Architecture and size relations: an essay on the apple (Malus × domestica, Rosaceae) tree†
The authors thank field technical staff at INRA Montpellier for tree management and two anonymous reviewers for helpful comments on a previous version of the manuscript. This research was partly funded by a predoctoral grant by the Département de Génétique et Amélioration des Plantes (DGAP), INRA, France.
Abstract
The influence of tree size independent of age on some architectural features (annual shoot length, lateral branching, flowering) was investigated on 4-yr-old apple (Malus × domestica) trees either own-rooted or grafted on the dwarfing rootstock M.9, giving rise to large and small trees, respectively. Tree size significantly affected the length of the first annual shoot of bottom branches with a lesser effect on the subsequent annual shoots of the same branches and on branches situated higher in the tree canopy. The linear regression parameters, i.e., slopes and intercepts, between annual shoot length and number of growing laterals were affected by the genotype and, depending on genotype, by tree size. Flowering was generally lower, delayed, and more irregular on large trees compared to small trees, with on average similar ranking of genotypes regardless of tree size. This study provides evidence for a specific effect of tree size, as affected by the root system, on architectural development of the apple tree regardless of the genotype. From an architectural viewpoint, the dwarfing mechanism could be interpreted as a faster physiological aging essentially related to the reduction in length of the first annual shoot of bottom branches and the high flowering on this shoot.
The criteria used to define tree architecture (sensu Hallé et al., 1978; Barthélémy et al., 1991; Bell, 1991) relate mainly to primary growth. They are usually classified into four groups: type of growth (rhythmic vs. continuous), branching pattern (e.g., sympodial vs. monopodial), morphological differentiation of branches (orthotropic vs. plagiotropic), and position of reproductive structures on scaffold branches (lateral vs. terminal). Although these criteria and the architectural models that they define are at first qualitative, the relationships to the size of organs (axis, leaf, inflorescence) have been addressed through “Corner's rules” (Corner, 1949; Hallé et al., 1978): axial conformity (the thicker the stem, the bigger the leaves and the more complicated their form) and reduction in size on ramification (the greater the ramification, the smaller the branches and their appendages). Various attempts have been made to study the correlations between dimensions of organs, at both the vegetative (e.g., leaf and stem size: Barcellos et al., 1986; Bond and Midgley, 1988; Brouat et al., 1998; Baret et al., 2003) and vegetative–reproductive (e.g., stem and inflorescence size: Lauri et al., 1996; Lauri and Trottier, 2004) levels.
Temperate fruit species including peach (Lauri, 1991), cherry (Lauri, 1992), and apricot (Costes, 1993) are usually well characterized by the architectural model of Rauh (see Hallé et al., 1978) with orthotropic shoots, rhythmic growth, and lateral flowering. Apple fits Rauh's and Scarrone's models (Scarrone's being characterized by orthotropic shoots, rhythmic growth, and terminal flowering; Lauri and Térouanne, 1995).
In apple, as with other fruit tree species, two main factors have been identified that modulate tree architecture and overall tree size. The first one is the genotype, which affects branching density (e.g., related to frequency of latent buds or to the physiological abortion of young growing points, referred to as the extinction phenomenon; Lauri et al., 1995), proportion of short vs. long shoots (e.g., type I tree characterized by high branching density and short branches vs. type IV cultivars characterized by scarce branching and longer branches; Lespinasse and Delort, 1986), and flowering pattern (e.g., lateral vs. terminal flowering; Forshey et al., 1992). A second factor is the root system, which has been used as an efficient although empirical means to control tree size, with variable results on flowering (i.e., dependent on the genotype; Webster et al., 1985). The effects of the root system have been studied using various plant materials: own-rooted trees (usually micropropagated; Webster et al., 1985; Zimmerman and Miller, 1991; Quamme and Brownlee, 1993; Hirst and Ferree, 1995a) and/or trees grafted on rootstocks (Lockard and Schneider, 1981; Cummins and Aldwinckle, 1983; Lehman et al., 1990; Warrington et al., 1990; Barritt et al., 1995; Ferree et al., 1995; Hirst and Ferree, 1995a; Costes et al., 2001) or on various rootstock–interstock combinations (Seleznyova et al., 2003). The root system, from “dwarf” to “vigorous,” strongly influences overall tree size, with own-rooted trees generally equivalent to or larger than trees of the same genotype grafted on the most vigorous clonal rootstocks (Ferree, 1988; Larsen and Higgins, 1990; Quamme and Brownlee, 1993) or on seedling rootstocks (Zimmerman and Miller, 1991). Trunk cross-sectional area (TCSA) and branch cross-sectional area (BCSA) are the main traits used to characterize the influence of the root system on tree growth (Westwood and Roberts, 1970; Moore, 1978). They are positively related to shoot length at trunk and branch levels (Hirst and Ferree, 1995a). A positive correlation probably exists between global growth of the root and shoot systems (Zimmerman and Miller, 1991; Fallahi et al., 2002).
Studies of rootstock–cultivar interactions show that the rootstock controls total growth, while the scion controls distribution of growth, such as short vs. long shoots (Ferree et al., 2001a, b), as well as the proportion of buds that become floral (Hirst and Ferree, 1995b, 1996). These results agree with earlier reports that flowering, especially precocity, and tree size, are controlled independently, precocity being more related to the genotype of the scion (Tubbs, 1974; Webster et al., 1985).
Our aim was to further document the relationships between the overall aboveground tree size, independent of age, and specific architectural features, including shoot length, branching, and flowering. Following Ferree (1988), the term “tree size” will hereafter be used instead of “vigor” (e.g., Seleznyova et al., 2004) because vigor also reflects growth dynamics and rhythmicity (Way et al. 1983), which was not considered here. To provide two different tree sizes, two root systems were used, the tree's own roots vs. a dwarfing rootstock, permitting comparison of the aerial development of a range of scion genotypes on either the same dwarfing root system with a graft union (both playing a role in the control of tree growth; Hirst and Ferree, 1995c; Zhu et al., 1999) or on their own roots without a graft union. We assumed that whatever the genotypic differences in root physiology, such as lower hydraulic conductivity in the roots of dwarfing rootstocks (Atkinson et al., 2003), and in architectural relationships between root and shoot systems (Groff and Kaplan, 1988), the experiment would provide contrasted tree sizes so that tree size per se was a variable (Ferree, 1988; Barritt et al., 1997).
A major difficulty with analyzing growth and flowering is their dynamic nature; the size of the morphological entities (internode, annual increment) decreases as the branch system increases in size and complexity with age. This decrease is positively and linearly related to the number of growing laterals for both trunk (Costes et al., 2001) and branch (Hirst and Ferree, 1995a; Seleznyova et al., 2003) with no influence of the root system and is referred to as aging (Wareing, 1970; Hackett, 1985). However, these relationships between axillary shoots and either length or number of nodes have been established between mean points, each one corresponding to a rootstock or rootstock–interstock combination. The linear regression parameters, i.e., slope and intercept, within each cultivar–rootstock combination have not been analyzed.
An experiment was designed based on the morphological analysis of consecutive annual shoots (Hallé et al., 1978; Bell, 1991) to answer three main questions: (1) Are the effects of tree size on the length of annual shoots constant over the first four consecutive years of tree growth? Does tree size affect (2) the relationship between length and number of growing laterals and (3) the flowering pattern?
MATERIALS AND METHODS
Plant material
The analysis was carried out on eight apple Malus × domestica Borkh. (Rosaceae) genotypes, three of which were recently named and selected at the Institut National de la Recherche Agronomique (INRA, France) for disease resistance and fruit quality traits. These genotypes offered a range of architectural features, including various distributions of branch length and orientation and relative proportion of lateral vs. terminal flowering (Fig. 1). They all belong to either type III or IV according to the typology of Lespinasse and Delort (1986), i.e., standard type or tip-bearing type, respectively, according to Forshey et al. (1992). To assure differences in tree size, they were either own-rooted (OT) through micropropagation or grafted onto the dwarfing rootstock M.9 Cepiland (GT) (Table 1; Fig. 1).
For each root system–genotype combination, 10 1-yr-old single-tree replications were planted in 2000 at the INRA station at Montpellier, in southern France, in rows running east to west at a spacing of 5 × 2 m. Trees were 1–1.5 m in height at planting. The experimental design consisted of two five-tree plots of each tree size–genotype combination. The trees were not trained or pruned during the trial, except for removal of branches along the trunk below 80 cm to avoid branches that would touch the ground. The trees grew in standard irrigation and fertilization conditions and were trellised to avoid leaning of the trunk and branch breakage by wind or fruit load, especially for grafted trees. No blossom or fruit thinning was performed.
Studied branches
In 2003, in their fourth year of growth, two branches per tree were chosen in the “bottom” (BB) and “top” (TB) positions of each tree. BBs originated on the 1999 annual shoot of the trunk, which had grown in the nursery, and were therefore composed of four consecutive annual shoots (ASs), 2000 to 2003. TBs were branched on the 2000 AS of the trunk and were composed of three consecutive ASs, 2001 to 2003 (Fig. 2). All branches were southwardly oriented toward the center of the inter-row. Only well-developed branches representative of the general habit of the genotype were selected, without excessive competition with neighboring ones.
Field measurements
In spring 2003, we measured trunk diameter at 20 cm above ground, i.e., 5 cm above graft union on grafted trees, and the basal diameter (2 cm from the trunk) of all lateral branches whose diameters were greater than 1 cm, including the above-defined BB and TBs. Diameters (D) were used to calculate both trunk cross-sectional area and branch cross-sectional area, considering trunk or branch cross-section as a disc (i.e., area = πD2/4). For each tree, BCSA of individual branches were summed to provide a single measurement per tree (SBCSA). TCSA and SBCSA were used as a measurement of overall aboveground tree size.
At the end of the growing season, the selected BBs and TBs were considered the main axis-bearing lateral branches and partitioned according to year of growth (Fig. 2). In case the main axis forked into two more or less equal halves, the longer one was considered as the main axis and the other as a lateral branch.
The length of each consecutive AS (LAS) of the main axis of all BBs and TBs was measured. On each AS, lateral branches were counted and the fate (either vegetative [V] or reproductive [R]) that the terminal bud had in each consecutive year was scored.
Data analysis
Five types of analysis were performed.
Trunk cross-sectional area and SBCSA were first compared through means comparison analysis between tree sizes within each genotype.
A second analysis investigated the effect of tree size on LAS. In a linear model framework, an analysis of variance was conducted to study the impact on LAS of several factors (genotype, year of growth, tree size, and branch position) and their interactions. By testing embedded models, we first selected our best model for data description and interpretation (e.g., model 6 in Table 2). Models were constructed starting from a simple model and adding successively significant factors or interactions (only first-order interactions were considered). Inside this model, we studied the importance of each factor differential effects (in particular tree size) and interaction effects (in particular between tree size and year of growth) (e.g., Table 3).
Third, the relationship between LAS in year Y vs. the number of growing laterals that developed in year Y + 1 was investigated. A linear relationship, through a covariance analysis, was assumed in studying the effects on slopes and intercepts of the genotype and tree size and their potential interactions. In this analysis the number of growing laterals was considered as the dependent variable, with three independent variables: LAS as a covariate and genotype and tree size as factors, with data pooled for all years of growth and branch positions.
Fourth, we studied the effects of factors (genotype, year of growth, tree size, and branch position) and their interactions on number of inflorescences that developed in the year following the year of annual shoot growth. For this, Poisson models were constructed using the logarithmic, i.e., canonical, link between the combination of factors and the Poisson parameter. Therefore, all factor effects were interpreted on this logarithmic scale (see differential effects in Table 7). We first selected the best model and then interpreted the importance of the effects inside it. The same model construction as in the second analysis was used.
The models have been investigated using R software, version 1.6.1 (http://www.r-project.org).
A fifth analysis compared the effects of tree size on the transition between lateral types from one year to the next (see methodology described in Lauri et al. [1995, 1997]). In the present study, we focus on transitions toward an inflorescence (I) in a given year from either a vegetative lateral (V to I transition; “trends toward flowering”) or an inflorescence (I to I transition; “return-bloom”) in the preceding year. On the 2000-AS of bottom branches transitions between 2001 and 2002 and between 2002 and 2003 were considered. On the 2001-AS of bottom and top branches, only transitions between 2002 and 2003 could be explored. The specific flowering patterns of the various tree size–genotype combinations led to insufficient sample sizes for many ASs; therefore, only a mean point per genotype was calculated and plotted on graphs comparing data for OT and GT.
RESULTS
Tree size and annual shoot length
Trunk cross-sectional area and SBCSA values were two- to three-fold as great for OT than for GT, except for the genotype Ariane, for which SBCSA was not significantly different between the two tree size classes (Table 1). These results are in agreement with visual observations that for all genotypes, tree height and overall volume were larger for OT than for GT (Fig. 1).
For all tree size–genotype combinations, LAS decreased as trees aged, except in a few cases (e.g., between 2001 and 2002 for top branches of X.3426 GT; Fig. 3). The distribution of length of consecutive ASs varied according to the tree size–genotype combination, with some genotypes (e.g., X.3423 OT) having a more progressive decrease in length on the consecutive ASs than others (e.g., Pitchounette OT) (Fig. 3). The interaction between tree size and the year of growth was significant (model 1, P = 2.34 × 10−4; Table 2) and was essentially explained by testing year 2000 against grouping of years 2001, 2002, and 2003 (model 2, P = 0.90; Table 2). The LAS was significantly higher for OT than for GT (+26.86, P < 10−4; Table 3), and, not considering the 2000-AS, was significantly lower for branches in top position than in the bottom position (−9.48, P = 3.50 × 10−3; Table 3). Considering the individual effect of year, the difference between the 2001-AS and the 2000-AS (−25.16, P < 10−4; Table 3) was greater than the difference between the 2002- and 2001-ASs and between 2003- and 2001-ASs (−11.14, P < 10−4; and −20.83, P < 10−4, respectively; Table 3). The influence of tree size on the length of consecutive ASs was greater in the first year of growth of bottom branches (2000), with a lower influence during the following three years (see interaction effects between OT and the grouping of years 2001–2002–2003: −12.85, P < 10−4; Table 3).
Length of annual shoots and number of growing laterals
The relationships between the two variables were clearly related to the genotype (Fig. 4). Pitchounette and Ariane were at the two extremes, with the highest correlation and slope for the former genotype vs. the lowest correlation and slope for the latter genotype. Considering all genotypes, there was a strong interaction between genotype and tree size for intercept (comparison of model 1a and model 0a, P < 5.19 × 10−3; Table 4A) and slope (comparison of model 2a and model 0a, P < 3.36 × 10−10; Table 4A) leading to model 0a. This interaction was mainly due to the two genotypes Pitchounette and Verline. Interaction was neither significant for intercept (comparison of model 1b and model 0b, P > 0.05; Table 4B) nor for slope (comparison of model 3b and model 1b, P > 0.05; Table 4B) for all genotypes but Pitchounette and Verline leading to model 3b. In model 0a (Table 5A), the effect of tree size on the intercept and the slope was essentially effective in the interaction with the genotype because the individual effect was not significant (P > 0.05), whereas in model 3b (without any kind of interaction; Table 5B) the effect of tree size on slope was highly significant (P < 5.49 × 10−7) and to a lesser extent on intercept (P = 0.01). These results therefore showed that both intercept and slope of the relationships between the number of growing laterals and the length of AS were significantly affected by the genotype and, and for some genotypes, by tree size (e.g., X.3263 vs. X.3305, without or with influence of tree size, respectively; Fig. 4).
Flowering in the year following annual shoot growth
Regardless of the genotype, the number of inflorescences did not vary linearly according to the AS position in the branch (Fig. 5). Two contrasted patterns were observed among genotypes. ‘Pitchounette' was characterized by almost no inflorescences in OT regardless of the AS, whereas there were many inflorescences on 2000-AS of GT, followed by a strong decrease in the following years (Fig. 5). By contrast, X.3423 had a high and more regular number of inflorescences on all ASs of both OT and GT (Fig. 5).
There was no significant interaction between branch position and tree size (P > 0.30; Table 6) on the number of inflorescences, with fewer inflorescences on top than on bottom branches (−0.39, P = 3.53 × 10−4; Table 7). There was no relationship between LAS and the number of inflorescences (data not shown). Rather, the number of inflorescences was significantly lower on OT than on GT (−2.66, P = 2 × 10−16; Table 7) and was significantly higher on the 2001- and 2002-ASs than on the 2000-AS (+1.35, P = 2.65 × 10−10, and +1.18, P = 4.82 × 10−8, respectively; Table 7). The negative effect of OT on flowering decreased with increasing year of shoot growth (see positive interaction between OT and years 2001 and 2002: +1.55, P < 2 × 10−16, and +1.10, P < 2 × 10−16, respectively; Table 7). A main difference between trees of contrasted sizes was in the flowering pattern of bottom branches: OT typically had a few or no inflorescences on the 2000-AS followed by a high number of inflorescences on the 2001-AS and a low or intermediate number of inflorescences on the 2002-AS, whereas GT typically had a high number of inflorescences on the 2000- and/or the 2001-AS with a lower number of inflorescences on the 2002-AS (Fig. 5).
Transitions to flowering in lateral development from the year following annual shoot growth onwards
Transitions from either a vegetative lateral (V) or an inflorescence (I) to an inflorescence (I) varied greatly depending on genotype. Pitchounette had the lowest values, around 0.15 for I to I for both tree sizes (Fig. 6a) and around 0.4 and 0.1 for V to I for GT and OT, respectively (Fig. 6b), whereas X.3423 had the highest values, between 0.8 and 1.0 for I to I and V to I for GT and OT (Fig. 6a, b). Transition to flowering was generally higher for GT than for OT for all genotypes, except for transition from I to I, for three genotypes that had similar values for both tree sizes, X.3423 and X.3424 with the highest values and Pitchounette with the lowest values.
DISCUSSION
Tree size and variations in annual shoot length
Data for the two variables taken as measurements of tree size, trunk cross-sectional area (TCSA) and sum of branch cross-sectional area (SBCSA), showed that, despite the genetic diversity of the root systems, the shoot growth of trees on their own roots was more influenced by a “own-root effect,” leading to significantly larger trees for all genotypes, than by an individual genotypic effect of the root system. Our results thus confirmed previous ones done from a more horticultural perspective and including total growth and leaf area and flowering precocity (Larsen and Higgins, 1990; Hirst and Ferree, 1995c). As previously shown in the same experiment (Maguylo and Lauri, 2004), for seven genotypes of eight, the ratio between SBCSA and TCSA varied between 1.3 and 1.9 and was not significantly influenced by tree size. This confirms that if a tree is left unpruned, the partitioning of biomass between the trunk, its lateral branches, and eventually the leaves follows a rather constant law whatever the tree size (pipe model theory of Shinozaki et al. [1964]).
Our results confirmed the positive relationship between tree size and annual shoot lengths of branches; branches on large trees are longer than on small trees (Ginkgo biloba, Gunckel et al., 1949; apple, Seleznyova et al., 2003), and the lengths of the consecutive annual shoots that compose a branch are reduced (Borchert, 1976). Costes et al. (2001) found an exponential decrease in the length of consecutive annual shoots, which was attributed to a reduction in the number of neoformed metamers (Seleznyova et al., 2003). In the present study the longer first annual growth of bottom branches explained a great part of the greater total branch length of large trees (Table 3). In many cases, the bottom branches developed in the terminal position on a short sylleptic branch in a lateral position on the first annual growth (1999) of the trunk (data not shown). This may explain the greater length of bottom branches (Remphrey and Powell, 1985).
Tree size and branching and flowering pattern
Complementing previous work (see the Introduction), our results clearly showed a linear relationship between the length of the annual shoot and the number of laterals that developed the next year and that there were significant differences between genotypes for both the slope and the intercept (model 3b, Table 5). This study highlights one more point: considering the entire population of annual shoots for all tree size–genotype combinations, and not just mean values (see Hirst and Ferree, 1995a; Seleznyova et al., 2003), tree size may significantly affect the linear regression parameters, i.e., slope and intercept, between the length of the annual shoot and the number of laterals, depending on the genotype.
Flowering in the lateral position the year following annual shoot growth (Lauri and Lespinasse, 2001; Lauri, 2002; Labuschagné et al., 2003; Planchon et al., 2003) and transition toward an inflorescence from either a vegetative shoot or an inflorescence in the preceding year (Looney and Lane, 1984; Lauri et al., 1995, 1997; Lauri and Trottier, 2004) vary with the genotype. At the tree level, although precocity of flowering may vary greatly among genotypes, whether it is enhanced (Quamme and Brownlee, 1993) or not (Tubbs, 1974) by a dwarfing influence of the root system is controversial. In our data, genotypes with a tendency for regular lateral flowering on consecutive annual shoots on large trees also showed this trend on small trees (e.g., X.3423; Fig. 5). Moreover, large trees were characterized by a lower frequency of lateral flowering, more specifically on the first annual shoot of bottom branches (Fig. 5), and also by lower values of transition to an inflorescence from either a vegetative shoot or an inflorescence the preceding year (Fig. 6). Therefore large trees tended to have a delayed as well as a lower and more irregular flowering than small trees, but the ranking of genotypes was unaffected. This agrees with previous observations that rootstock affects flowering the year after grafting, but not in the subsequent years (Hirst and Ferree, 1995c) and that tree size influences flowering mainly by affecting spur development and the proportion of shoots that undergo the transition from a vegetative to a floral state (Hirst and Ferree, 1996). However, it is likely that this pattern of flowering related to tree size, especially low and irregular flowering on large trees, partly depends on the architectural type of the genotype. As noticed by Barritt et al. (1997), unlike spur-type genotypes, in which an increase in tree size may contribute to more flowers, the large trees of standard and tip-bearing genotypes studied here have a more irregular pattern of flowering than do small trees, at least in their first year of growth.
The development of a plant follows a morphogenetic progression from the young, nonflowering individual to the senescing one and involves a succession of states, including reproductive and vegetative events (Nozeran et al., 1971; Nozeran, 1984; Borchert, 1976). The latter case may be illustrated well by morphological changes such as stem diameter (Allsopp, 1965), length and duration of growth units (Borchert, 1976), or the ratio between the axis and leaf components of the leafy shoot, i.e., axialization (Lauri and Térouanne, 1991; Lauri and Kelner, 2001; Baret et al., 2003). The concept of morphogenetic progression of interest in a modelling context (De Reffye et al., 1991; Barthélémy, 2003) may be illustrated by our results. We showed here that the relative position of the annual shoot within the tree, i.e, ontogenetic position taking into account time elapsed since the beginning of tree growth (Fortanier and Jonkers, 1976; Barthélémy, 2003), rather than position of the branch along the trunk, is the main determinant of vegetative growth and flowering. In our study, the 2000 annual shoot of the bottom branch and the 2001 annual shoot of the top branch both corresponded to the first year of growth of the branch (Fig. 2). However, we showed here that for both the length and flowering in lateral position, the 2001 annual shoot of top branches was more similar to the 2001 annual shoot of bottom branches than to the 2000 annual shoot of bottom branches.
This concept of morphogenetic progression would also apply to the comparison of trees with different sizes at the same age. Lespinasse (1980) established a general relationship between tree size, precocity of flowering, and distribution of flowering in the tree crown, i.e., the larger the tree, the later and the farther the flowering from the trunk. Our study compared trees of different sizes at similar “chronological age” (lato sensu since the plant material does not come from a seed; here, time elapsed was the same in respect to planting in the field). From an architectural viewpoint, our results suggest that the increase in size of a tree could be characterized by a swelling-like process that operates mainly in the first year of growth of bottom branches. This process implies long annual shoots, low flower initiation, and high number of either vegetative laterals (Chun et al., 2002; Pitchounette, data not shown) or latent buds (Ariane, data not shown) in the inner part of the large tree with a rapid decline of shoot length and an increased flowering thereafter (Fig. 7). Conversely a small—or “dwarf”—tree with shorter annual shoots of bottom branches and more precocious and higher flowering (having a reduced swelling-like process) would be “physiologically older,” i.e., it has a higher degree of meristem differentiation (Borchert, 1976; Fortanier and Jonkers, 1976; Barthélémy, 2003) than a large tree.