Testing the branch autonomy theory: a 13C/14C double-labelling experiment on differentially shaded branches


A. Lacointe. Fax: +33 (0)4 73 62 44 54; e-mail: lacointe@clermont.inra.fr


The impact of a heterogeneous within-crown light environment on carbon allocation was investigated on young walnut trees trained on two branches: one left in full sunlight, the other shaded until leaf fall resulting in 67% reduction in photosynthetically active radiation. In September, the two branches were separately labelled with 14CO2 and 13CO2, respectively, so that the photosynthates from each branch could be traced independently at the same time. Although some carbon movements could be detected within 5 d in both directions (including from the shaded branch to the sun branch), between-branch carbon movements were very limited: approximately 1% of the diurnal net assimilation of a branch. At this time of the year branch autonomy was nearly total, leading to increased relative respiratory losses and a moderate growth deficit in the shaded branch. The ratio of growth to reserve storage rate was only slightly affected, indicating that reserves acted not as a mere buffer for excess C but as an active sink for assimilates. In winter, branch autonomy was more questionable, as significant amounts of carbon were imported into both branches, possibly representing up to 10% of total branch reserves. Further within-plant carbon transfers occurred in spring, which totally abolished plant autonomy, as new shoots sprouted on each branch received significantly more C mobilized from tree-wide reserves than from local, mother-branch located reserves. This allowed great flexibility of tree response to environment changes at the yearly time scale. As phloem is considered not functional in winter, it is suggested that xylem is involved as the pathway for carbohydrate movements at this time of the year. This is in agreement with other results regarding sugar exchanges between the xylem vessels and the neighbouring reserve parenchyma tissues.


Among the environmental factors that affect tree growth and development, light plays a central role as the major factor involved in photosynthesis. Whole tree shading or canopy opening experiments have shown that shading results in decreasing, not only total growth, but also the root/shoot ratio (Van Hees 1997; Tognetti et al. 1998). Shading generally affects more readily radial growth than primary growth (Kappel & Flore 1983; Collet, Lanter & Pardos 2001), and may affect flowering and/or fruiting (Ryugo, Marangoni & Ramos 1980; Proctor & Crowe 1983; Marini & Sowers 1990).

Inside tree crowns, local light conditions can be very different among branches, so that the general effects of the shading level as mentioned above can be expected to result in differential growth and development patterns within the crown. This assumes, however, that the direct effect of light on photosynthesis – and hence on the local carbon balance between production and demand for growth and respiration – is not compensated for by photo-assimilate exchanges between branches of different light status. The latter assumption, called the ‘branch autonomy’ theory (Sprugel, Hinckley & Schaap 1991; Brisson 2001), has been invoked as a mechanism for differential branch growth and/or survival according to local light conditions, which in turn can explain specific tree shapes in relation to stand density or age. The ‘branch autonomy’ concept has been included in several tree growth models (Ford, Avery & Ford 1990; Takenaka 1994; Kellomäki & Strandman 1995).

To this day, experimental evidence in support of the branch autonomy theory has been mostly indirect, through its expected consequences. Defoliating or shading individual branches of pine (Honkanen & Haukioja 1994) or birch (Ruohomaki et al. 1997; Henriksson 2001) significantly affected their mid- or long-term growth or survival rate, often more severely for a branch-wide than for a tree-wide treatment. Looking at carbohydrate reserve dynamics in individual prune branches, Ryugo et al. (1977) reported at least partial autonomy of fruiting versus non-fruiting branches through local reserve mobilization; similarly, Langstrom et al. (1990) reported a significant decrease in starch reserves of partly defoliated individual pine branches whereas reserves in neighbouring branches were not affected.

However, direct experimental evidence through tracing photo-assimilate movements between branches of different light status is surprisingly limited: we are aware of only two investigations in this area. Using 14C labelling, Cregg, Teskey & Dougherty (1993) concluded that ‘loblolly pine shoots were usually autonomous with respect to carbohydrate supply, but that carbohydrate movement into the terminal shoot from subtending foliage could occur under conditions of very high stress’[i.e. high growth demand]. However, only mature needles, excluding growing organs and mature stem tissues, had been evaluated as possible sinks for between-branch transfers, so that conclusion should be considered partial. In the other experiment, on persimmon, Yamamoto et al. (1999) showed that ‘a very small quantity of 13C photosynthates was exported into neighbouring lateral branches for 44–72 h after a basal or terminal lateral branch was exposed to 13CO2’; unfortunately, due to individual variability the results were not conclusive about a possible effect of shading.

The first objective of the present investigation was to increase experimental knowledge about assimilate movements between differentially shaded branches. In order to evaluate the specific effects of shading, reciprocal transfer rates between a sunlit and a shaded branch were measured simultaneously in the same trees using double 13C/14C labelling, on the simplest possible experimental system: tree crowns consisting of only two branches.

Furthermore, beside current photo-assimilates soluble sugars can be found in the xylem sap of a number of species such as maple or walnut in winter (Améglio & Cruiziat 1992;Améglio et al. 2001), and at budbreak growth resumption can mobilize up to 50% of the tree's total carbon reserves (Lacointe et al. 1993, 1995a; Barbaroux, Bréda & Dufrêne 2003). The scope and significance of these potentially large scale movements with respect to branch autonomy was addressed by periodic sampling.


Plant material

Sixteen 2-year-old Juglans regia L. scions, cv. ‘Franquette’, were planted in 33 L containers. The soil was a peat : clay mixture (33% : 67%; vol : vol) complemented with 10 g NH4NO3, drip irrigated to field capacity. In April, the stem was partially disbudded, allowing only two branches to grow out. In early July, the lower branch was shaded with a shading net allowing approximately 33% of incident PAR, which resulted in a 30% decrease of its diurnal net assimilation. A few trees were kept in full sunlight as controls.

Isotopic labelling and sampling

In mid-September, each branch was separately, quantitatively labelled with *CO2 in 40 L chambers (for details see Kajji et al. 1993). The sunlit branch (open circle) was fed with 18 MBq 14CO2, the shaded one (closed circle) with 4 mmoles 13CO2. The shading system was removed at leaf shedding (November). Three trees were harvested at each of the following dates (four dates and 12 trees): 5 d after labelling (end of primary translocation, cf. Lacointe et al. 1995b); leaf shedding; just prior to budbreak (next spring); and in June, when new shoots displayed four fully grown leaves.

Biochemical analyses

After harvest, the plants were rapidly divided into coarse roots, fine roots, main stem, branches and new shoots when present, and immediately frozen in liquid nitrogen. After freeze-drying, each organ was ground to pass a 125-µm mesh. Soluble sugars were extracted in 80% ethanol at 80 °C, then purified successively on active charcoal + polyvinyl-polypyrrolidone (PVPP), cation exchanger and anion exchanger resins. Glucose, fructose and sucrose were assayed by the enzymatic method of Boehringer (1988). After enzymatic extraction from the pellet, starch was quantified as glucose equivalent (Boehringer 1988). The sum of soluble sugars plus starch is hereafter referred to as ‘mobilizable carbohydrate’ fraction, whereas the non-extractable fraction will be considered ‘structural carbon’.

Isotopic analyses

The 14C of the whole dry matter was determined at ‘infinite thickness’ in a low-background argon–methane flow counter (Numelec NU 20; Canberra Eurisys, Saint-Quentin-en-Yvelines, France). The 14C contents of the purified sugar extract and of the extracted starch were measured on all organs by liquid scintillation. For 13C, the total dry matter label was determined on all organs using an elemental analyser (Carlo Erba NA 1500; Thermo Electron Corporation, Milan, Italy) connected to an isotope mass spectrometer (Optima; Fisons, Villeurbanne, France); the 13C atom percentage excess for each organ was computed versus an unlabelled control tree. The same device was also used to determine the 13C content of soluble carbohydrates and of starch in each branch, after the same extraction procedure as for 14C, with a further purification step on PVPP + ion exchanger resins for starch extraction.

For both isotopes, the excess label recovered in each organ × biochemical fraction was expressed as ‘unit label’: 1 unit label = 1 per-cent (%) of the total amount of label initially fed to the tree. The resulting value (L) was then corrected for individual tree and organ size variability (and associated dilution effects) to yield a normalized value:


where ‘SM’ stands for perennial structural dry matter mass (g), and ‘average’ applies to all trees harvested at a given date.

In organs that lost some carbohydrates between the two sampling dates, the specific label (unit label per g carbohydrate) of the mobilized carbohydrates was estimated as the ratio of carbohydrate NL variation to the variation of ‘normalized carbohydrate amount’ (NC)


Results are expressed as average ± standard error values. Unless otherwise stated, significance of differences is assessed by Student's t-test (two-tailed, heteroscedastic version).


Effect of light on branch growth and carbohydrate content

The sunlit branch exhibited a slightly (but significantly: P < 0.05, non-parametric Wilcoxon paired test) higher total dry matter mass (DM) than the shaded one (Table 1). However, for practical reasons the sunlit branch was also the upper one whereas the shaded branch was in a lower position within the tree architecture, so that it was not clear which factor this effect was due to. To address this question, full-sun-grown trees were also investigated as controls. No such difference was found between the two branches of these control trees (g DM: 12 ± 2 for the upper branch versus 11 ± 2 for the lower one; P = 0.78, same test as above), so that the higher DM of sunlit versus shaded branches could actually be ascribed to the differential light environment between both branches rather than to their respective position within the tree architecture.

Table 1.  Dry matter mass (DM) and carbohydrate content of the different parts of the trees at leaf shedding, in November
OrganTotal DM
(mg g−1 DM)
  1. Carbohydrates (soluble sugars + starch) are expressed as mg glucose equivalent per g DM. *○ : • DM ratio is greater than 1 (non-parametric Wilcoxon significance level: α< 0.05)

Sunlit branch (○) 13 ± 2* 90 ± 5
Shaded branch (•)  9 ± 1* 88 ± 3
Main stem 60 ± 10108 ± 1
Coarse roots200 ± 30412 ± 10
Fine roots 40 ± 15147 ± 7

Regarding carbohydrate concentration, no difference was detected between both branches (Table 1), although the difference in total DM resulted in a similar difference in  total  (absolute)  carbohydrate  content  between branches.

Carbon allocation in September

A total of 52% of the total C assimilated by the shaded branch was lost from the tree within 5 d after labelling (Fig. 1a), versus only 35% for the sunlit branch (significance level of difference: P = 0.01). For both isotopes, less than 40% of the total label that was recovered within the source branch was incorporated in the ‘residual’ non-carbohydrate fraction, which is an indicator for anabolic activity-like growth, although the proportion was slightly higher in the shaded than in the sunlit branch (Fig. 1b). However, the spatial partitioning patterns were remarkably similar: in both cases, approximately 60% of the total C recovered in the whole plant was exported out of the source branch, the other 40% remaining within it. The patterns of partitioning among sink organs (Table 2) were also identical, as approximately 80% of the exported C was allocated to roots versus 20% allocated to the main stem, which was not surprising as they both reflected the same process (assimilate partitioning among the main sink organs), revealed by two independent tracers. However, the most interesting result was that from each source branch, regardless of light environment, a low but statistically significant proportion (approximately 0.2%) of the exported C was recovered in the opposite branch.

Figure 1.

Carbon allocation pattern in September, 5 d after assimilation. (a) Within-tree spatial distribution: for each source branch (○, sunlit branch; •, shaded branch), the figure shows the percentage of total photosynthates (‘normalized label’, see Material and methods) that are recovered either within the source branch (in grey) or in the rest of the tree (export, in black). Assimilates not recovered in either compartment are ascribed to respiratory losses. (b) Chemical partitioning of total carbon label recovered within each source branch, 5 d after assimilation in September: ○, sunlit branch (14C); •, shaded branch (13C). Vertical bars represent standard errors.

Table 2.  Spatial partitioning of the carbon (‘normalized label’, see Material and methods) exported from each source branch in September, expressed as percentage of the total exported C, namely the % of the total recovered outside of the source branch
 Source branch
Sunlit (○)Shaded (•)
  • *

    Value significantly > 0 (P < 0.05, Student's t-test, one-tailed distribution).

Opposite branch0.17 ± 0.04*0.27 ± 0.08*
Main stem21 ± 122 ± 4
Roots80 ± 178 ± 4

Autumn and winter dynamics

The sunlit branch-derived carbon incorporated in the structural fraction of dry matter exhibited no change in the roots or the main stem in autumn or winter (Table 3). However, there was an increase in both branches, particularly between November and April (significance levels: P = 0.08 for the source branch; P = 0.05 for the opposite branch).

Table 3.  Autumn and winter variations of sunlit branch-derived carbon in the extractable carbohydrate (= starch + soluble sugars) and the non-extractable (= ‘structural’) fractions of each part of the tree (excluding leaves in September)
 NL in non-extractable structuresNL in mobilizable carbohydrates
  1. Data are expressed as normalized label (NL) values (cf. Material and methods). *The difference between adjacent columns is significant (P = 0.05, Student's t-test). †The difference between September and April was significant (P = 0.05), although those between September and November and between November and April were not.

Source branch  1.0 ± 0.1 1.6 ± 0.2 2.5 ± 0.3  2.6 ± 0.1    *  1.2 ± 0.2  0.9 ± 0.2
Opposite branch0.015 ± 0.004  *0.05 ± 0.01  *0.09 ± 0.010.016 ± 0.001  *0.035 ± 0.003  *0.070 ± 0.006
Main stem  2.6 ± 0.3 2.6 ± 0.4 2.5 ± 0.1  5.6 ± 0.5    *  2.0 ± 0.5  1.5 ± 0.1
Coarse roots  5.4 ± 1.2 5.5 ± 1.1 4.5 ± 1.2   14 ± 2       8.6 ± 2.1      5.8 ± 1.4
Fine roots  6.8 ± 0.7   8 ± 2   8 ± 1  4.8 ± 0.7      2.8 ± 0.8      2.1 ± 0.3
Total   16 ± 1  18 ± 3  18 ± 2   27 ± 3     *   14 ± 3     *   10 ± 2

At the same time, the labelled carbohydrates decreased in most organs, particularly between September and November. However, the opposite (= shaded) branch did not fit in this pattern, as this was the one organ in which labelled carbohydrates did not decrease, but increased significantly, in the same way as labelled structural carbon, both in autumn (September to November: P = 0.01) and in winter (November to April: P = 0.01). When added, both fractions (i.e. the total dry matter label) in this branch exhibited a very significant increase of approximately 100% (representing approximately 0.075 unit label in absolute terms) between November and April (P = 0.01).

In the different organs that lost some labelled carbohydrates between November and April, and were thus potential carbon sources for the concomitant label increase in the shaded branch, the specific label of mobilized carbohydrates was approximately 0.4 label units per gram glucose equivalent as an order of magnitude (Table 4). This provides a rough estimate of the (minimum) amount of carbohydrate that had been mobilized and translocated into the shaded branch from other parts of the plant: 0.035/0.4 = 0.1 g carbohydrate as an order of magnitude.

Table 4.  Estimated specific label of 14C-carbohydrates mobilized in the whole tree between November and April (Normalized values, cf. Material and methods)
Label of mobilized carbohydrates (unit label)  4 ± 3
Amount of mobilized carbohydrates (g glucose equivalent) 16 ± 3
Specific label of mobilized carbohydrates (unit label g−1)0.4 ± 0.3

Furthermore, both branches appeared symmetrical in this respect (Fig. 2), although due to a higher error in 13C than 14C excess evaluation, the increase in 13C was not statistically significant (P = 0.22).

Figure 2.

Winter variations of the label derived from each branch that is recovered in the opposite branch. Total label is partitioned into extractable carbohydrates (= starch + soluble sugars) and non-extractable (=‘structural’) carbon. Data are expressed as normalized label (NL) values (cf. Material and methods). (a) Sunlit branch-derived carbon recovered in the shaded branch; (b) shaded branch-derived carbon recovered in the sunlit branch. Vertical bars represent standard errors.

Spring mobilization

Between November (leaf fall) and June (new shoots became self-sufficient), 75% of the total tree mobilizable sunlit branch-derived carbohydrates were used up (Fig. 3). Of that carbon, approximately 40% was recovered in the new shoot dry matter. However, most of the total consumption of labelled carbohydrates occurred in spring, between budbreak and new shoot self-sufficiency. This was massive mobilization, as 67% of the labelled reserves still present in April were lost in June, with approximately 62% of that spring-mobilized carbon recovered in new shoots.

Figure 3.

Winter and spring mobilization of sunlit branch-derived carbohydrate reserves: seasonal course of total reserves of whole tree (excluding new shoots in June), and reserve-derived carbon recovered in total dry matter of new shoots in June. Data are expressed as normalized label units (NL, cf. Material and methods).

Interestingly, the new shoots sprouted on the unlabelled branch got a much higher amount of label than expected from the pre-mobilization label content of their mother branch, and the same was true for the shaded branch-derived carbon (Table 5). As a result, in the sunlit branch-borne new shoots, the ratio of shaded branch-derived C (13C) to sunlit branch-derived C (14C) was increased by a factor of 9 as compared to their mother branch. A symmetric situation was observed in the shaded branch-borne new shoots versus their mother branch, although the factor was only 3.

Table 5.  Total label of carbohydrate reserves in both bearing branches just prior to budbreak (April) and of total dry matter in new shoots derived from each bearing branch a few weeks after budbreak (June)
  13C14Cratio 13C : 14C
  1. Symbols:, ○ ,bearing branch grown in full sun light and fed with 14C in the previous year; •, bearing branch shaded and fed with 13C in the previous year. Data are expressed as normalized label units (cf. Material and methods).

Total carbohydrates in0.06 ± 0.02 0.9 ± 0.20.07 ± 0.04
bearing branch (April) 0.6 ± 0.10.07 ± 0.01   9 ± 2
Total DM in new shoots sprouted 1.9 ± 0.2 3.4 ± 0.4 0.6 ± 0.1
from bearing branch (June) 2.1 ± 0.2 0.8 ± 0.2   3 ± 1

This dilution of ‘local reserve originating C’ by ‘whole-tree reserve originating C’ in new shoot dry matter could be viewed in a more quantitative way through a simple modelling approach, as the experimental design by double labelling provided two independent mobilization patterns, which allowed model fitting and model evaluation on two independent data sets.

A simplified model of spring mobilization and allocation of whole tree reserves to branches

This model was developed with the purpose of gaining a better understanding of the limiting factors and mechanisms involved in reserve mobilization associated with new shoot growth.

The total label of new shoots sprouted from each branch in June (NSL) is computed as the sum of a ‘global’ and a ‘local’ component. The ‘global’ component is assumed proportional to total label of carbohydrate reserves in main stem and coarse roots in April (global reserve label: GRL), possibly multiplied by new shoot mass (NSM) to account for a possible effect of new shoot growth rate. The ‘local’ component is assumed proportional to the label of carbohydrate reserves in bearing shoot in April (local reserve label: LRL), also possibly multiplied by new shoot mass (NSM).

As a first step, fitting was performed on the 14C data set, with the purpose of finding the best version of the model, constrained by the condition that the local component must not exceed the total amount of label actually mobilized from the bearing shoot. Least-mean-squared errors of fitting (MSE) – or highest coefficient of determination (R2), which is formally equivalent but easier to read— was used as the fitting criterion.

As expected, the above-mentioned predominance of ‘global’- over the ‘local’-originating C appeared as a higher contribution to goodness-of-fit of the global component, provided it included the NSM proportionality (Table 6a): in single-component versions of the model, it could account for over 80% of total as compared to approximately 60% for the local one alone. Furthermore, the specific contributions of both components appeared relatively independent as the full model accounted for over 90% of total variability.

Table 6.  Parameterization and evaluation of the model: NSL = Global + Local, where: NSL = total label of new shoots sprouted from each branch in June (norm. label units); Global = one of the following: {0, b, b × GRL, b × NSM, b × NSM × GRL}; Local = one of the following: {0, a × LRL, a × NSM × LRL}; with: a, b = constant parameters; NSM, total dry matter mass of new shoots (g); LRL, ‘local reserve label’ = total label of carbohydrate reserves in bearing branch in April (normalized label units); GRL, ‘global reserve label’ = total label of carbohydrate reserves in main stem + coarse roots in April (normalized label units). The model is constrained by the condition: ‘Local < total label actually mobilized from bearing shoot’. : (a) Parameters of the model as fitted on data relative to the mobilization of C originating from the sunlit branch, i.e. the 14C data. R2, coefficient of determination.
Local component
0 (none)a × LRLa × NSM × LRL
R2Parameter valueR2Parameter valueR2Parameter value
0 (none)− 1.330.69− 0.750.026
b (= constant)02.10.400.691.70.620.0261.5
b × GRL00.280.400.690.240.620.0260.21
b × NSM0.840.0720.920.690.0590.970.0260.052
b × NSM × GRL0.840.0100.920.690.00810.970.0260.0071

Although NSM (new shoot mass, i.e. cumulative growth) clearly contributed to the significance of the global component, this first step yielded no information about the significance of the global reserve store itself, as there was only one GRL value available in the 14C data set. This could be discriminated in the second step, the evaluation procedure, which was performed on the 13C data set, with the least mean squared error of prediction (MSEP, computed using the parameters resulting from the fitting step) as the evaluation criterion.

The evaluation (Table 6b) confirmed the significance of the local component beside the global, as it reduced MSEP by a factor of more than 2. The most interesting point, however, was that the best model (least MSEP) led to a relative error of prediction (the square root of MSEP divided by mean experimental value) that was less than 30%, i.e. only 1.5 times as high as the average relative error of fitting (the square root of MSE divided by mean experimental value). The corresponding model version involved NSM in the global but not in the local component, suggesting that the mobilization of local reserves could be less dependent on new shoot growth than that of global reserves.


Effect of shading on branch growth and fate of current photo-assimilates

Consistent with previous works on different species (Honkanen & Haukioja 1994; Ruohomaki et al. 1997; Henriksson 2001), differential shading had a slight but statistically significant effect on growth, as roughly measured by total dry matter. The use of isotopic tracers revealed another, major, effect: the carbon losses from the shaded branch-assimilated carbon were significantly higher than those from the sunlit branch-assimilated C. At this time scale, i.e. within 5 d after assimilation, carbon losses can be ascribed to respiration (Lacointe et al. 1995b), so this denotes higher relative respiratory losses in the former than in the latter case. As the metabolic requirements of the shaded branch can be assumed slightly lower than, or at most similar to, those of the sunlit branch, this is an indication for a lower absolute amount of available carbohydrate substrate, as expected if the lower photosynthetic production was not compensated for by import from other tree parts. However, the most direct evidence in support of branch autonomy was of course the very low level of between-branch photo-assimilate movements: from each source branch, regardless of light environment, only approximately 0.2% of the total exported C was recovered in the opposite branch. It can be calculated that this represented approximately 1% of the branch own diurnal net assimilation. Thus, it can be concluded that branch autonomy was nearly total regarding primary assimilate allocation.

In addition to the above-mentioned change in relative allocation to respiration, shading had another detectable effect on local carbon economy: the relative incorporation into the ‘structural’ fraction was slightly higher, which can be understood as a slight increase in the relative allocation of C to growth versus reserve storage activity. This change, however, was of limited extent, not significantly affecting the final relative carbohydrate content of the branch DM at the end of the season. In other words, the assimilate shortage affected both the growth and reserve storage rates, the latter being only slightly more affected. A similar ‘balanced covariation’ of both activities in response to different shading conditions was previously reported in young oak (Ziegenhagen & Kausch 1995) and beech (Gansert & Sprick 1998). Thus, although reserves certainly can act as short- or middle-term buffers to cope with temporary imbalance between C production and demand, such as during peak fruit growth (Ryugo et al. 1977), they should in many cases not be considered as mere passive buffers in the long term. Instead, they appear as a vital component that the tree will (if possible) ‘manage’ to keep above a critical level, possibly at the expense of growth in case of source limitation. As a practical consequence, source–sink based models of C partitioning in trees would likely benefit from assigning reserve storage a competitive ‘sink strength’, as is generally the case for growth (Lacointe 2000; Le Roux et al. 2001).

Autumn and winter dynamics

From 5 d after assimilation through leaf fall and winter, the total label incorporated in the structural fraction exhibited no change in most tree parts, namely the roots or the main stem. This is consistent with previous results (Lacointe et al. 1995b) showing that leaf export, which was the limiting step for the final pattern of spatial and biochemical partitioning of current assimilates, was completed by 90% within 5 d. In the present experiment, however, there was an increase in the structural fraction of branches, both between September and leaf fall, and even more after leaf fall. As it could in neither case be ascribed to current assimilate import  and  metabolism,  this  is  an  indication  that  some later anabolic activity occurred in current-year shoots, not only in the autumn but also in winter. Although this experiment provided no qualitative information in this respect, it might involve phenolic compounds, particularly phenol glycosides, which have been found in particular abundance in walnut tissues (Claudot, Drouet & Jay-Allemand 1992).

Regarding extractable carbohydrates, their total label decreased in most organs between September and November This is an indication for turnover, suggesting a role as mobilizable carbon source for the above-mentioned, post-primary export increase in source branch structural matter. However, as shown by the labelled carbohydrate increase in the opposite branch, part of that recirculating mobilized C was also re-deposited as carbohydrate reserves at a distance from the place of remobilization. When both fractions were added (extractable carbohydrates + structural C), the increase in the opposite branch total dry matter label exceeded 100%, so that branch autonomy before leaf fall did not appear so strict as it did when considering only primary assimilate partitioning.

Moreover, further and even more significant amounts of mobile carbon were imported into both branches during the winter, between November and April The amount of carbohydrate that has been mobilized and translocated into the shaded branch from other parts of the plant could be roughly estimated as 0.1 g carbohydrate as an order of magnitude, which is very low (approximately 0.1%) compared with the total tree reserves, but may be significant (approximately 10%) relative to the branches’ own reserves. Although this figure was yielded by a very rough calculation, it should be considered rather underestimated as it assumed that all of the imported carbon was recovered at harvest, thus ignoring any respiratory losses.

An interesting question is the pathway for carbohydrate movements in winter. As phloem is considered not functional in winter (Aloni 1991; Aloni & Peterson 1997), this suggests the involvement of the xylem pathway. This is in agreement with other results on walnut regarding sugar exchanges between the xylem vessels and the neighbouring reserve parenchyma tissues, with major consequences on the water status and likely spring development (Améglio et al. 2001; 2004).

Spring mobilization

A total of 75% of the total September-labelled reserves in the whole tree were used up between November (leaf fall) and June (new shoots became self-sufficient), with most of the total consumption of labelled carbohydrates occurring in spring, between budbreak and new shoot self-sufficiency. This is consistent with the mobilization rates of August-labelled reserves (45%) and October-labelled (80%) as previously found by Lacointe et al. (1993) in young walnut. Of the total mobilized carbon, approximately 40% was recovered in the new shoot dry matter, which again is consistent with the recovery rates of C derived from August-labelled reserves (60%) and October-labelled (15%) as found by Lacointe et al. (1993).

Regarding the branch autonomy issue, however, the most interesting point was that the new shoots sprouted on each branch got a much higher amount of C initially labelled in the opposite branch than expected from the pre-mobilization label content of their own mother branch. This resulted in significant dilution of ‘local reserve originating C’ by ‘whole-tree reserve originating C’ in new shoot dry matter. At this point, very little was left of branch autonomy.

This dilution effect could be simulated by a simple model of the respective contribution of ‘local’ (mother-branch) versus ‘global’ (tree-wide) reserves. After fitting on the 14C data set, the model fairly well simulated the 13C data set, which can be considered not a mere replication but an independent experiment in a different situation (regarding the light environment at labelling). The results suggested that the mobilization of ‘local’ (mother-branch) reserves could be less dependent on new shoot growth than that of ‘global’ (tree-wide) reserves, which would be consistent with  a  rather  ‘source-limited’ mobilization  for  local reserves whereas the tree-wide mobilization would be more ‘sink-limited’ or ‘sink-driven’. If so, the early stages of mobilization  just  before  or  at  budbreak  can  be  expected to involve preferentially ‘local’ reserves whereas the ‘global’ would be tapped on at later stages, in relation to actual growth rate; this hypothesis could be tested by frequent harvesting around and within a few weeks after budbreak.

Taken all together, these results do not support the idea of long-term branch autonomy, thus contradicting conclusions from previous defoliating or shading experiments, particularly in birch, as reported in the Introduction section. However, the contradiction may be only apparent, with at least three possible levels of explanation.

Firstly, some genetic factors, specific to each species, might be involved. Although there is a balance between growth and reserve storage in many species, including walnut, oak or beech as discussed above, this might not be the case in pioneer species such as birch. If growth indeed has priority over reserve storage as a sink in that species, local shade would lead to local significant reserve deficit (which did not occur in the present experiment in walnut). As a consequence, early bud growth might be affected, in turn hindering the subsequent growth-sink driven massive import of tree-wide originating resources as hypothesized above, etc., eventually leading to branch death as predicted by the branch autonomy principle.

Secondly, it should be emphasized that this experiment was carried out on young trees. However, as suggested by a few investigations into the response to pruning with respect to reserve dynamics (e.g. Clair-Maczulajtys & Bory 1988), within-tree C fluxes tend to become more ‘compartmentalized’ as trees grow older and larger. This would mean a higher significance of local versus tree-wide resources for local growth, resulting in better validity of branch autonomy in old than in young trees.

Thirdly, although shading systems were removed in winter both in Henriksson's (2001) and in the present experiment, light conditions were very different in spring. In the former experiment, branches were re-shaded at leaf flush. In contrast, in our experiment the shading system was not re-installed, so that not only bud break but also subsequent new shoot growth occurred in a sunlit environment, as would occur after a clearing. This could explain the observed difference, if the fate of resources from massive, tree-wide reserve mobilization is, directly or not, dependent on the local environment. Assuming sink-driven import, such dependence would arise from early sink demand of new shoots, or early growth rate, which could be stimulated by light through its own photosynthesis, or maybe through more qualitative effects on development. Through such a feedback loop, a clear light environment would not only promote import into sunlit branches, but also inhibit import of tree-wide originating resources into shaded branches, in relation to competition among sinks. This would further explain a seemingly paradoxical effect observed by Henriksson (2001), which true branch autonomy would not allow: the impact of shading, as evaluated by eventual death rate, was more pronounced on branches that were individually shaded, hence competing with sunlit ones, than on completely shaded trees.


In this experiment on young walnut, a heterogeneous within-crown environment yielded contrasting results regarding branch autonomy:

In summer, although some carbon movements could be detected in both directions (including from the shaded branch to the sunlit branch), between-branch carbon movements were very limited, being approximately 1% of the diurnal net assimilation of a branch. Thus, at this time of the year, branch autonomy was nearly complete, leading to increased relative respiratory losses and a moderate growth deficit in the shaded branch.

In winter, however, significantly more carbon was imported into the branches, probably representing up to 10% of total branch reserves, which makes branch autonomy more questionable at the seasonal time scale. It was even more so when including spring outgrowth, as new shoots sprouted from each branch got more carbon mobilized from whole tree-wide reserves than from the local reserves located in their own mother branch. However, there were indications that this massive import of tree-wide resources into the lateral branches in spring might be dependent on the local light environment at that time.

In this view, the dynamics of tree carbon economy and growth exhibits a great flexibility of response to environmental changes, as new shoots are granted a ‘new chance’ in spring if the environment is improved, as happens after a clearing or a storm, even if the bearing shoot was previously under adverse conditions.


This work was supported by the EU Project Fair CT96- 1887 (W-BRAINS), with Jean-Sylvain Frossard as the local co-ordinator. Additional technical assistance was provided by Christian Bodet, Maurice Crocombette and Stéphane Ploquin. We thank Pascale Maillard for helpful discussions.