|Home | About | Journals | Submit | Contact Us | Français|
The presence of fruit has been widely reported to act as an inhibitor of flowering in fruit trees. This study is an investigation into the effect of fruit load on flowering of ‘Moncada’ mandarin and on the expression of putative orthologues of genes involved in flowering pathways to provide insight into the molecular mechanisms underlying alternate bearing in citrus.
The relationship between fruit load and flowering intensity was examined first. Defruiting experiments were further conducted to demonstrate the causal effect of fruit removal upon flowering. Finally, the activity of flowering-related genes was investigated to determine the extent to which their seasonal expression is affected by fruit yield.
First observations and defruiting experiments indicated a significant inverse relationship between preceding fruit load and flowering intensity. Moreover, data indicated that when fruit remained on the tree from November onwards, a dramatic inhibition of flowering occurred the following spring. The study of the expression pattern of flowering-genes of on (fully loaded) and off (without fruits) trees revealed that homologues of FLOWERING LOCUS T (FT), SUPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1), APETALA1 (AP1) and LEAFY (LFY) were negatively affected by fruit load. Thus, CiFT expression showed a progressive increase in leaves from off trees through the study period, the highest differences found from December onwards (10-fold). Whereas differences in the relative expression of SOC1 only reached significance from September to mid-December, CsAP1 expression was constantly higher in those trees through the whole study period. Significant variations in CsLFY expression only were found in late February (close to 20 %). On the other hand, the expression of the homologues of TERMINAL FLOWER 1 (TFL1) and FLOWERING LOCUS C (FLC) did not appear to be related to fruit load.
These results suggest for the first time that fruit inhibits flowering by repressing CiFT and SOC1 expression in leaves of alternate-bearing citrus. Fruit also reduces CsAP1 expression in leaves, and the significant increase in leaf CsLFY expression from off trees in late February was associated with the onset of floral differentiation.
Citrus, like many other woody species, flower profusely. In the Mediterranean area, flower induction takes place in autumn/winter and the main blossoming in spring (Sherman and Beckman, 2003). Current knowledge discriminates between exogenous and endogenous components for regulating flowering biology in citrus (Krajewski and Rabe, 1995; Martínez-Fuentes et al., 2004). The former mainly includes climatic factors. Thus, environmental cues such as temperature or photoperiod, and stress conditions (e.g. water deficit or salinity), have been reported as modulators of flowering responses (Agustí, 1999; Valiente and Albrigo, 2004). On the other hand, endogenous factors are essentially genetic and/or hormonal. Phytohormones, such as cytokinins, polyamines or gibberellins, are known to participate to a certain extent in the physiological processes regulating flower induction or differentiation (Koshita et al., 1999). In particular, gibberellins have been traditionally considered as essential inhibitors of flower bud induction (Mutasa-Gottgens and Hedden, 2009).
In addition, the presence of fruit has also been found to be an inhibitor of flowering in fruit trees. The response of flowering to fruit load has been reported in numerous species (Garner and Lovatt, 2008; Spinelli et al., 2009; Rosenstock et al., 2010), including citrus (Monselise and Goldschmidt, 1982). This behaviour basically consists in the fluctuating production between heavy fruit yields (on year) followed by scarce ones (off year). Moreover, the inhibitory effect of fruit in flower formation depends on both the number of fruits developed and the harvest date (El-Otmani et al., 2004; Martínez-Fuentes et al., 2010). Hormonal factors, competition for nutrients or even changes in carbohydrate and mineral status (Goldschmidt et al., 1985; Valiente and Albrigo, 2004; Baninasab et al., 2007; Rohla et al., 2007) appear to participate in the regulation processes, although the way in which the presence of fruit affects flowering and the nature of the regulatory mechanisms remains unknown.
Recently, the development of genomic and transcriptomic tools has contributed to a better understanding of the metabolic and molecular processes involved in floral biology. Most of our knowledge about flower induction has come from studying flowering regulatory genes in Arabidopsis thaliana (Komeda, 2004). These genes appear to be extraordinarily conserved in woody species (Brunner and Nilsson, 2004), and previous research has demonstrated that many of them even share roles and/or metabolic pathways (Greenup et al., 2009). Moreover, current evidence assumes that flowering is the result of complex interactions at the metabolic and molecular level involving multiple promoter and inhibitor genes (Moon et al., 2005; Michaels and Michaels, 2009). Regarding citrus, several studies have recently been conducted to elucidate the molecular mechanisms involved in flower formation and differentiation. A first approach on citrus was developed by Peña et al. (2001), who demonstrated through plant transformation the role of APETALA1 (AP1) and LEAFY (LFY) genes in juvenile phase development. Later studies have identified and/or isolated in this species these (CsAP1, CsLFY; Pillitteri et al., 2004b) and other regulatory genes involved in both the determination of the flowering time and floral identity processes (see Dornelas et al., 2007).
The regulatory role of the FLOWERING LOCUS T (FT) gene has been identified in numerous species (Turk et al., 2008; Zhang et al., 2010) including citrus (Endo et al., 2005; Nishikawa et al., 2007), and the encoded protein (FT protein) associated with the mythic ‘florigen’ (Yu et al., 2006). Other genes, like CONSTANS (CO) or SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1) have been reported to integrate signals from different pathways and, at least in Arabidopsis, are involved in promoting vegetative to reproductive tissues (Borner et al., 2000; Lee et al., 2000; Onouchi et al., 2000; Komeda, 2004). Specifically, CO participates in upstream and complex regulatory pathways, whereas SOC1 appears to act downstream (Dornelas et al., 2007). In contrast, FLOWERING LOCUS C (FLC) represses flowering; high transcript levels have been found in late-flowering mutants, suggesting a central role of this gene in the control of flowering time, mainly through the vernalization pathway (Sheldon et al., 2000; Michaels et al., 2005). Even though other repressor genes of flowering, like TERMINAL FLOWER 1 (TFL1), have been isolated from ‘Washington’ navel sweet orange (CsTFL; Pillitteri et al., 2004a) and correlated positively with juvenility, rigorous information about the function of FLC in citrus is not yet available.
In summary, alternate bearing is the result of complex metabolic and molecular regulatory pathways affecting flowering induction and floral identity. The phenomenon has been largely explained as a crop load dependency and, although the factors modulating the processes involved have recently been studied for diverse species and key genes regulating flowering identified, we lack knowledge about the effects of crop load and the molecular mechanisms involved. In this work, for the first time, the effects of fruit on the expression of putative homologues of genes involved in flowering pathways were analysed to provide insight into the molecular mechanisms underlying alternate bearing in citrus due to crop load.
This study involved field grown 12-year-old trees of ‘Moncada’ mandarin [Clementina ‘Oroval’ (Citrus clementina Hort. ex Tan.) × ‘Kara’ mandarin [C. unshiu (Swingle) Marcow. × C. nobilis Lour.] trees, grafted onto ‘Carrizo’ citrange [C. sinensis Osbeck × Poncirus trifoliata (L.) Raf.] rootstock, planted 5 m × 5 m apart. Experimental fields were located in the IVIA Research Station (Moncada, Spain). Trees of this cultivar exhibit a marked alternate-bearing behaviour.
To examine the effect of fruit load on flowering, 25 trees were randomly selected for their uniformity in size and vigour at spring. Total yield per tree was determined by counting and weighing all fruits at harvest (April), and flowering intensity was evaluated in spring as follows. Four branches per tree of three ages (late spring, summer and autumn sprouts) with some 300 nodes per branch were previously selected. Both the number of sprouted nodes and sprouts were counted. The flowers per sprout were also counted, obtaining the results as the number of flowers per 100 nodes to compensate for the differences in size of the selected branches.
Defruiting experiments were performed on another set of 24 trees using four levels of fruit removal (0, 33 %, 66 % and 100 %). Treatments were performed at the onset of stage II of fruit development (July). A randomized complete-block design was used in the experiments.
From early September to the end of February, 30 fully developed mature adult leaves per tree from on (fully loaded) and off trees (without fruits) were collected for RNA extractions. Samples were pooled into three groups, and immediately ground and stored at –80 °C until analysed.
Total RNA was isolated from frozen tissue using the RNeasy Plant Mini Kit (Qiagen, Hilden, Germany). RNA samples were treated with RNase free DNase (Qiagen) through column purification following the manufacturer's instructions. RNA quality was tested by OD260/OD280 ratio and gel electrophoresis. RNA concentration was determined by fluorometric assays with the RiboGreen dye (Molecular Probes, Eugene, OR, USA) according to manufacturer's instructions. Three fluorometric assays per RNA sample were performed. Quantitative real-time RT-PCR was performed with a LightCycler 2·0 Instrument (Roche Diagnostics, Basel, Switzerland) equipped with LightCycler Software version 4·0. One-step RT-PCR was carried out. Reactions contained 2·5 units of MultiScribe Reverse Transcriptase (Applied Biosystems, Carlsbad, CA, USA), 1 unit of RNase Inhibitor (Applied Biosystems), 2 µL LC FastStart DNA MasterPLUS SYBRGreen I (Roche Diagnostics, Basel, Switzerland), 25 ng total RNA and 250 nm of the specific forward and reverse primers of each gene in a total volume of 10 µL. Incubations were carried out at 48 °C for 30 min, 95 °C for 10 min followed by 45 cycles at 95 °C for 2 s, 58 °C for 8 s and 72 °C for 8 s. Fluorescent intensity data were acquired during the 72 °C-extension step and transformed into relative mRNA values using a 10-fold dilution series of an RNA sample as a standard curve. Relative mRNA levels were then normalized to total RNA amounts as previously described (Bustin, 2002; Hashimoto et al., 2004) and an expression value of 1 was arbitrarily assigned to the first sample of the on trees. Specificity of the amplification reactions was assessed by post-amplification dissociation curves and by sequencing the reaction product.
Putative genes were identified through homology search with related genes from an EST database of a random 5′ ‘Clemenules’ mandarin (C. clementina Hort. ex Tan.) full-length cDNA library (Terol et al., 2007). Synthetic oligonucleotides were designed to amplify the gene from the selected clones and, as stated before, sequenced for confirmation. Details about the forward and reverse primers are listed in Table 1.
Parameters were statistically tested by analyses of variance (ANOVA), using the least significant differences (LSD) test for means separation. The experimental data were analysed with Statgraphics Plus 5·1 software (Statistical Graphics, Englewood Cliffs, NJ, USA).
A significant inverse relationship between preceeding fruit load and flowering intensity was found in ‘Moncada’ mandarin (r = –0·93; P ≤ 0·05; n = 25; Fig. 1). Considering the results, the larger the crop load the lower the flowering intensity, with a breaking point of about 50 kg tree−1. Above this fruiting value, flowering was independent of crop load and paralleled nearly nil values.
Although statistically significant, this relationship does not imply causality, therefore further experiments were carried out to demonstrate the direct effect of fruit removal upon flowering. As expected, control on trees (non-defruited, fully loaded) showed the lowest number of flowers (0·5 flowers/100 nodes, on average), in comparison with those completely defruited, which presented the highest number of flowers (142 flowers/100 nodes). Intermediate fruit loads (33 % and 66 % defruited trees) resulted in intermediate flowering intensities (Fig. 2).
The effect of fruit removal on flowering was also evident on fruit set and yield assessment. The higher the flowering intensity, the larger the crop load (data not shown). The magnitude of the response, however, depended on the time of fruit removal. Thus, removal of all fruits in August, September or October did not affect flowering in spring (100–135 flowers/100 nodes; Fig. 3); however, when fruit removal was performed from November onwards, a dramatic inhibition of flowering was observed the following spring (<10 flowers/100 nodes).
The time-course of CiFT expression in leaves throughout the study was strongly affected by fruit load (Fig. 4). Significant differences in mRNA transcripts were detected between on and off trees from October onwards. The expression in off tree leaves increased progressively, becoming more than 10-fold higher than that of on tree leaves in December, and remaining almost constant up to late in February. CiFT transcripts in on tree leaves did not significantly vary within the study period.
The time-course of SOC1 relative expression in leaves was also higher in off trees than in on trees (Fig. 5). Differences were statistically significant from early September to mid-December, becoming highest in November (close to 50 % higher for off trees). From early January onwards, differences between treatments were not statistically significant.
Figure 6 shows the time-course of FLC expression in leaves from on and off trees. From September to mid-December, no differences in gene expression between treatments were found. In January, however, levels of mRNA transcripts markedly increased in on trees. From January until the end of the study, expression in on tree leaves remained between 1·6- and 2·8-fold higher than in off tree leaves.
Expression of CsTFL in leaves was slightly reduced from September to mid-November; afterwards, it continuously increased up to the end of February, when it was >3 times the amount measured in November (Fig. 7). No significant differences were found between leaves from on and off trees during the study period.
CsAP1 and CsLFY revealed different patterns of change in response to fruit load (Fig. 8). CsAP1 expression remained almost constant from early October up to the middle of December. During this period, expression values were 3–5 times lower in on trees than in off trees. From December onwards expression increased in both cases and, although minor in magnitude, significant differences between on and off trees were also found until the end of February. On the other hand, CsLFY expression progressively increased from September to the middle of December (15- to 20-fold). After a transitory reduction in January, expression considerably augmented towards the end of February, reaching values between 30 and 50 times those registered in September. Differences between on and off trees were only significant at the end of the study. In late February, CsLFY expression in off trees was significantly higher than that found in on trees (close to 20 %).
Alternate bearing has been associated primarily with the presence of fruit (Verreynne and Lovatt, 2009; Martínez-Fuentes et al., 2010). According to the literature, it is assumed that crop load alters both flower induction and floral bud differentiation (Valiente and Albrigo, 2004). Particularly, in citrus, flower induction has been reported to occur late in autumn, whereas differentiation occurs afterwards (Iglesias et al., 2007). Recent molecular approaches support this scheme, distinguishing between genes regulating flowering induction and those regulating floral differentiation processes (Tadeo et al., 2008). The present results indicate that genes regulating flower induction are strongly related to fruit load in alternate-bearing citrus trees.
Fruit load acts as a strong inhibitor of flowering in numerous woody fruit tree species including citrus, although the responses vary among species and cultivars (Monselise and Goldschmidt, 1982; Valiente and Albrigo, 2004; Albrigo and Galán, 2004). The present data obtained for ‘Moncada’ mandarin confirmed, once again, a non-linear fruit load–flowering intensity relationship (Fig. 1). However, since this relationship does not necessarily imply causality, convincing evidence that fruit load is directly responsible for reducing flowering intensity was provided by means of defruiting treatments. Trees maintaining all developed fruits up to maturation hardly flowered (5 flowers/100 nodes, on average), whereas those fully defruited at the onset of stage II of fruit development flowered profusely (142 flowers/100 nodes, on average). Trees defruited at an intermediate intensity (33 % and 66 % of developing fruitlets) flowered intermediately as well. However, because no differences were found between 33 % and 66 % of fruit removal (16 and 22 flowers/100 nodes, respectively) the effect does not appear to be strictly quantitative. These results support the hypothesis that the fruit load–flowering intensity relationship is not a linear function and this coincides with previous reports suggesting that there might be a threshold value for crop load, dependent on the variety and physiological status, above which flowering is strongly inhibited (Agustí et al., 1992; Martínez-Fuentes et al., 2010).
It has been previously reported that flower inhibition due to crop load also depends on the length of time the fruit remains on the tree (see review by Monselise and Goldschmidt, 1982). Considering the present data for ‘Moncada’ mandarin, fruit remaining on the tree did not affect return bloom (100–120 flowers/100 nodes) up to November; however, fruit remaining on the tree from November onwards dramatically reduced flowering (values close to 0). Therefore, effective inhibition time of fruit load began between October and November. This period has been associated with the fruit reaching its maximum size (Martínez-Fuentes et al., 2010), or with peel ripening (García-Luis et al., 1986). In any event, fruit drastically inhibits flowering from November onwards, suggesting that there is a point of no return for induction of quiescent buds, and that some irreversible physiological, metabolic or molecular events induced by presence of fruit must be responsible for inhibition. This hypothesis is supported by the fact that bud sensitivity to gibberellic acid (GA3) inhibiting flowering occurs at this moment (García-Luis et al., 1986).
The mechanism whereby fruit load affects flowering intensity is not completely understood, although several regulatory factors have been described. Early observations linked carbohydrate and nitrogen metabolism to the process (Goldschmidt et al., 1985; Lovatt et al., 1988); however, recent studies demonstrated that carbohydrate or nitrogen status are involved in nutritional or storage adjustments rather than in the floral process directly (Reig et al., 2006; Martínez-Fuentes et al., 2010). Several studies have linked flowering intensity to bud sprouting, showing that changes in flowering intensity paralleled changes in the summer/autumn shoot number (Verreynne and Lovatt, 2009; Martínez-Fuentes et al., 2010), and also illustrated that the higher the fruit load the lower the number of sprouted nodes in spring (Martínez-Fuentes et al., 2010). Additionally, environmental factors can modulate flowering through modifications in the physiology of shoot development or even in key metabolic pathways (Agustí, 1999). Nonetheless, knowledge about factors affecting flowering has considerably increased, but data do not provide enough information to understand the mechanisms through which the fruit controls the process of flowering.
Recently, molecular and genomic approaches have been considered important tools to shed light on the complex physiological and metabolic pathways leading to flowering. In recent years, numerous flowering-gene promoters and inhibitors have been identified, isolated and characterized. Extensive research has been done in A. thaliana, for which the balance between promoters and inhibitors is decisive for the adequate determination of flowering time and floral identity (Chon and Yang, 1999; Kobayashi et al., 1999). However, in woody tree species, little has been done on this subject. The results presented herein regarding the expression of flowering genes affected by fruit load offer an insight into the molecular mechanisms underlying alternate bearing in citrus trees.
In several species, flowering ability has been demonstrated to be influenced by the integration of environmental signals from the photoperiod and vernalization pathways (Onouchi et al., 2000; Amasino, 2005; Sheldon et al., 2009), mainly modulated by two floral integrators, the FT and the SOC1 genes (Kardailsky et al., 1999; Borner et al., 2000; Lee et al., 2000; Samach et al., 2000). Both genes have been described as floral promoters and their overexpression induces early-flowering phenotypes (Lee et al., 2006; Sreekantan and Thomas, 2006; Zhang et al., 2010).
The FT gene has been demonstrated to be a pivotal factor controlling flowering period in numerous species (Faure et al., 2007; Chab et al., 2008; Hisamoto et al., 2008; Zhang et al., 2010). The present results support this hypothesis and further relate it to the effects of fruit load in citrus trees just as autumn/winter temperature enhances its expression (Nishikawa et al., 2007). Leaves from off trees (high-return bloom-flowering in spring) showed significantly increased mRNA transcript levels as compared with those from on trees (low return bloom). Thus, whereas on trees showed stationary basal levels of expression, a progressive increase in CiFT levels was observed in leaves from off trees (Fig. 4) until December, concomitantly with a higher flowering rate the following spring (Fig. 2). These results demonstrate, on the one hand, that an increased FT protein constitutes a signal per se that exports from leaf to the shoot apical meristem, where floral differentiation takes place (Notaguchi et al., 2008) and, on the other, that a translated FT protein can be translocated to the floral meristem at any time (Lin et al., 2007; Turck et al., 2008). Nevertheless, the translocation pathway followed by FT protein into the apex remains unknown.
Although not always significant, SOC1 also showed higher expression levels in leaves from off trees compared with those registered for on trees, particularly during the floral induction period (Fig. 5). Lee and Lee (2010) reported that constitutive expression of SOC1 promotes early flowering, while recessive loss-of-function seems to delay flowering (Onouchi et al., 2000). Although isolated in only a few plant species, highly conserved homologues of this gene have been identified in numerous species, including citrus, in which the constitutive expression of several SOC1-like homologues induces early flowering and delays senescence of floral organs (Tan and Swain, 2007). In this sense, the present results confirm the role of this gene in flowering resulting from alternate-bearing adult trees.
Unlike the functions attributed to SOC1 and FT genes, FLC has been described to encode a MADS-domain protein able to repress flowering (Michaels and Amasino, 1999). The present data showed no differences in FLC activity between on and off trees until December (Fig. 6). From this time onwards, there was a progressive increase in transcript levels in leaves from on trees, whereas those from off trees showed no change or even a slight decrease. Some authors have pointed out that vernalization promotes flowering through a permanent epigenetic repression of FLC (Michaels and Amasino, 1999), probably through histone methylation and changes in chromatin conformation (Bastow et al., 2004; Sung and Amasino, 2004). In the present study, trees with increased amounts of FLC mRNA transcripts near spring corresponded to on trees that did not flower. Additionally, previous reports have indicated antagonistic effects of flowering promoters and FLC on the expression of target genes leading to flowering. The study of regulatory pathways demonstrates that this gene is located upstream along the major pathways, since FT overexpression does not appear to affect FLC expression (Moon et al., 2005; Lee and Lee, 2010). Likewise, it has been proposed that elevated levels of FLC expression may be responsible for reductions in FT activity (Michaels et al., 2005). In this context, the marked increase in FLC measured in the on trees from December onwards (Fig. 6) might be related to the suppression of CiFT activity in their leaves compared with off trees (Fig. 4) and, therefore, with the inhibition of flowering.
Moreover, the results show that SOC1 expression increased in off trees before the sharp increase in CiFT, so the activation pathways appear to work, at least partially, autonomously; furthermore, from November onwards, SOC1 expression did not parallel the increased CiFT gene expression. This is not unusual since, at least partially, self-determining regulatory mechanisms for FT and SOC1 activity have been reported recently (Lee and Lee, 2010). Therefore, this hypothesis can be extrapolated to the molecular mechanisms affecting flowering through effects of fruit load. Moreover, whereas CiFT expression was significantly higher in off trees as compared with on trees until the end of February, significant differences in SOC1 expression disappeared by the beginning of January.
Although TFL1 has been described as a crucial floral timing regulator in several species (Shannon and Meeks-Wagner, 1991; Liljegren et al., 1999), in the present study no differences in CsTFL expression dependent on fruit load were found (Fig. 7). This gene was identified and isolated in Citrus sinensis (Pillitteri et al., 2004a) and has been proposed to participate in the development and maintenance of vegetative growth, repressing flowering in several plant species (Esumi et al., 2010; Mohamed et al., 2010). In addition Nishikawa et al. (2009) have reported no variations in the expression of this gene during induction and flower bud development, neither in trifoliate and deciduous Citrus-like species (P. trifoliata) nor in Satsuma mandarin (C. unshiu), indicating that this gene is related to juvenility processes rather than to annual floral transition. Since there were no differences between on and off trees regarding CsTFL expression, differences in flowering as a consequence of fruit load do not appear to be dependent on this gene.
Finally, AP1 and LFY genes have been reported to promote flowering in diverse species including citrus, although their primary role appears to be linked to the determination of floral identity (Pillitteri et al., 2004b). Although transgenic approaches have demonstrated that both genes strongly interact with other physiological processes, such as competition for nutrients, their constitutive expression per se reduces the juvenility period in transgenic citrus (Peña et al., 2001; Duan et al., 2010). In particular, AP1 seems to be more effective and dynamic than LFY in the induction of flowering (Peña et al., 2001), orchestrating floral initiation by repressing the action of inhibitors at the level of the meristem as stated by Kaufmann et al. (2010). The present results showed a reduced expression of CsAP1 in leaves in response to the presence of fruit (Fig. 8). Differences between the levels of expression in leaves from on and off trees were significant during the whole study period despite the reduction in these levels from January onwards. It is noteworthy that LFY activity has been associated with the control of floral meristem identity in Arabidopsis (Weigel et al., 1992). Moreover, recent studies on fruit trees, including citrus, demonstrated that once floral identity is determined, high LFY expression levels are almost limited to reproductive organs (Pillitteri et al., 2004b). This observation might also be linked to the absence of differences in CsLFY expression in leaves from on and off trees between September and January. The significantly higher values found in off trees at the end of February support the main role of this gene on bud differentiation and/or floral identity rather than on inductive processes.
In conclusion, in Citrus fruit load inhibits flowering by repressing CiFT and SOC1 expression in leaves during the floral bud induction period, whereas CsTFL and FLC expression does not seem to be associated with fruit load. Fruit load reduces CsAP1 gene expression in leaves during the floral bud induction period, although differences between on and off trees were attenuated from January onwards, indicating the onset of floral differentiation. This hypothesis is also reinforced by the significant increase in CsLFY expression found in leaves from off trees only in late February. The FLC-increased expression in on trees from early winter onwards might well be related to the suppression of CiFT expression in their leaves and, therefore, with inhibition of flowering.
We thank Dr D. Westall for her help in editing the manuscript. M.C. González was recipient of a contract by the Fundación Agroalimed (Conselleria d'Agricultura, Pesca i Alimentació, Generalitat Valenciana). This work was supported by a grant from the Instituto Nacional Investigaciones Agrarias, Spain (RTA2009-00147).