No correlation between inbreeding depression and delayed selfing in the freshwater snail Physa acuta.

Abstract

Inbreeding depression, one of the main factors driving mating system evolution, can itself evolve as a function of the mating system (the genetic purging hypothesis). Classical models of coevolution between mating system and inbreeding depression predict negative associations between inbreeding depression and selfing rate, but more recent approaches suggest that negative correlations should usually be too weak or transient to be detected within populations. Empirical results remain unclear and restricted to plants. Here, we evaluate, for the first time, the within-population genetic correlation between inbreeding depression and a trait that controls the amount of self-fertilization (the waiting time) in a self-fertile hermaphroditic animal, the freshwater snail Physa acuta. Using a large quantitative-genetic design (36 grand-families and 348 families), we observe abundant within-population family-level genetic variation for both inbreeding depression (estimated for survival, fecundity, and size) and the degree of behavioral selfing avoidance. However, we detected no correlation between waiting time and inbreeding depression across families. In agreement with recent models, this result shows that mutational variance rather than differential purging accounts for most of the genetic variance in inbreeding depression within a population.

Full text

Page 1

ORIGINAL ARTICLE
doi:10.1111/j.1558-5646.2007.00223.x
NO CORRELATION BETWEEN INBREEDING
DEPRESSION AND DELAYED SELFING IN THE
FRESHWATER SNAIL PHYSA ACUTA
Juan Sebasti´ an Escobar,1,2Guillaume Epinat,1,3Violette Sarda,1,4and Patrice David1,5
1Centre d’Ecologie Fonctionnelle et Evolutive (CEFE), UMR 5175 CNRS. 1919 Route de Mende, 34293 Montpellier Cedex 5,
France
2E-mail: escobar@cefe.cnrs.fr
3E-mail: epinat@cefe.cnrs.fr
4E-mail: sarda@cefe.cnrs.fr
5E-mail: pdavid@cefe.cnrs.fr
Received April 4, 2007
Accepted July 10, 2007
Inbreeding depression, one of the main factors driving mating system evolution, can itself evolve as a function of the mating
system (the genetic purging hypothesis). Classical models of coevolution between mating system and inbreeding depression
predict negative associations between inbreeding depression and selfing rate, but more recent approaches suggest that negative
correlations should usually be too weak or transient to be detected within populations. Empirical results remain unclear and
restricted to plants. Here, we evaluate, for the first time, the within-population genetic correlation between inbreeding depression
and a trait that controls the amount of self-fertilization (the waiting time) in a self-fertile hermaphroditic animal, the freshwater
snail Physa acuta. Using a large quantitative-genetic design (36 grand-families and 348 families), we observe abundant within-
population family-level genetic variation for both inbreeding depression (estimated for survival, fecundity, and size) and the
degree of behavioral selfing avoidance. However, we detected no correlation between waiting time and inbreeding depression
across families. In agreement with recent models, this result shows that mutational variance rather than differential purging
accounts for most of the genetic variance in inbreeding depression within a population.
KEY WORDS: Freshwater mollusk, hermaphrodites, mating system, self-fertilization, waiting time.
Mating system and inbreeding depression are two major deter-
minants of fitness in hermaphroditic organisms. Because self-
fertilization allows increased transmission of more genes per
offspring (i.e., the “cost of outcrossing”; Fisher 1941), classi-
cal models assume that the maintenance of outcrossing in self-
compatible species is due to the lower fitness of inbred offspring
relative to outbreds (i.e., inbreeding depression; Charlesworth
and Charlesworth 1987, 1999; Jarne and Charlesworth 1993).
In the last few decades, much debate in the field of mating
system evolution has revolved around the reciprocal evolution-
ary influences between mating system and inbreeding depression
(Lande and Schemske 1985; Holsinger 1988; Charlesworth et al.
1990;UyenoyamaandWaller1991a,b,c;SchultzandWillis1995).
This started with the Lande and Schemske (1985) model demon-
strating that long-term inbreeding enhances the removal of par-
tiallyrecessive,deleteriousallelesfromthepopulationandsubse-
quently,adecreaseininbreedingdepression(i.e.,geneticpurging;
Charlesworth et al. 1990; Byers and Waller 1999; Gl´ emin 2003).
Subsequently, mating system evolution has been considered as a
divergent process with two alternative evolutionary trajectories:
(1)highselfing/lowdepression,or(2)lowselfing/highdepression.
The bimodal distribution of selfing rates across plant species
(Schemske and Lande 1985; Goodwillie et al. 2005) seemed to
support this view, the animal dataset being much less clear (Jarne
2655
C ?2007 The Author(s). Journal compilation C ?2007 The Society for the Study of Evolution.
Evolution 61-11: 2655–2670

Page 2

JUAN S. ESCOBAR ET AL.
and Auld 2006). The predicted negative association between self-
ing rate and inbreeding depression at the species level has re-
ceived empirical support in plants, although not devoid of ambi-
guity (Husband and Schemske 1996). However, many questions
remain at lower taxonomic levels. Comparisons among popu-
lations within species are not conclusive (Johnston and Schoen
1996) and within-population genetic correlations between traits
controlling self-fertilization and inbreeding depression have been
rarely studied (Carr et al. 1997; Mutikainen and Delph 1998;
Chang and Rausher 1999; Vogler et al. 1999; Takebayashi and
Delph 2000; Stone and Motten 2002). Nevertheless, one of the
main scenarios for the evolutionary transitions from predominant
outcrossing to predominant selfing relies on the hypothesis that
alleles that promote selfing can become associated with purged
backgrounds and increase in frequency in an outcrossing popu-
lation with high average inbreeding depression (Holsinger 1988;
Uyenoyama and Waller 1991a). This scenario mainly emerged
from two- or three-locus models (Holsinger et al. 1984; Lande
and Schemske 1985; Holsinger 1988; Charlesworth et al. 1990;
Uyenoyama and Waller 1991a,b,c). More realistic simulations,
with multilocus determination of inbreeding depression, suggest
that, unless selfing modifiers and deleterious alleles have very
large effects, associations between individual inbreeding depres-
sionandselfingrateshouldbegenerallyweakwithinpopulations,
because the variation due to differences in purging and inbreed-
ing history among lines is overwhelmed by the variance due to
the random occurrence of mutations across generations (Schultz
and Willis 1995). Moreover, the existing associations may be
ephemeral because the fixation or elimination of selfing modi-
fiers is very rapid (Schultz and Willis 1995). Yet some transitions
betweenoutcrossingandselfingdooccur,asexemplifiedbythere-
centhistoryofArabidopsisthaliana(Shimizuetal.2004).During
thishistory,aself-compatiblelineageinvadedaself-incompatible
ancestralpopulation,andifsimilarinsionsarecurrentlyoccurring
in other species, genetic correlations between inbreeding depres-
sion and mating system traits are expected.
Apart from purging, another possible source of genetic cor-
relations, not considered in Schultz and Willis (1995), is the indi-
vidual adjustment of the selfing rate as a function of inbreeding
depression. Of course individuals are not able to evaluate their fu-
ture inbreeding depression directly; however, individual variation
in inbreeding depression is largely correlated to additive effects
of alleles (Schultz and Willis 1995), which can in principle be
perceived by an individual because they affect its own phenotype.
In other words, individuals with above-average fitness (or condi-
tion) generally have below-average (or even negative) inbreeding
depressionandcondition-dependentselfingcouldthereforebead-
vantageous in some circumstances.
For these reasons, theoretical predictions about genetic asso-
ciationsbetweeninbreedingdepressionandmatingsystemwithin
populations need to be confronted with empirical data. Such data
remain scarce because one needs to find large amounts of genetic
variation for both inbreeding depression and mating-system traits
within populations before testing for any association. Besides, all
studies so far concern plant systems (Carr et al. 1997; Mutikainen
and Delph 1998; Chang and Rausher 1999; Vogler et al. 1999;
Takebayashi and Delph 2000; Rao et al. 2002; Stone and Motten
2002). It is however important to test general ideas in a general
way, and to replicate results in systems that do not share the par-
ticulars of flowering plants (floral biology, self-incompatibility
system, coevolution with pollinators, sessile life). We here pro-
vide an evaluation of within-population genetic correlations be-
tween a mating-system trait and inbreeding depression in the
hermaphroditic freshwater snail Physa acuta. In this species, we
recently identified a behavioral trait (the waiting time; Tsitrone
et al. 2003b) by which individuals control their selfing rate within
populations. We set up a series of crosses to jointly estimate ge-
neticvariationforboththewaitingtimeandindividualinbreeding
depression(i.e.,theexpectedrelativedifferenceinfitnessbetween
outbred and inbred offspring of a single individual) and we per-
form a correlation analysis to test for the coevolution between the
two.
Materials and Methods
BIOLOGICAL MODEL AND REARING CONDITIONS
The species studied, P. acuta Draparnaud, is a simultane-
ously hermaphroditic freshwater snail that reproduces both uni-
parentally through internal self-fertilization and biparentally
throughcross-fertilization.Itisapredominantoutcrossingspecies
withlowselfingrates(0–20%,WethingtonandDillon1997;Jarne
et al. 2000; Henry et al. 2005) and its inbreeding depression was
estimated to be 0.90 over a full life cycle (Jarne et al. 2000). The
eggs are laid within capsules (clutches) typically containing a few
tens of eggs and hatching of juveniles (∼1 mm in length) occurs
6–7daysafteregglaying(Tsitroneetal.2003b).Inthelaboratory,
sexual maturity is attained six to eight weeks at 22–24◦C.
We sampled about 300 adult individuals (G0) of P. acuta
in the Mosson River (Mos 1), northwest of Montpellier, Southern
France(43◦40?N,3◦46?W)on18November2002.Thispopulation
isknowntohavealowselfingrate(0.02–0.03;Boussetetal.2004;
Henry et al. 2005). The individuals sampled were brought back
to laboratory, where they were maintained at 25◦C under a 12L:
12Dphotoperiodandfedwithboiledlettuce.Foodandwaterwere
changed twice a week.
MATING SYSTEM TRAIT AND EXPERIMENTAL DESIGN
One of our main objectives is to quantify genetic variation for
the tendency of an individual to self-fertilize or cross-fertilize
its eggs. A classical problem is that actual selfing rates are
2656EVOLUTION NOVEMBER 2007

Page 3

INBREEDING DEPRESSION AND DELAYED SELFING
often very dependent on external conditions, especially pollinator
(in plants) or mate (in animals) availability. Previous studies have
shownthatinP.acuta(Tsitroneetal.2003b;Henryetal.2005),as
in many other animal and plant species with low-to-intermediate
selfing rates (Njiokou et al. 1992; Goodwillie 2000; Kalisz et al.
2004; Coutellec and Lagadic 2006), selfing rates are plastic. In
these species, selfing represents a form of reproductive assurance
(Baker 1955) and occurs late in reproductive life, when external
pollenormatesarelacking;astrategyknownas“delayedselfing”
(Lloyd 1992; Jarne and Charlesworth 1993; Kalisz et al. 1999;
Tsitrone et al. 2003b). Tsitrone et al. (2003a) modeled the timing
of reproduction in preferentially outcrossing species as a plas-
tic life-history trait depending on mate availability, and predicted
an optimal strategy as a function of life-history parameters and
population density. Their model is adapted to animal situations
because they assume continuous growth and reproduction. This
model predicts a delay (or “waiting time”) between the age at first
reproduction in outcrossing conditions (with potential mates) and
the age at first reproduction in selfing conditions (without mates).
The model was validated by a laboratory experiment performed
withP.acuta(Tsitroneetal.2003b).Atanygivenpopulationden-
sity, the expected selfing rate of an individual is a monotonously
decreasingfunctionofitswaitingtime.Wethereforeusethewait-
ing time as our mating system trait. The most important point is
that Tsitrone et al. (2003b) found heritable variation for the wait-
ing time in two natural populations, and other studies (Jarne et al.
2000; Henry et al. 2003) observed variation in inbreeding depres-
sion among families, making P. acuta a good system to look for
genetic associations between the two. A potential problem with
this approach is that the waiting time is expressed only in the
absence of mates, in low-density populations, and may therefore
evolve near-neutrality if such contexts are infrequent. However,
giventhelifehistoryofP.acutalow-densityepisodesarenotrare:
theyoccur,forexample,duringtherecolonizationofpondsorseg-
ments of river beds that have dried up in summer, a frequent event
in the Mediterranean region. For instance, Henry (2002) points to
a period of several months each year in autumn and winter when
snails are so rare that we cannot detect them in the studied popu-
lation. This issue will be further considered in the Discussion.
Our experimental approach used a quantitative-genetic de-
sign with full-siblings. We focus on the estimation of two compo-
nents: (1) the waiting time, measured as the difference between
the age at first reproduction in selfing and outcrossing conditions;
and(2)inbreedingdepressiononthreelife-historytraits(survival,
fecundity, and growth).
Most previous studies (e.g., Jarne et al. 2000; Tsitrone et al.
2003b) used G0adult individuals (from the field) as parents and
measured traits on G1individuals. This has two drawbacks: (1)
the among-family variance contains an unknown part of mater-
nal effects especially for early expressed traits (Roach and Wulff
1987; Fenster 1991; Campˆ elo et al. 2004) and (2) the actual relat-
ednesswithinfamiliesisunknownbecausefield-caughtsnailscan
store sperm from an unknown number of mates (Jarne et al. 1993;
Dillonetal.2005).Toavoidthispitfall,weobtainedeggsfromG0
adult mollusks and started with approximately 100 unrelated G1
individuals kept virgin (isolated) until sexual maturity, then ran-
domly paired (Fig. 1). G1snails were paired for approximately
30 days, a largely sufficient time for many reciprocal copulations
to occur (Facon et al. 2006). After some mortality occurred, we
obtained offspring for 36 pairs, making 36 full-siblings, unrelated
G2families. Offspring of each G1pair were combined to produce
a G2family. Other G0individuals, not related to the experimental
families, were kept as a large stock colony.
Each G2family was randomly split into two treatments, out-
crossing and sequential mixed mating (hereafter SMM) (Fig. 1).
Outcrossers(∼threetofourindividualsperfamily;N =128)were
givenamaterandomlychosenfromthestockcolonyforthreeperi-
odsof3hperweek(9h/week).Thisprotocolminimizesdensityor
group effects (e.g., delayed growth or reduced fecundity in paired
individuals due to competition; Doums et al. 1994; Tsitrone et al.
2003b) while leaving time for many copulations to occur as soon
as snails are sexually mature (J. S. Escobar, pers. obs.). Partners
were marked on the shell with a dot of car-body paint (felt tip
pen from Sinto Peinture, France), and each G2in the outcross-
ing treatment received a different (randomly chosen) partner for
each 3-h period. Paint marks on the shell have been previously
shown to be harmless and have no effect on fitness (Henry and
Jarne 2007). SMM made up three-fourths of all G2individuals
(∼9–10 individuals per family; N = 348) and were kept isolated
until they self-fertilized. After the first clutch, we left them iso-
lated until at least 50 self-fertilized eggs were collected (left for
1–7daysapproximately),theneachofthemwasmatedfor24hto
a single, randomly chosen partner from the stock colony. We thus
obtained both self- and single-sire-cross-fertilized G3offspring
fromeachG2motherintheSMMtreatmenttoestimateinbreeding
depression.
Atallstages,offspringofthesameclutch(age=0day)were
raised together for two weeks. At 15 days they were transferred to
75-mLboxes(onegroupoffiveindividualsperbox)fortwomore
weeks,andsubsequently(age=30days,priortosexualmaturity)
isolatedin75-mLboxes.Thisprotocol,referredascommontreat-
ment in Figure 1, allows us to raise offspring together, keeping
the number of experimental units (75-mL boxes) at a reasonable
level, and prevents mating among siblings. The G2outcrossing
treatment started at 30 days.
TRAITS MEASURED
The shell height of G2individuals was measured with a binoc-
ular stereomicroscope micrometer when they first reproduced in
bothtreatments(outcrossingandSMM,whichatthisstageequals
EVOLUTION NOVEMBER 2007
2657

Page 4

JUAN S. ESCOBAR ET AL.
Figure 1. Schematic diagram of the experimental protocol. Girefers to generations. The waiting time was measured in G2individuals as
the difference in age at first reproduction in outcrossing and sequential mixed mating (SMM). Inbreeding depression was measured in G3
individuals for survival (early or ES, middle age or MAS, and late or RLS), fecundity (49 days), and size (30 and 80 days). Stars represent
pair mating with stock colony mollusks. More details are provided in text.
selfing). We also counted the number of eggs laid at this occasion
(1–2days)andrecordedtheirageatfirstreproduction.Thediffer-
ence between the average age at first reproduction in SMM indi-
viduals within a family and that of their outcrossing full-siblings
represents the family-level waiting time. Because full-siblings G2
families did not have significantly different ages at first repro-
duction in the outcrossing treatment (see Results), we also used
individual age at first reproduction in SMM as an individual mea-
sure of waiting time. This avoids the incorporation of additional
within-familyerrorvarianceassociatedwiththeoutcrossingtreat-
ment into the estimate of the waiting time.
In G3individuals (both selfed and outcrossed; Fig. 1) we ob-
tained an estimate of early survival (hereafter ES), including the
hatching rate, calculated for each capsule as the number of live
(free-moving) individuals at 15 days after egg-laying divided by
the number of eggs in the clutch. Note that no egg was ever ob-
served to hatch after 15 days. Shell height was measured in G3
mollusks at 30 days (i.e., prior to sexual maturity). At 45 days,
virgin G3individuals were pair-mated with paint-marked adults
partners from the stock colony during 24 h then isolated for 72
h; the eggs laid during this 72-h period constitute our measure of
fecundity (age = 49 days). After this (age = 50 days), G3mol-
lusks were individually tagged using a numbered, colored plastic
mark for honey bees (Ickowicz Apiculture?; diameter: 2.3 mm,
weight: 1.8 mg) and a dot of car-body paint on the shell (Henry
and Jarne 2007). This double marking provided unique combi-
nations for more than 2000 snails while limiting tag loss. We
computed survival before marking (hereafter middle age survival,
MAS) by dividing the number of survivors at 50 days by the num-
ber of individuals alive at 15 days. Marked individuals spent the
rest of their lives in 8.3-L plastic tanks at a constant density (200
individuals per tank). Density was adjusted within tanks every
week, using when necessary unmarked nonexperimental individ-
uals from the stock colony. Individuals were randomized across
tanks every week to avoid common-environment effects. Shell
height was measured at 80 days. Dead marked mollusks were in-
dividually recorded three times a week. Shell height at 30 days
in G3individuals was measured with a binocular stereomicro-
scope micrometer, whereas at 80 days shell height was measured
by means of pictures taken with a standard scanner and analyzed
withtheImageTool3.00software(UTHSCSA).Latesurvivalwas
quantified as the age at death minus 50 days. We consider this
measure as representative of the reproductive life span (hereafter
RLS), because the first reproduction occurred between 45 and
50 days for nearly all G3individuals, constrained by the fact that
we provided the mates only at that time.
2658EVOLUTION NOVEMBER 2007

Page 5

INBREEDING DEPRESSION AND DELAYED SELFING
Selfed and outcrossed ES was estimated from 48,160 G3
eggs; MAS and RLS were recorded from 2973 and 2179 G3mol-
lusks,respectively.Fecunditywasrecordedfrom2636individuals
and sizes were measured from 2069 and 1398 individuals at 30
and 80 days, respectively. The total duration of the experiment
was 379 days.
STATISTICAL ANALYSES
Size and age at first reproduction in G2individuals were analyzed
using ANOVA with treatment (outcrossing and SMM) as fixed
effect and family as random crossed effects. Normal distribution
of errors and homoscedasticity were checked. Additive genetic
variance (VA) was estimated for a full-sibling design assuming
no dominance (Falconer and Mackay 1996) as VA= 2(?f amilies).
Narrow sense heritability (h) and corresponding standard errors
were then estimated as suggested by Falconer and Mackay (1996,
p. 180).
ESTIMATES OF INBREEDING DEPRESSION AT THE
INDIVIDUAL AND FAMILY LEVEL
G3 individuals are characterized by treatment (selfed vs. out-
crossed), grand family of origin (corresponding to a pair of G1
grand parents) and full-sibling family within grand family (cor-
responding to one G2mother). For clarity we will refer to grand
familyandfull-siblingfamilyeffectsas“family”and“G2mother”
effects, respectively.
Inbreeding depression on G3individuals estimated for each
character and each G2mother was calculated using the classi-
cal equation (Charlesworth and Charlesworth 1987) ?j= 1 −
(Wi,j
?Wo,j), where Wi,jis the mean of a fitness trait of inbred
offspring from mother j, and Wo,jis the fitness of outbred off-
spring from the same mother. However this estimate has several
unsuitable statistical properties when it comes to testing for fam-
ily differences and combining various fitness traits (Johnston and
Schoen 1994). Using log-transformed traits is an efficient way to
express traits in both a relative and additive scale, as suggested by
Johnston and Schoen (1994). Thus we used the following mea-
sure, equivalent to the inbreeding load measured at the family
level, as defined in Charlesworth and Charlesworth (1987) and
Willis (1999)
?ln,j= −ln(Wi,j/Wo,j) = ln(Wo,j) − ln(Wi,j).
(1)
Note that ?lncan be back-transformed to recover classical
inbreeding depression, using ?j= 1 − Exp(−?ln,j).
We computed ln-inbreeding depression estimates for each
G2mother and family for each life-history trait measured in G3
individuals (three measures of survival, fecundity, and two size
measures). For survival traits (ES and MAS), in some cases, no
eggorindividualsurvived(W =0)andlogarithmictransformation
was impossible. However these data are important because they
represent extreme values of depression; instead of removing such
data,weusedthetransformationln(0.01+W)insteadofln(W)in
equation (1). The choice of the constant 0.01 is based on the fact
that given that most of the time we measure survival on the basis
of less than 100 eggs or individuals per capsule, we are unable to
detect survival differences below 0.01. This value thus represents
a reasonable “floor” for our survival measures.
Estimates of ln-inbreeding depression on survival obtained
for ES (0–15 days), MAS (15–50 days), and RLS (>50 days)
were combined into a single cumulative measure (inbreeding de-
pression for reproductive life expectancy, hereafter RLE). RLE
represents the expected number of days during which an individ-
ualisaliveandabletolayeggs,andwascalculatedforeachfamily
(or G2mother) as follows. Individuals may either die before re-
production or survive up to the age of first reproduction (50 days)
and have a non-null RLS. In a given group of individuals (e.g., a
family), the RLE is therefore:
RLE = ES × MAS × mean RLS,
(2)
where the mean RLS is over individuals that survived up to re-
production (50 days). Using equation (1), it is easy to show that
estimates of ln-inbreeding depression for ES, MAS, and RLS can
be simply added together to provide the ln-inbreeding depression
of RLE. Likewise, using our fecundity measure at 49 days as an
estimate of fertility during the reproductive period, the expected
lifetime reproductive success (LRS) is proportional to the product
of RLE and fecundity; so the ln-inbreeding depression for LRS
can be estimated as the sum of ln-inbreeding depression of RLE
and fecundity. Analyses of inbreeding depression on size (30 and
80 days) were performed separately. Size determines fitness only
viaitseffectsonfecundityandsurvival(i.e.,largeindividualsusu-
ally lay more eggs and survive better), but because fecundity and
survival were already taken into account in the LRS measure, size
was not directly included in it.
TESTS OF VARIATION OF INBREEDING DEPRESSION
AMONG FAMILIES AND G2MOTHERS
To test whether families and/or G2 mothers have significantly
different inbreeding depression, we performed deviance analyses
on G3traits (Crawley 2005) by fitting generalized linear models
(hereafter GLM) with R (R Development Core Team 2006). The
effects included in the models were treatment, family, G2mother
(nested within family), and their interactions. Proportions of dead
versus alive individuals (ES and MAS) were analyzed under a
GLM with logit link and binomial error distribution (Crawley
2005, pp. 247–262). Data on RLS are of a different kind, that is,
ages at death (minus 50 days, because all individuals were 50-day
old when this measure started) recorded for each G3individual.
EVOLUTION NOVEMBER 2007
2659