Introduction

Mast seeding or masting, the synchronous intermittent production of large seed crops in populations of perennial plants, is characteristic of many species, including tropical trees, temperate trees and herbs (Kelly 1994; Kelly et al. 2001; Kelly and Sork 2002). This phenomenon has attracted the interest of plant ecologists (Janzen 1971), with the majority of research addressing one of two distinct questions concerning mast seeding: (1) Which “ultimate factors” cause masting to occur? A large body of research has evaluated the evolutionary advantages or economies of scale of masting, revealing that individuals producing more flowers or seeds have lower costs per surviving offspring in years when other plants are also flowering or seeding heavily. The most accepted hypotheses are predator satiation and pollination efficiency, which have been demonstrated in many forests (Shibata et al. 1998; Rees et al. 2002; Kon et al. 2005). Several reviews have thoroughly addressed economies of scale (Kelly 1994; Kelly and Sork 2002; Piovesan and Adams 2005; Pearse et al. 2016). (2) How does masting occur? This question focuses on proximate factors involved in the physiological mechanisms associated with masting. In some studies, environmental cues have been found to correlate with mast seeding, including temperature (Koenig and Knops 2000; Kon and Noda 2007; Masaki et al. 2008; Kelly et al. 2013), rainfall (Piovesan and Adams 2001) and drought (Pérez-Ramos et al. 2010); however, environmental cues vary somewhat arbitrarily among studies, even within the same species. In the Japanese beech (Fagus crenata), for example, lower temperatures in the spring of the previous year may act as a cue for masting in the population on the Oshima Peninsula of Hokkaido, while high temperatures during the preceding summer were implicated for masting in populations in the Tohoku district of Honshu (Masaki et al. 2008).

Reproduction consumes substantial amounts of resources, which are often evaluated in terms of carbon currency (Obeso 2002). Focusing on annual patterns of individual resource storage and costs of reproduction, a resource budget (RB) model was first proposed by Isagi et al. (1997). The RB model assumes that an individual plant requires more resources to flower and fruit than it gains in a year, and therefore only flowers when a particular threshold amount of stored resource (LT) is surpassed. There are two key assumptions within the RB model at the individual level: (1) flower initiation is triggered when the levels of stored resources are higher than LT, and the amount of resources invested in flowering (Cf) exactly equals the surplus; (2) seed production consumes substantial resources (Ca), and the stored resources are depleted after seed production. When the ratio (Rc) of Ca to Cf, which is the determinant of reproductive behaviour in the RB model, is greater than 1, mast seeding occurs. Tree individuals within a population are linked via pollen supply and show a synchronous reproductive pattern in the absence of any environmental cues. Satake and colleagues further integrated this density-dependent pollen limitation into the RB model, referring to it as pollen coupling (Satake and Iwasa 2000, 2002; Satake and Bjørnstad 2008). The refined RB models incorporate the entire population in diverse dynamic behaviours: perfectly synchronized periodic reproduction, synchronized reproduction within a chaotic time series, clustering phenomena and chaotic reproduction without synchronization.

To what extent do these theoretical models predict the masting pattern in natural forests? Many studies have been conducted to collect empirical data to validate these RB models, most of which have focused on the assumption of resource depletion after reproduction. Although some studies have found evidence for non-structural carbohydrate (NSC) depletion (Miyazaki et al. 2002; Ichie et al. 2005), other studies indicate that reproduction has no effect on NSC (Hoch et al. 2003; Körner 2003). In this review, we discuss recent developments in these quantifications, where isotopic approaches that discriminate between stored NSC and current photosynthate have been used to elucidate the carbon sources used for reproduction (Hoch et al. 2013; Ichie et al. 2013; Han et al. 2016).

Following the publication of contradictory results based on empirical data validated with an NSC analysis about a decade ago, nutrients such as nitrogen (N) and phosphorus are now being examined, and the resultant empirical data show a depletion of nutrient resources after masting (Han et al. 2008a; Sala et al. 2012; Ichie and Nakagawa 2013; Han et al. 2014). In this review, we focus on N dynamics associated with mast seeding at organ and individual levels in temperate forests to examine the assumptions within the RB models. Synchronization among individuals was excluded from this review because two recent reviews discussed this issue in detail (Crone and Rapp 2014; Pearse et al. 2016); nevertheless, studies linking individual- and population-level behaviors are increasingly required to further our understanding of the proximate drivers of masting (Pesendorfer et al. 2016).

Carbon source for fruit production

Carbon constitutes about half the dry mass of plants and is considered their key resource. Carbon storage is a characteristic feature of most plants, particularly woody species, and the subject has been thoroughly reviewed with respect to chemistry, physiology, ecology and economics (Kozlowski and Keller 1966; Mooney 1972; Beck and Ziegler 1989; Chapin-III et al. 1990; Kozlowski 1992; Dietze et al. 2014; Hartmann and Trumbore 2016). The RB model was originally verified in relation to carbon currency and large amounts of experimental data have subsequently been collected. It is generally concluded that flowering, like leaf flushing, depends on stored NSC in masting trees, because both processes occur concomitantly in deciduous species. Most species seem to use NSC stored in branches for flowering, including fragrant snowbell (Styrax obassia) (Miyazaki et al. 2002), F. crenata (Han et al. 2016), European beech (F. sylvatica), common hornbeam (Carpinus betulus), sessile oak (Quercus petraea) (Hoch et al. 2003) and Borneo camphor (Dryobalanops aromatica) (Ichie et al. 2005). Flowering in some species seems to rely on NSC stored in trunks, for example in Guarumo peludo (Cecropia longipes) (Newell et al. 2002), or in roots, for example in Japanese cedar (Cryptomeria japonica) (Miyazaki et al. 2009).

Compared with flowering, the influence of seed production on NSC storage is inconclusive, as no NSC decrease was observed in some species (Hoch et al. 2003; Körner 2003) while organ-specific depletion detected in others (Miyazaki et al. 2002; Ichie et al. 2005). If NSC storage does not decline after seed production, this suggests that either NSC storage utilized for seed production is replenished quickly in the masting year or that the current photosynthate is directly allocated to seeds without the translocation of stored NSC. We used large-scale, continuous 13C labeling of mature deciduous trees in a temperate Swiss forest to investigate the extent to which fruit formation relies on NSC storage in three masting tree species: C. betulus, F. sylvatica and Q. petraea (Hoch et al. 2013). By exposing trees to 13C-depleted CO2 during eight consecutive years, stored NSC was strongly labeled to discriminate it from current-year photosynthate. Analyzing the 13C values in flowers and seeds in the first season following the termination of the 13C-depleted CO2 exposure, we demonstrated that seed maturation and winter bud formation depended primarily on current-year photosynthate, while flowering relied on stored NSC. We further investigated the extent to which immature fruits depend on NSC storage and/or current photosynthate in F. crenata in Japan (Han et al. 2016). By comparing seasonal variations in natural 13C values in leaves, shoots and fruits in both fruiting- and non-fruiting trees at a high temporal resolution, we found that NSC storage contributes to seed development until the middle of the growing season when the growth of woody organs and the cupules is complete, and thereafter current-year photosynthate provides the carbon for seed maturation. These results are consistent with the estimate of carbon age used for seed production (Ichie et al. 2013). Using a radiocarbon (14C) ‘bomb spike’ as a tracer of carbon substrate and age in fruits, Ichie et al. (2013) found that the carbon used to produce seeds was less than 1.4 years old in 10 masting deciduous tree species, and less than 1.0 years old in four species. It can be concluded, therefore, that stored NSC contributes to seed development earlier in the growing season when the vegetative organs are growing, and current-year photosynthate is the primary carbon source for seed maturation.

The extent to which stored NSC contributes to seed development depends on the abundance of fruits and the environmental factors that determine the amount of current-year photosynthate; thus, the dynamic balance between carbon demand and supply is the key determinant of the transition between carbon sources. This interpretation supports the concept of resource switching or resource accumulation (Kelly 1994; Pearse et al. 2016). In general, masting does not deplete NSC storage and carbon may not be a limiting factor, at least in woody species that possess a substantial NSC buffer. In this respect, the role of NSC in masting trees may differ from herbaceous species; for example, NSC controls flowering in the masting forb bitterroot milkvetch (Astragalus scaphoides) (Crone et al. 2009) and the annual herb Arabidopsis thaliana (Wahl et al. 2013). Reproduction occurs at the cost of the vegetative growth of leaves, branches and stems in woody species (Obeso 2002; Yasumura et al. 2006; Han et al. 2011, 2016); therefore, the cost of reproduction estimated using allometric methods may be underestimated because “somatic costs of reproduction” are not taken into account. These somatic costs may arise from energetic construction, the transport of metabolites and opportunity costs incurred through the losses of growth and competitive status as a result of resource allocation to reproduction instead of growth early in development (Reznick 1985; Ashman 1994; Bazzaz and Ackerly 2000; Obeso 2002; Weiner et al. 2009; Thomas 2011). It is unclear whether the trade-off between reproduction and vegetative growth results from a temporary imbalance in the carbon supply and demand, or in other resources such as N and phosphorus.

Nitrogen source for fruit production

Although the RB model was originally proposed for carbon resources, depletion of nutrient storage could also occur because seeds contain considerable amounts of protein. In this section, the seasonal dynamics of N, including its resorption, storage and remobilization, are summarized, and the current understanding of the influence of reproduction on N storage is reviewed with a focus on temperate woody species. Here, N storage is defined following Chapin et al. (1990): N is considered to be stored if “it can be mobilized in the future to support growth or other plant functions”. In this definition, a given compound may serve both storage and non-storage roles. For example, ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco), an essential photosynthetic protein, is also a major N store in leaves because it is recycled during leaf senescence (Millard and Grelet 2010).

Storage and remobilization of nitrogen

Nitrogen storage locations

To optimize growth and development when nutrients are limited, perennial plants have evolved special features at cellular, tissue, whole-plant and ecosystem levels (Millard 1996; Millard and Grelet 2010; Rennenberg and Schmidt 2010). At the whole-plant level, for example, N is withdrawn from senescing leaves in deciduous tree species in autumn and stored in woody organs, enabling it to be remobilized for new growth early in the next growing season to cope with the temporal uncoupling between nutrient supply from soils and growth demands. The locations and forms of stored N vary depending on leaf habit. In deciduous trees, woody organs are the main storage sites; young trees tend to store N in trunks and/or roots (Millard and Grelet 2010), while mature trees store N in branches, particularly young branches (Bazot et al. 2013; Han et al. 2014; Li et al. 2015). This storage distribution is generally considered advantageous because it reduces the distance required for N transport to sink organs (Staswick 1994; Tagliavini et al. 1997; Bazot et al. 2013). In this regard, reproductive organs are stronger N sinks than vegetative organs (Han et al. 2017).

Vegetative storage proteins (VSPs) represent the major form of N storage in the vegetative organs of both annual and perennial plants (Staswick 1994; Stepien et al. 1994). VSPs have been identified in the small vacuoles of inner bark parenchyma and xylem ray cells in the branches, main trunk and roots of woody species (Wetzel et al. 1989; Sauter and Cleve 1994; Tian et al. 2003, 2007; Wildhagen et al. 2010). In addition, N is also stored in the form of free amino acids, especially in roots (Geßler et al. 2004; Bazot et al. 2013). During dormancy, arginine is the most abundant amino acid in the bark and xylem sap (Geßler et al. 1998) because it contains four atoms of N per molecule, and thus is a suitable compound for N storage whilst limiting the associated carbon demand (Wildhagen et al. 2010).

In contrast, the main storage site for N in evergreen trees is the leaves (Fife and Nambiar 1984; Han et al. 2008b), often in the form of the photosynthetic protein Rubisco (Warren et al. 2003). Rubisco is considered a form of N storage because plants appear to synthesize more Rubisco than they need for carbon assimilation, particularly under N-replete conditions (Stitt and Schulze 1994). Specific adaptive traits can modify the storage site; for example, stem- or root-sprouting species generally allocate more N to the stems or roots, respectively, than non-sprouting species (Palacio et al. 2007).

Leaf senescence and internal N storage in autumn

Resorption of N from senescing leaves contributes to the build-up of stored N. Resorption efficiency, which is the percentage of original N withdrawn during senescence, ranges from 33 to 83% and is determined by diverse environmental factors, such as soil fertility and drought (Weih and Nordh 2002; Oleksyn et al. 2003; See et al. 2015), and internal factors, such as tree age and/or size (Gilson et al. 2014), genetic differences (Cooke and Weih 2005) and the functional group of the species (Aerts 1996; Killingbeck 1996; Eckstein et al. 1999). To date, there has been no definitive conclusion about whether reproduction influences leaf resorption efficiency in temperate trees, since there have been only three studies on two species: masting F. crenata (Yasumura et al. 2006; Han et al. 2014) and orchard-grown pistachio (Pistacia vera) trees (Rosecrance et al. 1998). Although reproduction was not found to influence leaf resorption efficiency in either F. crenata or P. vera, reproduction resulted in a decrease in the internal N storage pool because of an associated reduction in leaf biomass (discussed in the following sections).

Uptake from the soil also contributes to N storage, and has mostly been studied in orchard species and a few wild trees. In mature almond (Prunus dulcis) trees, the later an N fertiliser was applied during the growing season, the less it was used by the fruits and leaves, being instead allocated directly to internal N storage (Weinbaum et al. 1984). Autumnal N taken up from the soil and stored for growth during the subsequent spring has also been demonstrated in seedlings of peach (P. persica) and nectarine (P. persica var. nucipersica) (Tagliavini et al. 1999; Jordan et al. 2012). Moreover, reproduction is likely to affect the relative contribution of soil N uptake to the internal N storage pool; for example, in alternate-bearing pistachio trees, 61% of the N storage pool was derived from soil uptake in a non-reproductive year, whereas all the N was withdrawn from N storage in senescing leaves in a reproductive year (Rosecrance et al. 1998).

Remobilization in the following spring

In early spring, internally stored N is usually converted to asparagine and glutamine for remobilization to new growth in leaves and shoots to cope with the temporal uncoupling between nutrient supply from soils and growth demands (Millard et al. 2006). In a range of deciduous and evergreen species, N remobilization from internal storage typically occurs for 20–30 days before the roots actively take up N (Millard and Proe 1993; Tagliavini et al. 1997; Millard et al. 2001, 2006; Millard and Grelet 2010; El Zein et al. 2011). There is clear evidence demonstrating that flowering in orchard species relies primarily on the remobilization of internal N stores (Weinbaum and Van Kessel 1998; Policarpo et al. 2002; Cheng and Raba 2009); for example, using 15N labeling in 9-year-old walnut (Juglans regia cv. Hartley) trees, stored N was found to be the exclusive source for catkin production, and provided 64% of the N for mature fruits too (Weinbaum and Van Kessel 1998). In another case concerning 6-year-old apple (Malus × domestica) trees (Cheng and Raba 2009), internal N storage and the current season’s uptake from the soil each contributed about 50% to the total N in ripe fruits. These findings clearly demonstrate that internal N storage and recycling are important processes in supplying N for reproduction.

Influence of reproduction on N storage

It is clear from the above discussion that N storage and remobilization are ecologically and physiologically significant to trees. We now turn to discussing the influence of reproduction on N storage, especially examining the assumptions of the RB model, namely that reproduction depletes N storage.

Compared with NSC, there are only a handful of field studies that have addressed N depletion after reproduction (Han et al. 2008a; Sala et al. 2012; Ichie and Nakagawa 2013; Han et al. 2014). In F. crenata, both the N concentration in branches and the estimated whole-plant N storage pool decreased in the year following mast seeding (Han et al. 2014), indicating that masting depletes N storage. A similar depletion in branch N concentration was observed in F. sylvatica during a mast year, with additional depletions detected in the leaves (Han et al. 2011). In F. crenata, the N content in floral buds is twice that of the leaf buds, and the preferential allocation of N to seeds in masting years reduces its availability to form flower primordia (Han et al. 2008a). These results explain one of the main characters of masting: a good masting year is always followed by a poor masting year, according to two centuries of records for F. sylvatica (Hilton and Packham 2003), indicating that N is the limiting factor for mast seeding in Fagus species. Cones of the whitebark pine (Pinus albicaulis) mature over two years; in this species, N concentration was initially depleted in cone-bearing shoots in the masting year, and was depleted in all shoots at the end of the subsequent year, regardless of whether they bore a cone (Sala et al. 2012), suggesting individual-level N depletion caused by mast seeding. These reports provide indirect evidence to support the RB model with respect to resource depletion, which is consistent with the view that fruits deplete N stores in orchard trees (Birkhold and Darnell 1993; Muñoz et al. 1993; Rosecrance et al. 1998). In alternate-bearing pistachio trees, for example, heavy fruiting reduced the N storage pool to one-seventh of that of a non-reproductive year because there was no uptake from the soil to replenish the N storage in the reproductive year (Rosecrance et al. 1998).

Further direct evidence of the role of N in regulating mast seeding comes from experimental manipulation of resource levels (Miyazaki et al. 2014). In F. crenata trees, fertilizer addition increased N concentration in current-year shoots but the provision of soluble sugars or starch did not. Enhanced N concentration increased the expression of flowering genes, resulting in two consecutive years of flowering. For the same species, Abe et al. (2016) analyzed the resource contents in both floral organs (male inflorescence, cupule and pericarp) and fruits (seed), using both carbon and N currencies to estimate the Rc of the RB model. The estimated Rc using the carbon currency was less than 1, indicating an annual reproductive pattern, while the Rc of the N currency was higher than 1, indicating mast seeding. These results provide further evidence that, in F. crenata, N is a limiting factor in mast seeding but carbon is not.

Role of resource dynamics in mast seeding

The RB model assumes that floral initiation is triggered whenever the resource pool size surpasses a certain threshold and only the resource surplus is invested into flowering. Does resource level actually trigger floral initiation or is it merely a vital requirement for flowering and fruit production? The only empirical data about tree species is from the aforementioned study by Miyazaki et al. (2014). This study demonstrated that tree N status plays a role in regulating flower gene expression in F. crenata. This unique genomic-level study represents important first steps towards a more comprehensive understanding of the physiological mechanisms of the mast seeding phenomenon.

Other supporting data come from an alternate-bearing herbaceous legume; Crone et al. (2009) removed flowers from bitterroot milkvetch individuals to prevent seed set, which led to larger NSC storage and enabled them to flower again the following year. In contrast, plants that were allowed to set seeds had depleted NSC stores and did not flower the following year. This study provides evidence that NSC level is a key determinant of floral initiation. Overall, the dynamics of carbohydrates and N are involved in mast seeding, although the resource-type-specific roles remain unclear.

Conclusion and future prospects

Focusing on the key assumption of the RB models, N seems to be a more limiting resource than carbon for mast seeding in temperate tree species; however, it has only been investigated in a few species and deserves further study. Theoretically, masting could be driven almost entirely by either resource dynamics or by environmental cues (Kelly et al. 2013); in practice, however, specific conditions for both factors are required for masting to occur (Abe et al. 2016; Pesendorfer et al. 2016). In F. crenata, for example, the complex flowering dynamics of 170 trees observed over 13 years could be reproduced only when the interplay between N dynamics and temperature during floral initiation were taken into consideration (Abe et al. 2016). Plants may therefore respond to environmental cues only when a particular resource exceeds a certain threshold, as assumed in the RB model. How resources and environmental cues are linked in the underlying mechanism of mast seeding would be a valuable area for further investigation.

Although the carbon resource is less limiting than N in relation to mast seeding events in temperate woody species, does carbon cycling interact with N dynamics to impose an overall limitation on masting? We are far from being able to give a definitive answer since the RB model has only been empirically tested in very few species for both NSC and N. Overall, interdisciplinary approaches that combine molecular, physiological and ecological data would be very valuable in the elucidation of the mechanism driving mast seeding.