Ecology, 94(11), 2013, pp. 2575–2582
! 2013 by the Ecological Society of America
Adaptive paternal effects? Experimental evidence that the paternal
environment affects offspring performance
ANGELA J. CREAN,
1,2,4
JOHN M. DWYER,
1
AND DUSTIN J. MARSHALL
1,3
1
School of Biological Sciences, University of Queensland, Brisbane QLD 4072 Australia
2
Evolution and Ecology Research Centre and School of Biological, Earth and Environmental Sciences,
University of New South Wales, Sydney NSW 2052 Australia
3
School of Biological Sciences, Monash University, Melbourne VIC 3800 Australia
Abstract. The ability of females to adaptively influence offspring phenotype via maternal
effects is widely ack nowledged, but corresponding nongenetic paternal effects remain
unexplored. Males can adjust sperm phenotype in response to local conditions, but the
transgenerational consequences of this plasticity are unknown. We manipulated paternal
density of a broadcast spawner (Styela plicata, a solitary ascidean) using methods shown
previously to alter sperm phenotype in the field, then conducted in vitro fertilizations that
excluded maternal effects and estimated offspring performance under natural conditions.
Offspring sired by males from low-density experimental populations developed faster and had
a higher hatching success than offspring sired by males living in high densities. In the field,
offspring survived relatively better when their environment matched their father’s, raising the
possibility that fathers can adaptively influence the phenotype of their offspring according to
local conditions. As the only difference between offspring is whether they were artificially
fertilized by sperm from males kept in high- vs. low-density cages, we can unequivocally
attribute any differences in offspring performance to an environmentally induced paternal
effect. Males of many species manipulate the phenotype of their sperm in response to sperm
competition: our results show this plasticity can influence offspring fitness, potentially in
adaptive ways, raising the possibility that adaptive nongenetic paternal effects may be more
common than previously thought.
Key words: external fertilization; nongenetic inheritance; parental effect; paternal effect; sperm
phenotype; sperm plasticity; Styela plicata.
INTRODUCTION
Populations are linked across generations both
demogra phi ca lly (numer ically) and phenotypically.
While the demographic links among generations are
obvious and well studied (Ca ley et al. 1996), the
population consequences of phenotypic links between
generations are more subtle and have received less
attention (Marshall and Morgan 2011). The best known
sources of nongenetic phenotypic links between gener-
ations are maternal effects, whereby the environment
and/or phenotype of the mother affects the phenotype of
the offspring (Mousseau and Fox 1998). Maternal
effects can take a variety o f f orms ranging from
hormonal ly indu ced mor phol og ica l and behavi or al
changes, to transference of resistance to pollution and
pathogens, to epigenetic modification of gene expression
(Uller 2008, Jablonka and Raz 2009). They can buffer
offspring from environmental change, or act as conduits
whereby stress in one generation reduces productivity in
the n ext (Mousseau and Fox 1998). Surprisingly,
variation in offspring phenotype can have a greater
influence on population structure than variation in
offspring number (Burgess and Marshall 2011a), and
impacts of maternal effects on population dynamics can
persist for up to three subsequent generations (Benton et
al. 2005). Moreover, mothers can adaptively adjust
offspring phenotype to match local conditions, and
hence maternal effects can initiate and direct adaptive
evolution (Galloway and Etterson 2007, Uller 2008).
While the ecological role of maternal effects is
increasingly well recognized, other sources of nongenetic
phenotypic links among generations have largely been
ignored. In particular, the ecological consequences of
nongenetic paternal effects are poorly understood.
Traditionally, paternal effects (with the obvious excep-
tions of paternal care and paternal genetic effects) have
been assumed to be absent or negligible (Mousseau and
Fox 1998); but this assumption has been challenged by a
recent series of studies showing transgenerational
epigenetic effects transferred down the male line
(reviewed in Curley et al. 2011, Jenkins and Carrell
2012), and transmission of compounds in the sperm
cytoplasm and accessory gland products (Avila et al.
2011). In humans, for example, early paternal smoking
has been linked to an increased body mass index in sons,
and paternal grandfather’s food supply has been linked
Manuscript received 30 January 2013; revised 6 May 2013;
accepted 7 May 2013. Corresponding Editor: A. L. Shanks.
4
E-mail: a.crean@unsw.edu.au
2575
to the mortality risk of grandsons (Pembrey et al. 2006).
In mice, paternal obesity reduces sperm function (Bakos
et al. 2011), impairs embryo development and viability
(Mitchell et al. 2011), and diminishes the reproductive
health of offspring for two generations (Fullston et al.
2012). To date, studies of nongenetic paternal effects
have focused on the transgenerational effects of paternal
behavior, nutrition, toxins, and age (Curley et al. 2011).
Thus, with the exception of cases of paternal care, it is
unclear whether nongeneti c paternal effects can be
adaptive, whether they are vehicles for the transfer of
pathological effects, or whether they are simply physi-
ological inevitabilities (Marshall and Uller 2007).
Concurrently, evidence is accumulating that males can
adjust their sperm phenotype in response to their social
environment and perceived risk of sperm competition
(e.g., Cornwallis and Birkhead 2007, Simmons et al.
2007, Crean and Marshall 2008, Immler et al. 2010). The
functional consequences of this plasticity in sperm
phenotype have only been considered in terms of their
effects on fertilization success (Pizzari and Parker 2009),
and thus this potentially important source of paternal
effects remains unexplored. The recent evidence that
sperm can transfer more than just the paternal genome
raises the possibility that plasticity in sperm phenotype
may have transgenerational consequences for offspring
performance, and thus population dynamics (Zeh and
Zeh 2008). In biomedical studies of humans, there has
been increasing speculation regarding the existence of
‘‘adaptive’’ paternal effects (e.g., the ‘‘thrifty telomere
hypothesis’’; Eisenberg 2011), whereby males adjust the
phenotype of their sperm in response to their own
environment in order to increase offspring fitness. While
the potential for such effects is clearly exciting, it has not
been explored empirically.
External fertilizers are an ideal system in which to
empirically examine paternal effects, as eggs are released
before fertilization, precluding the possibility of females
biasing their investment in response to sperm quantity
or quality (Cunningham and Russell 2000, Sheldon
2000). In addition, sessile organisms cannot move after
settlement to cope with environmental heterogeneity,
and therefore developmental and tra nsgener ational
plasticity may be particularly important mechanisms to
ensure offspring success in these species (Galloway and
Etterson 2007, Uller 2008). However, unlike plants, as
both gametes are released prior to fertili zation in
broadcast spawners, paternal effects may be equally as
likely as maternal effects to develop in these systems.
Variation in conspecific density is an i mportant
ecological driver in marine sessile communities,
directly influenc ing the availability of food and space,
with strong effects on post-settlement survival and
growth (Allen et al. 2008). We have previously
demonstrated that the broadcast-spawning ascidian
Styela plicata adjusts the phenotype of their sperm
(size, motility, and longevity) in response to changes
in the density of co nspecifics (Crean and Marshall
2008). Furthermore, significant spatial genetic struc-
ture has been detected am ong S. plic ata individuals at
the same loca ti o n as the pres en t st ud y tha t wer e
0–5 m apart (David et al. 2010), suggesting that
dispersal distances are short an d thus the paternal
environment is likely to be a good predictor of the
offspring environment: a requirement for adaptive
transge ne ra ti on a l plastic it y (Gallow ay and Etterson
2007). Here , we manipulate the paternal environment
using the same field manipulation previously shown to
induce plasticity in sper m phenotype (Crean and
Marshall 2008). We use sperm collected from males
in these high- and low-density treatments to in vitro
fertilize a common pool of eggs, and track the
performance of offspring. Offspring developmental
performance is mea sured in the laboratory (due to
logistical constraints), and post-metamorphic perfor-
mance is measured in the field in both low- and high-
densit y co nditio ns t o te st for adap tive paternal
effects.
M
ATERIALS AND METHODS
Study species and field location
We used the solitary ascidian, Styela plicata, as a
mod el system to examine envir onmentally induced
paternal effects, as previous work has shown that they
exhibit plasticity in sperm size, motility, longevity, and
fertilization potential (Crean and Marshal l 2008).
S. plicata is a hermaphroditic broadcast spawner
commonly found in disturbed habitats such as marinas
and harbors throughout eastern Australia (Kott 1972).
We completed all field work at the East Coast Marina
(Manly, Brisbane, Australia; 27.4678 E, 153.1838 S), a
sheltered private-access marina consisting of floating
pontoons. At this site, S. plicata naturally occur in
densities ranging from isolated individuals to large
clumps (e.g., .50 adults on 1 m length of rope; A. J.
Crean, personal observation). Reproductively mature
individuals are present at this site most of the year,
excluding winter months (July to September; A. J.
Crean, personal observation).
Manipulation of the paternal habitat
The paternal envi ronment was manipulated using the
same density manipulation methods that were shown
previou sly to indu ce adapt ive plas tic ity in sper m
phenotype (Crean and M arshall 2008). Briefly, repro-
ductive adults were collected from the field site and
randomly allocated to either a high-density (15
individuals) or a low-density (single individual) treat-
ment cage (cage dimensions, 18 3 18 3 18 cm; mesh
size, 1 cm
2
). Treatment cages were suspended from the
pontoons and maintained in the field for one month,
and th en a randomly selected h igh-density in dividual
from each cage and all low-density individuals were
transported to the laboratory at the University of
Queensl and. Im portantly, no mortality of tr eatment
ani mals occurred in any of the high- or low-density
ANGELA J. CREAN ET AL.2576 Ecology, Vol. 94, No. 11
cag es, and thus dif ferences in offspring performance
cannot be explained by selection actin g on fathers in
each treatment group. We also collected randomly
selected non-treatment individual s from our field site to
be used as mothers in each as say. Data were collected
from three experimental runs beginning in January,
April, and June 2009, respe ctively. Measurements of
time to hatchi ng, hatching success, and larval size were
collected in run one, cell cleavage rate i n run two, and
post-metamorphic survival in runs two and three.
Within each run, multip le trials were conducted (run
one, six trials; run two, five tr ials; run three, five trials),
with each trial using a paired, split-clutch design: a
common pool of eggs was split into two groups and
then each group fertilized with sperm from either a
high-density or low-de nsity treatment male. Trials were
initiated consecutively over two days to minimize
temporal differences.
In vitro fertilization
Experimental in vitro fertilization roughly approxi-
mates natural reproduction for external fertilizers, and
can be tightly controlled such that all offspring are
reared under standardized conditions from fertilization
through to deployment in the field (Marshall et al.
2008). As we were interested in the nongenetic effects of
the paternal environment on offspring performance, the
sperm from males in density treatments was used in vitro
to fertilize eggs from non-treatment animals in stan-
dardized conditions. Gametes were extracted using
standard strip-spawning techniques (Crean and Mar-
shall 2008). Eggs were harvested from four randomly
selected non-treatment animals, combined together, and
5 mL of this combined egg solution was transferred into
each of two 35 mm diameter petri dishes, one for each
paternal treatment. Eggs from multiple females were
used in each trial to reduce maternal effects and male-
by-female interactions that could confound the inter-
pretation of results. A 5-mL control egg sample was also
put aside and not exposed to sperm to check for any self-
fertilization of eggs (,1% in all trials). We then collected
5 mL of sperm solution from one high-density and one
low-density treatment animal (order of treatments
randomized), and added the sperm to its assigned petri
dish, gently shaking each sample every two minutes to
mix the sperm and eggs. We washed eggs free of sperm
after 15 minutes (so exposure time was constant between
treatments) by gently rinsing each sample through a 100-
lm filter, and left the eggs to develop in 10 mL of filtered
seawater in a covered petri dish. The concentration of
sperm solution collected from each male was estimated
using a Neubauer improved hemocytometer under 4003
magnification (three replicate counts per sample). There
was no difference in the average concentration of sperm
collected from males in high- and low-density treatments
(high-density males, 1.719 3 10
8
6 1.861 3 10
7
sperm/
mL [mean 6 SE]; low-density males, 1.608 3 10
8
6 1.627
3 10
7
; paired t ¼ 0.075; df ¼ 15; P ¼ 0.941; Appendix:
Fig. A1).
Cell cleavage rate
To examine the effect of paternal environment on
embryonic cell cleavage rates, we placed each petri dish
with the developing eggs under a dissecting microscope
at 203 magnification, and used time-lapse photography
to record a digital image every 30 s with PixeLINK
Capture SE software (PixeLINK, Ottawa, California,
USA). As we did not know the exact time of
fertilization, cell development rate was estimated from
the time of the first cleavage to the four-cell stage (Rius
et al. 2010). The size of fertilized eggs was measured
from images before the first cleavage using Image-Pro
Express (version 5.1; Media Cybernetics, Silver Spring,
Maryland, USA). We estimated total egg area by
digitally tracing around the perimeter of the follicle
cells, and ovicell area by tracing around the intersection
of the follicle cells and ovicell (Crean and Marshall
2008). We measured 15 eggs per paternal treatment per
trial.
Time to hatching and hatching success
To measure development time and calculate hatching
success, fertilized eggs were individually collected with a
micro-pipette and transferred into individual 10 mm
diameter wells with 2 mL of filtered seawater in a 24-well
plate (one plate per replicate ¼ 24 subsamples). These
plates were left in a constant temperature (CT) cabinet
at 228C overnight. Developing embryos were observed
under a microscope (303 magnification) every 15
minutes (starting 10 hours post-fertilization) to measure
time to hatching. Eggs that had not hatched after 15
hours, and hatched larvae with highly abnormal
morphology, were classed as unsuccessful. Digital
images of successfully hatched larvae were recorded
under a microscope (453 magnification), and larval area
was measured by tracing around the perimeter of each
larva using Image-Pro Express. Larval area was
calculated from the mean of at least three traces from
different images of the same individual to minimize the
effects of measurement error.
Post-metamorphic survival
To measure the post-metamorphic performance of
offspring sired by fathers that experienced different
environments, we settled larvae sired by low- and high-
density fathers on to 35 mm diameter petri dishes and
then deployed them into the field for two weeks.
Although this may seem to be a short time frame to
measure offspring performance, a previous study on
Styela plicata at the same field site showed that 99% of
all post-settlement mortality occurred within the first
two weeks after settlement (Rius et al. 2009). We also
examined whether the effects of paternal environment
were context dependent by varying the density of settlers
in the field. Our analysis therefore included treatments
November 2013 2577ADAPTIVE NONGENETIC PATERNAL EFFECTS
applied at two scales: the paternal density treatment was
applied at the scale of trial, and the offspring density
treatment was applied at the scale of petri dish.
We fertilized eggs with sperm from males in low and
high densities, transferred the embryos from each
treatment into separate 250-mL beakers filled with
filtered seawater, and left them covered overnight in
the CT cabinet at 228C. Larvae were collected with a
pipette the following morning (approximately 11 hours
after fertilization), and transferred to pre-roughened and
bio-filmed petri dishes in a drop of water. For low-
density offspring treatments, we transferred five larvae
to each petri dish (six low-density offspring replicates
per paternal treatment per trial), to ensure that at least
one larva settled on the base of the dish. For offspring
allocated to the high-density treatment, we transferred
between twenty and thirty larvae to each petri dish
(depending on how many larvae successfully hatched),
with four high-density offspring replicates per paternal
treatment per trial. We first ensured larvae were free
swimming and not stuck in the water surface layer, and
then placed the covered petri dishes in the CT cabinet
for 24 hours to allow larvae to settle.
In low-density offspring treatments, we marked the
position of a haphazardly selected settler, and removed
all other settlers so that low-density treatments had a
single individual to mirror paternal density manipula-
tions. In high-density offspring treatments, all settlers
were marked and numbered (number of settlers per dish
ranged from 11 to 26; mean ¼ 20 individuals; mean
density ¼ 0.02 individuals/mm
2
). All replicates were then
transported to the marina in an insulated container filled
with seawater, and deployed in the field that afternoon.
Petri dishes were suspended vertically within a rigid
plastic mesh cage (dimensions: 44 3 28 3 18 cm length 3
width 3 height; mesh size 1 cm
2
), as the majority of
S. plicata were observed growing on vertical surfaces at
the field site. Replicates were enclosed within cages as
pilot studies showed that fish are attracted to field
equipment and preferentially eat S. plicata settlers (A. J.
Crean, personal observation). These cages were hung
from the pontoons in the marina, approximately 2 m
below the water surface, and left in the field for two
weeks. After this time, all settlers were transported back
to the laboratory and examined under a microscope
(303 magnification) to calculate survival.
Statistical analyses
Hatching success was modeled as a binary response
using mixed-effects logistic regression (logit-link func-
tion and binomial errors), with treatment (high or low
paternal density) included as a fixed effect and trial
included as a random effect. Eggs that did not hatch or
those that hatched with visible lethal deformities were
considered unsuccessful and given a value of zero in this
analysis. Time to hatching was analyzed using mixed-
effects Cox proportional hazards models, again with
paternal density as a fixed effect and trial included as a
random effect. Exploratory Kaplan-Meier plots indicat-
ed that these data met the assumption of proportional-
ity. We ran the Cox models using two ce nsoring
approaches that make different assumptions about the
fate of unhat ched eggs. First, we onl y included
individuals that hatched successfully and no individuals
were censored. This approach ignores the possibility that
some unhatched eggs could have hatched once moni-
toring ceased. Second, we included all individuals that
hatched successfully and those that did not hatch within
the monitoring period. The individuals that did not
hatch were censored. This approach assumes that all
unhatc hed eggs will eve nt ua ll y hatch, d es pit e the
prospect that ma ny of the unhat ch ed e g gs w ere
nonviable. Average estimates of sperm concentration,
cell development time, egg ovicell and total area, and
larval area were calculated for each trial, and paired t
tests used to test for consistent differences between low
and high-density paternal treatments.
Post-metamorphic survival was modeled as a binary
response using mixed-effects logistic regression (logit-
link function and binomial errors), with trial and dish
included as random effects. The number of larvae that
settled in high-density offspring treatments ranged from
11 to 26 individuals per dish, and for this reason we
chose to treat density as a continuous variable rather
than categorical to retain more of the information
contained within the dataset. Hence, fixed effects were
paternal densit y (h igh or low), offsprin g density
(continuous) and their interaction. As survival data
was collected in runs two and three, we checked for a
possible run effect by fitting models that allowed slopes
and intercepts to vary by run (as random effects). All
random-effect terms for run were not significant
(Appendix: Table A1) and the fixed-effect estimates
were very similar to those from models that excluded
run. We therefore present data pooled across both runs
for clarity. The significance of fixed-effect terms was
assessed using likelihood ratio tests (LRTs) by compar-
ing nested models with and without the variable of
interest. All analyses were completed using the R
statistical program (R Development Core Team 2013),
and the LME4 package (Bates et al. 2013) used to fit all
mixed effects logistic models.
R
ESULTS
Offspring developmental performance
Embryos fertilized by sperm from fathers in low-
density environments were more likely to successfully
complete development and hatch into larvae ( Fig. 1a),
with 69% (95% CI 55.2–79.5) of eggs fertilized by high-
density males successfully hatching into larvae, com-
pared with an 80% (95% CI 71.1–87.6) hatching success
rate of eggs fertilized by low-density males. Paternal
environment also influenced the time offspring took to
hatch into larvae, with offspring from high-density
males taking longer to complete development than
offspring from low-density males (Fig. 1b). This result
ANGELA J. CREAN ET AL.2578 Ecology, Vol. 94, No. 11
was confirmed by the mixed-effects Cox models,
although the effect of paternal density on hatching time
was more pronounced in the model that excluded all
unhatched eggs (low paternal density coefficient ¼ 0.79,
SE ¼ 0.154, v
2
1
¼ 26.4, P , 0.0001), compared to the
model that censored unhatched eggs (low paternal
density coefficient ¼ 0.50, SE ¼ 0.14, v
2
1
¼ 12.9, P ¼
0.0003). These coefficients can be interpreted as hazard
ratios to assist interpretation of relative treatment
differences. Using the latter model as an example, the
hazard ratio was e
0.5
¼ 1.65. Thus, by the time that 50%
of the eggs hatched in the high-paternal-density treat-
ment, 68% (1 " 0.5
1.65
) had already hatched in the low-
paternal-density treatment.
To determine what drove this difference in time to
hatching, we me asur ed the cleavage rate of eggs
fertilized by males in each environment. Eggs fertilized
by sperm from high-density males took 6% longer on
average to develop from the two-cell to the four-cell
stage than eggs fertilized with the sperm of low-density
males (mean time from two- to four-cell stage: high-
density father ¼ 882 s, low-density father ¼ 830 s; paired
t ¼ 3.104, df ¼ 4, P ¼ 0.036; Appendix: Fig. A1b). Thus,
it appears that differences in time to hatching are driven
by a faster developmental rate of offspring sired by
males in low densities. There was no significant
difference between paternal density treatments in the
size of eggs that were fertilized (total egg area, paired t ¼
"0.558, df ¼ 4, P ¼ 0.607; ovicell, paired t ¼"0.623, df ¼
4, P ¼ 0.567; Figs. A1b and c), and therefore differences
in developmental rate between paternal treatments are
unlikely to be driven by differences in egg size.
Offspring post-metamorphic survival
The effect of the paternal environment on the post-
metamorphic survival of offspring was context-depen-
dent, indicated by a significant interaction between
paternal and offspring densities (LRT, v
2
1
¼ 5.1, P ¼
0.020; Table 2; Fig. 2a). To investigate this interaction
further, we fitted two sets of post-hoc models; essential-
ly, simple main effects tests for exploring significant
interactions (see Quinn and Keough 2002). First, we ran
mixed-effects logistic regressions for each of the paternal
environments to assess the significance of slopes with
offspring density. This revealed that offspring density
had a significant negative effect on survival of offspring
of low-density males (LRT, v
2
1
¼ 15.2, P , 0.001), but
that offspring from high-density males survived equally
well at all densities (LRT, v
2
1
¼ 1.03, P ¼ 0.310). Second,
we categorized offspring density into thei r original
low- and high-density treatments and analyzed the
relative survival of offspring within each treatment; as
the relative fitness of offspring from each paternal
treatment within each offspring treatment can provide a
more evolutionarily infor mative estimate of fitness
(Kawecki and Ebert 2004, Stanton and Thiede 2005,
Burgess and Marshall 2011b). In low-density offspring
environments, offspring relative survival was higher for
individuals from low-density males, although this
difference was not statistically significant (LRT, v
2
1
¼
2.25, P ¼ 0.134; Fig. 2b). In high-density offspring
environments, the relative survival wa s l ower f or
individuals from low-density males (LRT, v
2
1
¼ 10.41,
P ¼ 0.001; Fig. 2b). As there was no difference in larval
size between paternal-density treatments (paired t ¼
1.397, df ¼ 5, P ¼ 0.221), these observed differences in
pos t-me tamorphic survival cannot be explained by
differences in offspring size.
D
ISCUSSION
Conspecific density in the paternal environment has
transgenerational consequences for offspr ing perfor-
FIG. 1. Effect of manipulating paternal density on offspring
developmental traits. Offspring sired from high-density males
are shown in black; offspring sired by low-density males are
shown in gray. Large points in panel (a) show the probability
(mean 6 SE) of hatching into a larva with a normal phenotype.
The small points linked by lines are the hatching probabilities
for each trial, as estimated by the mixed-effects logistic model.
The box and whisker plots in panel (b) represent raw values of
time from fertilization to hatching. The thick lines are medians,
boxes are 0.25 and 0.75 quantiles, and whiskers indicate
maximum and minimum values accounting for outliers (circles).
November 2013 2579ADAPTIVE NONGENETIC PATERNAL EFFECTS