|Home | About | Journals | Submit | Contact Us | Français|
Over a season, plant condition, amount of ongoing reproduction and biotic and abiotic environmental factors vary. As flowers age, flower condition and amount of pollen donated and received also vary. These internal and external changes are significant for fitness if they result in changes in reproduction and mating.
Literature from several fields was reviewed to provide a picture of the changes that occur in plants and flowers that can affect mating over a season. As flowers age, both the entire flower and individual floral whorls show changes in appearance and function. Over a season, changes in mating often appear as alteration in seed production vs. pollen donation. In several species, older, unpollinated flowers are more likely to self. If flowers are receiving pollen, staying open longer may increase the number of mates. In wild radish, for which there is considerable information on seed paternity, older flowers produce fewer seeds and appear to discriminate less among pollen donors. Pollen donor performance can also be linked to maternal plant age. Different pollinators and mates are available across the season. Also in wild radish, maternal plants appear to exert the most control over paternity when they are of intermediate age.
Although much is known about the characters of plants and flowers that can change over a season, there is less information on the effects of age on mating. Several studies document changes in self-pollination over time, but very few, other than those on wild radish, consider more subtle aspects of differential success of pollen donors over time.
Because resources, plant condition, pollinator service and mate availability change throughout a season, the patterns of mating and reproductive allocation that would maximize fitness probably change over time. At any point in time plants may allocate resources to immediate reproduction or to growth and maintenance functions that might increase future reproduction. The resources available for reproduction can be divided among pollen and ovule production followed by fruit and seed filling (reviewed in Bolmgren and Cowan, 2008; but see also Lacey and Pace, 1983; Case et al., 1996; Ollerton and Diaz, 1999; Picó and Retana, 2000; Wolfe and Burns, 2001; Galloway, 2002; Lacey et al., 2003). At any given time a plant might also have mechanisms that prevent or allow selfing (Seed et al., 2006; Qu et al., 2007), cause pollen to accumulate or germinate on the stigma (Herrero, 2003; Lankinen et al., 2007), and increase or decrease the selectivity among compatible mates (Dahl and Fredrikson, 1996). As plants age, the effectiveness and importance of these mechanisms may change. These changes can affect the quality and quantity of offspring produced and will, therefore, affect fitness (Kalisz and Vogler, 2003; Weekley and Brothers, 2006; Zhang and Li, 2008). Extensive experimental and theoretical work has been done to address these issues. Here this work is considered primarily from the perspective of changes in mating patterns as plants and flowers age.
Variation in mating patterns occurs across species, within species and within plants. We focus here on the kinds of variation that occur within plants as they age, as these differences can result from variation in conditions. Within plants, changes in mating in response to age can occur within or across seasons. To limit the scope of this review, we discuss the variation that occurs within a season, the only kind available to annual plants. We consider both the short-term changes in reproduction that occur as flowers age and the longer term changes as plants age over a season. Differences as flowers age are probably due primarily to physiological changes in the flower and availability of pollinators and mates. These may occur due to changes in environmental factors such as temperature, light, nutrients and water availability. Changes in maternal condition, developmental stage, resource allocation patterns, sink/source relationships and mating experience also occur as plants age during a season.
In reviewing the literature we were particularly interested in cases where changes in plants and flowers were tied to changes in the mating pattern. Therefore, particular attention was given to the body of work done on Raphanus sativus (wild radish) because there is a very large amount of information on seed paternity in this species (e.g. Marshall and Ellstrand, 1986; Marshall, 1988, 1991, 1998; Devlin and Ellstrand, 1990; Devlin et al., 1992; Marshall et al., 1996, 2000, 2007b). This is important because measurement of paternity is necessary to test whether success of pollen donors changes over time. Most previous studies of the effects of plant and flower age on mating evaluate differences in selfing and outcrossing, but scoring seed siring success allows evaluation of finer scale changes in mating patterns. In addition, a number of experiments with wild radish contain information about changes in mating over various time scales. Bringing all of these together in the context of the literature provides an unusually complete picture of how changes in plants and flowers over the season may alter mating.
Both physiological and architectural changes occur as plants and flowers age (Table 1). Within the lifetime of a flower there is a shift from specializing in receipt and distribution of pollen to fertilizing and nurturing ovules. In most species flowers senesce at a gradual pace that is genetically programmed but can be accelerated by pollination (reviewed in O'Neill, 1997; but see also Yasaka et al., 1998; Abdala-Roberts et al., 2007; Castro et al., 2008). Reproduction is resource intensive and reduces resources available for physiological processes that ensure survival, such as vegetative growth (Obeso, 2002; Hirose et al., 2005; Bolmgren and Cowan, 2008). Fruits and flowers that are further removed from resources (either because of distance or time) show a pattern of reduced investment; they may be smaller (Galen, 2005), or more likely to be aborted (Diggle, 1995). Maternal condition has also been shown to affect mate choice in the form of seed paternity (Marshall and Ellstrand, 1988; Marshall and Fuller, 1994; Levri, 1998; Marshall et al., 2007a). Other physiological constraints on reproductive output over time include genetically and environmentally determined changes in floral receptivity (O'Neill, 1997; Page et al., 2006; Lankinen et al., 2006, 2007).
Environmental factors also change as plants and flowers age (Table 1). Biotic factors that vary over time and can affect reproduction include pathogens (Shykoff et al., 1996), predators (Huth and Pellmyr, 1997; Mahoro, 2002; Lacey et al., 2003), herbivores (Vallius and Salonen, 2000, 2006), competition for pollinators (Wheelwright, 1985; Ishii and Sakai, 2001, 2002), and pollinator composition and abundance (Morinaga et al., 2003; Brunet and Sweet, 2006; Brunet, 2009). Abiotic factors that change over time and can affect reproduction include temperature (Chang and Struckmeyer, 1976; Kumar et al., 1995; Yasaka et al., 1998; Hedhly et al., 2003; He et al., 2006), and the availability of resources such as water (Galen, 2005), nutrients (Young and Stanton, 1990b; Hirose et al., 2005) and light (Schemske et al., 1978).
Conceptually, both internal and external conditions that change over time can affect mating (Fig. 1). Internal conditions that change over time include both genetic and plastic developmental changes, changes in the amount of garnered resources, which is usually reduced for older plants, and mating experience, which is detectable as the number of fruits maturing and their vigour, as well as pollen remaining in anthers or deposited on stigmas. External changes include biotic and abiotic factors (see above). At any point in time, plants can respond to the sum of information on internal and external conditions and alter investment in individual flowers, embryos or fruits as well as change allocation to female vs. male function. Improved mating success can also be achieved by altering the timing of the investments in reproduction such that access to a greater number and/or quality of mates occurs. Plants can alter the timing of events such as the onset of reproduction, the maturation and senescence of floral structures, stigma receptivity, and anther dehiscence and sex expression.
Flowers may be open from 1 or 2 d to a few months (Primack, 1985). Over even a short time period the physiological status of the flower may change, petals may open to their fullest extent, and the relative positions of stigmas and anthers may vary (Armbruster et al., 2009). While the stigma remains available, pollen may accumulate and germinate as it arrives or be restricted in germination time by stigmatic events. As the flower ages, changes in petal colour, condition and rewards may affect pollinator service (Devlin and Stephenson, 1985). And as flowers age, ovules age, leaving fewer ovules available for fertilization. The condition of the flower as well as availability of pollen and ovules may affect mating patterns and quality of seeds as flowers age.
Optimal floral longevity may be determined by the balance between costs of floral maintenance and construction of new flowers, and the rates of pollen dissemination and receipt (Ashman and Schoen, 1994). Long-lived flowers are predicted when floral maintenance costs and rates of pollen dissemination and receipt are low, and short-lived flowers are predicted when floral maintenance costs and rates of pollen dissemination and receipt are high (Schoen and Ashman, 1995). These predictions hold for Clarkia temblorensis, in which longer-lived flowers produced proportionally more nectar sugar but fewer seeds (Ashman and Schoen, 1997).
In some species, including many species of orchids, flowers can persist for a very long time. For example, Grammatophyllum multiflorum has a record floral age of 9 months in greenhouse conditions (Primack, 1985). In his review of floral longevity, Primack (1985) identified several patterns. With the exception of orchids, which can have flowers that last many months, tropical forest species generally have short-lived flowers, and temperate forest species have longer-lived flowers. In temperate forests, species that flower early in the growing season have longer-lived flowers, whereas late-season species have short-lived flowers. Flowers of inbreeding species tend to be short-lived, whereas those of outbreeding species typically last longer, perhaps because flowers of outbreeding species tend to have more ovules, requiring accumulation of more pollen grains (Cruden, 2000). Finally, female flowers generally last longer than male flowers.
Two kinds of changes in flower colour are known as flowers age. First, as flowers senesce they fade and wilt. Second, for some species, flowers, or parts of flowers, may change colour prior to senescence. These kinds of changes can affect mating patterns if they alter pollinator service and/or are affected by pollination events.
Changes in flower colour prior to senescence occur widely; they are found in at least 78 diverse plant families and have probably evolved multiple times (Weiss, 1995; Weiss and Lamont, 1997). These changes are likely to impact mating in that they are thought to aid long-distance pollinator attraction and, at the inflorescence level, to improve pollination efficiency. For example, deerweed (Lotus scoparius) flowers change from yellow to orange. The retention of these older flowers increased pollinator (bumble-bee) visitation and fruit production (Jones and Cruzan, 1999). And in an experiment with artificial floral arrays, a change in artificial flower colour for non-rewarding flowers reduced visits to those flowers (Kudo et al., 2007).
Changes (usually loss) in colour and declines in rewards for pollinators are normal parts of floral senescence. Unpollinated flowers that undergo senescence are unlikely to receive further visitors; pollinators tend to select younger flowers as older flowers tend to have fewer rewards. For example, nectar rewards are greater in younger flowers in Lobelia cardinalis (Devlin and Stephenson, 1985) and pollinators respond to these changes. Rosa virginiana blossoms fade as they age and honey-bees preferentially visit first-day flowers (MacPhail et al., 2007). This occurs even if nectar is added to the older flowers. Likewise, in Alkanna orientalis (Boraginaceae), bees preferentially visit bright yellow flowers rather than faded flowers (Nuttman et al., 2006).
As flowers age, the entire flower may not senesce at the same time. The whorls of floral parts may become active and senescent at different times and their relative positions may change. These changes lead to differences in floral attractiveness, changes in stigma receptivity and pollen availability, and changes in functional gender of the flowers.
Across a wide variety of plants, stigmas can remain receptive from 1 to 17 d, with peak receptivity often occurring within 1 week of flower opening (Sindhu and Singh, 1976; Egea and Burgos, 1992; Young and Gravitz, 2002; Masierowska and Stpiczyńska, 2005; Page et al., 2006; Waites and Ågren, 2006; Yi et al., 2006; Franchi et al., 2007). Stigma closure can be triggered by mechanical stimulation alone (Fetscher and Kohn, 1999; Richardson, 2004), or only by pollination (Waser and Fugate, 1986), but stigmatic reopening can occur, especially when pollen loads are low (Waser and Fugate, 1986; Richardson, 2004; but see Fetscher and Kohn, 1999), or little outcross pollen was received (Richardson, 2004). Delayed stigma receptivity can result in the accumulation of viable pollen on the stigma (Lankinen et al., 2007; but see also Galen et al., 1986; Murdy and Carter, 1987; Dahl and Fredrikson, 1996; Herrero, 2003) and could affect mating by increasing the number and/or variety of mates arriving on the stigma before germination can occur. Thus, delayed receptivity could increase the opportunity for both pollen competition and mate choice (Galen et al., 1986). In Collinsia heterophylla, timing of stigma receptivity varied among maternal plants and was affected by the identity of pollen applied to the stigma (Lankinen and Kiboi, 2007).
Pollen longevity and/or viability may change based upon environmental conditions and thus will change as the environment changes over a season. Changes in the cell membrane, carbohydrate levels, temperature, relative humidity and pollen morphology all appear to affect how long a pollen grain remains viable (reviewed in Dafni and Firmage, 2000) and most of these variables are in some way related to how much water the pollen grain needs to maintain viability (reviewed in Dafni and Firmage, 2000; Pacini et al., 2006). Pollen grains must have some moisture content at anthesis, and to maintain water levels they may manipulate carbohydrate levels such as metabolizing starches prior to anthesis and/or by maintaining sucrose levels in the cell membrane (reviewed in Dafni and Firmage, 2000; Pacini et al., 2006; Franchi et al., 2007). Pollen viability also appears to change depending upon temperature, which could be closely tied to water loss. For example, in Kochia scoparia, pollen was viable longer at lower temperatures and high relative humidity (Mulugeta et al., 1994). Low relative humidity has also been shown to reduce pollen longevity in species with desiccation-intolerant pollen such as Zea mays (Buitink et al., 1996; Fonseca and Westgate, 2005) and Pennisetum glaucum (Kumar et al., 1995), although this was not true of the dessication-tolerant pollen from Typha latifolia (Buitink et al., 1996). Conversely, increased pollen longevity with lower relative humidity was observed in the palm Trachycarpus fortunei (Guarnieri et al., 2006) and in Pinus taeda (Bohrerova et al., 2009). These data suggest that timing of flowering within a season could have a large impact on reproductive success when water availability is variable. Pollen size may also play a role in how long pollen survives, and it has been shown in five species of aroids that the larger pollen grains significantly reduced viability, while the smallest pollen had the longest viability (Barabé et al., 2008). Finally, pollen morphology may also have an effect on the length of pollen longevity. In species with trinucleate pollen, viability typically lasts a very short time, whereas it lasts considerably longer in species that produce binucleate pollen (reviewed in Brewbaker, 1967; Lora et al., 2009). Collectively, these environmental variables may have an effect on mating success over the season, and the timing of floral events may be adjusted in response to changes in a dynamic environment where selective pressures may vary over time.
Pollen longevity and viability may also change as flowers age. There are abundant examples of this in the literature. For example, in the orchid species Calopogon tuberosus and Pogonia ophioglosssoides, pollen aged up to 8 d outside of anthers was just as effective as fresh pollen (Proctor, 1998). However, in Cypripedium reginae, 8-d-old pollen produced lighter fruits with fewer embryonated seeds than pollen aged 0, 1, 2 and 4 d (Proctor, 1998). In Gentiana pneumonanthe, pollen viability declined sharply after 4–5 d (Petanidou et al., 2001). Additionally, 4-d-old pollen from Clarkia unguiculata produced markedly fewer seeds than fresh pollen (Smith-Huerta and Vasek, 1984). In longer lived species the pattern appears to be similar. The tree Austrocedrus chilensis also demonstrated a significant increase in aborted pollen as the trees age (Aizen and Rovere, 1995).
Patterns of presentation of pollen to pollinators may also affect pollen longevity. Pollen presentation theory (Lloyd and Yates, 1982; Thomson and Thomson, 1992) suggests that if pollinators are unpredictable, all pollen should be released simultaneously. However, if pollinator service is predictable, anthers should open sequentially, releasing smaller aliquots of pollen over a longer period of time, and allowing access to pollen for a larger number of pollinators. For example, early in the season when pollinators are numerous, Chamerion angustifolium releases its pollen sequentially, but as the number of pollinators declines later in the season, pollen presentation becomes simultaneous (Sargent, 2003). Additionally, some plants, such as members of the genus Penstemon, appear to have adjusted pollen release depending upon their most effective pollinators (Castellanos et al., 2006; Thomson, 2006). Limiting pollen availability per visit may compensate for the presence of an inefficient pollinator (Thomson, 2006; Castellanos et al., 2006). Differences in how pollen is released and the effectiveness of pollinators probably affect mating. For example, if a plant releases all its pollen at once, but there are few ovules available to fertilize, reproductive success through male function is decreased.
Different timing of maturation and senescence of floral whorls can also lead to changes in functional gender of flowers. And these changes can be affected by the availability of pollinators. Several examples of this are documented among protandrous species. The staminate phase of Lobelia cardinalis flowers was found to be longer than the pistillate phase in the field with pollinators present (Devlin and Stephenson, 1985); however, the pistillate phase of Campanula americana was longer than the staminate phase in the greenhouse without pollen removal or deposition (Evanhoe and Galloway, 2002). In the genus Campanula, the staminate phase has been found to be shortened by pollen removal (Evanhoe and Galloway, 2002; Giblin, 2005) and lengthened with low levels of pollen removal (Richardson and Stephenson, 1989). The pistillate phase has been found to be shortened by pollen deposition (Evanhoe and Galloway, 2002) and lengthened by fertilization failure (Richardson and Stephenson, 1989).
Flowers that remain open for longer periods may have an increased probability of attack by fungal pathogens. These pathogens may dramatically reduce reproductive success. For example, the flowers of the Caryophyllaceae are infected by an anther smut known as Microbotryum violaceum (a basidiomycete), which is transmitted through pollinators (Shykoff et al., 1996). Flowers that remain open longer are more susceptible to infection because they receive an increased number of insect visits. In addition, if the infection occurs in flowers that remain on the plant longer the chance that the fungus will be spread from the infected flower to the rest of the plant increases. These data suggest that species with a greater risk of pathogen attack through floral structures may experience selection pressure for shorter floral life spans.
The relationship between flower longevity and pathogen risk may be altered over the season because the risk of floral infection depends somewhat on environmental conditions. The fungus Sphaerotheca macularis f. sp. fragariae exhibits maximum germination and growth on strawberry flowers within a narrow range of temperature, irradiance and humidity (Amsalem et al., 2006), and high humidity and temperature were also found to increase the rate and severity of infection of Freesia hybrida flowers by the fungus Botrytis cinerea (Darras et al., 2006). The risk of infection by the bacteria Erwinia amylovora was affected by both time and environment in apple blossoms, infection being highest between 1 and 3d following antithesis (Thomson and Gouk, 2003), but increased with increasing flower wetness (Taylor et al., 2003) and decreased with increasing temperature (Pusey and Smith, 2008).
As flowers age, the kinds of changes documented above may lead to changes in mating patterns. In several species self-compatibility varies as flowers age. For example, early in floral lifetime outcrossed pollen outperforms self-pollen in Gentiana pneumonanthe, but as flowers age, discrimination between the two becomes much less pronounced (Petanidou et al., 2001). This pattern has also been observed in Lilium longiflorum (Ascher and Peloquin, 1966), Leptosiphon jepsonii (Goodwillie et al., 2004; Goodwillie and Ness, 2005) and Campanula rapunculoides (Vogler and Stephenson, 2001). Delayed selfing (Lloyd, 1979) occurs at the end of the floral lifetime in a wide variety of species with specialized pollination syndromes (Fenster and Martén-Rodríguez, 2007), but particularly in situations where pollinator number or diversity is limiting (Carrió et al., 2008; Etcheverry et al., 2008). In the monocot Cladium jamaicense, the timing of flowering within and between clones, rather than pollinator limitation, was the driving force for delayed selfing (Snyder and Richards, 2005).
The mechanism for delayed selfing varies and includes changes in stigma receptivity (Vogler and Stephenson, 2001), relative position of the stigmas and anthers (Seed et al., 2006), timing of male and female expression (Snyder and Richards, 2005), and corolla abscission (Qu et al., 2007). The physiological interactions between pollen and stigmas may weaken as stigmas age (Haber and Frankie, 1982). And the presence of outcross pollen on the stigma may prevent these developmental changes. In Hibiscus trionum, 50 or more pollen grains on a stigma prevented selfing (Seed et al., 2006).
Evidence that delayed selfing increases reproductive output when pollinators are unavailable has been found in a variety of plants across a wide array of families (reviewed in Fenster and Martén-Rodríguez, 2007). These include perennial dicots (Vogler and Stephenson, 2001; Kalisz and Vogler, 2003; Qu et al., 2007; Carrió et al., 2008; Etcheverry et al., 2008; Martén-Rodríguez and Fenster, 2008; Vaughton et al., 2008; Zhang and Li, 2008), clonal monocots (Snyder and Richards, 2005) and annuals (Ramsey et al., 2003). However, delayed selfing was ineffective in assuring reproduction in the short-lived perennial Polygala lewtonii (Weekley and Brothers, 2006). Selfing can result in inbreeding depression, as was seen in Hibiscus trionum (Ramsey et al., 2003; Seed et al., 2006), and may only provide a net reproductive gain if it does not pre-empt opportunities for outcrossing (Vaughton et al., 2008).
When flowers remain open for longer periods of time, they may receive larger numbers of pollinator visits, which could increase the number of pollen donors that sire seeds within fruits. Whether pollen deposited later can sire seeds depends on the rate at which pollen arrives, the number of ovules available and the amount of competition between pollen tubes (Mulcahy et al., 1983; Snow, 1986; Spira and Snow, 1996; Burkhardt et al., 2009). Two studies with species of Mimulus suggest that multiple paternity does increase in flowers that remain open for longer periods. In Mimulus gutatus, flowers that were allowed to retain their corollas for longer periods had 30 % more mates per fruit (Dudash and Ritland, 1991). For Mimulus ringens, flowers that received multiple pollinator visits had 58 % more mates per fruit than flowers that received single visits (Karron et al., 2006). In contrast, for other species such as Raphanus sativus most seeds are sired by pollen deposited on the first visit (Marshall and Ellstrand, 1985), so increasing floral life span would not increase mate number.
Just as the age of the flower can affect the outcome of mating, the act of mating, pollination in this case, can affect the longevity of flowers. For example, in seven early flowering woodland herbs observed over 3 years, unpollinated flowers lasted longer than pollinated ones (Schemske et al., 1978). This pattern of pollination-induced senescence is also found in Rhododendron calendulaceum, in which flowers receiving supplemental pollen senesced 3 d earlier than naturally pollinated flowers (Blair and Wolfe, 2007). In most orchids, pollination (insertion of a pollinium), not pollinia removal, triggers floral senescence (Proctor and Harder, 1995; Clayton and Aizen, 1996; Martini et al., 2003; Stpiczyńska, 2003; Abdala-Roberts et al., 2007).
The size of pollen load received can also affect floral senescence. For example, in Cleistes divaricata var. bifaria, flowers that were pollinated with small pollen loads did not fade as quickly as those with larger loads (Gregg, 1991). In addition, duration of stigma receptivity is often longer when pollen loads are small (Waser and Fugate, 1986; Neiland and Wilcock, 1995; Richardson, 2004). Fertilization success can shorten floral life span. For example, in Campanula rapunculoides, floral longevity is greater when the number of fertilized ovules within an ovary is low (Richardson and Stephenson, 1989). Within species, self-fertilized flowers may have greater longevity than cross-pollinated flowers (Singh et al., 1992; Espinoza et al., 2002; Sato, 2002; Clark and Husband, 2007; Weber and Goodwillie, 2007), although both self and outcross pollinations shortened floral lifetime in Ceiba chodatii and Ceiba speciosa (Gibbs et al., 2004) and in Hibiscus trionum (Seed et al., 2006). If pollination shortens floral life then the total number of mates that might contribute pollen to a stigma may be reduced.
Wild radish flowers live for only a few days. There are no striking colour changes, but flowers often fade as they age. Stigmas are receptive from the time flowers open, and even before, as bud pollinations are possible. However, relative positions of anthers and stigmas may change as stigmas often become more exerted over time. Sporophytic self-compatibility develops as the flowers reach anthesis. Flowers can be successfully self-pollinated 2 d before opening, but once flowers are open they are and remain self-incompatible (Sampson, 1957; Hinata and Nishio, 1980; Pundir et al., 1983). Thus, the effects of floral age on mating will be more subtle than a change in self-incompatibility.
In wild radish, we investigated the direct effects of flower age on mating in only one experiment. We performed experimental crosses on newly opened and 1-d-old flowers. Young flowers produced significantly more seeds per fruit (mean 6·45) than older flowers (mean 5·52) (Marshall et al., 2007a), suggesting that fewer ovules were available for fertilization in the older flowers. In addition, there were slight changes in seed paternity. One of the three pollen donors used sired fewer seeds on old than on young flowers. And seed paternity was more equal among the three donors on older than on younger flowers. These differences were significant in a categorical analysis and suggest that older flowers might discriminate less among pollen donors or favour different pollen donors than younger flowers (Marshall et al., 2007a). An alternative explanation is that the environment of older stigmas provides a different selective regime than the environment of younger stigmas. We know of no other studies that have compared the success of compatible pollen donors among flowers of different ages.
Plant architecture changes from vegetative to reproductive over the course of the season. There is variation in timing and size of floral organs from the top to the bottom of a plant, and from branch tip to base. Diggle (1995) provides a thorough review of architectural effects, so here we focus only on how architecture changes over the season in plants. Generally fruits and flowers open earlier and are larger closer to the centre of the plant. In addition, lower flowers tend to have greater nectar production and longevity (Clayton and Aizen, 1996; Stpiczyńska, 2003). Flower number, flower size, fruit number, fruit maturation, number of seeds per fruit, mass of seeds per fruit and number of good ovules all decrease in more distal flowers. Usually, this is interpreted as a response to declines in resource availability over time. Indeed, the sink strength of fruits increases with size and age (Marcelis, 1996), while the distance from a resource reduces its availability. For example, more distal flowers were shorter lived in Cohniella ascendens (Abdala-Roberts et al., 2007). In contrast, the apical flowers of Iris fulva were the highest performing no matter when they emerged (Wesselingh and Arnold, 2003). In this species floral display was important for attracting pollinators and hormones or a strong sink may draw nutrients toward the apical flowers.
When plants reproduce within a season can impact reproductive success. As plants age physiological information about current size and demands of existing reproductive structures may affect future investment (Obeso, 2002). Earlier reproduction may leave fewer resources for maternal growth, and fewer accumulated resources at the time of reproduction. In contrast, earlier flowering could also allow for longer development time for seeds and more time within a season for offspring germination and growth. Although later reproduction may allow time for more resources to accumulate, time for seed maturation is shorter and predators and pathogens may have accumulated. Across a wide array of annuals (572 species), earlier flowering plants had higher seed mass (Bolmgren and Cowan, 2008). Thus, a trade-off between maternal size at reproduction and offspring size may exist. Environmental factors may also influence this relationship; for example, in Plantago lanceolata larger plants flowered first and had higher seed mass; large plants also had reduced predation by grasshoppers (Lacey et al., 2003). There were also differences in seed germination among early- and late-flowering plants that might have been due to seasonal changes in mating, and the authors suggest that differences in pollinator behaviour might have caused differential paternity of seeds. This hypothesis is supported across a variety of species in which later flowering plants tend to escape predation while early- or peak-flowering plants are favoured by pollinators (Elzinga et al., 2007).
As plants age, they may also change allocation to reproduction via male or female function and these changes are likely to affect fitness (Charnov, 1979, 1982; Charlesworth and Charlesworth, 1981). Female function is usually thought to be more ‘costly’ than male function. For example, in sexually dimorphic species females use proportionately more resources for reproduction than males (reviewed in Obeso, 2002; Bañuelos and Obeso, 2004) and females usually delay reproduction until they reach a larger threshold size than males and have lower flowering frequency than males (Delph, 1999; Rocheleau and Houle, 2001; Obeso, 2002). However, strong selection against a disproportionate cost of reproduction may be expected and female plants may exhibit compensatory mechanisms that offset their higher costs (reviewed in Obeso, 2002; Rakocevic et al., 2009) even in the face of resource limitations (Sakai et al., 2006). It has been suggested that the cost of male function may be low enough that it is negligible in large plants (Caesar and Macdonald, 1984; Goldman and Wilson, 1986; Rameau and Gouyon, 1991; Delesalle and Mooreside, 1995; Karlsson et al., 1996; Henriksson and Ruohomäki, 2000; Houle, 2001). However, significant costs of male function may be incurred (Rameau and Gouyon, 1991; Delesalle and Mooreside, 1995), particularly during flowering (Rocheleau and Houle, 2001) or when male function is accompanied by nectar production (Southwick, 1984; Ornelas and Lara, 2009), as evidenced by the fact that nectar can be reabsorbed following pollination (Koopowitz and Alejandro-Marchant, 1998). The relative costs of nectar production can be altered depending on the intensity of pollen receipt (Ornelas and Lara, 2009), which may change over the life of the plant or flower. In addition, because more visits from pollinators are required to complete male vs. female function, nectar production can differ in timing, duration or quantity in male (or male phase) vs. female (or female phase) flowers (Devlin and Stephenson, 1985; Carlson and Harms, 2006).
Because the costs of reproduction probably differ in amount and timing for male and female function, there may be differences in allocation to these functions over a season. Allocation to gender in hermaphroditic plants has received a tremendous amount of attention in the literature. In hermaphrodites, there may be differences in male and female phases of individual flowers (see ‘Changes in stigma receptivity as flowers age’) and there may also be changes in gender in sequential flowers of an inflorescence. The latter may occur in response to differences in pollen receipt, removal and deposition (‘mating environment’, Brunet and Charlesworth, 1995) and to differences in the resources available as a plant ages. Female function (ovule number, fruit number and/or seeds produced) typically declines from the basal to the distal end of an inflorescence and male function (pollen number) remains constant or only changes slightly (Ehlers and Thompson, 2004; Guitián et al., 2004; Guitián, 2006; Cao et al., 2007; Hiraga and Sakai, 2007). The decline in female function coupled with stable male function results in flowers becoming functionally more male later in the season. In contrast, for Aquilegia yabeana and Helleborus foetidus, both protogynous species, ovule number increased or remained constant and pollen number decreased in distal flowers (Huang et al., 2004; Guitián, 2006). These changes in functional gender may play an important role in reproductive success. Although many studies have examined how male and female function change over time, few have specifically addressed whether this is a successful strategy. To our knowledge, none has investigated whether changes in allocation that lead to functionally male flower production actually lead to increases in the number of seeds sired, a more direct measure of fitness.
Pollinator behaviour may influence changes in functional gender and can therefore affect mating patterns. For example, the staminate phase of Lobelia cardinalis, found in flowers at the top of inflorescences, had more nectar. Its hummingbird pollinators tend to forage from the middle to the tops of inflorescences (Devlin and Stephenson, 1985). Bees are known to forage from the basal to the distal end of inflorescences in both Digitalis purpurea and Chamaenerion angustifolium (Best and Bierzychudek, 1982; Galen and Plowright, 1985). The effect of these foraging patterns on mating depends on the functional gender of flowers along inflorescences.
For animal-pollinated plants, changes in the quality and quantity of pollinators over the season may also affect the outcome of mating. Specific pollinators may have preferences for particular flower morphologies (Ashman and Stanton, 1991; Brunet and Sweet, 2006), flower colours (MacPhail et al., 2007), flower positions (Devlin and Stephenson, 1985) and inflorescence sizes (Ishii and Sakai, 2001; Mitchell et al., 2004). These preferences may change the pattern of mating within and among populations. For example, in Aquilegia caerulea, outcrossing rates increased with greater hawkmoth abundance, but not with the increased abundance of any other pollinators (Brunet and Sweet, 2006). Similar results have been reported in Heloniopsis orientalis (Morinaga et al., 2003) and in Phyllodoce aleutica (Kameyama and Kudo, 2009). However, a shift in the relative abundance of pollinator types did not influence reproduction in other studies (Schmidt-Adam et al., 2000; Eckert, 2002).
Some pollinators are more effective than others at removing pollen from anthers or delivering pollen to receptive stigmas (Schemske and Horvitz, 1984; Young and Stanton, 1990b; Fishbein and Venable, 1996; reviewed in Herrera, 2000). Differences in foraging patterns are a factor in determining pollinator effectiveness (Herrera, 1987; Eckert, 2002; Karron et al., 2004; Young et al., 2007) and may influence mating patterns (reviewed in Mitchell et al., 2009). In general, large bees carry more pollen than butterflies, small bees or flies. However, they tend to visit flowers on fewer plants and fly shorter distances (Herrera, 1987). The quality of the pollen deposited on the stigma may decrease as a result of pollinators visiting fewer flowers and/or by delivering less pollen per visit (reviewed in Ashman et al., 2004).
As pollinators vary in their effectiveness, differences in pollinator assemblages over time may affect reproduction. Both the composition and the abundance of pollinators can change over time (Rush et al., 1995; Waser et al., 1996; Brunet, 2009). Thus, flowers produced at different times may experience dissimilar pollinator service. For example, in the alpine shrub Phyllodoce aleutica, the timing of pollinator availability during a season was largely responsible for variation in seed set and outcrossing rates (Kameyama and Kudo, 2009). Similar results were found in Rhododendron aureum where changes in the rate of selfing and seed set were associated with later flowering dates and a seasonal change in the amount and type of pollinators (Hirao et al., 2006). Even over short time periods, differences in pollinator assemblages have been shown to affect seed production and progeny quality. In Lavandula latifolia, a Mediterranean shrub, differing pollinator composition between early morning and afternoon time periods resulted in differences in seed production (Herrera, 2000). Flowers visited by small bees and butterflies in the early morning experienced greater pollinator visitation rates than those visited by large bees later in the day. The early pollinator visits produced more seeds per inflorescence and the resulting seeds were more likely to produce seedlings when planted in the field. Taken together, differences in pollinator composition, behaviour, abundance and effectiveness are likely to alter reproductive success across a season (Schemske and Horvitz, 1984; Ashman and Stanton, 1991; Conner et al., 1995; Ivey et al., 2003; Brunet and Sweet, 2006).
Changes in flowering phenology have been shown to result in differential paternity in several perennial herb species. In Primula sieboldii, a clonal herb, synchronous flowering was shown to increase pollen flow between close neighbours and to alter paternity. However, there was no overall effect of phenology on pollen flow distance because of high variability among genets (Kitamoto et al., 2006). In another species of Primula, P. cuneifolia, flowering took place over a 50-d period and there was a higher degree of kinship among co-flowering pairs of plants than among non-co-flowering pairs (Hirao and Kudo, 2008). The same effect of phenology has been observed in the annual crop species Zea mays, in which alteration of sowing date significantly delayed flowering and reduced the number of seeds fathered by genetically modified genotypes (Palaudelmàs et al., 2008).
Effects of phenology on paternity have also been observed in long-lived woody species. In two hybridizing species of ash which have early and late phenology but form hybrids with intermediate phenology, Fraxinus angustifolia and F. excelsior, respectively, isolation by time was shown to be a factor in maintaining the two non-hybrid phenotypes (Gérard et al., 2006b). Additional effects of phenology were observed on paternity. Individual trees were most successful at fertilizing ovules on trees that had overlapping but slightly later flowering, suggesting that earlier flowering plants may be functionally male (Gérard et al., 2006a).
In addition, temporal and spatial patterns may be confounded in many cases, especially when seed dispersal distance is short, resulting in individuals that are closer in space being more closely related and thus more likely to flower synchronously. Mating with closer neighbours could alter the total number of mating opportunities for individual plants, as has been observed in two herbaceous species of Primula (Kitamoto et al., 2006; Hirao and Kudo, 2008). Patchy spatial distribution of individuals with similar phenology has also been observed in the woody species Fraxinus (Gérard et al., 2006a). In the perennial rye grass Lolium perenne, distance from the maternal plant explained a significant proportion of the observed variation in paternity, but the addition of flowering time to a model comparing predicted and observed paternity did improve predictions by 0·77 % (van Treuren et al., 2006). A significant effect of the interaction between flowering time and distance was also detected in Pseudotsuga menziesii, with both early- and late-flowering phenotypes having higher relative mating success than those with intermediate phenology when they were located less than 20 m from a maternal tree, while males at greater distances showed higher mating success if their flowering time was intermediate (Burczyk and Prat, 1997).
Wild radish grows first as a rosette, storing resources in the roots and then bolts after a few weeks or a few months of growth. Once they have bolted, plants can reproduce for months, but they cannot make new, large rosette leaves or return to the vegetative growth form. Some plants reproduce early, maturing seeds quickly, which would be a successful strategy during a short season. Season length is often determined by the onset of summer drought. Other plants that remain vegetative for months accumulate more resources in the roots and can eventually produce many more flowers than plants that reproduce early. However, in a season that is truncated early, these plants may not reproduce at all.
This variation in reproductive stage over the season is demonstrated in an experimental garden study that examined the progeny of crosses among eight maternal plants and ten pollen donors. Seven weeks after planting, 28 % of the plants remained vegetative, 20 % had buds but no flowers and 52 % had flowers. The vegetative group included slow bolting and damaged plants, so is not informative. However, at the end of the study, plants that were flowering at 7 weeks weighed less and had less total reproduction (flowers plus fruits plus pedicels) than those that had been in the bud stage (above-ground mass: 61·8 and 69·2 g for early and late reproducers, respectively; total reproduction: 604 and 735 for early and late reproducers, respectively). In both cases the means were significantly different (F2,1052, P < 0·0001).
Reproductive stage was also measured in a separate experimental garden study of plants derived from crosses between three pollen donors and ten maternal plants. At 7 weeks, 16 % remained vegetative, 45 % had flowers but no fruits and 38 % had fruits. The plants that had fruits at 7 weeks tended to be smaller at harvest than those only producing buds and flowers at 7 weeks (Marshall et al., 2007a).
Variation in timing of flowering must also be important in natural populations of wild radish as seed production of plants through male and female function varied over time in the field (Devlin and Ellstrand, 1990; Devlin et al., 1992). Maximum seed initiation occurred 2 weeks prior to peak flowering (Stanton, 1987a, b; Devlin and Ellstrand, 1990; Ashman et al., 1993). Interestingly, peak pollinator foraging occurred later in the season, long past the period of greatest seed initiation time and after maximum flower production (Ashman et al., 1993).
As a season progresses, R. sativus experiences a decline in flower production both in the field and in the greenhouse, but the decline is not linear; there are oscillations during the season (Mazer et al., 1989; D. Marshall, pers. obs.). In addition to a decline in flower production, R. sativus also exhibits reductions in fruit and seed set (Mazer et al., 1989; Marshall and Oliveras, 2001) presumably due to limits on resources for the plants, as these plants are not typically pollen limited in the field (Stanton, 1987a, b). There are also declines in ovules (to a slight degree), pollen production and pollen size in both the field and the greenhouse (Stanton, 1987b; Mazer et al., 1989; Young and Stanton, 1990a; Ashman et al., 1993, Marshall et al., 2007a). Table 2 details reductions in a number of components of reproduction in a greenhouse experiment. Both male and female components of reproduction declined over the experiment.
Over a season, these changes in reproduction may lead to differences in functional gender. In the greenhouse, wild radish plants that had set many fruits showed a decline in fruit-set. Fruit-set declined from 85 % in the first quarter of the experiment to 65 % in the second and third quarters, and then to <50 % in the last part of the experiment (Marshall and Oliveras, 2001). Thus, many of the flowers were functionally male. However, it is possible for plants to regain female function after some of the developing fruits mature and no longer draw resources because plants can continue to make new inflorescence branches (D. Marshall, pers. obs.).
Changes in allocation to male and female function may also occur within flowers over the lifetime of the plant as production of pollen and ovules within flowers changes over the season (Young and Stanton, 1990b). These changes may vary among plants; in a greenhouse study, one plant showed the highest amount of pollen and the sharpest decline in ovules over time while another plant showed the opposite pattern (Young and Stanton, 1990a). Interestingly, the plant with the fewest pollen grains per flower also produced the largest pollen grains. In wild radish larger pollen grains are more likely to sire seeds in the basal region of fruits (D. Marshall, pers. obs.) and these embryos are less likely to be aborted (Marshall and Ellstrand, 1988). The plant with the fewest pollen grains per flower also produced the fewest ovules per flower and the most flowers overall (Young and Stanton, 1990a).
How might these characteristics of R. sativus affect mating? Wild radish is self-incompatible so changes in selfing over the season are not an issue. However, the identity of pollen donors that sire a plant's seeds may change over time due to changes in pollinator identity and/or behaviour (Stanton, 1987a, b; Stanton et al., 1989, 1991; Ashman et al., 1993), availability of mates and changes in ability of particular pollen donors to compete for ovules.
Over a season, variation in pollinator availability can change mating patterns because the pollinators have different foraging preferences and behaviours. Comparison of 15 genera of floral visitors to the closely related R. raphanistrum revealed that visitation rate, pollen removal and effect on seed set varied considerably (Sahli and Conner, 2007). Although that study showed no temporal variation in the pollinator fauna, other work with R. raphanistrum does show temporal change in the pollinator community (Rush et al., 1995). These seasonal changes in the pollinator fauna may affect mating because, in natural populations, pollinators show preferences for particular flower colours of wild radish. For example, honey-bees preferred white and yellow flowers while syrphid flies preferred pink flowers. Solitary bees appeared to prefer yellow or pink flowers (Stanton, 1987a; Stanton et al., 1991). In R. raphanistrum, nectar-feeding butterflies do not necessarily prefer a particular flower colour, but have been shown to be more efficient pollinators than bumble-bees (Conner, 1997). Additionally, in R. raphanistrum, pollinators preferred yellow flowers, and this translated into increased seed siring success for yellow-flowered plants (Stanton et al., 1986). Although pollinator preferences might affect the identity and amount of the pollen transported, there does not seem to be an effect on female fertility (fruit- and seed-set) in R. sativus or in R. raphanistrum, probably because they are not pollen limited (Stanton et al., 1986, 1991). Thus, seasonal effects of changes in the pollinator community are most likely to be on patterns of seed paternity.
Studies using isozyme markers for seed paternity of wild radish reveal that time in the season and plant age affect mating patterns. In the field, the relative ability of plants to produce seeds as seed and pollen parents was measured for six weeks. Seed production through male function declined in the middle of the season and then recovered later in the measurement period (Devlin and Ellstrand, 1990). Female fertility also declined because the available flowers had received pollen and were filling fruits while new inflorescences had not yet had time to fully develop. Early in the season, fewer flowers were produced (Devlin et al., 1992), but most of those flowers probably received pollinator attention as they were the only flowers present. Later in the season, it is more likely that pollinators may have missed flowers as the overall number of flowers increases dramatically.
The most direct observation of the effect of plant age on mating comes from a greenhouse study of R. sativus using ten maternal plants and three pollen donors (Marshall et al., 2007a). As plants aged, pollen donor performance changed, resulting in differences in seed paternity across young and old plants. The most striking difference was for pollen donor ‘A’ that sired 26 % of seeds on young plants and 62 % of seeds on old plants. Plants sired by this pollen donor flowered later and were larger in size at the end of a 10-week experimental garden study than those sired by the other pollen donors (Marshall et al., 2007a).
Changes in mating pattern across time were also seen in a study using four pollen donors of wild radish on 16 maternal plants. The pollen donors were used in four pairwise combinations and 20 replicates of each cross were performed across 3 months. Comparing the percentages of seeds sired by the more successful donor in each cross (Fig. 2), it is clear that seed siring success changed over time. The more successful donor usually showed a dip in performance in replicates 7–9 and then regained relative ability to sire seeds in the later replicates. This was a significant effect; the number of seeds sired varied with replicate in each of the four pairwise crosses (χ2 < 0·01 in all four crosses).
In the study mentioned above where components of reproduction varied over time, the proportion of seeds per fruit sired by the 12 pollen donors also varied across the time periods (ANOVA, pollen donor × time in experiment interaction, F11, 3402 = 1·93, P = 0·0317). The mean proportion of seeds sired per fruit by these donors when in competition with a pair of test donors (for an explanation of test donors see Marshall et al., 2007b) increased from 60 to 68 %. And the coefficient of variation in performance among the pollen donors declined from 13·9 to 7·6 %, suggesting that there was less discrimination among the 12 pollen donors.
Additional effects of plant age on mating are revealed in studies of the pattern of seed paternity along branches. One study examined patterns of seed siring success over a plant's life span for three pollen donors. Across time, variation in pollen donor success was lowest on very young maternal plants (abundant resources, no mating experience), slightly greater on older plants (many resources, some mating experience) and reduced again on the oldest plants (few remaining resources due to extensive fruit filling) (Marshall and Oliveras, 2001).
The conclusion that declining resource availability affects the outcome of mating in R. sativus is also supported by studies showing that stress on maternal plants alters mating patterns. Generally, when maternal plants are under stress, the degree of sorting among compatible mates is reduced. For example, in a study with four pollen donors where half of the 16 maternal plants were subjected to low water availability, the variance in the proportion of seeds sired among the pollen donors was 0·0078 on control maternal plants and 0·0049 on stressed maternal plants (Marshall and Diggle, 2001). However, in a separate study using three pairs of pollen donors, differences in pollen donor performance were greater on stressed than on control maternal plants in two of three pairs (Shaner and Marshall, 2003).
Additional evidence also suggests that mating experience (which accumulates over time) can affect the outcome of subsequent crosses. For example, when long series of crosses were performed along branches, fruits sired by a rare male on each branch had more and larger seeds than those sired by the common donor, regardless of the identity of the rare and common pollen donors (Marshall and Oliveras, 1990). This kind of selective fruit filling could not occur until several fruits are forming on branches. In addition, a more complicated series of crosses showed that the identity of the pollen donors that sire neighbouring fruits can affect the filling of a particular fruit, but this effect is weaker (Marshall and Oliveras, 2001).
More indirect evidence also suggests that changes in pollen availability over the season for wild radish could affect the outcome of mating. When pollen loads of two disparate sizes (23 vs. 200) were applied to wild radish plants, measurement of progeny grown in the greenhouse suggested that the outcomes of mating were different (A. Harbison, University of New Mexico, Albuquerque, USA, pers. obs.). The progeny of low pollen loads produced more pollen grains per flower in the greenhouse than the progeny of high pollen loads (75 300 pollen grains per flower in the progeny of low pollen loads vs. 71 200 pollen grains per flower in the progeny of high pollen loads (ANOVA, F1,762 = 4·08, P = 0·0438). This result is consistent with studies of dioecious plants in which low pollen loads result in the production of a higher proportion of male progeny than do high pollen loads (e.g. Taylor et al., 1999; Stehlik and Barrett, 2006; Stehlik et al., 2008).
The morphological and physiological changes that occur as flowers and plants age are fairly well known. These occur in response to both internal and external changes that develop over time. Less is known about the consequences of variation in the timing of reproductive events. The first level of consequence would be variation in the amount of reproduction; reductions in success through female function over time are often observed. However, if the mating pattern changes as plants age, then a second kind of consequence is possible. The quality as well as the quantity of offspring could be altered. Changes in the amount of selfing are known over time, usually increasing as plants and flowers age. However, more subtle variation in the mating pattern can also occur. For example, flowers that remain open longer may accumulate pollen from a greater variety of mates. Although studies of seed paternity are becoming more common (reviewed in Bernasconi, 2003), few of them have sufficient data to address changes in paternity over a season. This is worth addressing because data from wild radish show that the identity of the compatible donors that sire seeds differs among young and old flowers and young and old plants. Offspring sired by these pollen donors also show differences in growth across the season. Therefore, in order to understand whether the changes that occur as plants and flowers age affect fitness, we need more studies that measure variation in reproduction through both male and female function, more studies that examine changes in seed paternity over time and additional studies that include measurement of the effects of this variation on offspring success.