Separating parental environment from seed size effects on next generation growth and development in Arabidopsis


T. L. Durham Brooks. e-mail:


Plant growth and development is profoundly influenced by environmental conditions that laboratory experimentation typically attempts to control. However, growth conditions are not uniform between or even within laboratories and the extent to which these differences influence plant growth and development is unknown. Experiments with wild-type Arabidopsis thaliana were designed to quantify the influences of parental environment and seed size on growth and development in the next generation. A single lot of seed was planted in six environmental chambers and grown to maturity. The seed produced was mechanically sieved into small and large size classes then grown in a common environment and subjected to a set of assays spanning the life cycle. Analysis of variance demonstrated that seed size effects were particularly significant early in development, affecting primary root growth and gravitropism, but also flowering time. Parental environment affected progeny germination time, flowering and weight of seed the progeny produced. In some cases, the parental environment affected the magnitude of (interacted with) the observed seed size effects. These data indicate that life history circumstances of the parental generation can affect growth and development throughout the life cycle of the next generation to an extent that should be considered when performing genetic studies.


A major goal in plant biology research is elucidation of gene function and characterization of phenotypes due to genetic variation. Despite Arabidopsis thaliana being a popular model plant for such studies, the functions of most of its genes are still undetermined in part because most mutants have no associated phenotype (Bouché & Bouchez 2001; Alonso & Ecker 2006). There are many potential explanations for why genetic variation, either natural or induced by mutation, does not produce a readily detectable or even a subtle phenotype. One of them, the subject of this report, is that variability in function of the reference genotype (the wild type in mutant studies) masks the effect of a genetic alteration. From some perspectives, this variability is viewed as noise to be suppressed, but from others, it is a source of information about how development responds appropriately to changes in environmental input, or phenotypic plasticity (Sultan 2005; Tonsor, Alonso-Blanco & Koornneef 2005). When the goal of a study is to detect a difference in developmental response due to a particular difference in genetic sequence, the potential masking effect of response variability is typically managed by controlling environmental inputs as well as possible. But parameters other than those that are commonly controlled (e.g. light, temperature, and mineral content) could be just as confounding to genetic studies.

One source of variability that may not be obvious is parental environment. Two genetically identical plants derived from seed produced in different conditions may grow, develop and respond to external stimuli differently because influences of the parental environment were carried over through the seed to the next generation (Roach & Wulff 1987). Parental environment can affect the early stages of the next generation, such as seed size and primary root growth in Arabidopsis (Andalo et al. 1998); it can affect seedling height, cotyledon area, leaf area and weight of seed produced in velvet leaf (Wulff & Bazzaz 1992) and its effect on Scots pine tree height can persist for at least 6 years after germination (Lindgren & Wei 1994). Even herbivory experience in the parental generation can be passed on to influence defence responses of the next generation (Agrawal, Laforsch & Tollrian 1999).

Another and potentially interrelated source of variability is the size of the seed. A single genetically homozygous Arabidopsis plant produces a wide distribution of seed sizes. It should not be assumed that all of this seed, while genetically identical, will be at equivalent starting points. Indeed, seed size has been found to significantly affect multiple stages of growth and development in other plants such as the two oak species Quercus rogosa and Quercus laurina (Bonfil 1998), wild radish (Stanton 1984, 1985) and the clover Trifolium subterraneum (Black 1956). In all of these studies, seed size positively correlated with growth and seedling establishment. In general, species exposed to more stressful environments such as shade or drought preferentially produce larger seed (Jurado & Westoby 1992; Leishman & Westoby 1994). Production of fewer but more advantaged offspring appears to be the strategy at least some plants employ when faced with stressful environments.

In studies of A. thaliana aiming to accurately describe the effects of genes on organismal growth, parental environment and seed size could potentially confound a genetic effect. A central question addressed in this study is how seed size, independent of the parental environment, and the parental environment, independent of seed size, can affect seedling growth and development in the next and subsequent generations. This study contributes to the recasting of the phenotype as a process that occurs over developmental time within a dynamic environmental context and moves away from the conception of the phenotype as an end point as called for by Sultan (2004). The main findings suggest a way in which dynamic environmentally contingent variation can be used as information in a reverse-genetic mutant characterization. This study led to more detailed morphometric characterization of the plasticity of root gravitropism (Durham Brooks, Miller & Spalding 2010), which further led to the characterization of a subtle single-gene effect on root gravitropism (Miller et al. 2010). This study along with the supporting studies provide a case in which conceptualizing response variability as plasticity enabled detection of a single-gene effect that would have otherwise been lost in the ‘noise’.


Experimental overview

A pair of experiments was performed to investigate the influence of parental environment and seed size on growth and development of the next generation from germination through reproduction (Fig. 1). It is important to note that experiments 1 and 2 were not duplicate trials, but rather a pair of distinct studies of parental and seed size effects on the next generation. They differed in the following aspects. First, different starting stocks of wild-type Col-0 seed were used and different chambers were used to produce the seed used in the rest of the experiment, which was termed the test seed (Fig. 1, top). Second, the common environments (starting with germination and continuing through flowering and seed production in each experiment) in which those test seed were subsequently cultured were different (Fig. 1, middle). Third, a system for measuring root gravitropism from digital images of plates was developed between experiments; therefore, this parameter was measured in experiment 1, but not experiment 2 which was completed 1 year earlier. Finally, seed production parameters were not measured in experiment 2 due to a fungal infection that occurred after the plants had bolted. This greatly reduced the total seed yield, and therefore, these data were not included.

Figure 1.

Schematic of the experiment designed to quantify parental environment and seed size effects on growth and development of individuals in the next generation.

Outlined below are the methods involved in the progression of each experiment as depicted in Fig. 1.

Producing the test seed

For experiments 1 and 2, a different stock of wild-type A. thaliana (Col-0 ecotype) was used to initiate the activities outlined in Fig. 1. First, seed was surface sterilized with 70% ethanol, 2% Triton X-100 and then divided into seed size classes using a 212 µm, 250 µm, 300 µm and 355 µm test sieve. Seed smaller than 212 µm and larger than 355 µm was not used in this study. Seed was planted on 1/2X MS plates supplemented with 1% sucrose and adjusted to pH 5.7 with Bis-tris propane (BTP). Plates were stratified for 2 d at 4 °C and then grown in long days (16 h of 50 µmol m−2 s−1 cool white fluorescent light, 8 h dark) at approximately 25 °C for 1 week. Fifteen to 25 seedlings were then transferred to each of the different growth chambers (described below, Supporting Information Table S1). The proportions of each sieve class transferred to each chamber were identical to ensure that seedling populations between chambers started out as uniform as possible. The plant culture methods employed by the regular caretakers of each chamber were maintained so soil type, pot size and configuration, water regime, pest control and drying procedures contributed to the parental environment diversity.

Chambers 1–3 were used in experiment 1, while chambers 4–6 were used in experiment 2 (Supporting Information Table S1). The chambers chosen were all regularly used by three separate laboratories for growth of A. thaliana. Chambers 1, 2 and 5 were all managed by a single laboratory and chambers 2 and 5 are the same chamber used 1 year apart. Seedlings grown in these chambers were transferred to a soil consisting of 1:1:1 peat : perlite : vermiculite. Pots were watered with 10% Hoagland's solution (Hoagland & Arnon 1950) using a wicking system. Chambers 3 and 6 are the same chambers used 1 year apart in a second laboratory. Seedlings were planted using the Arasystem components ( and potted with Jiffy Mix (Jiffy Products of America, Pots were watered on a weekly to biweekly basis. Chamber 4, used by a third laboratory, did not have humidity control and temperature was not well controlled during the hottest summer months when the chamber was used for this experiment. After transplanting into Fafard 3B mix (, plants were watered as needed, weekly or biweekly. Approximately 2 months after flowering, seed from each chamber was collected from each individual plant and seed characteristics were determined as outlined below.

Measurement of seed size and weight

Morphological seed characters were determined by scanning 500–1500 seed spread across the bottom of a dish at 3200 dpi. A custom image analysis program written in Matlab ( thresholded the images and used blob detection/analysis routines to quantify the area of each seed as well as lengths of the major and minor axes. The seed-size frequency histograms indicated that the scanned populations were normally distributed so the means and standard deviations of each scanned population were used to construct the best fitting normal distribution. The seed in the scanned dish were weighed. Dividing this mass by the number of seed detected by the image-analysis program gave the per seed weight.

Preparing and sowing the test seed

For each experiment, a total of 15 populations of seed (collected from 15 individual plants) were chosen by selecting five seed populations from each of the three chambers that had an average seed area as close as possible to 280 µm2. Each population was subdivided into large and small seed size classes with U.S. Standard Brass Test Sieves (Fisher Scientific Co., Seed that passed through a 250 µm mesh but was retained on a 212 µm mesh was defined as small, while those passing through a 355 µm mesh and retained on a 300 µm mesh were defined as large. Seed was surface sterilized and then sown on 1% agar plates containing 1 mM KCl, 1 mM CaCl2, 5 mM 2-(N-morpholino)ethanesulfonic acid, adjusted to pH 5.7 with BTP.

Each sieved seed population (a total of 30 for each experiment, Fig. 1) was represented by two plates, resulting in a total of 60 plates per experiment. For logistical reasons, the test seed entered the common culture environments over a period of time rather than all at once. To reduce any temporal bias, the planting regimen was randomized such that seed size and parental environment was represented equally across experimental time. The plates were stratified for 2 d at 4 °C and then grown vertically under continuous 50 µmol m−2 s−1 white light for 4 d.

Germination, root growth and gravitropism

Images of the seed planted and developing seedlings on the plate were taken every 12 h. Germination was scored as the time after transfer to the light that the radical first emerged. Seedling images were taken every 12 h from germination to the end of 4 d after transfer to light and root length was measured using ImageTool (

In experiment 1, at the end of 4 d, the plate was rotated 90 degrees and images were collected at 0, 3 and 6 h time points. The root tip angle was determined automatically using a custom image analysis program designed for quantifying the gravitropic response. The algorithm used a corner detector with a disk size larger than the root width to find regions of the image that possibly contained a root tip. The disk image segments were computationally compared to a manually constructed library of root tip images. The disk image segment that most closely matched the library was defined as the root tip. To find tip angle, the disk was thresholded and principal components analysis (PCA) was performed. Because the disk size is larger than the root width, the first principal component (PC1) points in the direction along the root length. The tip angle is defined as the angle formed by the first principal vector and the camera's horizon. The seedlings were returned to the vertical position and allowed 12 h of recovery before being transferred to soil.

Common environments for rosette growth through seed production

After transferring to soil, plants were grown in a common growth chamber. For experiment 1, plants were grown on a 16 h light (100 µmol m−2 s−1 white light provided by cool white fluorescent bulbs) and 8 h dark schedule. The average temperature was 21 °C and the average humidity was 81%. Plants were transferred to Fafard's 3B mix ( in Arasystem pots ( and watered as needed, usually once every two weeks. For experiment 2, plants were also grown under a 16 h light, 8 h dark cycle, but the photon fluence rate delivered by the cool white fluorescent bulbs was 50 µmol m−2 s−1. The average temperature was 23 °C and average humidity was 69%. Plants were transferred to Fafard's 3B mix in three-inch pots. Trays were watered as needed, usually once every 1–2 weeks.

Rosette growth, flowering time and seed set

Rosette growth measurements (measured as the largest diameter from the tip of one rosette leaf to the other) were made weekly until the first floral bud appeared. From that point, rosette measurements were made daily until the bolt reached 1 cm. Flowering was measured in units of days after germination (DAG). DAG to the appearance of the first floral buds was measured for both experiments. DAG to the first open flower was measured in experiment 1. In experiment 1, plants were allowed to mature and seed was collected. Morphometric seed parameters were determined via the algorithm described above. Seed was stored in a common dessicator for approximately 3 months before seed measurements were collected.

Statistical analysis

For the statistical analyses in Supporting Information Tables S1 and S2, one-way analysis of variance (anova) was used in tandem with a Tukey test to identify mean differences that were significantly different. Temperature and humidity data were collected with a portable datalogger (HOBO, Onset, every 2 min for 4–7 d. The full raw recordings from each chamber were used to calculate the values and statistics. For Supporting Information Table S2, the average seed area was determined by pooling all seed measurements by chamber. The average seed number and average seed weight are measurements made per plant; therefore, there are approximately 15 measurements per chamber.

For analysis of the effects of seed size and parental environment on growth and development (Supporting Information Table S3), two-way anova was performed. For root growth and gravitropism, the PC1 value was used as a single value descriptor of the time course for each individual. For rosette growth, a logistic function was fit to each individual growth curve to convert the variously timed rosette measurements into a uniformly spaced data set. PCA was run on the fits and PC1 was used for ANOVA analysis. The PC1 of an individual response was chosen over the response mean since this parameter contains information about variation from the mean time course and in that sense maintains time course information. Secondly, the PC1 is a more unique descriptor of an individual response since it is less likely that two different responses will have identical PC1 than identical means. Two-way ANOVA was run using the PC1 (for root growth, root gravitropism and rosette growth) or single number descriptor (for germination time, flowering time and seed production) both with and without interaction between seed size and chamber. The model that described the most variance in the data is shown in Supporting Information Table S3. For Supporting Information Tables S1, S2 and S4, Tukey tests were used to quantify the magnitudes and significance of mean differences.


Design of experiments to test effects of seed size and parental environment

Two experiments consisting of thousands of measurements outlined in Fig. 1 was designed to investigate the influence of parental environment and seed size on growth and development of the next generation from germination through reproduction (Materials and Methods). Both studies address the question of whether seed size and parental environment affect growth and development in the next generation; however, the experiments differed in several aspects (outlined in Materials and Methods) making it logical to accept the results of each experiment individually.

The first step in both experiments was production of seed populations that differed mainly in the parental environment they experienced during their production (these populations are referred throughout as test seed). The environments chosen varied non-uniformly (Supporting Information Table S1), but all were controlled spaces used routinely for growth of Arabidopsis. Some environments were managed by different laboratories, others were different chambers managed by the same laboratory, and two chambers were used in both experiments, 1 year apart (see Materials and Methods). Therefore, this study sought to determine whether small differences in parental environment as could be expected to occur within and across laboratories could impact growth and development of the next generation.

Wild-type seed (Col-0 ecotype) from two separate stocks was planted in six different growth chambers normally used for raising Arabidopsis (Fig. 1). The environmental conditions of these chambers were characterized by real-time measurements of temperature and humidity averaged over multiple day/night cycles to enable statistical tests of significant differences. No two chambers had the same humidity; chamber 4 was significantly warmer than the other chambers. These data are shown in Supporting Information Table S1 along with the type of light sources, day lengths and photosynthetically active photon fluence rates. Characteristics and numbers of seed produced in the six different parental chambers are shown in Supporting Information Table S2. Within both experiments, seed number differed significantly between plants grown in all chambers and plants from one chamber in each experiment had significantly different seed weight. Seed area differed significantly between seed from all chambers in experiment 2. These data show that the variation between the chambers was sufficient to produce a wide range of seed sizes, weights and numbers even though seed from a common stock was used in each experiment. This variation was observed across environments that were fairly constrained given that they all were being used for laboratory culture of Arabidopsis.

From each of the six parental chambers, the seed from five individual mother plants was chosen. All seed populations chosen had a similar average seed size and distribution. Each individual population was then divided into large and small populations by mechanical sieving (Fig. 1). After sowing on agar medium, germination of each population was scored as the time of radical emergence after transfer to light. Root length was measured from images of seedlings collected daily from germination through day 4. At the end of day 4, the plate was rotated 90 degrees in experiment 1. Root tip angle was determined using custom image analysis software in order to quantify the gravitropic response. Seedlings were then transferred to soil and cultured in a single, common environment (growth chamber) that was different between experiments 1 and 2. Rosette growth and flowering events were measured in both experiments and seed was collected in experiment 1 (Fig. 1). In experiment 2, gravitropism was not measured since the image analysis tools for making this measurement were not completed at the time this experiment was performed. Seed production in experiment 2 was stunted by a fungal infection; therefore, seed parameters were not measured in this experiment. In total, approximately 12 000 measurements were made on the progeny of seed that differed primarily or even solely in parental history and size.

Effects of seed size and parental environment on development throughout the life cycle

The chamber (parental environment) and seed size effects on the next generation were separated and quantified by a two-way anova in both experiments. The F-ratios and P-values in Supporting Information Table S3 show the size and statistical significance of these two factors, along with evidence of interaction between them in three of the measured processes. Seed size was shown to be able to affect all of the measured stages of development except for seed production. In experiment 1, seed size significantly affected germination, root gravitropism and the time after germination to the first open flower. In experiment 2, seed size affected rosette growth and the time after germination to the first floral bud. In both experiments, root elongation was affected by seed size. The significance of seed size effects were stronger during seedling development and tapered off as development progressed, with the smallest significant seed size effect displayed in the time to the first open flower (Supporting Information Table S3). Significant interaction between seed size and parental environment was found during root gravitropism and in the time to the first open flower in experiment 1 and in germination time in experiment 2.

Parental environment affected germination time in both experiments, but did not affect seedling development in either. An effect of parental environment was detected during rosette growth and in the time to the first floral bud in experiment 2 and in the seed weight of progeny of the test seed in experiment 1. The effects of parental environment were less significant overall than those of seed size. The significance of parental environment did not change in a regular way as development progressed as was observed with seed size effects.

These data indicate that both seed size and parental environment can have significant effects on seedling growth and development from germination through seed production. The following sections provide reaction norm-style comparisons of the means of the specific measurements taken throughout development.


Germination was significantly affected by both seed size and parental environment in experiment 1 (Fig. 2a) and by parental environment in experiment 2 (Fig. 2b). The largest difference, between chambers 2 and 3, was approximately 20%. Supporting Information Table S4 shows the magnitude and statistical significance of each inter-chamber and seed-size difference in population mean. Subsequent developmental times were measured using germination as t = 0 so the differences in Supporting Information Table S4 and Fig. 2 were taken into account.

Figure 2.

Effects of seed size and parental environment on germination time. Germination time of test seed by parental chamber and seed size for experiment 1 (a) and experiment 2 (b). Error bars indicate standard error of the mean.

Root growth

Figure 3a shows the time course of root elongation for seedlings from large and small seed produced in all three chambers of experiment 1 and Fig. 3c shows the same data for experiment 2. Roots produced by seedlings from large seed grew faster and reached a length that was approximately 60% longer than roots from small seed. To describe each root elongation result with a single number, PCA was performed on the individual time courses. The PC1 is a more accurate descriptor of an individual response for two reasons. First, PC1 maintains some of the time course information in the respect that this number represents a deviation from the average response of the entire population. Second, it is less likely that two different responses will have the same PC1 value than that they will have the same mean. Therefore, PC1 can more uniquely describe an individual response than the mean. The resulting PC1 coefficients were used for statistical analysis. The effect of seed size on this measure of root elongation was highly statistically significant (Supporting Information Table S3). No inter-chamber differences were statistically significant (Supporting Information Table S4) so no evidence of parental environment influencing root elongation was detected. Because measurements began after the radicle emerged from the seed, differences in germination time were not a factor in these results.

Figure 3.

Effects of seed size and parental environment on root growth and gravitropism response of seedling roots from large (black) and small (grey) seed. (a) Time course of root elongation for seedlings from large or small seed in experiment 1. (b) Principal components analysis of experiment 1 responses generated a single value descriptor of the root growth [first principal component (PC1)] that showed a large effect of seed size and a much smaller influence of parental environment (chamber). (c) Time course of root elongation of seedlings from small and large seed in experiment 2. (d) First principal component coefficients for root growth of individual seedlings from experiment 2 were averaged by parental environment and seed size. Seed size had a large effect on root growth for all chambers of experiment 2. (e) Time course of root gravitropism for seedlings from large or small seed. (f) Principal components analysis generated a single value descriptor of the gravitropic response (PC1) that showed an effect of seed size and a particularly strong influence of parental chamber 3. Error bars are standard error of the mean.

Root gravitropism

Root gravitropism, like root growth, was affected by seed size. Within 6 h, tips of roots from large seed had achieved an angle of approximately 75 degrees, while those from small seed had reached 60 degrees, on average (Fig. 3e). In order to determine statistical significance of the seed-size versus chamber comparisons given multipoint time-course data, PCA was again used. PC1 provided a useful single-number descriptor of each individual trial. Figure 3f shows the results in reaction-norm style. The seed size effect is evident. Its statistical significance and the lack of significance of parental environment effects are shown in Supporting Information Table S4. An interaction between the chamber and seed size, such that the seed size effect was greatest for seedlings from chamber 3 parents and smallest for seedlings from chamber 2 parents, is evident in Fig. 3f and shown to be statistically significant in Supporting Information Table S3.

Rosette growth

Rosette growth was examined for evidence of previous generation effects. Figure 4a, as an example, shows that rosettes of plants from chamber 3 large seed grew faster than rosettes of plants from chamber 3 small seed when both were cultured in the experiment 1 common chamber. There was not an overall effect of seed size on rosette growth for plants cultured in chambers 1 or 2 (Fig. 4c, Supporting Information Fig. S1). Figure 4b shows that a seed size effect was apparent in experiment 2 (across all chambers) despite the slower growth rate in its common environment. As in the root growth and gravitropism studies, the data were reduced down to a single-value descriptor by PCA, in this case, after fitting a logistic (sigmoid) function to the raw time courses. The results showed a significant effect of seed size only in the case of seed from chamber 3 in experiment 1 (Fig. 4c). When plants from all chambers in experiment 1 were pooled, the seed size effect was not significant (Supporting Information Table S3). In the common environment of experiment 2, however, seed size had a very significant effect on rosette growth (Supporting Information Table S3) regardless of parental environment (Fig. 4d).

Figure 4.

Effects of seed size and parental environment on rosette growth. (a) Time course of rosette diameter growth for plants from large or small seed produced in parental chamber 3. This condition was chosen as an example of seed size greatly affecting an adult plant process. (b) Time course of rosette diameter growth for plants from large or small seed produced in all parental chambers employed in experiment 2. Combining the results from all chambers in experiment 2 did not eliminate a seed size effect on rosette growth. (c) Principal components analysis produced a first principal component (PC1) descriptor of the time course that showed a large effect of seed size in seed from chamber 3. (d) PC1 descriptor of rosette growth showed a large effect of seed size in all parental chambers in experiment 2. Error bars are standard error of the mean.

Flowering time

Interestingly, seed size significantly affected the time to the first open flower by an average of 2.9 d (Supporting Information Table S4) or approximately 10%, but not the time to the first floral bud in experiment 1 (Fig. 5a, b, Supporting Information Table S3). Put another way, seed size did not affect the time to form the first flower but the first flower opened almost 3 d later in plants from small seed than plants from large seed. Conversely, in the common chamber used in experiment 2, seed size did significantly affect the time to the first floral bud (Fig. 5c, Supporting Information Table S3) with plants from small seed being slower by 2.4 d (Supporting Information Table S4). Parental environment (chamber) had a strong effect on time to form the first floral bud in experiment 2. The seed size effect was seen at different stages of floral development between experiments. However, in both experiments, the effect of seed size on flowering was manifest consistently between 32 and 37 d after germination, possibly indicating that seed size effects on flowering are more temporally dependent than developmentally dependent (Fig. 5b, d). Importantly, these data show the effects of seed size persisting long after the seed ceased to exist.

Figure 5.

Effects of seed size and parental environment on flowering time and seed weight. (a) The time to produce the first floral bud in experiment 1 was not greatly influenced by seed size or parental environment. (b) The time to first flower opening showed a strong seed size effect in experiment 1. (c) The time to produce the first floral bud in experiment 2 showed an influence of parental environment and seed size. (d) Effects of seed size and grandparental environment on seed weight in experiment 1. Seed weight was effected more by grandparental environment than seed size. Error bars are standard error of the mean. DAG, days after germination.

Seed production

At the completion of experiment 1, seed was collected from each individual in the common chamber. Seed weight was found to be significantly affected by the grandparental environment (Fig. 5d, Supporting Information Table S3). The difference in average seed weights was 3.8 µg between grandparental chambers 2 and 3, a 20% difference (Supporting Information Table S4). The size of the test seed sown in the common chamber did not significantly affect weight of the seed produced (Fig. 5d, Supporting Information Table S3) or seed size (data not shown). Like seed size, the parental environment can effect development at late stages of the life cycle, even affecting reproduction and progeny weight.


Dissection of parental environment and seed size on development

The effects of parental environment on development of the next generation must be carried through the seed. Many studies have described the effects of parental environment (Donohue 2009), but most of them do not isolate the effect of seed size (Roach & Wulff 1987). The design of the present study, namely generating small and large size test seed populations by sieving seed from six chambers, enabled this separation. This study was completed in two separate experiments aimed at determining the influence of seed size and parental environment on development through reproduction in the next generation. Because these separate experiments differed in the initial seed stocks used, the common environment of the test seed populations, and the analysis methods available, it is most appropriate to consider the results of these experiments additively and not as duplicate trials. Together, they show the potential for seed size and parental environment to affect growth and development throughout the entire life cycle.

This study showed clearly that seed size and environment could work both independently and together to affect development. In both experiments, it was found that the chamber (parental environment), independent of seed size, influenced growth and development throughout the life cycle, even to the extent of affecting the weight of the seed those plants produced, as was found in experiment 1 (Supporting Information Table S3). Clearly, this result indicates that information about the parental environment can be carried to the next generation in a seed-size independent manner. Parental environment was also found to affect germination of small and large seed differently giving evidence of interaction as was seen in Fig. 2b. Additional evidence of interaction between parental environment and seed size can be seen in the way chamber 3 conditions exacerbated the effects of seed size on the gravitropic response (Fig. 3f) and rosette growth (Fig. 4c). When interaction was pervasive across environments, such as for gravitropism and flowering time, it could be detected as significant overall through anova (Supporting Information Table S3).

Seed size by itself was also found to have effects on growth and development independent of the parental environment and like parental environment, these effects could be detected throughout development (Supporting Information Table S3). For example, regardless of parental environment, primary root growth was faster in seedlings coming from large seed (Fig. 3a, b). Seed size effects tended to predominate early in development and became less significant later on. However, seed size effects were much more significant overall than were the effects of parental environment (Supporting Information Table S3).

Seed size and parental environment can act as independent factors affecting growth and development. However, it is also true that parental environment can directly affect seed size, as was found in the production of the test seed (Supporting Information Table S2, experiment 2). Additionally, it was found that seed size and parental environment can interact to affect development as described above. It may therefore be more accurate to think of seed size and parental environment as separate but related factors mediating transgenerational effects on growth and development. Mousseau & Fox (1998) explain how this applies to propagules in general, and therefore is not plant-specific.

In a comprehensive study of gravitropism spawned from the present research, the responses of roots from small seed became much more like those of large seed when the growth medium was supplemented with mineral nutrients (Durham Brooks et al. 2010). This seemed consistent with small seed being deficient in one or more important nutrients relative to large seed, and thus, the behavioural differences between the seedlings were nutritionally based. However, direct elemental analysis of the seed did not provide good support for this idea. The mineral content analyses showed similar amounts of 19 elements on either a per tissue or per seed basis (Durham Brooks et al. 2010). One potential clue produced by the elemental analysis was that sodium levels were disproportionately high in small seed (Durham Brooks et al. 2010).

Interestingly, parental environment during production of the test seed and grandparental environment in experiment 1 were both found to significantly affect seed production parameters (Supporting Information Tables S2 & S3). On the other hand, the seed size of the parental generation did not affect the measured seed parameters in the next generation in experiment 1. In fact, the effect of seed size on progeny seed weight was highly insignificant (Supporting Information Table S3). These data may indicate different underlying mechanisms for the effects of parental environment and seed size on transgenerational development.

What has been called seed size in this report may be more specifically called seed area, as that is the parameter specifically measured and shown to have a transgenerational effect. Seed weight is yet another seed character. This parameter was not explicitly tested for its effects on development in the next generation, although it could have contributed to the observed effects of parental environment. Seed area and seed weight as measured in this study did not always correlate with each other, as shown in Supporting Information Table S2. Seed produced in chamber 3 was lighter and seed from chamber 5 was heavier than their areas would predict. From these results and the chamber characteristics in Supporting Information Table S1, differences in seed density resulting from differences in chamber humidity can be inferred. Differences in water content may explain these differences in seed density; however, seed was dried extensively by storage in a common dessicator for approximately 3 months before measurements were made or seed planted (Materials and Methods). Different protein : lipid ratios are another possible explanation for the observed variation in seed density.

Practical consequences to plant genetic studies

The results presented here have implications for reverse genetic studies in which the primary goal is to discover and describe the phenotypic effects of a mutation. Seed size, for example, affected root growth and gravitropism as much as some important mutations. At least one seed size effect persisted through to the time of first floral bud opening (Supporting Information Table S4, Fig. 5), indicating that no process in the life cycle may be considered far enough removed from the seed to be exempt from a potential influence. Furthermore, the environments utilized in this study were routinely used for laboratory culture of Arabidopsis. With large and persistent effects of seed size within the wild type, the potential to confound genetic studies is very real, and may be expected to increase as phenotyping tools achieve higher and higher resolution.

Fortunately, the present study does more than highlight the problem. It also provides information that will help researchers minimize the effects of seed size through experimental design (e.g. sieve-selecting uniform seed populations for comparison) or incorporate them into the design by using seed size as a condition variable. The same is true of parental environment. Even though the range of parental environments used in this study was relatively constrained due to the fact that normal laboratory environments were used, they still had significant effects on growth and development in the next generation and even beyond. Effects of parental environment persisted well past the time the embryo and the few micrograms of seed content were a significant portion of the growing plant. Adult plant processes such as rosette growth, time to flower and even weight of seed produced were affected by parental environment (or grandparental environment in the case of seed weight). To produce seed stocks less influenced by differences in parental environment, randomized placement of plants in a common environment for at least two generations is recommended, since over a single generation seed weight was significantly affected by the environment of the grandparental generation (Supporting Information Tables S3 & S4).

Interesting questions about the persistence in subsequent generations of the non-genetic or epigenetic effects described here are more exposed for investigation as a result of this study. For example, whether or not seed size effects will persist through culturing in a common environment to affect root growth behaviour in one or more subsequent generations can now be directly tested using the characterization presented here as background. One or more of the effects documented here could become model systems for mechanistic studies of parental effects on plant development. A better understanding of the effects of seed size and parental environment may lead to better control of variability in phenotype studies, ultimately resulting in more useful information about gene function being extracted from mutant populations. Indeed, the results of this study directly informed a reverse genetics study that described a subtle but persistent effect of a glutamate-like receptor gene mutation on root gravitropism (Miller et al. 2010)

Extrapolating to outside of the growth chamber

Although far from any natural setting, the growth chambers and Petri plates used for plant culture in this study generated results that may nonetheless have natural relevance. The effects of seed size and parental environment described here would be expected to add variability to the behaviour of offspring in a natural setting through mechanisms mediating phenotypic plasticity, such as epigenetic effects. Variability arising from these processes is not strictly linked to gene sequence variation, but it has genetic underpinnings, and where it has adaptive significance, it may influence selection and contribute to phenotype evolution (Schlichting & Pigliucci 1998). Laboratory-based investigations of conditional phenotype expression in a model organism, when integrated with field-based studies, may provide unique insight into natural phenotype expression beyond what could be achieved with either approach alone (Tonsor et al. 2005). For example, the higher temperature and lower humidity of chamber 4 was associated with large seed area in experiment 2 (Supporting Information Table S2), consistent with previous observations that species from dry, warm and shaded environments produce larger seed (Jurado & Westoby 1992; Leishman & Westoby 1994; Moles & Westoby 2004). Figure 2 shows seed size, and parental environment affected germination time. If amplified in some natural scenarios, this parental effect on germination time could translate into an advantage among seedlings competing for establishment (Donohue 2009) and even result in a qualitative shift in life history (Wilczek et al. 2009).

The present results support the findings of previous laboratory and field studies. For example, Andalo et al. (1998) found a parental environment effect on seed size that was correlated with primary root growth and root branching. Also, parental nutrition and parental environment were positively correlated with seed mass, germination and seedling growth (Aarssen & Burton 1990; Schmitt, Niles & Wulff 1992). In stressful conditions, plants have been shown via their seed to prepare or predispose progeny towards a strategy that promotes their fitness in that condition. For example, in a poor nutrient environment, plants may produce smaller seed that results in slower developing progeny equipped to adopt a ‘wait and tolerate’ strategy until more favourable conditions develop (Aarssen & Burton 1990). Or a different strategy is to produce fewer but larger, more robust seed in a stressful environment (Smith & Fretwell 1974; Westoby et al. 2002). The large seed area associated with chamber 4 (high temperature, low humidity) may be the case of the latter strategy (Supporting Information Table S2), while the slower rosette growth and flowering time of plants from the smaller and lighter chamber 6 seed may be evidence of a ‘wait and tolerate’ strategy (Figs 4d & 5b,c).

The adaptive significance of the transgenerational effects observed in this study were not explicitly tested; however, the plasticity observed in the growth and development of the next generation brings forth the possibility that genetic programs are differentially expressed or regulated between populations from different seed sizes or parental environments. Recent reports illustrate this concept. The effect of parental environment on germination frequency was determined for two phytochrome mutants, phyB and phyD (Donohue et al. 2008). The mutant seed germinated normally when the parent was grown in a warm environment, but not in colder parental conditions, suggesting that these genes are recruited to mediate normal development in a changing environmental context. Furthermore, a study of the role of enhancers in the expression of a gene required for trichome development of Drosophila larvae found that these regulatory regions were required for normal trichome development at extreme temperatures, but not at the optimal temperature (Frankel et al. 2010). These data suggested that the temperature-dependant enhancer regions of this gene helped mediate phenotypic robustness in the face of environmental variability. Parental environment appears to determine the degree to which genes and regulatory elements mediate organismal development, and therefore determines scenarios in which regions of the genome may experience selective pressure. Any genes differentially regulated between the test seed populations used here could be exposed to natural selection to a degree that depends on parental environment. Long-term exposure of the populations used in this study to environments that cause divergent developmental (ontogenetic) trajectories could eventually lead to genetic variation between the populations that would stabilize (canalize) their development in the new environment.


This work was supported by the National Science Foundation Plant Genome Research Program (grant no. DBI-0621702) and by the Department of Energy (grant no. DE-FG02-04ER15527).