Content uploaded by Joseph Travis
Author content
All content in this area was uploaded by Joseph Travis on Dec 12, 2013
Content may be subject to copyright.
q
2005 by the American Society of Ichthyologists and Herpetologists
Copeia, 2005(3), pp. 649–656
Population Differences in Pre- and Post-fertilization Offspring
Provisioning in the Least Killifish, Heterandria formosa
M
ATTHEW
S
CHRADER AND
J
OSEPH
T
RAVIS
We used data from a long-term field study of two populations of the Least Killi-
fish, Heterandria formosa, to examine whether genetically based population differ-
ences in offspring size at birth are mainly due to differences in pre-fertilization
offspring provisioning (i.e., differences in egg mass) or differences in post-fertiliza-
tion offspring provisioning (i.e., the degree of matrotrophy). We found differences
between populations in pre-fertilization offspring provisioning and larger differenc-
es in post-fertilization provisioning. These results establish population variation in
H. formosa as a potential model for studying the costs and benefits that could mod-
ulate the evolution of matrotrophy. In addition, our results illuminate some of the
costs and benefits associated with matrotrophy and offer insight into how matrotro-
phy influences the expression of other life history relationships, including that be-
tween female size and fecundity, the trade-off between offspring number and size,
and the population-specific responses of offspring number and size to variation in
population density.
P
ARENTAL care can be defined as any activ-
ity of a parent that is likely to benefit the
offspring of that parent (Clutton-Brock, 1991).
In its broadest sense, this definition includes
several parental behaviors including nest build-
ing and defense, parental investment in eggs,
and the provisioning of offspring before and af-
ter birth (Clutton-Brock, 1991; Roff, 2002). Ma-
trotrophy, the provisioning of offspring between
fertilization and birth via a maternal-embryo
connection such as a placenta, is a complex
mechanism of parental care. The evolution of
matrotrophy as a form of parental care presents
a number of challenges. First, matrotrophy rep-
resents a major life history transition because it
requires a shift in the timing of maternal re-
source allocation to reproduction from pre- to
post-fertilization. The evolution of matrotrophy
also requires changes in traits that influence,
among other things, the transfer of nutrients
between the mother and the embryo, gas ex-
change, the disposal of waste products, and the
suppression of immunological interactions be-
tween the mother and the embryo (Haig, 1993;
Lombardi, 1996). Nevertheless, matrotrophy is
a widespread process found in several taxa (e.g.,
some fish, amphibians, reptiles, and nearly all
mammals) and has evolved independently sev-
eral times within at least one genus of fish (ge-
nus Poeciliopsis, Reznick et al., 2002).
There has been much interest in the ecology
and evolution of matrotrophy as a component
of parental care (Thibault and Schultz, 1978;
Lombardi, 1996; Trexler and DeAngelis, 2003)
and as a model system for studying the evolu-
tion of novel, complex adaptations (Reznick et
al., 2002). However, biologists have only recent-
ly begun to investigate the costs and benefits of
this type of parental care. Several potential costs
have been proposed. For example, compared to
lecithotrophy (embryo nourishment by yolk de-
posited in the egg prior to fertilization), matro-
trophy increases the duration of maternal-em-
bryonic nutrient transfer from a period of a few
days to a period of several weeks (Thibault and
Schultz, 1978; Lombardi, 1996). In addition,
matrotrophy is likely to be energetically costly
and may enhance predation risk on the mother,
especially as large embryos approach the end of
gestation (Trexler et al., 1994). Matrotrophy is
also likely to increase prenatal competition be-
tween siblings for resources (Mock and Parker,
1997). Finally, Reznick et al. (1996) have sug-
gested that obligate matrotrophy is a constraint
on the subsequent evolution of certain life his-
tory traits; in particular matrotrophy may pre-
vent females from adaptively adjusting offspring
size when resources fluctuate.
Conversely, there are several potential bene-
fits of matrotrophy. First, compared to lecitho-
trophy and release of undeveloped eggs, matro-
trophy may increase offspring survivorship
through the earliest developmental stages (ex-
tending Shine’s 1995 discussion of embryo re-
tention and viviparity). Second, matrotrophy,
like other forms of post-fertilization care (e.g.,
post-natal parental care; Burley, 1986, 1988; Ko-
tiaho et al., 2003), may allow females to invest
more efficiently in offspring and modify their
investments as a function of paternal genotype.
Third, Trexler (1997) and Trexler and De-
Angelis (2003) have proposed that matrotrophy
650 COPEIA, 2005, NO. 3
allows a female to circumvent the trade-off be-
tween offspring size and number, facilitating in-
creased fecundity for a given reproductive ef-
fort. Finally, compared to lecithotrophy, matro-
trophy allows a greater regulation of resource
allocation throughout embryonic development,
a form of maternal control that is accentuated
in species with clutch overlap (superfetation)
(Travis et al., 1987). Lecithotrophic species are
not capable of such refined control (Meffe and
Vrijenhoek, 1981). In fact, in viviparous poeci-
liids, superfetation is largely restricted to matro-
trophic species (Turner, 1937; Scrimshaw, 1944,
1945), although there are exceptions (e.g., Thi-
bault and Schultz, 1978; Reznick and Miles,
1989; Arias and Reznick, 2000).
It is unclear which costs are most critical and
which benefits are most important for setting
the value of matrotrophy. Nor is it clear how
matrotrophy affects the relationships among
other facets of the life history. Almost all of our
current understanding of matrotrophy is based
on broad comparisons between lecithotrophic
and matrotrophic species (Thibault and
Schultz, 1978; Reznick and Miles, 1989). These
comparisons indicate the circumstances in
which matrotrophy has been favored but not
why or how it has evolved. Poeciliid fishes are
well suited for studying the cost-benefit equa-
tions that surround hypotheses about matrotro-
phy. Phylogenetic relationships among poeciliid
taxa are becoming clearer (Reznick et al.,
2002). This group exhibits ample interspecific
variation in offspring provisioning (from pure
lecithotrophy to mixed provisioning to pure ma-
trotrophy) (Thibault and Schultz, 1978; Reznick
and Miles, 1989; Reznick et al., 2002), and there
is considerable knowledge of the anatomical
and physiological bases of offspring provision-
ing (Grove and Wourms, 1991, 1994; Marsh-
Matthews et al., 2001). What is lacking is a thor-
ough scrutiny of intraspecific variation in ma-
trotrophy and the selective agents that maintain
that variation through specific costs and bene-
fits. Intraspecific variation in matrotrophy has
been described in Sailfin Mollies (Poecilia lati-
pinna), a species with mixed provisioning (Trex-
ler, 1985). However, the mechanism by which
female mollies provision their young after fer-
tilization is unknown. Nothing is known about
variation in matrotrophy in any of the species
that rely upon it as the primary mechanism of
offspring provisioning.
The Least Killifish, Heterandria formosa,isan
excellent candidate for testing hypotheses about
the evolution of matrotrophy and for under-
standing the consequences of matrotrophy for
overall life history evolution. Among poeciliids,
H. formosa has one of the highest levels of ma-
trotrophy, displaying up to a 30-fold increase in
embryo mass between fertilization and parturi-
tion (Reznick and Miles, 1989). In addition, H.
formosa exhibits a high level of superfetation,
with females simultaneously carr ying up to six
broods of young at different stages of develop-
ment (Travis et al., 1987). These fish inhabit a
variety of habitats with different physical, chem-
ical, and biotic conditions, which include long-
standing differences in population densities
(Leips and Travis, 1999; Soucy and Travis,
2003). In addition, populations in northern
Florida exhibit substantial variation in life his-
tories, including variation in female sizes, levels
of superfetation, brood sizes, and offspring sizes
at birth (Leips and Travis, 1999). Life history
differences between two of the most divergent
populations, Trout Pond and Wacissa River,
have been shown to have a genetic basis (Leips
et al., 2000).
The variation in offspring size among popu-
lations of H. formosa could represent variation
in the level of matrotrophy, variation in pre-fer-
tilization offspring provisioning (i.e., ovum size)
that is propagated throughout development, or
the combined effects of variation in matrotro-
phy and ovum size. If the variation in offspring
size is purely a function of pre-fertilization off-
spring provisioning, then differences among
populations should appear prior to fertilization
and embryonic gains in mass between consec-
utive stages of development should be compa-
rable in magnitude in different populations. If
the variation in offspring size is attributable to
different levels of matrotrophy, then there
should be no differences among populations in
ovum size, and variation in embryo size should
appear later in development and embryonic
gains in mass between at least some pairs of con-
secutive stages will be larger in one population
than the other. In this study we show that the
well-documented differences in offspring size
between two populations of H. formosa are in
part the result of differences in the sizes of ma-
ture ova and, in larger part, differences in ma-
trotrophy during the late stages of develop-
ment. In addition, we show that understanding
these differences in pre- and post-fertilization
offspring provisioning illuminates other life his-
tory relationships, including different norms of
reactions of life history traits to variation in pop-
ulation density.
M
ATERIALS AND
M
ETHODS
Study populations.—Trout Pond is a small (5 ha)
lake located in the Apalachicola National For-
651SCHRADER AND TRAVIS—MATROTROPHY IN H. FORMOSA
est, Leon County, Florida. The Wacissa River is
a spring-fed river located in Jefferson County,
Florida. The two study sites are separated by ap-
proximately 50 km, and data from allozymes
and microsatellites suggest there is little or no
gene flow between these two populations (Baer,
1998; Soucy and Travis, 2003). Trout Pond and
the Wacissa River differ in several abiotic and
biotic conditions. These differences are de-
scribed in detail by Leips and Travis (1999);
however, we will summarize some of the major
differences here. Trout Pond has consistently
lower pH and lower conductivity than the Wac-
issa River. In addition, water temperature is
more variable throughout the year in Trout
Pond (range 14–38 C) than in the Wacissa River
(17–28 C). Densities of H. formosa are consis-
tently lower in Trout Pond than in the Wacissa
River, and it appears that per capita predation
pressure is greater in Trout Pond.
Sample collection.—Up to 20 haphazardly chosen
adult females were collected from Trout Pond
and the Wacissa River during each month of the
breeding season (March–September) between
September 1999 and September 2003 and dur-
ing April and May 2004. These fish were eutha-
nized with an overdose of anesthetic (ms222)
and preserved in 10% formaldehyde. All fe-
males were dissected to determine the number
of embryos in each developmental stage. Em-
bryos were scored for their developmental stage
using a classification scheme modified from
Reznick (1981). In this study, stages 2, 3, 4, and
5 correspond to Reznick’s (1981) early-eyed,
mid-eyed, late-eyed, and very late-eyed stages,
respectively.
Pre-fertilization offspring provisioning.—We tested
for population differences in pre-fertilization
offspring provisioning by comparing the num-
ber and mean mass of mature ova carried by
females in each population. Mature ova were
removed from each female collected in April
and May 2004 and counted. The entire mass of
a female’s mature ova was then freeze-dried for
24 hours and weighed to the nearest 0.001 mg
on a Cahn C-31 microbalance. For each female,
we calculated her mean ovum mass by dividing
the dry-mass of all of her mature ova by the
number of ova in the brood. For each popula-
tion, we tested for significant correlations be-
tween mean ovum mass and ovum number, fe-
male standard length and mean ovum mass,
and female standard length and ovum number
using Pearson correlations. We then tested for
population differences in mean ovum mass and
ovum number using ANOVA.
Post-fertilization offspring provisioning.—We tested
for population differences in post-fertilization
offspring provisioning by comparing the varia-
tion in embryo mass as a function of develop-
mental stage for both populations. The most de-
veloped brood from each female collected be-
tween September 1999 and August 2003 was re-
tained, freeze-dried for 24 hours, and weighed
to the nearest 0.001 mg on a Cahn C-31 micro-
balance. The average mass of embryos in each
brood was determined by dividing the dry-mass
of the entire brood by the number of embryos
in the brood. Our sampling protocol between
September 1999 and August 2003 left us with a
small sample size for stage 2 embryos. To in-
crease our sample size for early stage embryos,
we retained, dried, and weighed stage 2 embry-
os from additional females collected in April
and May 2004. No other embryos from these
females were included in the analysis of post-
fertilization offspring provisioning. In the fol-
lowing analyses, each female is represented by
a single datum (for females collected prior to
April 2004, the average mass of an embryo in
her most developed brood; for females collect-
ed in April and May 2004, the average mass of
her stage 2 embryos). Consequently, the data
are independent.
We tested whether population differences in
offspring size at birth are due to differences in
matrotrophy using a series of two-way ANOVAs,
with embryo mass as the dependent variable
and population and stage of development as the
factors. Each individual ANOVA examined the
effects of population and stage of development
on embryo mass for two consecutive stages of
development. In these analyses, a significant
population by stage interaction indicates that
the populations differ in the amount of mass
that embryos gain between two consecutive stag-
es of development. Differences between popu-
lations in the amount of mass that embryos gain
between consecutive developmental stages can
be interpreted as population level differences in
the degree or duration of matrotrophy.
We employed this approach instead of an
analysis of covariance (ANCOVA) approach
(e.g., Trexler, 1985; Reznick et al., 2002) for two
reasons. First, the analysis of covariance requires
that the covariate (here developmental stage)
be either a continuous, metric, or meristic var-
iable (Snedecor and Cochran, 1980). Develop-
mental stage as used here is an ordinal variable–
the values (e.g., ‘‘2,’’ ‘‘3’’) are arbitrary desig-
nations and ANCOVA is inappropriate. We
cannot convert developmental stage to ‘‘devel-
opmental age,’’ which would be an appropriate
covariate, because we do not know the relation-
652 COPEIA, 2005, NO. 3
Fig. 1. The relationship between the number of
mature ova in a female and her standard length
(mm) for the Trout Pond population (triangles) and
the Wacissa River population (squares). The dashed
and solid lines indicate the linear least squares re-
gression lines for the Trout Pond and Wacissa River
populations, respectively.
ships between embryo age and developmental
stage. The second reason we chose not to use
ANCOVA is that it requires a linear relationship
between predictor (developmental stage) and
response (embryo mass) that is uniform
throughout the range of both variables and, if
not linear, capable of being made linear
through transformation. Differential matrotro-
phy would be manifested as a change in the re-
lationship between stage of development and
mass at some point in development. Such a
change could create patterns that do not match
the assumptions of ANCOVA. We believe that
our method is more suitable for detecting the
level of variation that might be expected in in-
traspecific comparisons and is more robust
when fewer developmental stages are being rec-
ognized. An additional advantage to using a se-
ries of two-way ANOVAs is that it gives us the
ability to estimate when during development
(i.e., between what developmental stages) dif-
ferences in matrotrophy arise, if they exist. One
drawback of using a series of two-way ANOVAs
to test for interpopulation differences in matro-
trophy is that this approach increases our prob-
ability of making a Type-1 error. This problem
arises because the amount of mass gained be-
tween two consecutive stages of development
(e.g., between stages 2 and 3) is not indepen-
dent of the amount of mass gained between the
next two stages of development (e.g., stages 3
and 4) because the data in the middle stage
(stage 3) are common to both of the two-way
ANOVAs. In order to reduce our probability of
committing a Type-1 error, we used an experi-
mentwise error rate (
a9
) of 0.006, where
a9 5
1–(1–
a
)
1/k
(Sokal and Rohlf, 1995) for k
5
9
significance tests (three two-way ANOVAs with
three significance tests per ANOVA).
R
ESULTS
Pre-fertilization offspring provisioning.—Females
from the different populations differed substan-
tially in the relationship between the number of
mature ova and female body size but differed
only slightly in the average mass of those mature
ova and the relationship of ova mass to female
body size. Trout Pond females carried signifi-
cantly more mature ova than Wacissa River fe-
males (Trout Pond: mean
5
11.94, SE
5
0.94,
n
5
29; Wacissa River: mean
5
4.37, SE
5
0.45,
n
5
30; F
(1, 57)
5
48.64, P
,
0.0001). The num-
ber of ova increased with female standard
length in both populations; however, ovum
number increased more rapidly with female size
in the Trout Pond population (slope
5
0.91, SE
5
0.25) than in Wacissa River population (slope
5
0.42, SE
5
0.10; Fig. 1). This slope difference
was not statistically significant (ANCOVA, pop-
ulation by standard length interaction, F
(1, 53)
5
3.62, P
5
0.062), however we had low power
(0.46) to detect an interaction of this magni-
tude or lower, and we believe that our failure to
achieve significance here is an issue of sample
size. Given that Trout Pond females were larger
on average than Wacissa River females (in this
study and in prior work; Leips and Travis,
1999), it is not surprising that Trout Pond fe-
males had substantially higher average ovum
numbers.
The average mass of a female’s mature ova
was not correlated with her number of mature
ova in either population (Trout Pond, r
5
2
0.067, P
5
0.73, n
5
29; Wacissa River, r
5
0.24, P
5
0.20, n
5
30). In addition, there was
no correlation between the average mass of a
female’s mature ova and her standard length in
either population (Trout Pond: r
5
0.093, P
5
0.63, n
5
29; Wacissa River: r
5
0.28, P
5
0.25,
n
5
30). The ova of Wacissa River females were,
on average, 20% heavier than the ova of Trout
Pond females (Trout Pond: mean
5
0.015 mg,
SE
5
0.0013, n
5
29; Wacissa River: mean
5
0.018 mg, SE
5
0.0011, n
5
30). This difference
was not statistically significant (F
(1, 57)
5
3.30, P
5
0.075), however we had lower power (0.43)
to detect an effect of this magnitude.
Post-fertilization offspring provisioning.—Wacissa
River embryos were significantly heavier than
Trout Pond embryos at all stages of develop-
ment (63% heavier at stage 2, 29% at stage 3,
653SCHRADER AND TRAVIS—MATROTROPHY IN H. FORMOSA
Fig. 2. Mean (
6
SE) dry mass of embr yos (mg),
from the Wacissa River and Trout Pond, in stages 2
through 5. Upper numbers indicate Wacissa River
sample sizes; lower numbers indicate Trout Pond sam-
ple sizes. Lines connecting means are shown to illus-
trate differences between the two populations in the
amount of mass gained between consecutive stages of
development.
T
ABLE
1. T
HE
E
FFECTS OF
P
OPULATION
,S
TAGE OF
D
EVELOPMENT
,
AND THEIR
I
NTERACTION ON THE
M
ASS
(
MG
)
OF
H. formosa E
MBRYOS
. The table shows Fand Pvalues from three separate two-way ANOVAs. For each two-
way ANOVA, the dependent variable was embryo mass and the factors were population (Trout Pond or Wacissa
River) and stage of development where the two levels were consecutive stages of development (e.g., 2, 3).
Asterisks indicate significance at
a9 5
0.006.
Source
Comparison
Stages 2, 3
df FP
Stages 3, 4
df FP
Stages 4, 5
df FP
Population 1 8.52 0.004* 1 50.02
,
0.0001* 1 87.45
,
0.0001*
Stage 1 108.6
,
0.001* 1 306.25
,
0.0001* 1 88.40
,
0.0001*
Population
3
stage 1 1.74 0.189 1 1.22 0.271 1 10.99 0.0010*
22% at stage 4, and 38% at stage 5; Fig. 2, Table
1). The increases in average dry mass between
successive stages were comparable across popu-
lations and stages between stages 2 and 4 (i.e.,
there were no significant population by stage
interactions, Table 1). However, the populations
differed dramatically in the increase in average
dry mass between stages 4 and 5 (Fig. 2). The
increases in dry mass between stages 2 and 3
were about 0.19 mg in both populations and the
increases between stages 3 and 4 were about
0.24 mg in both populations. The increase in
dry mass between stages 4 and 5 was compara-
ble to the preceding increases in Wacissa River
embryos (approximately 0.24 mg), but Trout
Pond embryos exhibited a dramatically smaller
gain in dry mass between stages 4 and 5 (ap-
proximately 0.11 mg). These numbers repre-
sent about a 33% gain in average dry mass be-
tween stages 4 and 5 in Wacissa River embryos
but only a 17% gain in Trout Pond embryos.
D
ISCUSSION
Our results suggest that consistent, genetical-
ly based differences in offspring size between
the Wacissa River and Trout Pond populations
are due to differences in both pre-fertilization
(ovum size) and post-fertilization (matrotro-
phy) provisioning. Although the population dif-
ference in ovum mass was not significant (P
5
0.075), we had low power to detect an effect of
the magnitude suggested by our data. We be-
lieve that the early differences in embryonic size
are best explained as a result of differences in
pre-fertilization offspring provisioning. The sub-
stantial differences between the two populations
in the amount of mass that embryos gained be-
tween stages 4 and 5 can only be explained by
differential matrotrophy, although the mecha-
nism through which this occurs is unclear (see
below).
Our results suggest a mechanism for the pop-
ulation-specific responses of offspring number
and size to population density reported in ear-
lier work. Leips et al. (2000) showed that the
Wacissa River and Trout Pond populations dis-
played different reaction norms to population
density. In that study, higher population density
led to a larger decrease in offspring number in
Trout Pond stock than in Wacissa River stock
and a larger decrease in offspring size in Wac-
issa River stock than Trout Pond stock. We
found that the slope of the relationship be-
tween ovum number and female standard
length was steeper in the Trout Pond popula-
tion than in the Wacissa River population (Fig.
1). This suggests that changes in female size will
have a greater effect on offspring number
(number of mature ova) in Trout Pond females
than in Wacissa River females. Field and meso-
cosm studies have shown repeatedly that female
654 COPEIA, 2005, NO. 3
body size decreases substantially as population
density increases, regardless of population or
genetic stock (Leips and Travis, 1999; Leips et
al., 2000). Together, these results suggest that
the effects of density on female body size will
exert a stronger indirect effect on offspring
number in Trout Pond fish because of their
steeper relationship of offspring number and
body size.
Female size is not likely to mediate the pop-
ulation-specific reaction of average offspring
mass to density because we found no correla-
tion between female size and offspring mass
and prior studies have not found consistent cor-
relations either (Travis et al., 1987; Leips and
Travis, 1999). We suggest that the different lev-
els of matrotrophy in the two populations ex-
plain this result. In obligate matrotrophs such
as H. formosa, embryo mass is directly affected
by the female’s access to food (Reznick et al.,
1996; Rodd and Richardson, unpubl. data). The
differences in matrotrophy that we detected,
along with Leips et al.’s (2000) confirmation of
the genetic basis of offspring size differences,
suggest that Wacissa River females divert more
of their immediate food intake to developing
embryos than Trout Pond females. If this is the
case, then changes in food level will have a
greater impact on the mass of Wacissa River em-
bryos than on the mass of Trout Pond embryos.
This prediction is eminently testable with feed-
ing experiments similar to those of Reznick et
al. (1996).
There are two ways that the interpopulation
differences in matrotrophy that we observed
could arise. First, assuming that gestation time
is the same in both populations, the differences
in matrotrophy may be due to differences in the
rate of maternal investment to late-stage embry-
os. Alternatively, assuming that the rate of ma-
ternal investment to late-stage embryos is the
same in both populations, the differences in
matrotrophy may be due to differences in ges-
tation time. Our data do not allow us to distin-
guish between these two possibilities; however,
transfer rates of labeled tracers (e.g., Marsh-
Matthews et al., 2001) could be used to deter-
mine whether females from Trout Pond and the
Wacissa River differ in the proportion of their
food intake that is passed to their embryos. In
addition, it might be possible to estimate the
entire length of the gestation period using the
otoliths of newborn fish, or late stage embryos
(e.g., Schultz, 1990).
It is possible that the differences in matrotro-
phy that we observed in these field samples are
due to differences between the environments
experienced by mothers; however, it is likely
that the difference in matrotrophy is genetically
based. The mean sizes of stage 5 embryos in this
study are similar to estimates of offspring size at
birth found in prior studies of these populations
(Trout Pond means: 0.48–0.55 mg, Travis, un-
publ. data; Schrader, unpubl. data; Wacissa Riv-
er means: 0.79–1.17 mg, Cheong et al., 1984;
Henrich and Travis, 1988). In addition, the
mean sizes of stage 4 embryos in this study are
consistent with previous studies (Trout Pond
means: 0.43–0.53 mg, Wacissa River means:
0.58–0.67 mg, Leips and Travis, 1999; Leips et
al., 2000) and are known to have a genetic basis
(Leips et al., 2000). Regardless of whether or
not the different patterns of offspring provision-
ing have a genetic basis, it is interesting that the
differences in matrotrophy are only present in
the late stages of development. This suggests
that offspring provisioning in the later stages of
development is either more plastic or more evo-
lutionarily labile than offspring provisioning
during the early stages of development. The
higher lability of size later in development con-
forms to predictions offered by Leips and Travis
(1994) in their verbal model for resource allo-
cation during amphibian development. This
model suggests that most resources during the
early stages of development are allocated to de-
velopmental processes, with a gradual increase
in the allocation to growth as development ad-
vances and the formation of key structures is
completed. Such an allocation pattern creates
apparently greater plasticity in offspring size in
later stages of development (e.g., Twombley,
1996).
Another possible explanation for the appear-
ance of differential matrotrophy late in devel-
opment is a decline in the importance of ma-
ternal effects and an increase in the importance
of offspring genome expression. Maternal ef-
fects on offspring traits are generally thought to
be strongest early in development (Bernardo,
1996a), and studies of lecithotrophic fish (e.g.,
Chinook Salmon, Heath et al., 1999; Brook
Charr, Perry et al., 2004) have shown that the
importance of maternal effects on offspring size
declines throughout development. These stud-
ies also suggest that major transitions during de-
velopment (e.g., the resorption of the yolk sac
and the initiation of exogenous feeding) mark
major changes in the genetic architecture of off-
spring size. It is possible that the differences in
matrotrophy that we observed late in develop-
ment are due to the waning of maternal effects
and an increase in offspring gene expression,
and that there may be a developmental transi-
tion around the late-eyed stage (i.e., stage 4)
that marks this change in embryonic growth.
655SCHRADER AND TRAVIS—MATROTROPHY IN H. FORMOSA
Maternal effects are thought to be extremely
important in viviparous, matrotrophic species
because of the nature and duration of the ma-
ternal-embryo relationship (Bernardo, 1996b).
However, it is possible that paternal alleles allow
offspring to exploit mothers more or less effi-
ciently, thereby increasing the potential for off-
spring to control their prenatal growth rate.
The importance of the maternal and offspring
genome to prenatal growth rates in H. formosa
is currently unknown. However, it may be pos-
sible to estimate relative importance of mater-
nal and paternal alleles to prenatal growth us-
ing interpopulation crosses between the Wacissa
River and Trout Pond populations.
A
CKNOWLEDGMENTS
We thank M. Aresco, M. Gunzburger, and ev-
eryone else who helped in the field and the lab.
J. Hereford, K. McGhee, M. Pires, D. Reznick,
and A. Winn provided helpful comments on a
previous version of this manuscript. This work
was performed under Florida State University’s
ACUC Protocol 9321. Fish were collected under
USFWS permit number 0007. This study was
supported by a National Science Foundation
grant to JT (DEB 99–03925).
L
ITERATURE
C
ITED
A
RIAS
,A.L.,
AND
D. R
EZNICK
. 2000. Life history of
Phalloceros caudiomaculatus: a novel variation on the
theme of livebearing in the family Poeciliidae.
Copeia 2000:792–798.
B
AER
, C. F. 1998. Species-wide population structure in
a southeastern US freshwater fish, Heterandria for-
mosa: gene flow and biogeography. Evolution 52:
183–193.
B
ERNARDO
, J. 1996a. The particular maternal effect of
propagule size, especially egg size: patterns, mod-
els, quality of evidence, and interpretations. Am.
Zool. 36:216–236.
———. 1996b. Maternal effects in animal ecology.
Ibid. 36:83–105.
B
URLEY
, N. 1986. Sexual selection for aesthetic traits
in species with biparental care. Am. Nat. 127:415–
445.
———. 1988. The differential allocation hypothesis:
an experimental test. Ibid. 132:611–628.
C
HEONG
, R. T., S. H
ENRICH
,J.A.F
ARR
,
AND
J. T
RAVIS
.
1984. Variation in fecundity and its relationship to
body size in a population of the least killifish, Het-
erandria formosa (Pisces, Poeciliidae). Copeia 1984:
720–726.
C
LUTTON
-B
ROCK
, T. H. 1991. The Evolution of Paren-
tal Care. Princeton University Press, Princeton, New
Jersey.
G
ROVE
, B. D.,
AND
J. P. W
OURMS
. 1991. Follicular pla-
centa of the viviparous fish, Heterandria formosa 1.
Ultrastructure and development of the embryonic
absorptive surface. J. Morphol. 209:265–284.
———,
AND
———. 1994. Follicular placenta of the
viviparous fish, Heterandria formosa 2. Ultrastructure
and development of the follicular epithelium. Ibid.
220:167–184.
H
AIG
, D. 1993. Genetic conflicts in human pregnancy.
Q. Rev. Biol. 68:495–532.
H
EATH
, D. D., C. W. F
OX
,
AND
J. W. H
EATH
. 1999.
Maternal effects on offspring size: variation
throughout early development of Chinook salmon.
Evolution 53:1605–1611.
H
ENRICH
, S.,
AND
J. T
RAVIS
. 1988. Genetic variation in
reproductive traits in a population of Heterandria
formosa (Pisces, Poeciliiade). J. Evol. Biol. 1:275–
280.
K
OTIAHO
, J. S., L. W. S
IMMONS
,J.H
UNT
,
AND
J. L. T
OM
-
KINS
. 2003. Males influence maternal effects that
promote sexual selection: a quantitative genetic ex-
periment with dung beetles Onthophagus taurus.
Am. Nat. 161:852–859.
L
OMBARDI
, J. 1996. Postzygotic maternal influences
and the maternal-embryonic relationship of the vi-
viparous fishes. Am. Zool. 36:106–115.
L
EIPS
, J.,
AND
J. T
RAVIS
. 1994. Metamorphic responses
to changing food levels in two species of hylid frogs.
Ecology 75:1345–1356.
———,
AND
———. 1999. The comparative expres-
sion of life history traits and its relationship to the
numerical dynamics of four populations of the least
killifish, Heterandria formosa. J. Anim. Ecol. 68:595–
616.
———, ———,
AND
F. H. R
ODD
. 2000. Genetic influ-
ences on experimental population dynamics of the
least killifish. Ecol. Monogr. 70:289–309.
M
ARSH
-M
ATTHEWS
, E., P. S
KIERKOWSKI
,
AND
A. D
E
-
M
ARAIS
. 2001. Direct evidence for mother-to-em-
bryo transfer of nutrients in the livebearing fish
Gambusia geiseri. Copeia 2001:1–6.
M
EFFE
,G.K.,
AND
R. C. V
RIJENHOEK
. 1981. Starvation
stress and intra-ovarian cannibalism in live-bearers
(Antheriniformes, Poeciliidae). Ibid. 1981:702–705.
M
OCK
,D.W.,
AND
G. A. P
ARKER
. 1997. The Evolution
of Sibling Rivalry. Oxford University Press, Oxford.
P
ERRY
, G. M. L., C. A
UDET
,B.L
APLATTE
,
AND
L. B
ER
-
NATCHEZ
. 2004. Shifting patterns in genetic control
at the embryo-alevin boundary in Brook Charr.
Evolution 58:2002–2012.
R
EZNICK
, D. 1981. ‘‘Grandfather effects:’’ the genetics
of inter-population differences in offspring size in
the mosquito fish. Ibid. 35:941–953.
———,
AND
D. B. M
ILES
. 1989. A review of life history
patterns in Poeciliid fishes, p. 125–148. In: Ecology
and Evolution of Livebearing Fishes (Poeciliidae).
G. K. Meffe and F. F. Snelson, Jr. (eds.). Prentice-
Hall, Englewood Cliffs, New Jersey.
———, H. C
ALLAHAN
,
AND
R. L
LAUREDO
. 1996. Ma-
ternal effects on offspring quality in poeciliid fish-
es. Am. Zool. 36:147–156.
———, M. M
ATEOS
,
AND
M. S. S
PRINGER
. 2002. Inde-
pendent origins and rapid evolution of the placen-
ta in the fish genus Poeciliopsis. Science 298:1018–
1020.
656 COPEIA, 2005, NO. 3
R
OFF
, D. A. 2002. Life History Evolution. Sinauer As-
sociates, Inc., Sunderland, Massachusetts.
S
CHULTZ
, E. 1990. Daily otolith increments and the
early life history of a viviparous fish, Micrometrus
minimus. Copeia 1990:59–67.
S
CRIMSHAW
, N. S. 1944. Superfetation in poeciliid fish-
es. Ibid. 1944:180–183.
———. 1945. Embryonic development in poeciliid
fishes. Bio. Bull. 88:233–246.
S
HINE
, R. 1995. A new hypothesis for the evolution of
viviparity in reptiles. Am. Nat. 145:798–823.
S
NEDECOR
,G.W.,
AND
W. G. C
OCHRAN
. 1980. Statisti-
cal Methods, 7
th
edition. Iowa State University Press,
Ames, Iowa.
S
OKAL
,R.R.,
AND
F. J. R
OHLF
. 1995. Biometry, 3
rd
edi-
tion. W. H. Freeman and Company, New York.
S
OUCY
, S.,
AND
J. T
RAVIS
. 2003. Multiple paternity and
population genetic structure in natural populations
of the poeciliid fish, Heterandria formosa. J. Evol.
Biol. 16:1328–1336.
T
HIBAULT
, R. E.,
AND
R. J. S
CHULTZ
. 1978. Reproduc-
tive adaptations among viviparous fishes (Cyprino-
dontiformes: Poeciliidae). Evolution 32:320–333.
T
RAVIS
, J., J. A. F
ARR
,S.H
ENRICH
,
AND
R. T. C
HEONG
.
1987. Testing theories of clutch overlap with the
reproductive ecology of Heterandria formosa. Ecology
68:611–623.
T
REXLER
, J. C. 1985. Variation in the degree of vivi-
parity in the sailfin molly Poeciliia latipina. Copeia
4:999–1004.
———. 1997. Resource availability and plasticity in
offspring provisioning: embr yo nourishment in sail-
fin mollies. Ecology 78:1370–1381.
———,
AND
D. L. D
E
A
NGELIS
. 2003. Resource alloca-
tion in offspring provisioning: an evaluation of the
conditions favoring the evolution of matrotrophy.
Am. Nat. 162:574–585.
———, R. C. T
EMPE
,
AND
J. T
RAVIS
. 1994. Size selective
predation on sailfin mollies by 2 species of herons.
Oikos 69:250–258.
T
URNER
, C. L. 1937. Reproductive cycles and super-
fetation in poeciliid fishes. Biol. Bull. 72:145–164.
T
WOMBLEY
, S. 1996. Timing of metamorphosis in a
freshwater crustacean: a comparison with anuran
models. Ecology 77:1855–1866.
F
LORIDA
S
TATE
U
NIVERSITY
,D
EPARTMENT OF
B
I
-
OLOGICAL
S
CIENCE
,T
ALLAHASSEE
,F
LORIDA
32306–1100. E-mail: schrader@bio.fsu.edu.
Send reprint requests to MS. Submitted: 18
Aug. 2004. Accepted: 24 Jan. 2005. Section
editor: C. M. Taylor.