Transgenerational effects of plant sex and arbuscular mycorrhizal symbiosis

Authors


Summary

  • In gynodioecious plants, females are predicted to produce more and/or better offspring than hermaphrodites in order to be maintained in the same population. In the field, the roots of both sexes are usually colonized by arbuscular mycorrhizal (AM) fungi. Transgenerational effects of mycorrhizal symbiosis are largely unknown, although theoretically expected.
  • We examined the maternal and paternal effects of AM fungal symbiosis and host sex on seed production and posterior seedling performance in Geranium sylvaticum, a gynodioecious plant. We hand-pollinated cloned females and hermaphrodites in symbiosis with AM fungi or in nonmycorrhizal conditions and measured seed number and mass, and seedling survival and growth in a glasshouse experiment.
  • Females produced more seeds than hermaphrodites, but the seeds did not germinate, survive or grow better. Mycorrhizal plants were larger, but did not produce more seeds than nonmycorrhizal plants. Transgenerational parental effects of AM fungi were verified in seedling performance.
  • This is the first study to show transgenerational mycorrhiza-mediated parental effects in a gynodioecious species. Mycorrhizal symbiosis affects plant fitness mainly through female functions with enduring effects on the next generation.

Introduction

Parents determine the genotype of their offspring, but can also influence them beyond direct genetic contributions in response to environmental heterogeneity. The ability of seeds to germinate, grow, survive and reproduce may be influenced not only by the particular environment the seedling is growing in, but also by the environmental conditions the parents experienced during ovule fertilization and posterior seed development and maturation, which determines seed quality (Rossiter, 1996). Thus, transgenerational phenotypic plasticity or parental effects occur when the phenotype or performance of the offspring is affected by the phenotype or the environment of its parents (Roach & Wulff, 1987; Donohue, 2009). Parental effects may be divided into maternal and paternal effects. Maternal parents provision the seeds, and this is of course strongly linked to the environment in which the mother grows. Furthermore, in angiosperms, mothers transmit two-thirds of the endosperm genetic material and the cytoplasmic DNA, and environmentally influenced gene expression may therefore disproportionately reflect the maternal environment (Mazer & Gorchov, 1996). Paternal environment influences both pollen quantity and quality (Young & Stanton, 1990; Delph et al., 1997). Paternal environmental effects are considered only prezygotic (Lacey, 1996) and therefore the paternal effects on offspring may be less strong than those of the maternal environment (Schmid & Dolt, 1994; Mazer & Gorchov, 1996). Generalizations about paternal effects should be made with prudence, because in contrast to the large body of evidence supporting an environmental maternal influence on offspring (Roach & Wulff, 1987; Rossiter, 1996), we know very little about paternal environmental effects (Young & Stanton, 1990; Schmid & Dolt, 1994; Lacey, 1996).

Parental effects on offspring can also be modified by antagonistic biotic interactions, including interactions with herbivores or pathogens experienced by the parents (reviewed in Holeski et al., 2012). Whether mutualistic biotic interactions such as those with arbuscular mycorrhizal (AM) fungi may influence parental effects is largely unexplored. Approximately 80% of land plants grow in symbiosis with AM fungi in their roots (Wang & Qiu, 2006), forming the so-called AM symbiosis. This association is regarded as mutualistic because there is a bidirectional transfer of resources between host plant and fungus, even though the relative benefits gained by the two partners may range from negative to positive depending on the cost/benefit ratio (Johnson et al., 1997; Jones & Smith, 2004). As a result of improved nutrient acquisition, mycorrhizal plants are usually larger and better able to tolerate environmental stress than nonmycorrhizal plants (Smith & Read, 2008). In addition, some studies have shown that AM fungi have effects beyond the parental generation: not only are AM plants able to mother more and/or better quality seeds than nonmycorrhizal plants (reviewed in Koide, 2010), but AM plants may also father more and/or better quality offspring through symbiotic effects on pollen quantity and quality (reviewed in Varga, 2010). The beneficial effects of AM fungi on plant reproductive parameters have been linked to the better acquisition of nutrients, especially phosphorus (P), in mycorrhizal plants (Koide, 2010; Varga, 2010).

Gynodioecy is a common genetic polymorphism found in c. 7% of plants, where females (male-sterile plants) coexist with hermaphrodites (Jacobs & Wade, 2003). In gynodioecious populations, females must compensate for not contributing genes through pollen, as hermaphrodites do, in order to be maintained in the same population (Lewis, 1941; Lloyd, 1975). The compensation is generally considered to take place through producing more and/or better seeds. The level of the compensation needed depends on the mechanism of sex determination involved (Charlesworth, 1981): if male sterility genes are transmitted only by nuclear genes, females must have at least double the reproductive success of hermaphrodites to compensate, but if the sex determination system is under cytoplasmic-nuclear control, the level of female reproductive advantage may be lower. The female advantage may be achieved through different mechanisms, including inbreeding depression avoidance, resource compensation mechanisms or less detrimental interactions with pathogens and herbivores (reviewed in Dufaÿ & Billard, 2012).

Even though the data are very novel and scarce, AM symbioses have been shown to provide sex-specific growth benefits to sexually dimorphic plants (Varga & Kytöviita, 2008, 2010a,b; Vega-Frutis & Guevara, 2009) as well as reproductive benefits in terms of flowering frequency or seed production (Varga & Kytöviita, 2010a,b). These sex-specific effects have been linked to the different resource needs and allocation patterns found in the different sexes (Varga & Kytöviita, 2008; Vega-Frutis & Guevara, 2009; Vega-Frutis et al., 2013). Moreover, AM symbioses may affect inbreeding depression (Nuortila et al., 2004; Botham et al., 2009), an important parameter in gynodioecious plants, where offspring produced by females will always benefit from being outcrossed compared with offspring produced by hermaphrodites, where the possibility of selfing exists in self-compatible species. Unfortunately, knowledge about the effects of AM symbiosis on inbreeding depression is too limited to allow firm conclusions to be drawn (Nuortila et al., 2004; Botham et al., 2009; Collin & Ashman, 2010). In view of the differences in root colonization observed by Collin & Ashman (2010) in selfed versus outcrossed plants, these authors hypothesized that AM could decrease inbreeding depression under benign conditions, whereas an increase in inbreeding depression may be predicted in harsher environments. Nevertheless, to our knowledge, no studies have been performed in sexually dimorphic plants evaluating simultaneously the effects of AM fungal symbiosis of both parental sexes on offspring performance. In this study we evaluated the influence of AM fungal symbiosis on maternal and paternal effects and that of host sex on seed production and posterior seedling performance in Geranium sylvaticum, a gynodioecious plant species. We performed a glasshouse experiment where we hand-pollinated cloned female and hermaphrodite plants inoculated with either Glomus claroideum or Glomus hoi or in nonmycorrhizal conditions. We measured the seed number and seed mass and posterior seedling survival and growth of the seedlings. Therefore, the aim of this study was to examine the following specific hypotheses: AM fungal symbiosis will improve host growth; AM fungal symbiosis of the parents will also have beneficial transgenerational effects and will enhance seedling performance; and female plants will produce more and/or better offspring (i.e. with enhanced survival and/or growth) than hermaphrodites.

Materials and Methods

Study material

Geranium sylvaticum L.(Geraniaceae) is a protandrous, self-compatible perennial plant with a Eurasian distribution found in herb-rich forests and meadows and along roads. Populations of G. sylvaticum consist mainly of female (male-sterile) and hermaphrodite individuals with ten functional stamens, but intermediate plants are also present in some populations (Asikainen & Mutikainen, 2003). The intermediate plants have one to nine functional stamens per flower or a mixture of female, intermediate and hermaphrodite flowers in the same plant (S. Varga, unpublished). Even though the frequency of the intermediate plants may be up to 84% (Vaarama & Jääskeläinen, 1967; Asikainen & Mutikainen, 2003; Volková et al., 2007), in the available studies they are usually not distinguished from hermaphrodites because both morphs produce pollen, in contrast to female plants. In the field, G. sylvaticum starts flowering in mid-June; the plants are pollinated by bumblebees, syrphid flies and other dipterans (Varga & Kytöviita, 2010b). Both sexes are usually colonized by similar amounts of AM fungi in the field (Varga et al., 2009).

The material used in this study was raised as described in Varga & Kytöviita (2010a,b) (see Supporting Information, Notes S1). In short, we raised 306 cloned individuals from nine hermaphrodite (referred to as ‘H’ hereafter) and five female (referred to as ‘F’ hereafter) plants. When 4 yr old, the plants were inoculated with the AM fungus Glomus claroideum (referred to as ‘CLA’ hereafter) or Glomus hoi (referred to as ‘HOI’ hereafter) or left noninoculated (referred to as ‘NM’ hereafter). We used two different Glomus species which provided different mycorrhizal benefits to the host plants, at least in terms of mass and P accumulation, in two previous experiments with similar environmental conditions (Kytöviita et al., 2003; Varga & Kytöviita, 2010a) and both AM fungal species have been observed in soils where G. sylvaticum grows (S. Varga, pers. obs.). Four to 11 plants within fungal treatments represented each plant genotype. The treatments formed a factorial experiment with two fixed factors (plant sex: F or H; and fungal treatment: CLA, HOI or NM) and one random factor (individual plant genotype: nine hermaphrodite and five female genotypes).

Data collection from the parental lines and hand-pollinations

After mycorrhizal inoculation, the number of rosette leaves was noted weekly for each plant to estimate plant size. Flowering phenology was followed by recording when each individual started and finished flowering. Also, the number and sex of each open flower were noted every second day. Some plants originally classified as hermaphrodites when collected in 2002 turned out to be intermediates; in this work we refer to all pollen-bearing individuals as hermaphrodites. No female plant produced any flower bearing pollen during the experiment.

We performed 1579 hand-pollinations from 29 June until 30 November 2005. During that time, 230 (out of 306 plants) plants flowered and we used 103 of them as pollen donors and 209 (75 F and 134 H) as pollen receivers. We pollinated flowers with self-pollen (self-pollination) or with pollen from another individual (cross-pollination; Fig. S1). On each pollination day, pollen donor/recipient pairings were chosen arbitrarily from the plants that had flowers presenting pollen (pollen donors) and receptive stigma (pollen receivers). Each flower received pollen from only one donor, but on a whole-plant basis plants received pollen from multiple donors from different mycorrhizal treatments (Fig. S1). Similarly, the same flower could be used as a pollen donor in multiple pollinations as only one stamen was used per pollination and hermaphrodite G. sylvaticum flowers normally produce 10 functional stamens.

Pollinations were conducted in an arbitrary order by rubbing one mature anther on the stigma of the recipient flower. Pollinated flowers were individually marked using a cotton thread after checking that the stigma was covered by pollen and bagged individually with a mesh bag to prevent the seeds being ejected. The species is not apomictic (S. Varga, pers. obs.) and pollination is necessary for seed production. After fertilization takes place in Geranium flowers, locules become more visible to the naked eye and the style elongates. We define this stage as ‘fruit’.

Seeds were collected when ripe, counted, weighed individually and kept at 4°C. All flowers (whether or not hand-pollinated) and fruits were collected at the end of the flowering period, counted and measured for flower length (distance between the receptacle and the stigma; Fig. S2). Nonpollinated and thus unfertilized flowers were on average 9 ± 1 mm (mean ± SD) long and hand-pollinated flowers measured on average 15 ± 1 mm (Fig. S2). We use this limit of 15 mm to define whether fertilization was successful or not (successfully fertilized: flower length ≥ 15 mm; not successfully fertilized: flower length < 15 mm; Fig. S2). Among successfully pollinated flowers, the fruits can further mature and contain ripe seeds, but not all fruits contain mature seeds. This was used to define seed success of the fruits (yes: fruits with at least one mature seed; no: fruits without mature seeds; Fig. S2). Seed set was calculated as the proportion of fruits that contained at least one mature seed. Seed number per fruit was also recorded.

Offspring parameters

Altogether, 1103 seeds were collected from the hand-pollinations and kept at 4°C for 2 yr without stratification. In August 2007, seeds were sown into pots containing heat-sterilized sand, watered with fertilized water (2.5 ml of Ingestad solution l−1 water; Ingestad, 1979) and stratified at 4°C for 5 months. In January 2008, seeds were moved to a glasshouse and germination was followed for 3 months. The germination rate was 40% at this point. Reported germination rates in G. sylvaticum range from 10 to 71% (Vaarama & Jääskeläinen, 1967; Asikainen & Mutikainen, 2003). In order to induce further germination, we stratified the remaining seeds again at 4°C for a further 5 months. In August 2008, the seeds were moved back to the glasshouse and germination was followed for another 3 months. The cumulative germination rate at this point was 50%. In November 2008, we performed a third stratification treatment and followed germination, but this time no more seeds germinated. Inspection showed that the remaining seeds were dead (decayed) at that point.

The germination percentage was calculated as the final cumulative germination percentage. Seedlings were potted individually in pots containing 200 ml of sterilized sand when the two cotyledons were fully expanded. From that point on, the seedlings were watered using dilute fertilizer solution (0.2 ml of Ingestad solution l−1 water). When each seedling was 2 months old, we calculated the relative growth rate (RGR) by gently washing the seedlings (shoots and roots) clear of potting medium and weighing them. RGR was calculated as (log seedling fresh weight – log seed fresh weight)/60 d. Survival of the seedlings until 2 months was also noted.

Statistical analyses

In all cases, we first performed a graphical data exploration (Zuur et al., 2010) to assess the potential best model with which to analyze each type of data. All interaction terms were included in the initial models and nonsignificant terms were then removed through a series of simplifications until the minimal adequate model for each variable measured was achieved using the Akaike information criterion (AIC) for comparison between models (Crawley, 2007). After fitting an adequate model, we checked the assumptions of each model employed. All statistical analyses were conducted in R (R Development Core Team, 2011).

Parental plant parameters

We used generalized linear mixed-effects models (glmer, library lme4) with a Poisson error structure to explore the differences in the number of leaves at the time of AM inoculation, and the maximum number of rosette leaves produced during the experiment. A generalized linear mixed-effects model with a binomial error structure was used to explore the differences in whether or not plants flowered among fungal treatments and between the sexes. In all models, the fixed factors were fungal treatment (CLA, HOI or NM), plant sex (F or H) and the interaction between these two factors. Plant sex was nested within genotype and was included as a random factor. The models were fitted by Laplace approximation. Differences between levels of a significant factor (i.e. fungal treatment) were tested with a posteriori contrast (Crawley, 2007).

Linear mixed-effects models (lme, library nlme) with the same structure in their fixed and random components as described in the previous paragraph were used to explore differences in the timing of the start of flowering, flowering duration, the number of flowers produced per plant and the number of hand-pollinated flowers per plant. To meet the model assumptions, the response variables were averaged and rank transformed, and the variance was modeled with varPower structure. In addition, we used a continuous correlation structure (corCAR1, continuous-time) only for the number of pollinated flowers (Pinheiro et al., 2011). The models were fitted by restricted maximum likelihood (REML).

We used generalized linear mixed-effects models with binomial error structure to explore the differences in fertilization success (yes or no) and seed success (yes or no); a generalized linear mixed-effects model with a Poisson error structure to explore the differences in seed set and in seed number per fruit; and a generalized linear mixed-effects model with a Gaussian error structure to explore the differences in seed mass. In all models, the fixed components were fungal treatment of the pollen donor (CLA, HOI or NM), fungal treatment of the pollen recipient (CLA, HOI or NM), sex of the plant (F or H) and type of pollination (self or cross). In the random component of models, we included plant sex nested within genotype. The models were fitted by Laplace approximation. Differences between levels of a significant factor (i.e. fungal treatment) were tested with a posteriori contrasts (Crawley, 2007). The relationship between seed number and seed mass was explored with Spearman's correlation.

Offspring parameters

Generalized linear mixed-effects models with a binomial error structure were used to explore the differences in the proportion of seeds that germinated and survived until 2 months, while generalized linear mixed-effects models with a Gaussian error structure were used to explore the differences in RGR and seedling mass. The fixed components of the models were fungal treatment of the pollen donor (CLA, HOI or NM), fungal treatment of the pollen recipient (CLA, HOI or NM), plant sex (F or H) and type of cross (self or cross). In addition, the seed mass was included as a covariate in the models used for seed germination proportion, seedling survival and seedling mass. In the random component of models, we included plant sex nested within genotype. The models were fitted by Laplace approximation.

Results

Parental plant parameters

At the time of AM inoculation, plant size (inferred from the number of rosette leaves) was not significantly different between sexes (inline image = 0.22; = 0.64) or fungal treatment (inline image = 0.64; = 0.72; Fig. 1). However, during the experiment AM fungi improved plant growth. Plants produced on average 9.4 ± 0.2 (mean ± SE) leaves irrespective of their sex (inline image = 1.22; = 0.28), but inoculated plants had 1.4 times more leaves than noninoculated plants regardless of the AM species used (inline image = 68.62; < 0.01; Fig. 1).

Figure 1.

Number of rosette leaves at the time of arbuscular mycorrhizal (AM) inoculation (11 April 2005) and the maximum number of rosette leaves present during the experiment in female (white bars) and hermaphrodite (black bars) Geranium sylvaticum plants inoculated with Glomus claroideum (CLA) or Glomus hoi (HOI) or noninoculated (NM). Mean ± 1 SE is given; = 306. Different letters above the bars indicate significant differences among fungal treatments at < 0.05.

Flowering took place from 3 May 2005 until 22 December 2005. During that period, 75% of plants flowered regardless of their sex (inline image = 1.66, = 0.20) or fungal treatment (inline image = 0.76; = 0.68). The sex of the plant did not have any statistically significant influence on the start (F1,12 = 1.55; = 0.23) or the duration of the flowering period (F1,12 = 1.65; = 0.24), but fungal treatment had a significant effect on the start of flowering (F2,208 = 3.04; = 0.05). Plants inoculated with G. hoi started flowering 20 d earlier than plants in the other two fungal treatments. Nevertheless, flowering duration was not significantly affected by the fungal treatment (F2,208 = 0.14; = 0.87). During the flowering period, plants produced on average 12 ± 1 flowers irrespective of sex (F1,12 = 0.64; = 0.43) or fungal treatment (F2,208 = 2.76; = 0.06).

We pollinated on average half of all flowers per plant, resulting in 1579 hand-pollinations, of which 19% were self-pollinations. The number of hand-pollinated flowers per plant was not related to any of the experimental factors considered (F1,12 = 0.30; = 0.59 and F2,208 = 2.36; = 0.10, for the effect of plant sex and fungal treatment, respectively). Fertilization success was higher in female mothers compared with hermaphrodites (68% versus 51%, respectively) and was independent of the fungal treatment of the pollen donor or receiver or the type of pollination (Table 1). Among these 935 successfully fertilized flowers, 55% produced at least one seed and, similarly to fertilization success, seed success was higher in female mothers compared with hermaphrodites (46% versus 24%, respectively) and was independent of the fungal treatment of the pollen donor or receiver or the type of pollination (Table 1). Moreover, female mothers produced more seeds per fruit than hermaphrodite mothers regardless of the fungal treatment used or the type of pollination (Table 1, Fig. 2a). Seed mass was on average 3.97 ± 0.05 mg regardless of any of the parameters considered in the statistical models (Table 1, Fig. 2b) and there was no relationship between seed number and seed mass (Spearman's rho: 0.016; significance = 0.75). Differing from the AM symbiosis effect on the mother, the fungal treatment of the pollen donor did not have any statistically significant effect on any of the parameters measured related to seed production (Table 1).

Table 1. Summary of the results from the generalized linear mixed-effects models for several parental plant parameters analyzed in Geranium sylvaticum
 dfFertilization successSeed successNumber seeds/fruitSeed mass
χ2 P χ2 P χ2 P χ2 P
  1. Significant values (< 0.05) are shown in bold. See the 'Materials and Methods' section for more details.

Sex of the pollen receiver110.43 <0.01 8.42 <0.01 4.87 0.03 2.470.29
Fungal treatment of the pollen receiver20.770.680.530.773.450.181.460.48
Fungal treatment of the pollen donor21.790.413.190.200.260.880.001.00
Pollination type10.240.622.000.161.910.170.490.48
Figure 2.

(a) Number of seeds per fruit and (b) mass of the seeds in mg produced by female and hermaphrodite Geranium sylvaticum plants inoculated with Glomus claroideum (CLA) or Glomus hoi (HOI) or noninoculated (NM) when hand-pollinated with pollen from a plant inoculated with G. claroideum (gray bars) or G. hoi (black bars) or noninoculated (white bars). Mean ± 1 SE is given. Different letters above the bars indicate a statistically significant difference between sexes at < 0.05 (see Table 1).

Offspring parameters

Seed mass was a significant covariate in all models explaining variation in all traits (Tables 2, 3), although the principal effects on seedling parameters were mostly attributable to the fungal treatments of the pollen receivers. Offspring parameters were not related to the sex of the mother or the type of pollination (Tables 2, 3). Germination rate was on average 50%, with a significant interaction between sex and fungal treatment of the pollen receiver (Table 2). In females, a larger proportion of seeds from CLA-inoculated mothers germinated compared with HOI and NM mothers, but in hermaphrodites, the largest proportion of seed germination was from NM mothers (Tables 2, 3; Fig. 3a). Of the 569 seeds that germinated, 81% survived until the age of 2 months. Mycorrhizal mothers produced seedlings with a higher survival probability compared with NM plants (Table 3, Fig. 3b). These seedlings also had higher RGR (Table 3, Fig. 3c) and, consequently, larger seedling mass when they reached the age of 2 months than seedlings produced by NM plants (Table 3, Fig. 3d). In direct contrast to the beneficial mycorrhizal effects exerted through the pollen receiver, the fungal treatment of the pollen donor did not improve seedling germination, survival or seedling mass (Tables 2, 3; Fig. 3a,c,d). However, seedlings fathered by HOI plants had the highest RGR (Fig. 3b).

Table 2. Summary of the results from the generalized linear mixed-effects models for the proportion of seed germination in Geranium sylvaticum
 Seed germination
dfχ2 P
  1. Significant values (< 0.05) are shown in bold. See the 'Materials and Methods' section for more details.

Sex of the pollen receiver10.290.59
Fungal treatment of the pollen receiver22.730.26
Fungal treatment of the pollen donor25.610.06
Pollination type15.800.08
Sex of the pollen receiver × fungal treatment of the pollen receiver26.94 0.03
Seed mass124.51 <0.01
Table 3. Summary of the results from the generalized linear mixed-effects models for several offspring parameters analyzed in Geranium sylvaticum
 dfSeedling survivalSeedling FW at 2 monthsSeedling RGR
χ2 P χ2 P χ2 P
  1. Significant values (< 0.05) are shown in bold. See the 'Materials and Methods' section for more details.

  2. RGR, relative growth rate; FW, fresh weight.

Sex of the pollen receiver10.000.960.001.000.001.00
Fungal treatment of the pollen receiver215.55 <0.01 45.14 <0.01 57.00 <0.01
Fungal treatment of the pollen donor25.160.085.160.087.05 0.03
Pollination type10.001.000.001.000.190.66
Seed mass123.46 <0.01 10.56 <0.01
Figure 3.

(a) Proportion of seeds that germinated, (b) proportion of seedlings that survived until 2 months, (c) seedling relative growth rate (RGR) in μg mg−1 d−1 and (d) seedling fresh weight (FW, in mg) produced by female and hermaphrodite Geranium sylvaticum plants inoculated with Glomus claroideum (CLA) or Glomus hoi (HOI) or noninoculated (NM) when hand-pollinated with pollen from a plant inoculated with G. claroideum (gray bars) or G. hoi (black bars) or noninoculated (white bars). Mean ± 1 SE is given. The small graphs on the right show the mean values among fungal treatments of the pollen donor and receiver. Different letters above the bars indicate statistically significant differences among fungal treatment of the pollen receiver at < 0.05 (see Tables 2, 3) except in (a) where significant differences attributable to the interaction between arbuscular mycorrhizal (AM) treatment of the pollen receiver and sex of the pollen receiver are indicated.

Discussion

This is the first work to report transgenerational mycorrhizal maternal and paternal effects in next-generation sporophytes in a gynodioecious species. AM fungal symbiosis improved host growth, but did not increase offspring production (seed number). We observed mycorrhizal maternal and paternal effects on the quality of offspring, and these effects were AM species-specific. Moreover, female plants produced more numerous, but not better offspring than hermaphrodites, suggesting that the sex of the mother plant did not affect the early life history performance of offspring.

Parental AM effects on offspring performance

Transgenerational mycorrhizal maternal effects have been examined in three hermaphroditic species previously, Abutilon theophrasti (Shumway & Koide, 1994a,b, 1995; Heppell et al., 1998), Avena fatua (Lu & Koide, 1991; Koide & Lu, 1992) and Campanula rotundifolia (Nuortila et al., 2004), and paternal effects have been examined in one species, C. rotundifolia (Nuortila et al., 2004). Species-specific mycorrhizal effects on the parental generation have been shown before (Oliveira et al., 2006), but this is the first study to reveal mycorrhizal species specificity in transgenerational effects. Beneficial mycorrhizal maternal effects were observed on the quality of seeds, and mycorrhizal mothers produced seeds that germinated better and seedlings with higher survival, RGR and mass. Similarly, beneficial mycorrhizal paternal effects were only detected in a seedling parameter (seedling RGR). There are two possible, not mutually exclusive pathways of maternal mycorrhizal effects: the seeds may have a higher nutrient content as a direct consequence of AM colonization or through an indirect pathway. Seed size was not affected by parental symbiotic interactions; however, the better quality of the seedlings produced was related to variation in seed size. Besides seed size, nutrient provisioning of seeds may affect posterior seedling performance (Fenner & Thompson, 2005). We could not investigate seed nutrient contents because measurements could not have been performed without destroying the seeds, but increased seed nutrient contents in response to AM inoculation are commonly reported (reviewed in Varga, 2010). The seed is composed of the sporophyte, the endosperm and the seed coat. Seed nutrient content refers to the nutrient reserves in the seed endosperm that the sporophyte will use. Even though AM plants may produce seeds with higher nutrient reserves, the link between seed provisioning and germination rate and seedling performance is not always straightforward (Lewis & Koide, 1990). Interestingly, beneficial maternal AM effects on seed germination were sex-specific. In females, AM increased the proportion of seeds that germinated regardless of the AM species considered, but in hermaphrodites, seeds produced by mycorrhizal mothers showed less germination than seeds produced by nonmycorrhizal mothers.

The mycorrhizal symbiosis of the pollen receiver did not influence offspring number, but increased seedling RGR in plants in symbiosis with one of the two fungal species examined and had marginally significant effects on all other seedling parameters investigated. Compared with knowledge of AM fungal effects on female function, relatively little is known about how AM fungi influence pollen. Paternal environmental effects have been little studied in general. It has been shown previously that the paternal light and nutrient environment affects percentage seed germination and days to germination, but not later life traits (Galloway, 2001). Overall, pollen production depends on resource availability during pollen development (Delph et al., 1997). Mycorrhizas enhance plant nutrient acquisition, which should result in improved pollen production and performance. Enhancement of pollen tube growth and siring success by AM fungi has been reported both in vitro and in vivo (Poulton et al., 2001). One explanation for the mycorrhizal paternal effect on seedling RGR observed in the present study is higher pollen competition among mycorrhizal pollen. The pollen competition hypothesis predicts that, when the number of pollen grains deposited onto stigmas exceeds the number of ovules, selection operates in the time frame between pollination and fertilization. Pollen competition has been shown to improve seed fitness (Mulcahy & Mulcahy, 1975), and mycorrhizal pollen could yield a higher germination probability compared with pollen of nonmycorrhizal plants (Poulton et al., 2001). However, tests of the pollen competition hypothesis are not conclusive (Mitchell, 1997) and the effects of pollen competition on seedling performance may become apparent only later in life and when assessed in different resource environments (Kalla & Ashman, 2002).

Effects of maternal sex on offspring performance

Generally, in gynodioecious species, females set more seeds than hermaphrodites (Shykoff et al., 2003). In this study, females produced 1.2 times more seeds than hermaphrodites. Furthermore, female flowers were more likely to be fertilized and to set more fruits with more seeds than hermaphrodites. This is the first study showing that female and hermaphrodite flowers differ in the probability of being fertilized after pollen application. The reasons for this difference need to be explored further, but one explanation could be the difference in pistil length between flowers produced by the two sexes. As in many other gynodioecious plants (e.g. Phacelia linearis; Eckhart, 1992), female G. sylvaticum have shorter pistils than hermaphrodites (S. Varga, unpublished). Pistil length has been shown to act as a selective agent for pollen competition (Lankinen & Skogsmyr, 2001; Ramesha et al., 2011) and the shorter pistil in females may discriminate less pollen and thus facilitate higher fertilization rates. This would be contradictory to females producing high-quality offspring unless a targeted post-zygotic abortion mechanism exists (Teixeira et al., 2009). Selective embryo abortion is common in plants (see Korbecka et al., 2002). Theoretically, AM fungi could also influence selective embryo abortion by increasing the amount of available maternal resources, but we did not find any evidence for that; the AM symbiotic status of the pollen donor and receiver did not affect seed success. Also, Nuortila et al. (2004) did not find any AM effect on embryo abortion rates in C. rotundifolia.

In the present study, seed mass was similar regardless of the sex of the mother, as previously reported by Asikainen & Mutikainen (2003). Sexual dimorphism in seed size has been reported for several gynodioecious species, suggesting female advantage gained from larger seeds that germinate better than seeds from hermaphrodites (reviewed in Shykoff et al., 2003). However, seed size is one of the least variable traits in plants (Marshall et al., 1986) and it does not always correlate with early seedling performance in gynodioecious plants (Shykoff et al., 2003), suggesting that other seed parameters such as nutrient contents may better reflect offspring quality. Seed P content is similar between the sexes in G. sylvaticum (S. Varga et al., unpublished), which may explain the lack of sexual dimorphism in seed germination, seedling survival and RGR observed. Taken together, these results suggest that the sex of the mother plant did not affect the early life history performance of offspring in the present study.

Effect of AM on parental performance

The positive effects of AM fungi on plant growth and reproduction are well established (Smith & Read, 2008). Contrary to expectations, the positive mycorrhizal effect on parental size did not translate into more numerous offspring in mycorrhizal mothers. Some studies have shown that plants gain benefits from AM fungi mainly as a result of improved acquisition of nutrients, particularly P, which usually translates into greater growth (Smith et al., 2010). Plant growth is normally positively correlated with flower and seed production, and larger seed output in plants is sometimes reported in response to AM (reviewed in Varga, 2010). However, species (Oliveira et al., 2006) and sexes within dimorphic species (Varga, 2010) are known to vary in their response to AM fungi. We did not measure AM colonization in the plants to avoid disturbing the root systems while plants were flowering. However, we did measure AM colonization in the following years in the same plants. The two sexes had similar amounts of hyphae and vesicles (Varga & Kytöviita, 2010a,b), similar to field observations (Varga et al., 2009). AM increased plant size and affected the onset of flowering regardless of the sex of the plant, but did not affect flowering probability or the number of flowers produced. This may be explained by the short time (a few months) that elapsed between AM inoculation and the flowering measurements. Colonization is normally observed a few days after inoculation in the field (Gay et al., 1982). However, the establishment of functional nutrient exchange may take longer. We measured flowering in the 2 yr following this experiment in the same parental plants, and found that AM inoculation increased flowering probability compared with nonmycorrhizal plants (Varga & Kytöviita, 2010a,b). Moreover, in the present study with 2-yr-old plants, AM decreased the time needed to initiate flowering in plants inoculated with G. hoi. This AM effect was present in the following 2 yr (Varga & Kytöviita, 2010a,b), even though the direction was not always the same. For example, when they were 3 yr old, the nonmycorrhizal plants started flowering 12 d earlier than AM plants, but when they were 4 yr old, both nonmycorrhizal plants and G. hoi-inoculated plants started flowering at the same time (Varga & Kytöviita, 2010a,b). This suggests that other factors, such as trade-offs between current and future reproduction or trade-offs between reproduction and plant functions, may influence the symbiotic relationship. AM fungi have been shown to influence flowering phenology in other species in contrasting ways (Bryla & Koide, 1990; Nakatsubo, 1997; Philip et al., 2001; Poulton et al., 2002). Taken together, these findings suggest that plant age affects the benefit the host derives, although it is difficult to fully evaluate the net effect of AM symbiosis on G. sylvaticum, considering that it is a long-lived perennial plant.

Inbreeding depression is theoretically considered an important force maintaining females in gynodioecious species (Charlesworth & Charlesworth, 1987), and symbiosis with AM fungi has been reported to affect the magnitude and direction of inbreeding depression (Nuortila et al., 2004; Botham et al., 2009). We found no strong evidence of inbreeding depression, although various levels of inbreeding depression have been reported in other gynodioecious species (Webb, 1999). Inbreeding depression may be manifested after the early life stages (Melser et al., 1999). Botham et al. (2009) observed that the effects of AM on Fragaria virginiana were trait- and life-stage dependent. Collin & Ashman (2010) showed that selfed plants were less likely than out-crossed plants to be colonized by AM fungi. Changes in the expression of maternal effects through time have been reported in perennial plants (Latzel & Klimešová, 2010), but the longevity of parental environmental effects, including those exerted by mycorrhizal symbiosis, remains unclear.

Conclusions

Mycorrhizal symbiosis of the parents may significantly affect offspring through female and male functions. Maternal AM benefits were not only related to larger seed mass, suggesting that either better seed provisioning or epigenetics may play an important role in determining seedling performance. The less pronounced paternal symbiosis-mediated effects may indicate that the reduced size of the male gametophyte does not facilitate transfer of paternal environmental information to the next generation. In evolutionary ecology, plant fitness is usually measured as plant size and offspring number. However, we show that symbiosis may affect not the offspring number but the offspring quality. In contrast to the beneficial mycorrhizal effects, plant sex had no effect on offspring quality, although this is usually considered to be a component of mechanisms maintaining females in gynodioecious populations. Different results may be obtained when different AM species from those used here, or plants growing with more complex AM fungal communities are evaluated, and revealing the full ecological and evolutionary importance of the parental legacy remains an exciting challenge for the future.

Acknowledgements

We thank the staff at the Botanical Gardens, University of Oulu, for the facilities provided, Carolin Nuortila and David Carrasco for helping us with the practical work and for valuable discussions, and the three anonymous reviewers for valuable suggestions. This study was financially supported by the Finnish Cultural Foundation and Finnish Academy of Science (S.V.) and by Consejo Nacional de Ciencia y Tecnología (CONACyT, R.V-F.).

Ancillary