0.002for N = 1 to − 0.067 for N = 16. Consequently, in theabsence of biases introduced by undetected null alleles,the expectation for the mean IR of healthy, outbred animalsin our study is likely to be below zero at around − 0.04.

A second question to ask is whether there is arelationship between the course of the epizootic and theapparent level of inbreeding of the animals who died. Todiscover whether IR varied over the course of our study,animals were grouped by year of sampling. Since preand post-epizootic sample sizes were small, these wereplaced into just two groups, yielding four classes: pre-epizootic (n = 3), 1990 (n = 25), 1991 (n = 35) and post-epizootic (n = 34). An Anova of IR value against yearindicated significant among-year effects (F = 4.11, d.f. =93,3, P = 0.009). A graph of IR value against year is givenin Fig. 2(a). The most striking feature is an apparent dipin value during the height of the epizootic in 1990/91.

To look at this in more detail, this portion of the graphhas been replotted in Fig. 2(b), using exact date of sam-pling. A significant decline in IR value is apparent (r =− 0.323, n = 61, P = 0.011), and there is no obvious dif-ference between the sexes (estimated slopes: male slope =− 0.42, female slope = − 0.38: Figs. 2(c) and 2(d)). Byinspection, the strongest effect seems to be at the start ofthe epizootic. From the fitted line, mean IR values declinefrom around 0.1 at the start of the outbreak to somewhatbelow zero at the end. However, by the late 1990s, themean IR value of stranded animals had risen significantlyback above zero to a mean of 0.065. Finally, there appearsto be a temporal progression of the disease from area toarea. Replotting the data with each area contributing one

Dolphin mortality and inbreeding 143

–0.6

–0.4

–0.2

0

0.2

0.4

0.6

0.8

87 88 89 90 91 92 93 94 95 96 97 98 99

pre-epidemic

epidemic post-epidemic

(a)

(b)

1990 1991 –0.6

–0.4

–0.2

0

0.2

0.4

0.6

0.8

0,1 0,2 0,3 0,5

epidemic - all

–0.6

–0.4

–0.2

0

0.2

0.4

0.6

0.8(d) epidemic - females

–0.6

–0.4

–0.2

0

0.2

0.4

0.6

0.8(c) epidemic - males r2 = 0.076 r2 = 0.088

r2 = 0.104

1990 19911990 1991

0 100 200 300 400 500

LS

LS GV

GV NT

NT

NI

NI

SA SA

SI

100 200 300 4000

0.1

0.2

–0.2

–0.1

–0.3(e)

Fig. 2. Internal relatedness (IR) values for stranded dolphins: (a) across the whole study interval (10/88–08/99); (b) during the epidemicperiod (08/90–09/91); (c) during the epidemic period for males only; (d) during the epizootic period for females only and (e) during theepidemic showing average dates of stranding for each sex (squares and continuous line, males; diamonds and broken line, females) andregion classes. For sampling location codes, see the legend to Fig. 1.

mean value each for IR and date of sampling for each sex,reveals similar downward slopes (Fig. 2(e)), suggesting arelationship between the progression of the disease andthe average IR values for animals in each region.

DISCUSSION

Striped dolphins from six different Mediterraneanlocations, including specimens stranded during a viralepidemic, were surveyed for 10 microsatellite markersand compared with Atlantic conspecifics from the NorthSea. The Mediterranean sample appears to be relativelyisolated from our Atlantic sample, in agreement with pre-vious molecular (Garcia-Martinez et al., 1999) and mor-phological (Di Meglio, Romero Alvarez & Collet, 1996)evidence that revealed differences between Mediterraneanand Atlantic dolphins sampled near the Gibraltar opening.

The Mediterranean sample also revealed reduced levels ofvariability in both nuclear (this study) and mitochondrial(Garcia-Martinez et al., 1999) markers. Within theMediterranean, the distribution of nuclear genotypessuggests that any restriction of gene flow between differentbasins that may exist is slight at most. During the epizooticin 1990/91, the inbreeding coefficient of stranded dolphinswas at first high and then declined, suggesting that asthe outbreak spread, increasingly outbred individuals weresuccumbing to infection.

Despite their obvious ability to move over largedistances, striped dolphins of the Mediterranean appearto be largely isolated from conspecifics in the Atlanticand the assignment tests suggest slight differences evenamong regions within the Mediterranean. However, oursample sizes are low and it is unrealistic to try to quantifysuch low levels of differentiation and to identify possible

144 E. VALSECCHI ET AL.

Mediterranean ‘stocks’ until many more samples becomeavailable.

Overall, our sample has rather high mean IR values,suggesting that stranded striped dolphins tend to beenriched for relatively inbred individuals. Although mostof our Mediterranean samples were affected by a disease,individuals stranded both before and after the epidemicalso showed higher levels of IR than we would haveexpected by chance and, indeed, the six individualsstranded on Atlantic coasts showed similar levels (meanIR = 0.08). Although we do not have control samplescollected from healthy adult dolphins with which tocompare our stranded material, there are several reasonswhy we believe the interpretation that stranded dolphinshave high IR values is correct. First, the Amos laboratoryhas genotyped samples of healthy animals from manyspecies, including a range of marine mammals. Withoutexception, these data reveal mean IR values that areextremely close to and, if anything, slightly below zero(pers. obs.). Second, with the dolphin data we tried tobe conservative by using region-specific allele frequencydistributions to calculate IR. Simulations indicate that thisapproach biases IR downwards and the more subdivideda sample, the lower the expectation for IR. Finally, duringthe middle/end of the epizootic, mean IR falls significantlyto below zero, showing that the high IR values are nota general feature of the whole population. Since it isdifficult to see why animals with high IR values shoulddie preferentially at the end of an epidemic, the mostparsimonious explanation is that stranded animals arenormally enriched for individuals with high average IRvalues (i.e. relatively inbred). During an epidemic, themean IR value measurable in stranded animals is forceddown towards the population average as large numbers of‘normal’ animals succumb to the virus.

If stranded dolphins are more inbred than average,this has implications for how individuals behave whennear death. To explain the pattern we observe, twoconditions are required. First, having a high IR valuemust correlate with some causes of mortality (for exampleby being associated with increased susceptibility toinfectious diseases) but not others (for example predatorattacks or old age). Second, there must be some formof behavioural differences between these two classes ofanimal. Specifically, animals who die from causes notassociated with high IR values must be less likely to endup recorded as stranded, possibly because they tend todie further from shore and decompose or sink before theyhave a chance to wash up. An anecdotal observation thatcould help explain this pattern is the apparent tendencyfor diseased and injured animals to use shallow water asa place to rest (and die). If this is the case and strandedanimals are indeed more homozygous than the populationthey represent, this needs to be considered when usingstranded cetaceans as a source of biological samples inpopulation genetic studies.

The possibility that relatively inbred individuals aremore susceptible to infectious diseases is supported byseveral recent studies using alternative genetic measuresof inbreeding. For example, in an unmanaged population

of primitive Soay sheep, the burden of nematode gutparasites is greater in inbred compared with outbredindividuals (Coltman et al., 1999). More generally, it hasbeen observed that inbreeding influences juvenile survivaland a component of this is likely to be disease tolerance(Coltman et al., 1998; Coulson et al., 1998). Of particularrelevance is a recent study by Acevedo-Whitehouse et al.(2003) on stranded sea lions that shows how high IR valuesare linked to susceptibility to a wide range of challenges,from algal poisoning through bacterial and parasiticinfections to carcinoma. Our data lend further supportin the context of an epizootic. During the 2 peak years ofstrandings, a significant fluctuation in mean IR value wasobserved, with early mortality being characterised by highmean IR value and later mortality showing a significantfall. Such a pattern would be expected if relatively inbredindividuals either contract the disease preferentially or diesooner (or both). As the outbreak progresses, two factorsmay cause a fall in IR. On the one hand, the numberof infected individuals is likely to be greater, increasingthe amount of virus in the environment and, perhaps,thereby reducing the importance of genetic factors. On theother hand, if animals with high IR values die with highprobability, there may be local depletion of such animals,forcing the average down.

Interestingly, although the change in mean IR value overtime appears to be relatively smooth, suggesting that allregions sampled responded similarly and simultaneously,in reality the disease appears to spread from region toregion. Consequently, the shape of the decline in IR valueseems to depend less on the way the disease spreads withina region and more on the order in which different regionsbecame infected. For example, the pattern could havearisen if the disease entered the Mediterranean by gaininga foothold in a group of unusually inbred individuals in thewest, namely in GV and LS. Plotting region averages fortiming of stranding and IR appear to support this, showinga similar negative slope for both sexes (see Fig. 2(e)).Such patterns argue against a process in which the virusspreads solely according to geographical proximity, butinstead that most regions were exposed early, with theepizootic taking off earlier in regions containing unusuallylarge numbers of animals with high IR values.

The relationship between IR and date of death issignificant at P < 0.01, but the proportion of variationexplained is small, of the order of 7–10% (see Figs 2(b)–(d)). However, when based on only 10 or so microsatellites,IR can be expected to correlate only weakly with trueparental relatedness. Indeed, it has been argued that thecorrelation between heterozygosity, as calculated from ahandful of markers, and parental similarity is so tenuousthat studies such as this are not measuring inbreeding atall (Hansson & Westerberg, 2002). Instead, the correlationbetween disease susceptibility and IR might be driven byone or a few markers that lie close enough to a geneexperiencing balancing selection to show heterozygoteadvantage themselves (Heath et al., 2002). We consideredthis possibility by testing each marker separately for adifference in the frequency of heterozygotes between 1990and 1991, the 2 years that show the biggest change in IR.

Dolphin mortality and inbreeding 145

None of the loci show an independently significant effect,although again our limited sample size means that wecannot preclude the presence of single locus effects.

The pattern we observe adds two interesting new factorsthat may influence the course of similar outbreaks. First,infections may tend to enter a population through itsweakest members. Stress, for example through starvation,is one well-known factor that may create a populationof unusually susceptible individuals. Our data point to asecond possibility whereby genetic factors may providea similar opportunity, either interacting with stress, oracting independently. Second, the change in IR over thecourse of the outbreak suggests that populations are notgenetically homogeneous, but instead may contain subsetsof individuals with inherently greater susceptibility. Thedistribution of these individuals clearly has the potentialto influence both the rate and direction of spread of thedisease and, in extreme cases, whether or not an outbreaktakes off.

Acknowledgements

We thank Giuseppe Coci, Simon Goodman, LetiziaMarsili, Lucio Rositani, Nicola Zizzo and the “CentroStudi Cetacei”, for providing samples. We are grateful toArnaud Estoup for his valuable comments at the earlystages of the study. This research project was supportedby the Australian Research Council, BBVA Foundationand the Spanish Ministry of the Environment.

REFERENCES

Acevedo-Whitehouse, K., Gulland, F., Greig, D. & Amos, W. (2003).Inbreeding-dependent pathogen susceptibility in California sea lions.Nature 422: 35.

Aguilar, A. (1991). Calving and early mortality in the westernMediterranean striped dolphin, Stenella coeruleoalba. Can. J. Zool.69: 1408–1412.

Aguilar, A. & Borrell, A. (1994). Abnormally high polychlorinatedbiphenyl levels in striped dolphins (Stenella coeruleoalba) affectedby the 1990–1992 Mediterranean epizootic. Sci. Total Environ. 154:237–247.

Aguilar, A. & Raga, J. A. (1993). The striped dolphin epizootic in theMediterranean Sea. Ambio 22: 524–528.

Allen, P. J., Amos, W., Pomeroy, P. P. & Twiss, S. D. (1995).Microsatellite variation in grey seals (Halichoerus grypus) showsevidence of genetic differentiation between two British breedingcolonies. Mol. Ecol. 4: 653–662.

Amos, W., Worthington Wilmer, J., Fullard, K., Burg, T. M., Croxall,J. P., Bloch, D. & Coulson, T. (2001). The influence of paternalrelatedness on reproductive success. Proc. R. Soc. Lond. Ser. B 268:2021–2027.

Bortolotto, A., Casini, L. & Stanzani, L. A. (1992). Dolphin mortalityalong the southern Italian coast (June–September 1991). Aquat.Mamm. 18: 56–60.

Cebrian, D. (1995). The striped dolphin Stenella coeruleoalba epizooticin Greece, 1991–1992. Biol. Conserv. 74: 142–145.

Coltman, D. W. & Slate, J. (2003). Microsatellite measures ofinbreeding: a meta-analysis. Evolution 57: 971–983.

Coltman, D. W., Bowen, W. D. & Wright, J. M. (1998). Birth weight andneonatal survival of harbour seal pups are positively correlated withgenetic variation measured by microsatellites. Proc. R. Soc. Lond.Ser. B 265: 803–809.

Coltman, D. W., Pilkington, J. G., Smith, J. A. & Pemberton, J. M.(1999). Parasite-mediated selection against inbred Soay sheep in afree-living island population. Evolution 53: 1259–1267.

Coulson, T. N., Pemberton, J. M., Albon, S. D., Beaumont, M., Marshall,T. C., Slate, J., Guinness, F. E. & Clutton-Brock, T. H. (1998).Microsatellites reveal heterosis in red deer. Proc. R. Soc. Lond. Ser.B 265: 489–495.

Di Meglio, N., Romero Alvarez, R. & Collet, A. (1996). Growthcomparison in striped dolphins, Stenella coeruleoalba, from theAtlantic and Mediterranean coasts of France. Aquat. Mamm. 22:11–19.

Domingo, M., Ferrer, L., Pumarola, M., Marco, A., Plana, J., Kennedy,S., McAlister, M. & Rima, B. K. (1990). Morbillivirus in dolphins.Nature 348: 21.

Domingo, M., Visa, J., Pumarola, M., Marco, A. J., Ferrer, L., Rabanal,R. & Kennedy, S. (1992). Pathological and immunocytochemicalstudies of morbillivirus infection in striped dolphins (Stenellacoeruleoalba). Vet. Pathol. 29: 1–10.

Domingo, M., Van Bressem, M. F. & Kennedy, S. (2001). Marinemammal mass mortalities. In: Marine mammals biology andconservation: 425–456. Evans, P. G. E. & Raga, J. A. (Eds). London:Kluwer Academic/Plenum Press.

Duignan, P. J., Geraci, J. R., Raga, J. A. & Calzada, N. (1992). Pathologyof morvillivirus infection in striped dolphins (Stenella coeruleoalba)from Valencia and Murcia, Spain. Can. J. Vet. Res. 56: 242–248.

Forcada, J., Aguilar, A., Hammond, P. S., Pastor, X. & Aguilar, R.(1994). Distribution and numbers of striped dolphins in the westernMediterranean Sea after the 1990 epizootic outbreak. Mar. Mamm.Sci. 10: 137–150.

Garcia-Martinez, J., Moya, A., Raga, J. A. & Latorre, A. (1999).Genetic differentiation in the striped dolphin Stenella coeruleoalbafrom European waters according to mitochondrial DNA (mtDNA)restriction analysis. Mol. Ecol. 8: 1069–1073.

Goodman, S. J. (1997). RST CALC: a collection of computer programsfor calculating unbiased estimates of genetic differentiation anddetermining their significance for microsatellite data. Mol. Ecol. 6:881–885.

Hansson, B., Bensch, S., Hasselquist, D. & A�

kesson, M. (2001).Microsatellite diversity predicts recruitment of sibling great reedwarblers. Proc. R. Soc. Lond. Ser. B. 268: 1287–1291.

Hansson, B. & Westerberg, L. (2002). On the correlation betweenheterozygosity and fitness in natural populations. Mol. Ecol. 11:2467–2474.

Harder, T., Willhaus, T., Frey, H. R. & Liess, B. (1990). Morbillivirusinfections of seals during the 1988 epidemic in the Bay ofHeligoland III. Transmission studies of cell culture-propagatedphocine distemper virus in harbour seals (Phoca vitulina) and agrey seal (Halichoerus grypus): clinical, virological and serologicalresults. J. Vet. Med. B 37: 641–650.

Heath, D. D., Bryden, C. A., Shrimpton, J. M., Iwama, G. K., Kelly,J. & Heath, J. W. (2002). Relationships between heterozygosity,allelic distance (d2), and reproductive traits in chinook salmon,Oncorhynchus tshawytscha. Can. J. Fish. Aquat. Sci. 59: 77–84.

Hedrick, P., Fredrickson, R. & Ellegren, H. (2001). Evaluation of d2,a microsatellite measure of inbreeding and outbreeding, in wolveswith known pedigree. Evolution 55: 1256–1260.

Hoelzel, A. R., Dahlheim, M. & Stern, S. J. (1998). Low geneticvariability among killer whales Orcinus orca in the eastern NorthPacific and genetic differentiation between foraging specialists.J. Heredity 89: 121–128.

Krutzen, M., Valsecchi, E., Condor, R. C. & Sherwin, W. B. (2001).Characterization of microsatellite loci in Tursiops aduncus. Mol.Ecol. Notes 1: 170–172.

Loose, L. D., Pitman, K. A., Bentitz, K. F. & Silkworth, J. B. (1977).Polychlorinated biphenyl and hexachlorobenzene induced humoralimmunosuppression. J. Reticuloendo. Soc. 22: 253–271.

Osterhaus, A. D. M. E. & Vedder, E. J. (1988). Identification of viruscausing recent seal deaths. Nature 335: 20.

146 E. VALSECCHI ET AL.

Paetkau, D., Calvert, W., Stirling, I. & Strobeck, C. (1995).Microsatellite analysis of population structure in Canadian polarbears. Mol. Ecol. 4: 347–354.

Paetkau, D., Waits, L. P., Clarkson, P. L., Craighead, L. & Strobeck, C.(1997). An empirical evaluation of genetic distance statistics usingmicrosatellite data from bear (Ursidae) populations. Genetics 147:1943–1957.

Queller, D. C. & Goodnight, K. F. (1989). Estimating relatedness usinggenetic markers. Evolution 43: 258–275.

Raga, J. A. & Aguilar, A. (1991). Mass mortality of striped dolphinsin Spanish Mediterranean waters. In: Proceedings of the Mediterra-nean striped dolphin mortality – International Workshop: 21–25.Pastor, X. & Simmonds, M. (Eds). Palma de Mallorca, 4–5 November1991.

Rice, W. B. (1989). Analyzing tables of statistical tests. Evolution 43:223–225.

Schlotterer, C., Amos, W. & Tautz, D. (1991). Conservation of polymor-phic simple sequence loci in cetacean species. Nature 354: 63–65.

Shinohara, M., Domingo-Roura, X. & Takenaka, O. (1997).Microsatellites in the bottlenose dolphin Tursiops truncatus. Mol.Ecol. 6: 695–696.

Slate, J., Kruuk, L. E., Marshall, T. C., Pemberton, J. M. & Clutton-Brock, T. H. (2000). Inbreeding depression influences lifetimebreeding success in a wild population of red deer (Cervus elaphus).Proc. R. Soc. Lond. Ser. B 267:1657–1662.

Slate, J. & Pemberton, J. M. (2002). Comparing molecular measures ofdetecting inbreeding depression. J. Evol. Biol. 15, 20–31.

Slatkin, M. (1995). A measure of population subdivision basedon microsatellite allele frequencies. Genetics 139: 457–462.

Valsecchi, E. & Amos, W. (1996). Microsatellite markers for the studyof cetacean populations. Mol. Ecol. 5: 151–156.

Van Bressem, M. F., Visser, I. K. G., Van de Bildt, M. G. W., Teppema,J. S., Raga, J. A. & Osterhaus, A. D. M. E. (1991). Morbillivirusinfection in striped dolphins (Stenella coeruleoalba). Vet. Rec. 129:471–472.

Wright, S. (1921). Systems of mating. Genetics 6: 111–178.