Resource manipulation through experimental defoliation has legacy effects on allocation to reproductive and vegetative organs in Quercus ilex

Abstract

Background and Aims

In plants, high costs of reproduction during some years can induce trade-offs in resource allocation with other functions such as growth, survival and resistance against herbivores or extreme abiotic conditions, but also with subsequent reproduction. Such trade-offs might also occur following resource shortage at particular moments of the reproductive cycle. Because plants are modular organisms, strategies for resource allocation to reproduction can also vary among hierarchical levels. Using a defoliation experiment, our aim was to test how allocation to reproduction was impacted by resource limitation.

Methods

We applied three levels of defoliation (control, moderate and intense) to branches of eight Quercus ilex trees shortly after fruit initiation and measured the effects of resource limitation induced by leaf removal on fruit development (survival, growth and germination potential) and on the production of vegetative and reproductive organs the year following defoliation.

Key Results

We found that defoliation had little impact on fruit development. Fruit survival was not affected by the intense defoliation treatment, but was reduced by moderate defoliation, and this result could not be explained by an upregulation of photosynthesis. Mature fruit mass was not affected by defoliation, nor was seed germination success. However, in the following spring defoliated branches produced fewer shoots and compensated for leaf loss by overproducing leaves at the expense of flowers. Therefore, resource shortage decreased resource allocation to reproduction the following season but did not affect sex ratio.

Conclusions

Our results support the idea of a regulation of resource allocation to reproduction beyond the shoot scale. Defoliation had larger legacy effects than immediate effects.

INTRODUCTION

Climate change is currently affecting the reproductive success of trees (Pérez-Ramos et al., 2010; Sánchez-Humanes and Espelta, 2011; Caignard et al., 2017; Bogdziewicz et al., 2020a) and their allocation of resources to reproduction (Monks et al., 2016; Gavinet et al., 2019). While the frequency of reproductive failures has been increasing in some species and areas (Bogdziewicz et al., 2020a), jeopardizing the natural regeneration of the forest, fruit production has been increasing in others (Caignard et al., 2017). Climate change effects on tree growth can be either positive due to CO₂ fertilization, nitrogen deposition (Fernández-Martínez et al., 2017) and the lengthening of the growing season (Menzel and Fabian, 1999; Piao et al., 2007; Delpierre et al., 2009), or negative because of more stressful conditions, especially in water-limited ecosystems, where aggravated droughts are expected (Dai, 2013; IPCC, 2013). These climate change effects are likely to affect both the carbon source through photosynthesis (Luyssaert et al., 2007; Biederman et al., 2016) and the carbon sink through cambial activity (Babst et al., 2013; Lempereur et al., 2017), and to modify carbon allocation to the different organs, especially the reproductive structures (Gavinet et al., 2019; Bogdziewicz et al., 2020b). Understanding strategies of resource allocation to reproductive functions versus other functions, as well as the environmental determinants of reproductive success, is therefore essential to grasping and predicting how the reproductive success of trees and regeneration will be affected by future climatic conditions.

Trade-offs in resource allocation arise from the fact that when limiting resources are allocated to one function, such as reproduction, they become unavailable for others (Bazzaz et al., 1987; Stearns, 1989; Obeso, 2002), thereby creating the need for priority rules of allocation (Suzuki, 2001; Wiley and Helliker, 2012). Reproduction being costly for plants, it might result in trade-offs with other functions, such as growth, survival, defence against pests and herbivores, and resistance to extreme abiotic conditions (Obeso, 2002; Barringer et al., 2013), as well as with subsequent reproductive efforts. In trees, the trade-off between growth and reproduction, generally studied at the stand scale, has been the focus of most investigations and has found some support in some cases (Han et al., 2011; Sánchez-Humanes et al., 2011; Martín et al., 2015; Vergotti et al., 2019), but not in all (Dick et al., 1990; Yasumura et al., 2006; Knops et al., 2007; Alla et al., 2012; Redmond et al., 2019; Vergotti et al., 2019).

Understanding how reproduction may be limited by other functions or, on the contrary, limit them is of particular importance in mast-seeding species. Mast-seeding species show years of massive and synchronous production of seeds that alternate with one or more years of negligible production. Fruit production during a mast year consumes a significant amount of resources (Janzen, 1971; Kelly, 1994; Kelly and Sork, 2002) as fruit biomass amounting up to 79 and 52 % of annual wood biomass production has been reported during a mast year in beech and sessile oak, respectively (Mund et al., 2010; Delpierre et al., 2016). Different mechanisms of allocation to reproduction have been proposed to explain mast seeding (Pearse et al., 2016). Among them, two have found some support in oaks. First, the resource depletion (or storage) hypothesis proposes that the tree’s reserves are depleted during mast years and that the tree needs to stock resources for several years before it can invest strongly in reproduction again (Sork et al., 1993; Sánchez-Humanes et al., 2011; Pearse et al., 2016). Second, the resource-switching hypothesis proposes that resources are shifted from vegetative growth to reproduction in mast-seeding years (Norton and Kelly, 1988; Hirayama et al., 2008; Sánchez-Humanes et al., 2011). Whichever hypothesis, or their combination, is correct, the nature of the limiting resource(s) inducing the mast-seeding behaviour remains unknown in many species (Han and Kabeya, 2017).

Studies investigating mast seeding are usually based on correlative analyses of fructification time-series at the tree or plot scale. Understanding the physiological mechanisms responsible for mast seeding, and more generally the allocation of resources to reproduction in trees, requires a deeper understanding of the regulation of reproduction all along the reproductive cycle from bud initiation to fruit maturation (Miyazaki et al., 2002; Bañuelos and Obeso, 2005; Ichie et al., 2013; Allen et al., 2017). The goal of this study was thus to better understand how allocation to reproduction is regulated at the different steps of the reproductive cycle, from flowering to seed germination.

A key question in understanding mast seeding, and more generally the inter-annual variation in tree fecundity, is which resource is most limiting to reproduction and how this resource is allocated to the different functions. Phosphorus, nitrogen and non-structural carbohydrates are known to be involved in the proximate mechanisms driving mast seeding (Han et al., 2011; Sala et al., 2012; Miyazaki et al., 2014; Allen et al., 2017; Han and Kabeya, 2017; Fernández-Martínez et al., 2019). Most of the carbon used for fruit growth, either in forest trees or in fruit trees, has been shown to derive from short-term photoassimilates mostly produced by leaves in the close vicinity of the fruit (Hasegawa et al., 2003; Volpe et al., 2008; Hoch et al., 2013; Ichie et al., 2013; Han et al., 2016). This suggests that foliated branches are autonomous for carbon for most of the growing season to produce the fruits, as proposed by the branch autonomy theory (Watson and Casper, 1984; Sprugel et al., 1991). However, other studies have also shown that individual fruit-bearing branches are sometimes unable to supply all the carbon required for the development of their fruits (Newell, 1991; Miyazaki et al., 2007; Pasqualotto et al., 2019), which suggests that whole-tree regulation or physiological integration among branches is sometimes necessary. Therefore, it seems that the scale of the regulation of carbon allocation to sexual reproduction in woody plants can range from the branch to the whole individual (Ushimaru and Genkai-Kato, 2011).

The degree of physiological autonomy of shoots for reproduction varies among species and situations (Henriksson, 2000; Hasegawa et al., 2003; Díaz et al., 2004; Sánchez-Humanes et al., 2011). Therefore, costs of reproduction may vary in contradictory ways if studied at the tree, the branch or the shoot level (Obeso, 1997), but few studies have examined reproduction costs and allocation trade-offs between reproduction and other functions at multiple hierarchical levels (Sánchez-Humanes et al., 2011; Alla et al., 2012; Barringer et al., 2013; Hossain et al., 2017). Furthermore, very few of these studies are explicitly related to mast seeding (Miyazaki, 2013). Here we aimed to bring new insights to the elucidation of the mechanisms of mast seeding by investigating the investment in reproduction at different physiological scales. More precisely, we aimed to investigate allocation to reproduction at the branch and shoot levels from fruit initiation to fruit initiation of the next season in a mast-seeding species.

One method of studying the allocation relationships between different sinks is to manipulate sink–source relationships (Iqbal et al., 2012; Bogdziewicz et al., 2020c). Manipulation of sink–source relationships by defoliation can help to determine whether branches or individual plants are able to compensate for the loss of photosynthetic capacity and nutrient storage to achieve their reproduction. Compensation may happen by either changing allocation from other functions to reproductive organs (growth, survival, storage, future reproduction) (Sprugel et al., 1991; Obeso, 1998; Hoch, 2005) or by increasing the photosynthetic activity of the remaining leaves in the case of carbon (Iqbal et al., 2012).

Most defoliation and herbivore exclusion experiments on temperate forest trees and fruit trees have shown a negative effect of defoliation on fruit set (Mehouachi et al., 1995; Obeso, 1998; Iglesias et al., 2003; Frioni et al., 2018) and on fruit size (Obeso, 1998; Hoch, 2005; Matsumoto et al., 2017). In oaks, most studies concerned natural herbivory and showed also negative impacts on fruit set and total fruit production (Crawley, 1985; May and Killingbeck, 1995; Hochwender et al., 2003; Nakajima, 2015; Pearse et al., 2015; Canelo et al., 2018). However, other studies have found no effect of herbivory on fruit set, fruit size or yield in other species (Obeso and Grubb, 1993; Mehouachi et al., 1995; Tamura and Hiura, 1998; Ezzahouani and Williams, 2003; Bañuelos and Obeso, 2005; Frioni et al., 2018; Pasqualotto et al., 2019). In some cases, overcompensation, i.e. a positive effect of herbivory or leaf removal (Agrawal, 2000; Iqbal et al., 2012), has been observed on fruit size in Vitis vinifera (Ezzahouani and Williams, 2003) and on flower production in Quercus ilex (Díaz et al., 2004). However, it is difficult to draw general conclusions from these studies because they differ in many respects: defoliation intensity, defoliation extent (branch or whole tree), moment of the reproductive cycle, targeted variables (fruit number, fruit set, total seed biomass, yield, etc.). Moreover, most studies have looked at the immediate effect of defoliation on ongoing reproduction, while few of them have explored the legacy effect of defoliation on resource allocation to different functions in the subsequent seasons (Noyce et al., 2016; Wiley et al., 2017). The majority of defoliation studies on forest trees have focused on deciduous species, which sustain lower leaf construction cost than evergreen species (Villar and Merino, 2001) and might be more tolerant of defoliation (Piper and Fajardo, 2014), but also have a shorter photosynthetic activity period. Determining what impact defoliation has on resource allocation in evergreen species is thus not straightforward.

Here, we studied the regulation of allocation to reproduction relative to other functions in the monoecious evergreen tree Q. ilex, which is the most widespread forest tree species in the Mediterranean Basin, using experimental defoliation to create a situation of resource limitation. The originality of our study is to cover the main steps of the reproductive cycle, at different scales within the tree and with a large range of defoliation intensities, from 0 to >80 % defoliation, and to quantify its impact not only on allocation to reproduction in the same year but also on allocation to vegetative and reproductive organs the following year. More specifically, we aimed to answer the following questions:

• (1) Are branches and shoots able to compensate for resource limitation to maintain fruit production (fruit survival, growth and germination potential) and production of vegetative and reproductive organs the following year?

• (2) Are there trade-offs of allocation at the shoot scale between reproductive and vegetative organs, between subsequent years and between male and female reproductive organs, and how are these trade-offs affected by resource limitation?

• (3) Are there some steps of the reproductive cycle that are more sensitive to resource limitation than others?

We applied three defoliation treatments (0 %, control; 50 %, moderate; 85 %, intense leaf removal) to six branches of eight trees and monitored the impact it had on allocation to reproduction at different steps of the reproductive cycle as presented in Fig. 1. First, we expected that shoots would be less able to compensate for loss of leaves necessary for fruit production in the intense defoliation treatment compared with the moderate defoliation treatment. Second, we expected reduced germination success for seeds produced by defoliated branches due to a reduced amount of reserves. Third, we expected strongly reduced, or even suppressed, production of flowers the following year on the defoliated branches because of priority investment in leaves to compensate for defoliation, as observed by Wiley et al. (2017) in Quercus velutina. Finally, since sex allocation theory predicts that female function is more expensive (Charlesworth and Morgan, 1991), we expected a shift towards maleness the following year on the defoliated branches.

Fig. 1.

Diagram summarizing the main results of this study. Each box corresponds to a statistical analysis and the numbers following dependent variables correspond to the model numbers in Tables 1–3. Only defoliation effects and significant covariables are shown here. The sign of the relationship between dependent variables and fixed effects is expressed as follows: NS, no significant difference from control; +, positive effect; −, negative effect.

MATERIALS AND METHODS

Study species and area

Quercus ilex is an evergreen wind-pollinated monoecious tree that usually flowers in May in the study area. The male inflorescences, called catkins, bear around 20–25 staminate flowers (Yacine and Bouras, 1997; Gómez-Casero et al., 2004) (Fig. 1). Catkins develop in the axils of lower leaves of the current-year shoot or in separate buds bearing only catkins. Female pistillate flowers mature a few days after staminate flowers and are located on an inflorescence at the upper part of the current-year shoot. Female inflorescences can bear one to six pistillate flowers (Fig. 1). Fertilization occurs in late June and early July, leading to fruit initiation. Fruits (acorns) achieve maturation in November–December (Yacine and Bouras, 1997).

The experimental plot was located in Montpellier, France (latitude 43.64°N, longitude 3.86°E, altitude 76 m). The climate is Mediterranean with an annual rainfall of 629 mm and mean annual temperature of 15.2 °C. The soil of the field site is a rendzina-like silty clay soil, with a pH of 8 and a depth varying from 150 to 200 cm. The trees used for the experiment were planted in 1998 from fruits collected from nearby natural populations. Trees were on average (± s.d.) 4.3 ± 0.3 m tall with a mean basal area of 124 ± 39 cm² during the experiment.

Experimental setting

In 2018, eight trees bearing fruits were selected for the experiment. On each tree, six branches bearing at least 10–15 initiated fruits (i.e. fertilized pistillate flowers) were selected in different and distant parts of the tree crown. The shoot is defined here as the growth unit of the current year (or the spring and summer growth units in cases when two growth flushes happened due to polycyclism). The branch refers to a ramified structure consisting of several shoots and including all the ramifications above the lowest one carrying at least one fruit (Fig. 2). The number of 2018 shoots per branch varied from 18 to 192 among all trees, and the number of fruit-bearing 2018 shoots per branch varied from four to 17 among all trees.

Fig. 2.

Schematic representation of the defoliation protocol.

For each tree, branches were randomly assigned to one of the three following treatments (two replicates per tree): no defoliation (control); moderate defoliation; and intense defoliation. Defoliation in the moderate treatment consisted of removal of half of the leaves and in the intense defoliation treatment it consisted of removal of all leaves except the uppermost leaf of each shoot, which led to removal of 85 ± 3 % of the leaves. For this last treatment, one leaf per shoot was left to allow minimal transpiration flow in the shoot (Figs 1 and 2).

Branches were defoliated between 21 and 27 June 2018, shortly after fruit initiation and once the leaves of the spring flush were mature. On 28 out of the 48 monitored branches, a second flush of leaves (called ‘summer flush’ hereafter) happened in early July. These shoots underwent the same defoliation treatment as their respective spring shoot in mid-July.

In order to estimate the mean individual leaf area for each tree, we randomly selected and scanned around 200 leaves per tree using the image analysis software ImageJ, and used the scan to calculate the total leaf area of every shoot.

Fruit growth monitoring and germination the year of defoliation

Fruit growth and survival between initiation and maturity were followed on a total of 930 acorns that were individually tagged and monitored. Up to six fruits could grow at the axil of the same leaf and this information was recorded for each fruit as a single fruit or two or more fruits growing at the axil of the same leaf. From July 2018 until November 2018, fruit survival and size was monitored six times [1 July (day 0), 23 August (day 53), 14 September (day 75), 5 October (day 96), 19 October (day 110) and 8 November (day 130)]. Fruit size was measured with an electronic calliper. During the early stage of fruit development, we measured fruit diameter because its shape is round (the embryo is invisible, inside the acorn cup). As soon as the shell protecting the embryo protrudes from the cup, we measured fruit length from the base of the cup to the top of the fruit excluding the remains of the style. Fruit mass at maturity was strongly correlated to fruit size (Supplementary Data Fig. S1B), so we assume that seed growth in mass and volume occurred simultaneously. Shoots that had died or were broken between July and November (3 %) were removed from the dataset so that fruit survival could be calculated without interference of shoot mortality (Canelo et al., 2018). After the last fruit size measurement in November, we monitored fruit maturation (desiccation) twice a week. Once they started turning brown, fruits were collected and weighed.

In order to determine their capacity to germinate, all collected fruits were placed in closed Magenta™ boxes filled with 18 g of vermiculite imbibed with 50 mL of distilled water and kept at 25 °C in the dark in germination stoves (LMS™). Fruits infested by insects (19 out of 247) were discarded, and germination success was recorded after 4 weeks if a radicle had emerged. Holm oak acorns are recalcitrant seeds, meaning that they are extremely sensitive to dehydration (Joët et al., 2013). For this reason, they have no dormancy and have to germinate rapidly as germination probability decreases with desiccation (Joët et al., 2016).

The number of leaves and the basal diameter of all the monitored 2018 spring shoots were counted and measured in December 2018 and January 2019, i.e. after fruit maturity and before the 2019 spring flush. In addition, on each of the 48 manipulated branches we tagged five 2018 shoots that had not borne a single fruit in 2018, on which we performed the same leaf and basal area measurements. Although we did not count the total number of leaves that 2018 shoots carried before defoliation, we checked that shoot basal area and the number of spring leaves were very strongly correlated in the control treatment (Supplementary Data Fig. S1A).

Mean leaf area per fruit for each 2018 shoot was obtained by multiplying the number of leaves per shoot by the mean leaf area for that particular tree, and dividing by the number of fruits that were initiated by the shoot in July.

Photosynthesis and predawn leaf water potential in the year of defoliation

In order to check the treatment effect on photosynthesis, we measured leaf gas exchange on current-year leaves adjacent to the fruits. We measured gas exchange on one leaf per treatment (i.e. three leaves per tree) of six trees, except the moderate defoliation treatment on the sixth tree due to accessibility constraints (17 leaves in total). Leaf gas exchange was measured on 4 July 2018, 1–2 weeks after defoliation and after the spring leaves had reached maturity. Measurements were carried out with two portable photosynthesis systems (Li-6400, Li-Cor, Lincoln, NE, USA) equipped with a light source (6200-02B LED, Li-Cor). Leaves were first acclimated in the chamber for >20 min at ambient temperature, ambient CO₂ concentration (400 ppm) and a saturating photosynthetic photon flux density (PPFD) of 1500 µmol m⁻² s⁻¹.

Tree water stress at the end of summer was determined by the predawn leaf water potential measured on 5 September and 4 October 2018 with a pressure chamber (PMS 1000, PMS Instruments, Corvallis, OR, USA). Leafy shoots were collected before sunrise and stored in airtight bags in a refrigerator until measurements 2 h later. All trees were sampled, including two leafy shoots per tree, and the difference between the two shoots never exceeded 0.2 MPa.

Shoot, leaf, flower and initiated fruit production the year following defoliation

On each of the 48 branches that had been manipulated in 2018, we randomly selected five 2018 shoots that had borne at least one fruit and on which at least one bud was starting to break. On each of these 2018 fruit-bearing shoots and on each of the tagged 2018 non-fruit-bearing shoots (produced either during the spring flush or the summer flush), we counted the number of 2019 spring shoots, the number of leaves per 2019 spring shoot, the number of catkins per 2019 spring shoot and the number of female flowers per 2019 spring shoot. Note that 2019 spring shoots could be composed exclusively of catkins (in this case there is no twig), of both leaves and flowers (catkin or female inflorescence) or exclusively of leaves.

For each 2018 shoot of seven of the trees (out of eight), we collected two catkins (if existing) out of all the catkins produced by 2019 shoots. On these catkins, we counted the total number of staminate flowers that they bore and we selected one staminate flower in the middle of each to count the number of stamens per flower.

The sex ratio of the shoots was calculated as the proportion of female flowers produced by the shoots out of the total number of inflorescences of the shoot (female flowers plus catkins).

The number of female flowers that had developed into fruits (fruit set) was counted on 17 July 2019. We could not monitor fruit growth and maturation in summer and autumn 2019 as an early extreme heat wave significantly damaged the leaves of the tagged shoots on 28 June 2019.

Statistical analyses

All statistical analyses and visual representations were conducted using the software R version 3.6.1 (2019) and the ggplot2 package (Wickham, 2009). We used the following packages for data analysis: lme4, car, multcomp and MuMIn (Hothorn et al., 2008; Bates et al., 2015; Fox and Weisberg, 2019; Bartoń, 2019).

We studied the effects of defoliation and additional covariates on reproductive allocation with 12 generalized linear mixed models (GLMMs) (Table 1, Fig. 1). Four of these models are dedicated to allocation to reproduction in the year of defoliation (2018), and eight are dedicated to allocation to reproduction the following year (Table 1, Fig. 1). All continuous covariables were standardized prior to analysis to compare model estimates between variables. The analyses of Models (5), (11) and (12) (number of 2019 shoots, 2019 sex ratio and 2019 fruit set, respectively) took into account the number of 2018 shoots with an offset. The analyses of Models (6), (7) and (10) (number of leaves, number of catkins and number of female flowers produced in 2019, respectively) took into account the number of 2019 shoots with an offset. We only included the interactions between defoliation and the other covariates in the complete models. For each response variable, we then applied a simplification of the model by sequentially removing the insignificant interaction terms, starting with the weakest and least significant interaction. We considered our level of significance to be P < 0.05.

Table 1.

Detail of the structure of GLMMs used to establish the effect of defoliation on the allocation to reproduction in the year of defoliation and the following year: response variable, distribution of the model, set of covariates and random effects, number of observations and residual degrees of freedom

Model	Response variable	Model distribution	Set of covariables1	Random effects	No. of observations (d.f.)
Allocation to reproduction in the year of defoliation (2018)
(1)	Fruit survival	Binomial	A	Tree|branch|shoot	918 (907)
(2)	Photosynthesis rate	Gaussian	None	tree	17 (12)
(3)	Mature fruit mass	Gaussian	A	Tree|branch|shoot	247 (233)
(4)	Fruit germination success	Binomial	B	Tree|branch|shoot	228 (221)
Allocation to reproduction the year following defoliation (2019)
(5)	No. of 2019 spring shoots per 2018 shoot	Poisson	C	Tree|branch	241 (230)
(6)	No. of leaves per 2019 spring shoot	Poisson	C	Tree|branch	241 (232)
(7)	No. of catkins per 2019 spring shoot	Poisson	D	Tree|branch	241 (230)
(8)	No. of staminate flowers per catkin	Gaussian	None	Tree|branch	190 (184)
(9)	No. of stamens per staminate flower	Gaussian	None	Tree|branch	190 (184)
(10)	No. of female flowers per 2019 spring shoot	Poisson	D	Tree|branch	241 (228)
(11)	Sex ratio of 2019 spring shoots	Binomial	D	Tree|branch	241 (228)
(12)	Fruit set on 2019 spring shoots	Binomial	E	Tree|branch	140 (133)

¹In addition to the effect of defoliation, the set of covariates shared by the models is indicated as follows: (A) 2018 shoot basal area, mean leaf area per fruit on 2018 shoot, summer flush in 2018 (yes/no), more than one fruit initiated per leaf (yes/no), minimum predawn potential of the tree (measured on 4 October); (B) fruit mass; (C) 2018 shoot basal area, summer flush in 2018; (D) 2018 shoot basal area, summer flush in 2018, number of mature acorns on 2018 shoot, number of leaves per 2019 shoot; (E) 2018 shoot basal area, number of leaves per 2019 shoot; None: no covariable added.

We used tree, branch and shoot as random factors. For measurements at the shoot scale (e.g. fruit survival), we considered tree, branch and shoot together as nested random factors. Thus, the model estimated the variability of the data due to the differences between trees, between the branches of a tree and between the shoots of a branch. For variables analysed at the branch scale (e.g. the number of leaves per shoot in 2019), we nested branches within tree as random factors. For random sampling at the tree scale (e.g. the rate of photosynthesis), we used trees as a random factor to take into account the between-tree heterogeneity (Table 1).

We fitted GLMMs with a Gaussian distribution for the response variables presenting a normal error structure, with a binomial distribution and its logit link function for dichotomous response variables and with a Poisson distribution and its log link function for count data (Table 1). The significance of the studied effects was determined through an ANCOVA of type II using a Wald χ² test followed by a Tukey post hoc test with Bonferroni–Holm correction to perform pairwise comparisons.

We completed our analyses by characterizing fruit length and survival rate at each measurement date over the course of the year for the different treatments of defoliation with GLMMs (Gaussian and binomial distribution respectively, treatment as the fixed effect and tree/branch/shoot as a random effect).

We checked for dataset homogeneity (Supplementary Data Table S1) and evaluated collinearities between variables by calculating the variance-inflated factor (Dormann et al., 2013). This index is calculated as the inverse of the proportion of variance specific to each explanatory variable. It was calculated as 1/(1 − R²), where R² is the coefficient of determination of the linear regression between a given explanatory variable and the remaining variables. Marginal (R²_m) and conditional (R²_c) coefficients were calculated with the package MuMIn according to Nakagawa’s method (Nakagawa and Schielzeth, 2013) to estimate the variance explained by fixed effects and fixed plus random effects, respectively.

RESULTS

Fruit growth showed a linear progression in 2018, although it slowed down strongly from early September to early October when drought stress was maximum (measured predawn water potential of −3.1 ± 0.2 MPa on 5 September and −3.8 ± 0.3 MPa on 4 October). Growth resumed after the first heavy rains of autumn, which occurred on 6 and 7 October (Fig. 3A, C). Less than half of the fruits reached maturity whatever the treatment (Fig. 3B). Like almost all the studied variables, mature fruit mass and fruit survival varied strongly between trees, branches (Supplementary Data Fig. S2) and shoots, which also explains the fact that the marginal R² (only fixed effects) was much lower than the conditional R² (fixed + random effects, here shoot nested in branch nested in tree) of models (1) and (3) [0.07 and 0.33 for model (1), 0.07 and 0.73 for model (3); Table 2]. Besides, fruit-bearing branches were usually larger than non-fruit-bearing branches (Supplementary Data Fig. S3).

Fig. 3.

Variation of (A) mean fruit size (± 95 % confidence interval, raw data), (B) fruit survival rate for each treatment (all trees pooled together, raw data) and (C) daily precipitation from 1 July 2018 to 7 November 2018. GLMM testing the effect of treatment alone as fixed effect and with Gaussian distribution for (A) and binomial distribution to test the survival of each acorn for (B): *significant difference between treatments; ns, no significant difference (P < 0.05).

Table 2.

Summary of GLMMs testing the effects of defoliation and other biological covariates on fruit survival, growth and germination, as well as photosynthetic rate in 2018. The table reports statistics, P-values and marginal (R²_m) and conditional (R²_c) R² for the reduced, final model. Bold font indicates significant effects (P < 0.05)

Model	Response variable	Predictor	Wald χ 2 (P-value)	R  2  m (R2c)
(1)	Fruit survival between initiation and maturation	Defoliation treatment	8.7 (0.01)	0.07 (0.33)
		2018 shoot basal area	0.3 (0.6)
		Mean leaf area per fruit on 2018 shoot	0.2 (0.7)
		Summer flush in 2018 (yes/no)	8.1 (0.004)
		More than one fruit initiated per leaf (yes/no)	4.0 (0.04)
		Minimum predawn potential	0.6 (0.4)
(2)	Photosynthesis rate	Defoliation treatment	0.3 (0.9)	0.01 (0.44)
(3)	Mature fruit mass	Defoliation treatment	4.1 (0.1)	0.07 (0.73)
		2018 shoot basal area	1.8 (0.2)
		Mean leaf area per fruit on 2018 shoot	0.005 (0.9)
		Summer flush in 2018	0.02 (0.9)
		More than one fruit per leaf	0.2 (0.6)
		Minimum predawn potential	0.9 (0.3)
		Defoliation × more than one fruit per leaf	9.5 (0.009)
(4)	Fruit germination success	Defoliation treatment	0.4 (0.8)	0.12 (0.51)
		Fruit mass	11.6 (<0.001)

Effect of defoliation on fruit growth, fruit survival, seed germination and photosynthetic activity

The effect of experimental defoliation on fruit survival depended on defoliation intensity: the moderate defoliation treatment had a negative effect on fruit survival compared with the control (coefficient parameter estimate ± s.e.: −0.67 ± 0.3), but not the intense defoliation treatment [Table 2, Model (1); Supplementary Data Table S2]. Fruit survival decreased with the second summer flush [Table 2 Model (1), −0.94 ± 0.3; Supplementary Data Table S2] but increased with the number of fruits growing in the axil of the same leaf [Table 2, Model (1), 0.37 ± 0.2; Supplementary Data Table S2]. The probability of fruit survival was not affected by 2018 shoot basal area, mean leaf area per fruit or minimum predawn water potential of the tree [Table 2, Model (1)]. Defoliation did not affect the photosynthetic rate measured between 1 and 2 weeks after defoliation [Table 2, Model (2)].

The interaction of defoliation with the number of acorns per leaf was significant [Table 2, Model (3); Supplementary Data Table S2]. Mature fruit mass increased with both moderate and intense defoliation only in the case of more than one fruit per leaf (0.22 ± 0.09 and 0.21 ± 0.09, respectively). Mature fruit mass was not affected by 2018 shoot basal area, mean leaf area per fruit, summer flush or minimum predawn water potential of the tree [Table 2, Model (3); Supplementary Data Table S2].

Mean germination rate was low in the three treatments (52, 52 and 56 % in the control and moderate and intense defoliation treatments, respectively) and strongly positively correlated to fruit mass [Table 2, Model (4), coefficient parameter estimate ± s.e., 0.84 ± 0.3]. Defoliation did not affect the seed germination rate [Table 2, Model (4); Supplementary Data Table S2].

Effect of defoliation on the production of vegetative and reproductive organs the following year

Shoots and leaves. The number of shoots produced per 2018 shoot in spring 2019 was higher on branches that had reflushed during the summer of 2018 [9.7 shoots on average on reflushed branches compared with 4.0 on branches with no 2018 reflush; Table 3, Model (5); Fig. 4A]. It was also positively correlated with 2018 shoot basal area whatever the treatment [Table 3, Model (5); Supplementary Data Table S3], although defoliation affected this relationship [Table 3, Model (5); Supplementary Data Table S3]. For an equivalent 2018 basal area increment, fewer shoots were produced during spring 2019, when the branch was intensely defoliated compared with control (−0.39 ± 0.08). The effect was in a similar direction in moderately defoliated shoots, although it was only marginally significant (−0.18 ± 0.09). Overall, the number of 2019 spring shoots per 2018 shoot was lower in the intense defoliation treatment (3.9 shoots on average compared with 5.3 shoots in the control treatment; Fig. 4A), while the number of leaves supported by 2019 spring shoots increased both with defoliation [Table 3, Model (6); Fig. 4B; Supplementary Data Table S3] and with the basal area of 2018 shoots [Table 3 Model (6), 0.08 ± 0.02].

Table 3.

Summary of GLMMs testing the effects of experimental defoliation and other biological covariates on growth parameters in 2019. The table reports statistics, P-values and marginal (R²_m) and conditional (R²_c) R² for the reduced, final model. Bold font indicates significant effects (P < 0.05)

Model	Response variable	Predictor	Wald χ 2 (P-value)	R  2  m (R2c)
(5)	Number of 2019 spring shoots per 2018 shoot	Defoliation treatment	12.0 (0.002)	0.40 (0.63)
		2018 shoot basal area	103.3 (<0.001)
		Summer flush in 2018 (yes/no)	23.7 (<0.001)
		Defoliation × 2018 shoot basal area	29.0 (<0.001)
		Defoliation × summer flush in 2018	8.0 (0.02)
(6)	Number of leaves per 2019 spring shoot	Defoliation treatment	22.7 (<0.001)	0.21 (0.90)
		2018 shoot basal area	26.9 (<0.001)
		Summer flush in 2018	3.0 (0.08)
		Defoliation × summer flush in 2018	17.0 (<0.001)

Fig. 4.

Effects of defoliation treatment and the second growth flush during summer 2018 on (A) number of 2019 spring shoots produced per 2018 shoot on a log₁₀ scale and (B) number of leaves per 2019 spring shoot. Black points indicate group means and grey points represent raw data. Different letters correspond to treatment effect in pairwise comparisons using Tukey’s post hoc test in the GLMM [Table 1, Models (5) and (6), respectively] (P < 0.05). Within treatments: *significant effect of the summer flush in 2018; ns, no significant difference (Tukey post hoc test with Bonferroni–Holm correction, P < 0.05).

Male flowers. The number of catkins per 2019 spring shoot was lower in the moderate defoliation treatment than in the control, and even more so in the intense defoliation treatment [Table 4, Model (7); Fig. 5A; Supplementary Data Table S4]. Thus, defoliation reduced the number of catkins and it also reduced the number of staminate flowers per catkin [Table 4, Model (8), −2.8 ± 1.2 for moderate defoliation, −3.7 ± 1.2 for intense defoliation] but had no effect on the number of stamens per staminate flower [Table 4, Model (9); Supplementary Data Table S4]. The 2019 spring shoots carried by large 2018 shoots produced fewer catkins [Table 4, Model (7); coefficient parameter estimate ± s.e., −0.05 ± 0.02]. Catkin number was negatively related to leaf number in the moderate and intense defoliation treatments but not in the control treatment (−0.33 ± 0.05 for moderate defoliation, −0.35 ± 0.07 for intense defoliation; Supplementary Data Table S4).

Table 4.

Summary of GLMMs testing the effects of defoliation and of other biological covariates on flowering and fruit set parameters in 2019. The table reports statistics, P-values marginal (R²_m) and conditional R² (R²_c) for the reduced, final model. Bold font indicates significant effects (P < 0.05)

Model	Response variable	Predictor	Wald χ2 (P-value)	R  2  m (R2c)
(7)	Number of catkins per 2019 spring shoot	Defoliation treatment	31.8 (<0.001)	0.43 (0.83)
		2018 shoot basal area	4.8 (0.03)
		Summer flush in 2018 (yes/no)	1.6 (0.2)
		Number of mature acorns on 2018 shoot	0.1 (0.3)
		Number of leaves per 2019 shoot	37.9 (<0.001)
		Defoliation × leaves per 2019 shoot	17.9 (<0.001)
(8)	Number of staminate flowers per catkin	Defoliation treatment	10.5 (0.005)	0.05 (0.62)
(9)	Number of stamens per staminate flower	Defoliation treatment	0.01 (1.0)	0.001 (0.52)
(10)	Number of female flowers per 2019 spring shoot	Defoliation treatment	12.6 (0.002)	0.40 (0.76)
		2018 shoot basal area	3.9 (0.05)
		Summer flush in 2018	1.0 (0.3)
		Number of mature acorns on 2018 shoot	23.4 (<0.001)
		Number of leaves per 2019 shoot	89.2 (<0.001)
		Defoliation × 2018 shoot basal area	7.9 (0.02)
		Defoliation × number of mature acorns	11.7 (0.003)
(11)	Sex ratio	Defoliation treatment	2.5 (0.3)	0.12 (0.32)
		2018 shoot basal area	28.4 (<0.001)
		Summer flush in 2018	1.0 (0.3)
		Number of mature acorns on 2018 shoot	13.6 (<0.001)
		Number of leaves per 2019 shoot	59.8 (<0.001)
		Defoliation × leaves per 2019 shoot	20.1 (<0.001)
		Defoliation × summer flush in 2018	8.9 (0.01)
(12)	Fruit set in 2019	Defoliation treatment	2.3 (0.07)	0.05 (0.35)
		2018 shoot basal area	0.1 (0.5)
		Number of leaves per 2019 shoot	5.1 (0.02)

Fig. 5.

Effect of defoliation treatment on (A) the mean number of catkins and (B) the mean number of female flowers per 2019 spring shoot. Black points indicate group means and grey points represent raw data. Different letters indicate significant differences between treatments [Table 1, Models (7) and (10), respectively]. The statistical significance threshold is 0.05. For each group, the shaded area represents the right side of the kernel density plot.

Female flowers. The number of female flowers per 2019 spring shoot was reduced in both the moderate and the intense defoliation treatment [Table 4, Model (10); Fig. 5B]. The number of female flowers was positively correlated to the number of leaves in all treatments [Table 4, Model (10); coefficient parameter estimate ± s.e., 0.66 ± 0.07]. The number of female flowers was also positively related to the 2018 shoot basal area in the control and intense defoliation treatments [0.25 ± 0.11 and 0.09 ± 0.04, respectively; Table 4, Model (10)]. The number of female flowers was negatively related to the number of fruits produced the year before in both the moderate and the intense defoliation treatment [−0.46 ± 0.11 and −0.47 ± 0.12, respectively; Table 4, Model (10); Supplementary Data Table S4], but not in the control treatment.

Sex ratio

The sex ratio (defined here as the proportion of female flowers produced by the shoot out of the total number of inflorescences of the shoot, i.e. female flowers plus catkins) was clearly biased towards male on all trees (13, 21 and 21 % of female flowers in the control, moderate and intense defoliation treatments, respectively; Supplementary Data Fig. S4A). On average the sex ratio was not affected by defoliation because both female and male flower production was decreased on defoliated branches [Table 4, Model (11); Supplementary Data Table S4]. The sex ratio was positively related to the 2018 shoot basal area (coefficient parameter estimate ± s.e., 0.28 ± 0.05) and negatively related to the number of mature fruits produced in 2018 [−0.27 ± 0.07; Table 4, Model (11)]. The interaction term between treatment and a second flush in summer 2018 was significant [Table 4, Model (11); Supplementary Data Table S4], with an effect of intense defoliation significant and positive only in the absence of a summer flush.

There was no relationship between female flower production and catkin production per 2019 spring shoot in the control and intense defoliation treatments, but a negative relationship appeared in the moderate defoliation treatment (Supplementary Data Fig. S5).

Fruit set

Fruit set in 2019 (proportion of female flowers that succeeded in initiating a fruit) was relatively low (33, 48 and 30 % in the control and moderate and intense defoliation treatments, respectively; Supplementary Data Fig. S4B). Fruit set did not differ between treatments [Table 4, Model (12)] and was not affected by 2018 shoot basal area [Table 4, Model (12)]. However, fruit set was positively related to the number of leaves produced in 2019 [coefficient parameter estimate ± s.e., 0.40 ± 0.18; Table 4, Model (11)].

Interactive effect of defoliation and reproductive status at defoliation

Defoliation had similar effects on shoots on defoliated branches that did not bear a single fruit in 2018 and shoots that bore a fruit in terms of shoot production and catkin production the following year (Supplementary Data Table S5, Supplementary Data Fig. S6A, C). Leaf production per 2019 shoot was higher for fruit-bearing shoots compared with non-fruit-bearing shoots in the control and moderate defoliation treatments, but not in the intense defoliation treatment (Supplementary Data Table S5, Supplementary Data Fig. S6B). Female flower production in 2019 was higher on 2018 fruit-bearing shoots compared with non-fruit-bearing shoots in the control treatment, but this difference disappeared in the defoliated treatments (Supplementary Data Table S5, Supplementary Data Fig. S6D).

DISCUSSION

In this study we investigated the impact of resource limitation induced by defoliation on the main steps of the reproductive cycle, from flower bud development to seed germination (Fig. 1). We found that defoliation applied shortly after fruit set had limited effects on fruit development, had no effect on fruit final mass and germination success, and did not upregulate photosynthesis. In the following spring, we found that defoliated branches produced fewer shoots, fewer flowers per shoot and more leaves per shoot. We also found negative relationships between staminate flower and leaf production in defoliated treatments, as well as between fruit production and subsequent flower production. Finally, we found that defoliation did not affect the sex ratio the following year.

Branches upregulate leaf production relative to flowers following defoliation

As we expected, defoliation reduced the number of catkins per shoot, the number of staminate flowers per catkin and the number of female flowers per shoot in the following year, while it increased the number of leaves. This allocation shift did not impact the number of stamens per flower, which might be more developmentally constrained. We observed a reduction, but not an interruption, of female flower production the year following defoliation. This contrasts with findings by Wiley et al. (2017) after whole-tree defoliation in Q. velutina, probably because, in our case, defoliated branches could rely on resources from non-defoliated branches further away.

The intense defoliation induced a reduction in the number of spring shoots the following year, consistent with the usually observed growth reduction after experimental defoliation in Q. ilex and other species (Vanderklein and Reich, 1999; Piper and Fajardo, 2014; Schmid et al., 2017; Wiley et al., 2017), although an increase in shoot production after defoliation has sometimes been observed (Cherbuy et al., 2001). Quercus ilex has preformed buds (Montserrat-Martí et al., 2009), which are already formed at the time of defoliation. Thus, our experimental defoliation could not affect bud set, but might have affected the allocation of resources to buds during their development from the summer that followed defoliation to the next spring. The 2019 spring shoots bore on average more leaves in the defoliation treatments compared with control, consistent with previous studies on Q. ilex (Cherbuy et al., 2001; Schmid et al., 2017) and other species (Iqbal et al., 2012; Nakajima, 2015). Quercus ilex is an evergreen species in which two or three cohorts of leaves usually coexist, and which has been shown to compensate for leaf loss in the following year (Cherbuy et al., 2001; Limousin et al., 2012). Therefore, a larger allocation to vegetative organs was expected to compensate for the previous leaf loss.

We observed that branches favoured the completion of fruit development during the year of defoliation, but favoured the production of leaves relative to flowers the following year. Our results at the branch level are consistent with the few studies that investigated the impact of resource manipulation on the different reproductive steps. In Q. velutina, Wiley et al. (2017) observed that production of second-year acorns was not significantly reduced after whole-tree defoliation but that production of flowers was suppressed the following year on the defoliated branches because of priority investment in defoliation recovery. We could hypothesize that in the event of resource limitation preferentially takes place before flower initiation and/or development than after a significant amount of resources, especially nutrients, has already been invested in flowers. As both carbohydrate and nitrogen availabilities have been identified as potentially involved in the initiation of flowers (Miyazaki, 2013; Miyazaki et al., 2014), such a regulation might be linked to the carbohydrate and nitrogen content of the branch.

To sum up, defoliated branches compensated for leaf loss the year following defoliation by increasing the number of leaves per shoot consistently with defoliation intensity. The production of shoots, staminate flowers and female flowers was, however, reduced compared with the control treatment. The loss of resources induced by defoliation was compensated primarily for achieving fruit development during the year of defoliation, and for recovering the leaf area in the following year.

Compensation for resource limitation and branch autonomy

Branch defoliation had contrasted effects on fruit and seed development during the months that followed (Fig. 1). First of all, defoliation, either moderate or intense, did not decrease the mass of the mature fruits, contrary to our expectations. Second, fruit abortion increased with moderate defoliation but not with intense defoliation. Third, defoliation had no effect on seed germination success, which was only positively related to fruit mass, as previously observed in other oak species (Bonfil, 1998; Huerta-Paniagua and Rodríguez-Trejo, 2011; Sánchez-Montes de Oca et al., 2018; Shi et al., 2019). Results of the moderate defoliation treatment are consistent with the generally negative impact of natural herbivory on oak fruit production, even though natural herbivory intensity is usually, but not always, less than removal of half of the leaves (Crawley, 1985; May and Killingbeck, 1995; Pearse et al., 2015; Nakajima, 2015; Canelo et al., 2018). The absence of the effect of intense defoliation on fruit production is, however, more surprising but is consistent with the recent studies of Pasqualotto et al. (2019) and Wiley et al. (2017) on hazelnut development and oak acorn production (initiated before defoliation) in Q. velutina, respectively. However, our results contrast with those of the few experiments that tested how different defoliation intensities impacted fruit production, and either found an increasingly negative effect with increasing defoliation intensity (Mehouachi et al., 1995; Kaitaniemi et al., 1999; Hoch, 2005) or no effect of defoliation at all (Obeso, 1998; Tuomi et al., 1988; Tamura and Hiura, 1998). Our results suggest that at the level of the branch the compensation for resource loss in order to maintain fruit growth might differ depending on defoliation intensity.

The branch autonomy theory suggests that foliated branches are carbon-, but not nutrient-, autonomous for most of the growing season (Watson and Casper, 1984; Sprugel et al., 1991). At the shoot scale, acorn growth in oaks is thought to be mainly based on carbohydrates produced during their development by the adjacent leaves (Hoch et al., 2013; Ichie et al., 2013; Fernández-Martínez et al., 2015) and also by the photosynthetic cells of their own pericarp until it dehydrates (Hoch and Keel, 2006). At the branch scale, shoot-bearing fruits are known to obtain part of their resources from adjacent shoots that do not bear fruits (Miyazaki et al., 2007; Sánchez-Humanes et al., 2011; Xie and Guo, 2015). In Q. ilex, Alejano et al. (2008) observed that acorns from the southern side of trees were significantly heavier than those at other positions, suggesting a local regulation of resource allocation to fruits.

Fruit growth in the intense defoliation treatment was similar to that in the control, although we did not observe any increase in photosynthetic rate in the remaining leaves to compensate for assimilate loss, consistently with most previous studies on oak saplings (Lovett and Tobiessen, 1993; Vanderklein et al., 2001; Wiley et al., 2013). Therefore, carbon used for fruit filling either originated from recent photoassimilates produced by non-defoliated branches further away (Oitate et al., 2011), from local reserves in the shoot and its vicinity, or from reserves in distant storage organs such as the stem and below-ground parts. Additional measurements of the amount of stored non-structural carbohydrates in some small branches at the time of defoliation suggest that they would not be sufficient to fill all the acorns that reached maturity (results not shown). This suggests that the carbon contained in the mature fruits of highly defoliated branches probably came from further away, although girdling experiments would be necessary to strictly assess branch autonomy for fruit filling. Our experiment shows, however, that the higher fruit abortion rate on moderately defoliated branches results from a branch allocation strategy rather than from impossibility of sustaining fruit development despite lower leaf area. The year 2018 was an intermediate year in terms of fruit production for Q. ilex in the area, i.e. neither a mast-seeding year nor a year with considerable reproductive failure (data not shown). If the same experiment had been performed during a mast year, the effect of defoliation might have been stronger because of increased competition for carbon between fruits. More experimental work is needed to understand how fruiting intensity might interact with the defoliation effect by repeating the experiment in multiple years, and by coupling them to chemical analysis of both carbohydrate and nutrient reserves in branches in order to understand the physiological basis of allocation regulation.

Interestingly, shoot basal area did not affect mature fruit mass in 2018 and fruit set in 2019, contrary to what has been observed in some fruit trees (George et al., 1996). The mean leaf area per fruit did not affect fruit mass either, consistent with observations at the branch scale in hazelnut trees (Pasqualotto et al., 2019). Therefore, fruit development might not be as dependent on local leaf photosynthesis as we initially expected. The ability of the branches to rely on distant sources of carbon to sustain fruit development might explain the absence of a clear defoliation effect in our experiment.

The production of shoots and flowers the year following defoliation appeared to be resource-limited. The limiting resources might have been carbon, as leaf removal prevents local production of non-structural carbohydrates. However, leaf and flower production might depend more on nutrients, such as phosphorus and nitrogen, that are known to be involved in the proximate mechanisms driving mast seeding (Han et al., 2011; Sala et al., 2012; Miyazaki et al., 2014; Allen et al., 2017; Han and Kabeya, 2017; Fernández-Martínez et al., 2019). As evergreen species store part of their nitrogen and phosphorus reserves in their foliage (Chapin et al., 1990; Cherbuy et al., 2001), trees lose a substantial portion of their nutrients with defoliation (Millard et al., 2001; Iqbal et al., 2012), although Q. ilex also stores a non-negligible part of the nutrient reserves in the shoots (Palacio et al., 2018). Defoliation treatment did not affect fruit set the following year. Besides, other factors known to impact fruit set in Q. ilex, such as pollen limitation or spring drought (Bogdziewicz et al., 2017), probably had minor impacts on fruit set in 2019 because spring was neither particularly dry nor particularly rainy during pollination.

Increasing resource limitation generates allocation trade-offs

Resource allocation trade-offs in trees may exist between vegetative growth and reproduction (Obeso, 2002; Barringer et al., 2013), between current and future reproduction as currently assumed in mast seeding species (Koenig and Knops, 2000) and between male and female flowering in monoecious species (Charlesworth and Morgan, 1991). In this study, we found no clear evidence for a trade-off between vegetative growth and female reproduction at the shoot scale. On the contrary, we found that shoots that had initiated a fruit in 2018 were larger than those that had not, as already observed in Q. ilex (Sánchez-Humanes et al., 2011; Alla et al., 2012), and that large 2018 shoots produced 2019 shoots with more leaves and more female flowers. During morphogenesis, there might be a developmentally constrained positive relationship between growth and female function at the shoot scale, probably because large shoots can provide more nutrients to fruits. However, the summer flush in 2018 had a negative impact on fruit growth, which suggests that the summer shoot growth was a competing resource sink for fruits, and that there is a trade-off between shoot growth and fruit production when the two occur simultaneously.

The resource budget model hypothesis predicts that resources would be depleted after fruiting, which would limit the number of flowers produced the following year (Crone et al., 2009). A negative correlation between seed production one year and the next has indeed often been observed at the tree level in oaks (Sork et al., 1993; Kelly and Sork, 2002; Pérez-Ramos et al., 2010). Consistently, we found a trade-off, at the shoot scale, between mature fruit production and female flower production the following year in defoliated branches. This result thus highlights the need for evaluating reproductive costs at both the modular (shoot and branch) and the individual level, and over multiple years throughout the individual’s lifespan (Genet et al., 2010; Sánchez-Humanes et al., 2011; Sala et al., 2012; Bogdziewicz et al., 2019).

Sex allocation

Factors determining sexual allocation in natural tree populations are still poorly known, especially in mast-seeding species (Kazuhiko, 2007; Knops and Koenig, 2012; Rapp et al., 2013). The sex allocation theory assumes the existence of a trade-off between male and female functions, and that if there is resource shortage monoecious plants should shift towards maleness because maintaining the male function usually requires less investment than maintaining the female function (Charlesworth and Morgan, 1991). This hypothesis might, however, not apply to mast-seeding species, in which increased pollination efficiency requires synchronous investment in male and female function during mast years (Rapp et al., 2013). Sex allocation is very much male-biased in Q. ilex, but we found no correlation between the number of female flowers and the number of catkins at the shoot scale in the control treatment, as observed in previous studies on Q. ilex (Pulido et al., 2014) and other oak species (Knops and Koenig, 2012). Our results provide no support for the sex allocation theory, even in a context of resource limitation by defoliation, as defoliation reduced allocation to pistillate and staminate flowers similarly. Sex allocation shift towards maleness therefore might occur after a resource limitation imposed by a defoliation at the plant scale (e.g. Narbona and Dirzo, 2010), but not when defoliation is only applied at the shoot or branch scale (e.g. Wang et al. 2016).

Conclusions

Our results strongly suggest that the regulation of resource allocation to reproduction occurs at a larger scale than the shoot scale, and that flower production is more sensitive to resource fluctuation than fruit development and seed germination success. Most importantly, our results also reveal the complexity of strategies of resource allocation to the different plant functions over two consecutive years depending on resource availability. Climate change is currently significantly modifying water and carbon availability through altered phenology and suboptimal temperature and soil moisture conditions, subsequently modifying the allocation of resources to the different organs. Our results show that it is essential to explore more deeply the complexity of resource allocation to flowers to propose robust projections of tree fecundity, and subsequently of forest regeneration in future climatic conditions.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Figure S1: relationships between shoot basal area and leaf number and between fruit length and biomass. Figure S2: variation of fruit survival and mass within trees. Figure S3: effect of reproductive status of the 2018 spring shoot on its basal area. Figure S4: variation of sex ratio and fruit set across treatments. Figure S5: relationship between number of catkins and number of female flowers within treatments. Figure S6: effect of defoliation and of having borne or not borne fruits in 2018 on organ production in 2019. Table S1: dataset characteristics across treatments. Tables S2, S3 and S4: estimates corresponding to Tables 2–4. Table S5: effect of the interaction between defoliation and of bearing or not bearing a fruit in 2018 on organ production in 2019.

FUNDING

This work was supported by LabEx CeMEB, an Agence Nationale de la Recherche (ANR) ‘Investissements d’avenir’ programme (ANR-10-LABX-04-01), the ANR FOREPRO (ANR-19-CE32-0008) and Agroparistech and the French Ministry of Environment to I.L.R.