Variation among populations of Diodia teres (Rubiaceae) in environmental maternal effects


J. Hereford, Department of Biological Science, Florida State University, Tallahassee, FL, USA 32306-1100.
Tel.: 850 644 9822; fax: 850 644 9829; e-mail:


Previous studies have quantified variation in environmental maternal effects (EME) within populations, but these effects could differ among populations as well. In this study we grew clonal replicates of individuals from three populations of the annual plant Diodia teres in their native and non-native environments. Our goal was to estimate the effects of maternal environment and maternal population on seed and seedling traits. Seeds that were produced in this field study were then planted in two soil types to quantify effects of the offspring environment on seedling traits. There was substantial variation among populations for seed weight. We found population variation for EME, and maternal environment by offspring environment interactions. We conclude that variation among populations in EME may be an unrecognized component of local adaptation, and that attempts to control maternal effects by statistically accounting for variation in seed weight may be ineffective.


Parental environment effects can have a profound influence on offspring phenotype (Kirkpatrick & Lande, 1989). These effects operate beyond the influence of Medelian inheritance and include the effects of the parental environment as well as genetic control of parental environment on offspring phenotype (Lynch & Walsh, 1998). Two patterns that have emerged from studies of parental effects are that environmental maternal effects (EME) are stronger than paternal environment effects (Mazer, 1987; Mazer & Wolfe, 1992; Platenkamp & Shaw, 1993; Montalvo & Shaw, 1994; Fox et al., 1995; Byers et al., 1997), and that EME are stronger than genetic maternal effects (Platenkamp & Shaw, 1993; Schmid & Dolt, 1994; Byers et al., 1997; Hunt & Simmons, 2002). Therefore, the parental effects most likely to have the strongest effect on offspring phenotype are EME.

Environmental maternal effects are a form of across generation phenotypic plasticity in that offspring trait values depend on the maternal environment (Bernardo, 1996; Mousseau & Fox, 1998). These effects have been shown to strongly influence offspring phenotype, which is a major determinant of offspring fitness (Endler, 1986; Kingsolver et al., 2001). For example, EME have a strong influence on seed (Roach & Wulff, 1987; Lacey, 1998) and egg (Bernardo, 1996; Rossiter, 1996) size, both of which have been shown to be positively correlated with offspring fitness components (Schaal, 1980; Dolan, 1984; Stratton, 1989; Rossiter, 1991; Azevedo et al., 1997). Additionally, maternal effects may cause short-term maladaptive responses to selection or could possibly slow the response to selection (Falconer, 1989; Kirkpatrick & Lande, 1989).

Given the role of offspring in dispersal and gene flow, it is surprising that the evolutionary consequences of EME have been considered primarily within populations (Roach & Wulff, 1987; Bernardo, 1996; Lacey, 1998). Dispersal of seeds (Latta & Mitton, 1999) or deposition of eggs is a component of gene flow and colonization, which largely determine the extent of population differentiation. Environmental maternal effects could influence the success of colonization or gene flow if the maternal environment influences offspring post dispersal success. Variation among populations in EME may be fairly common given that genetic variation among populations in other types of phenotypic plasticity is common (Winn & Evans, 1991; Donohue & Schmitt, 1999; Weinig, 2000). This type of variation could lead to increased population divergence and possibly speciation (Wolfe & Brodie, 1998).

Population differentiation for EME has been shown in vertebrates (Badyaev et al., 2002) and invertebrates (Mousseau, 1991; Mousseau & Dingle, 1991), but there is no evidence in plants. Badyaev et al. (2002) showed that divergence in maternal effects on hatching order has helped to promote adaptive divergence of populations of house finches. Maternal effects could promote divergence in plants in a similar fashion. Plants often show high levels of genetic differentiation among populations (Loveless & Hamrick, 1984), and express large maternal effects on offspring phenotype (Roach & Wulff, 1987). Therefore variation among populations in EME in plants is plausible.

In addition to genetic differentiation among populations in EME, there could be purely environmental interactions between maternal and offspring environments on offspring phenotype. For example, offspring environment may influence consequences of the effects of the maternal environment. This type of environment-by-environment interaction has rarely been studied (but see Rossiter, 1998). The likelihood of successful dispersal from a given environment may depend not only on the maternal environment, but also on the environment into which the offspring disperse.

The primary mechanism for EME in plants is thought to operate through the weight of seeds (Roach & Wulff, 1987). Endosperm from seeds provide nourishment to seedlings and the seed coat can control timing of germination. Maternal effects may also influence offspring phenotype through seed quality, such as concentration of nutrients provided in the endosperm (Baskin & Baskin, 1998). Therefore, there may be maternal effects that occur through variation in seed qualities other than seed weight. Given that seed weight is often used as a covariate in statistical analyses to remove the influence of EME, it is important to understand maternal effects that operate by mechanisms other than seed weight.

Here we report a test for population differentiation for EME in an annual plant. We address three questions, is there population differentiation for EME on seed and seedling traits, are there maternal environment by offspring environment interactions for seedling traits, and do these effects occur when the influence of seed weight is removed. We grew clonal replicates from three populations in three field environments to examine the effects of population of origin and maternal environment on seed weight. Additionally, we planted a subset of the offspring of this generation to quantify the interaction between the effects of offspring and maternal environments on offspring traits. Finally we quantify maternal effects beyond the influence of seed weight by performing statistical analyses with seed weight as a covariate. Maternal provisioning includes not only the weight of seeds produced by the maternal plant, but also other qualities of those seeds, therefore we were interested in these effects beyond the influence of seed weight.

Materials and methods

Diodia teres (Rubiaceae) is a self-compatible annual plant, distributed from Panama to the North-eastern United States, west to Michigan, USA (Kearney & Peebles, 1964). In northern Florida, D. teres is found in contrasting habitats including coastal sand dunes with sandy soils, little herbaceous cover and no tree canopy, in inland sites characterized by clay soils, more herbaceous growth and a partial tree canopy, and in sandhills habitats with sandy soils and more herbaceous growth than dunes sites (J. Hereford, personal observation). Jordan (1992) demonstrated local adaptation in D. teres to similar habitats in North Carolina.

Experimental procedures

In December of 1998, seeds from 20 to 21 individuals of D. teres were collected from one source population in each of the three habitat types described above. The dunes population seeds were collected at Saint Joseph Bay State Park in Bay County, Florida. The sandhills population seeds were collected at Dog Lake in Leon County, Florida. The seeds from the inland population were collected at Pebble Hill Plantation in Thomas County, Georgia. All seeds were germinated and one individual from each original maternal plant was grown to adult size in an environmentally controlled growth chamber under a15/9 h. light/dark cycle and a 30 °/20 ° temperature cycle.

To determine the effects of maternal environment on offspring phenotype, we planted clonal replicates of the plants grown in the growth chamber into each of three maternal environments in the field. Maternal genotypes were replicated by making cuttings from plants grown in the growth chamber. The cuttings were 5–8 cm in length and always included at least two internodes. To aid in the establishment of new roots, each cutting was treated with a commercial rooting agent. Cuttings were planted in potting soil in 72-well flats and allowed to take root in the greenhouse. Approximately 95% of the cuttings established roots within 14 days. After making sure that the cuttings had taken root and had begun to grow new leaves, cuttings were planted in three field sites. They were planted at the sites of two of the populations from which maternal genotypes were collected, Pebble Hill Plantation and Dog Lake. Cuttings could not be planted at St Joseph Bay State Park because there was limited space where human disturbance could be avoided. Consequently, they were planted at a different dunes site, St George Island State Park, Franklin County, Florida, which is similar to St Joseph Bay in soil texture, and D. teres density (J. Hereford, unpublished data). A total of 718 cuttings were planted with an average of three replicates from each genotype at each planting site in May of 1999.

In December 1999 when most transplants had senesced, we collected and weighed a sample of the seeds produced by each transplant. Eight randomly selected seeds produced by each cutting were individually weighed to the nearest 0.001 mg using a Cahn microbalance.

The transplants at all sites were open pollinated, and it is possible that paternal effects also influenced seed and seedling traits. We consider this to be unimportant for two reasons. First, EME are typically much more important than paternal genetic effects for early life-cycle traits such as seed weight and germination (Platenkamp & Shaw 1993; Byers et al., 1997). Second, D. teres is likely to be highly selfing because it produces inconspicuous (∼3 mm in diameter), self-compatible flowers. Therefore, most seeds were likely selfed progeny with no influence of a foreign paternal genotype.

To test for maternal effects on seedling traits and the interaction between maternal environment and offspring environment, we planted seeds collected from field-grown cuttings into two offspring environments. The eight seeds collected from each cutting were divided, with four seeds placed in each of two soil types. The soil types were a sandy soil collected from St Joseph Bay and clay soil collected from Pebble Hill Plantation. In some instances we were not able to collect eight seeds from a plant, then we collected as many seeds as possible and divided them evenly between soil types. Both soil types were collected in areas where D. teres was present, and the soil was sieved to remove wild D. teres seeds. We planted the seeds into 72 well flats, with each well having a volume of 80 mL. Flats were watered daily so that the surface-soil layer was constantly moist.

We measured whether or not each seed germinated, time to germination, and seedling biomass after 8 weeks. Because germination occurred frequently at the start of the experiment, germination was recorded twice a day and time to germination was measured in units of half days. After 8 weeks, germination had ceased for three consecutive days, and we harvested all the seedlings. Seedlings were oven dried, and the belowground and aboveground tissues of all plants that germinated were weighed to the nearest 0.01 mg. Seedlings that died before the end of the experiment were collected and then dried and weighed along with those that had survived.

Data analysis

Fixed effect analysis of variance (anova) was used to quantify the effects of maternal environment, population, and their interaction on seed weight. Population and maternal environment effects on seed weight describe the influence of source population and maternal planting site, respectively. Another set of analyses was performed on germination, time to germination, and seedling biomass to quantify the influence of the maternal and offspring environments. The interpretation of the effects of this set of analyses is the same as the analyses on seed weight except there is the additional factor of offspring environment. Differentiation among populations in the expression of EME is indicated by significant population by maternal environment interactions. Effects of maternal environment on offspring environments are indicated by significant maternal environment by offspring environment interaction. To quantify maternal effects beyond the influence of seed weight, we used seed weight as a covariate in analyses of time to germination and seedling biomass in ancova. Seedling biomass was natural log transformed to meet assumptions of normality and homoscedasticity. We did not include plants that failed to germinate in analyses of time to germination and seedling biomass.

Germination was treated as a categorical variable, with two levels, germinated or not germinated, and was analysed using a generalized linear model (McCullagh & Nelder, 1989). This procedure used maximum likelihood to estimate effects of seed weight, population, maternal environment, offspring environment, and all interaction terms on germination. We used a logit link function (Log odds of germination) and a binomial distribution function in the model.

Statistical analyses were carried out using the statistical package SAS, version 8 (SAS Institute, 1999). The procedure GLM was used to perform all anovas, and ancovas. The procedure genmod was used to analyse generalized linear models.


We detected large differences in seed weight among maternal environments and populations (Table 1). The dunes population produced the largest seeds, and the sandhills population produced the smallest (Fig. 1), indicating genetic differences among populations for offspring traits. Mean seed weight was much larger in plants from the dunes population (4.28 mg ± 0.05 SE) than the sandhills and inland populations (2.81 mg ± 0.05 SE, 3.54 mg ± 0.04 SE respectively). Maternal environments differed significantly as well, indicating significant EME, with the largest seeds being produced by cuttings planted in the sandhills site (3.83 mg ± 0.04 SE) and the smallest produced by transplants planted in the dunes site (3.55 mg ± 0.05 SE). The magnitude of the effect of maternal environment was smaller than the population effect (Table 1).

Table 1. anova of the effects of planting site, source population, and their interaction on seed weight.
Source of variationd.f.MSF-ratioP-value
Maternal environment248.2923.39<0.0001
Maternal environment × population427.6013.27<0.0001
Figure 1.

Interaction between source population and planting site on seed weight. The dunes population is indicated by the triangles, the inland population is indicated by the solid circles, and the sandhills population by the open circles.

The significant maternal environment by population interaction confirms that there is population differentiation for EME on seed weight (Table 1). Cuttings from the dunes population produced their largest seeds at the dunes site, and the inland and sandhills populations produced their smallest seeds at that site (Fig. 1).

The experiment-wide proportion of seeds that germinated was 0.21 and did not differ between offspring environments (Table 2). A subset of ungerminated seeds were judged viable by checking for a moist endosperm, all seeds checked were judged to be viable. Larger seeds had a significantly higher probability of germinating than smaller seeds (logistic regression β = 0.14, d.f. = 1, P < 0.0001), but other factors influenced probability of germination as well. After accounting for variation in seed weight, seeds produced in different maternal environments and by different populations varied in probability of germination (Table 2). The significant maternal environment by population interaction shows that populations differed in how maternal environment affected probability of germination. Seeds produced by cuttings from the inland population maintained a constant, high germination rate regardless of the maternal environment, but the sandhills and dunes populations’ germination rates depended on the planting site of their mothers (Fig. 2).

Table 2.  Effects of seed weight, maternal environment, population, offspring environment, and interactions on germination.
Source of variationd.f.χ2P-value
  1. Results are from a generalized linear model of these effects on germination. Germination was treated as a categorical variable with two levels, germinated or not germinated.

Seed weight141.05<0.0001
Maternal environment25.600.0609
Offspring environment10.780.3764
Maternal environment × population420.12<0.0001
Maternal environment × offspring environment217.310.0002
Population × offspring environment218.46<0.0001
Maternal environment × population × offspring environment43.870.4243
Figure 2.

Interaction between source population and planting site on seed germination probability. The dunes population is indicated by the triangles, the inland population is indicated by the solid circles, and open circles show the germination of the sandhills population.

The maternal environment also influenced how the offspring environment affected probability of germination, as reflected in the significant maternal environment by offspring environment interaction (Table 2). Germination probability was similar in the clay offspring environment regardless of maternal environment. In the sand offspring environment, seeds from mothers grown in the sandhills site had significantly greater germination probability than seeds from mothers grown in inland and dunes sites (Fig. 3).

Figure 3.

Interaction of planting site and soil treatment on germination probability. Seeds from the dunes site are indicated by the triangles, the inland site by the solid circles, and seeds from the sandhills site by the open circles.

Maternal environment and population influenced time to germination as well (Table 3). Large seed weight was associated with earlier germination (linear regression b = −0.86, d.f. = 1, P < 0.01). There was no significant effect of offspring environment on time to germination and no significant interactions between main effects (Table 3). Plants from the sandhills population germinated earliest (8.7 days ± 0.61 SE), the dunes population was intermediate (10.2 days ± 0.59 SE), and the inland plants had the longest mean time to germination (12.4 days ± 0.55 SE). In the overall statistical model, and in subsequent analyses with seed weight and time to germination as covariates, all assumptions of equality of slopes were met.

Table 3.  Analysis of covariance on time to germination.
Source of variationd.f.MSF-ratioP-value
  1. Seed weight was used as the covariate.

Seed weight1693.869.100.0027
Maternal environment2249.823.240.0384
Offspring environment12.820.040.8475
Maternal environment × population4138.621.820.1236
Maternal environment × offspring environment274.020.970.3793
Population × offspring environment250.000.660.5194
Maternal environment × population × offspring environment464.560.850.4958

Maternal environment had no effect on seedling biomass, but offspring environment did have a significant effect (Table 4). The biomass of plants grown in clay soil was larger (0.14 g ± 0.003 SE, 0.06 g ± 0.003 SE). This result contrasts with the results for germination probability and time to germination, where maternal environment had stronger effects and the offspring environment was unimportant. Seed weight and time to germination also influenced seedling biomass. Larger seeds grew into larger plants (linear regression b = 0.15, d.f. = 1, P < 0.0001), and plants that germinated earlier were larger (linear regression b = 0.05, d.f. = 1, P < 0.0001).

Table 4.  Analysis of covariance of seedling biomass.
Source of variationd.f.MSF-ratioP-value
  1. Seed weight and time to germination were used as covariates. Biomass was natural log transformed prior to analysis.

Seed weight15.619.220.0025
Time to germination1107.24176.33<0.0001
Maternal environment20.270.440.6418
Offspring environment1164.96271.25<0.0001
Maternal environment × population40.460.760.5532
Maternal environment × offspring environment20.250.430.6539
Population × offspring environment20.831.370.2560
Maternal environment × population × offspring environment40.300.490.7409

Not including seed weight or time to germination as covariates led to similar qualitative results to analyses that included these covariates. The only exception was a significant effect of population on time to germination and seedling biomass. Consequently we do not present these results.


We found significant population variation for EME on seed weight and germination indicated by the maternal environment by population interaction (Tables 1 and 2). We observed large differences among populations in the weight of seeds produced and in rates of seed germination. Maternal environment and population effects were not significant for time to germination and seedling biomass. Finally, maternal effects that influenced seedling traits acted beyond the influence of seed weight.

Seed weight and germination have been shown to have substantial effects on offspring fitness (Stratton, 1989; Kelly, 1992). Therefore, population differentiation for EME could have a substantial effect on divergence or local adaptation among populations. Interactions between population and environment for EME have been found in other studies (Mousseau, 1991; Mousseau & Dingle, 1991; Badyaev et al., 2002), though not in all (Schmitt et al., 1992). Studies that have found these interactions have shown them in populations that occur in different environments and are separated by large geographical distances. Schmitt et al. (1992) studied populations that occurred at the same latitude and the environmental difference was because of human disturbance. Thus, there may not have been much time for populations to have diverged. Given that our study also found population level variation for EME, this may be a common feature of divergent populations. Not only do they differ in phenotypes, but also in maternal effects on offspring phenotype.

The interaction between maternal environment and population on seed weight shows population differentiation for EME. The inland and sandhills populations produced their smallest seeds at the dunes maternal environment, but the dunes population produced its largest seeds at that site (Fig. 1). The pattern of this interaction suggests that the differentiation of the dunes population could be adaptive. Larger seeds have been shown to be positively associated with components of offspring fitness (Schaal, 1980; Dolan, 1984; Stratton, 1989). The differences in effects of maternal environment on seed weight suggest that the sandhills and inland populations are maladapted to the dunes environment, and that the dunes plants are adapted to the environment. Other traits influenced by EME have been shown to be adaptive (Kalisz, 1986; Winn, 1988; Fox et al., 1997). The divergence in seed weight could also represent an unmeasured component of local adaptation. Assuming the dunes population is locally adapted, foreign individuals invading the dunes population would not only produce fewer seeds than native individuals, but they would also produce smaller seeds. Thus these individuals produce both fewer and poorer quality offspring. The importance of this aspect of local adaptation depends on the pattern of selection on seed weight in the field (e.g. Kalisz, 1986; Winn, 1988; Gomez, 2004).

The maternal environment also affected the influence of the offspring environment on germination, as indicated by the maternal environment by offspring environment interaction (Table 2). This effect has been noted for germination in another species (Schmitt et al., 1992), and may be common on other traits in plants as well as animals (Rossiter, 1998). Because we planted clonal replicates at all sites, our study can differentiate the purely environmental interaction between maternal and offspring environments from the three-way interaction that signifies genetic variation within populations for the maternal by offspring environment interaction.

One important difference between our study and similar studies that have found EME, and maternal environment by offspring environment interactions for germination (e.g. Schmitt et al., 1992) is that we show that these effects can occur beyond the influence of maternal effects on seed weight. Seed weight was used as a covariate in analyses of germination, and as a result linear effects of seed weight were accounted for in the analyses. Therefore there must be other aspects of seeds, such as endosperm quality or variation in the composition of the seed coat, that affect rates of germination in addition to the influence of seed weight. These results show that the influence of maternal effects on the results of experiments may not be easily eliminated by using seed weight as a covariate in data analysis.

Environmental maternal effects had no influence on time to germination or seedling biomass. There were no significant effects of maternal environment or interactions between maternal environment and population on time to germination or seedling biomass, but this could be the result of the relatively benign environment provided in the greenhouse. Weller (1985) showed that plants grown with supplemental water showed effects of seed weight on emergence time and seedling biomass, but not on survival. Thus providing supplemental water may not eliminate EME on some traits, but could decrease the influence that EME have on fitness components. Additionally, EME have been shown to be more important in the presence of competition than without competition (Stratton, 1989; Miao et al., 1991). We provided seedlings with a constant source of moisture and eliminated competition, which could have prevented maternal effects from affecting seedling traits.

The lack of maternal effects on seedling biomass appear to indicate that maternal effects on offspring fitness are not important, but this conclusion is unlikely. The trait most likely to be correlated with fitness in our design is germination, because germination was under less influence of the benign offspring environment than any other trait besides seed weight. Additionally, germination is an important trait in annuals with seed dormancy because there is a fitness cost associated with delaying germination (Cohen, 1968).

This study is a first step in determining the extent of larger scale variation in EME than within population variation. Further work will have to determine the extent that variation in traits such as seed weight and germination probability affect not only establishment of seedlings in foreign environments but also on rates of colonization and gene flow. Estimating these effects may require a combination of ecological genetic and demographic methods, such as methods that combine phenotypic selection estimates with analyses of population growth (e.g. van Tienderen, 2000; Coulson et al., 2003). These types of studies will show whether variation in maternal effects could affect macroevolutionary processes or is a source of statistical error that only slows microevolution.


We thank A.A. Winn for direction and helpful discussions of the experimental design and manuscript. R.C. Fuller also provided helpful discussion of the experimental design. We thank J.H. Burns for help in the greenhouse and two anonymous reviewers for helpful comments on the manuscript. M.S. Schrader provided helpful discussion of maternal effects. S. Herman and C. Martin of Tall Timbers research station provided valuable logistical support. The Florida Department of Environmental Protection and the U.S. Forest Service provided access to St George Island and Dog Lake, respectively. This work was supported by NSF grant DEB-9903878 awarded to A.A. Winn.