Can plasticity compensate for architectural constraints on reproduction? Patterns of seed production and carbohydrate translocation in Perilla frutescens


  • Katherine A. Preston

    Corresponding author
      *Present address and correspondence: Department of Biological Sciences, Stanford University, Stanford, CA 94305, USA (e-mail
    Search for more papers by this author

*Present address and correspondence: Department of Biological Sciences, Stanford University, Stanford, CA 94305, USA (e-mail


1 The architecture of Perilla frutescens was manipulated by growing plants in a greenhouse under two photoperiodic regimes, representing conditions near the northern and southern ends of the species’ range, to investigate its effect on patterns of seed production. Within each photoperiodic regime, plants were labelled with 14C at one of three developmental stages at one of three leaf positions to assess the effects of architecture on carbohydrate translocation.

2Perilla flowers in response to short days, i.e. southern-treatment plants flowered earlier and were smaller and more sparsely branched than northern-treatment plants. Branch number limits inflorescence number in Perilla, and southern-treatment plants therefore made fewer inflorescences. Inflorescences are, however, indeterminate, and the southern-treatment plants made more flowers per inflorescence than the northern-treatment plants, such that architecture did not constrain flower number.

3 As a result, plants in both treatments made a similar number of seeds and did not differ in total seed mass.

4 Increased seed production per inflorescence was associated with reduced carbohydrate movement between branches and more 14C-labelled assimilate per seed at the labelled node.

5 In summary, plasticity at the level of the inflorescence largely compensated for architectural constraints on inflorescence number, and was associated with differences in carbon translocation patterns between the photoperiod treatments.


Plant architecture has been shown to affect several ecologically important functions, including reproduction ( Lechowicz 1984; Maillette 1985; Geber 1990; Schmitt 1993), interactions with herbivores ( Mopper et al. 1991 ; Marquis 1992; Rosenthal & Welter 1995; Alonso & Herrera 1996) and regrowth following damage ( Watson & Casper 1984; Watson 1995). Individual architecture is partly based on genotype – it is a particular instantiation of an underlying genetically based developmental programme – but it also has a strong ontogenetic component ( Tomlinson 1983) and so reflects the developmental history of each plant. An individual’s architecture depends on which of its meristems remain dormant, and, when active, whether they give rise to branches, stolons, flowers, etc. Because the growth of meristems is sensitive to the environment ( Harper & Bell 1979; Watson et al. 1995 ), architecture can be a highly plastic trait. Given the correlation between architecture and aspects of ecological performance, individual variation in architecture is potentially an important source of ecological variation.

This effect of architecture on other traits can be traced to the fact that architecture is not only a result of development, but may also be a constraint on further development ( Watson & Casper 1984; Diggle 1994). For example, Impatiens plants from a shaded population have a fixed pattern of meristem allocation that limits the plants’ plasticity in response to increased light ( Schmitt 1993). In Ipomopsis, the level of compensatory growth in response to herbivory depends on the architecture of the plant at the time it is damaged ( Maschinski & Whitham 1989). Plant architecture may also constrain flower production by limiting the number of meristems available for flowering ( Maillette 1985; Geber 1990). It is important to note, however, that architecture can act as a constraint on plant development without conclusively determining its outcome. For example, although meristem number limits inflorescence number in Perilla frutescens ( Preston 1998a), inflorescence number does not necessarily limit flower or seed number: inflorescences in this species are indeterminate, thus an individual could, in theory, compensate for a limited number of inflorescences by increasing the number of flowers and seeds produced by each inflorescence.

In addition to its effects on whole-plant characteristics, such as seed production or the potential for future development, architecture affects the functional relationships that exist among parts within plants. In an individual such as a young plant with very simple architecture, carbohydrates assimilated by a given leaf may move freely to all parts of the plant, whereas when the plant has developed a complex architecture, carbon translocation is typically restricted to a subset of parts ( Watson & Casper 1984). In some cases, the functional relationships among plant parts manifest ecological effects at the level of the whole plant; thus, whereas cultivated maize is severely affected by stemborer infestation, one of its wild relatives is more tolerant of stemborer damage owing to its highly branched and physiologically subdivided architecture ( Rosenthal & Welter 1995). Patterns of carbohydrate translocation also influence growth following defoliation ( Thomas & Watson 1988; Price & Hutchings 1992) and the ability to tolerate marginal ( Larson et al. 1994 ), resource-poor ( Jónsdóttir & Watson 1997) or heterogeneous habitats (reviewed in Alpert 1996).

The study reported here examined some ecological consequences of variation in architecture as a function of flowering time in Perilla frutescens Britt. (Lamiaceae). Perilla is a non-native annual distributed widely in the eastern United States, from Connecticut to Florida ( Deam 1940). It is a short-day photoperiodic plant and, because of its wide distribution, grows under a range of photoperiodic environments. Floral induction occurs in autumn, at which time all active meristems produce indeterminate inflorescences bearing perfect flowers. The shoot consists of an upright axis with opposite leaves, each pair emerging at right angles to the last (decussate phyllotaxis; Fig. 1a). Axillary meristems give rise to branches with the same phyllotaxis as the main stem. Like other members of the mint family, Perilla has a simple vascular architecture. Leaves and branches on the same side of the square stem share vascular bundles (i.e. are in the same orthostichy); leaves and branches on different but adjacent sides share half of their bundles; those on opposite sides of the stem have no bundles in common ( Fig. 1b; see also Price et al. 1992 for a diagram of an analogous vascular system in Glechoma).

Figure 1.

Perilla frutescens. (a) Habit sketch, showing upper three nodes of a plant in fruit. (b) Cross-section of a stem, showing the placement of vascular bundles and the decussate arrangement of leaves. Numbers refer to node position of leaves. Each leaf is supplied by the pair of vascular bundles located on its side of the stem. (c) Developmental stage and location of labelled leaves (shaded) for each labelling treatment. The second plant of each pair has been rotated counter-clockwise relative to the first. By the third labelling stage, leaves at the first two nodes had abscised.

In a common garden experiment designed to relate seed production in Perilla to architecture at flowering, as determined by latitude and source population ( Preston 1998a), architecture was manipulated by growing plants from two different latitudes together in gardens at both latitudes. Perilla flowers in response to photoperiod, and flowering precludes further vegetative growth. As a result, plants grown in the south (regardless of origin) flowered earlier, having made fewer branches than those grown in the north. Total flower production per plant depended strongly on branch number, and there was no evidence that plants compensated for this architectural constraint. It was not clear from this experiment, however, whether the observed differences in plant architecture were due primarily to a difference in flowering time, or whether uncontrolled factors such as temperature or humidity were more important. Moreover, even if photoperiod had been the only variable in the experiment, differences in plant architecture could have accumulated over the season due to the longer days in the north. In the greenhouse experiment reported here, architecture was again manipulated by growing plants under northern and southern day-length regimes. The controlled conditions of the greenhouse allowed the effects of photoperiod to be separated from other environmental effects on plant growth. In addition, the influence of photoperiod on flowering date could be distinguished from its effects on developmental rate, either of which could affect architecture at flowering.

Manipulating architecture enabled the following questions to be asked: (i) Is total seed production constrained by meristem number, and therefore a function of architecture at flowering? (ii) Alternatively, are plants with fewer inflorescences able to compensate for this kind of architectural constraint by increasing the seed production of each inflorescence?

Another set of questions, regarding patterns of carbon translocation, were similarly motivated by results from a previous study. In that experiment ( Preston 1998b), the extent of carbon movement (integration) and the degree to which it was restricted by vascular tissue to vertical sectors or orthostichies (sectoriality) was shown to depend on developmental stage and the location of the leaf that was the source for assimilates. The current experiment addressed whether these basic allocation patterns are further influenced by plant architecture: (iii) Are differences in architecture correlated with differences in patterns of carbon allocation in Perilla? (iv) If so, are carbon translocation patterns related to the distribution of seed production among branches at each node?


Seeds were collected from individuals that had been transplanted from northern (Indiana, USA) and southern (Georgia, USA) sites and grown in a reciprocal common garden experiment at the northern site ( Preston 1998a). All parent plants had thus experienced the same environment during seed maturation. The seeds were separated into two groups based on the source populations of their maternal parents (northern vs. southern natives) and were placed on wet filter paper. Seeds germinating within a week were transplanted into flats containing equal parts perlite and a peat-lite potting mix. Three weeks later, the largest seedlings were transplanted individually into 7.5-cm plastic pots filled with peat-lite potting mix. Fertilizer was added to each pot weekly (25 ml standard Peter’s solution).

Plants from each maternal source latitude were divided between two day-length treatments (‘northern’ and ‘southern’) that recreated the photoperiodic conditions of the previous garden experiment, i.e. those of 3 June to 29 October in Indiana (39°N) and Georgia (33°N) ( Preston 1998a). As seeds from southern maternal plants were underrepresented among the most robust seedlings, each day-length treatment group contained 12 seedlings from southern maternal plants and 24 from northern maternal plants. In both day-length treatments, photoperiod was adjusted every fourth day to track changes in day-length over the growing season. During the light phase of the cycle, plants received ambient light supplemented by arc lights. Because the dark phase was controlled by covering plants with a black cloth supported by a frame, plants in a given photoperiod treatment were grouped together.

Architecture and reproduction

Growth curves for each day-length treatment were generated from weekly measurements of plant height, basal stem diameter (BSD), number of primary and secondary branches, and node number. Secondary branch primordia that failed to elongate were not counted. Beginning at week 6, leaf senescence was recorded as the number and nodal positions of leaves that had abscised in the previous week.

Plants were harvested at week 20. Roots and main stems were separated and weighed after drying at 60 °C. Fine-scale architectural and reproductive measurements were taken just before harvest on 24 of the plants (six plants from each source latitude from each day-length treatment): the number of secondary branches on each primary branch, the number of flowers on each primary and secondary branch and apical inflorescence, and the seed set and mean seed weight of each flower were recorded. In Perilla, each flower produces a maximum of four seeds (nutlets), which remain enclosed in an enlarged urn-shaped calyx. Occasionally, seeds were lost from the calyx, but seed set could still be determined by counting the scars left by mature seeds on the receptacle. In these cases, mean seed weight per flower was based on the remaining seeds, under the assumption that seed loss was random with respect to seed weight.

Carbohydrate translocation

At the beginning of the experiment, plants within each day-length treatment were divided among six labelling treatments and arranged in randomized blocks along a greenhouse bench. Each of the labelling treatments consisted of four randomly assigned replicates from southern maternal plants and eight from the northern source group. Treatments consisted of combinations of leaf position and developmental stage at which 14C labelling occurred ( Fig. 1c). Plants labelled at the vegetative stage had at least six nodes but had not begun flowering; at the first reproductive stage all treated individuals had visible flower buds; at the second reproductive stage, treated individuals had opened about half the flowers in their apical inflorescences. Plants to be treated at the vegetative stage were labelled at one of two leaf positions, the third or the sixth leaf above the cotyledonary node (leaf 3 or leaf 6). Either leaf 6 or leaf 9 (near the apex) was labelled at the two reproductive stages. No single plant was labelled at more than one time or at more than one leaf position. Because plants in the two day-length treatments differed in flowering time (by about 2.5 weeks), the date on which they reached the two reproductive stages also differed. Therefore plants in different day-length treatments were labelled not on the same date, but at the same time relative to their flowering date.

Labelled leaves were exposed to 9.35 × 105 Bq of 14CO2 for 15 min in a clear plastic enclosure, as in Landa et al. (1992) . After harvest at week 20 all plants were divided into stems, branches, roots and apical inflorescences. Stems and roots were dried at 60 °C, weighed, and ground individually in a Wiley mill (A. H. Thomas Co., Philadelphia, PA, USA) fitted with a no. 20 mesh. All seeds were separated from their calyces, grouped by branching order and orthostichy, and dried. Seeds from the orthostichy containing the labelled leaf (the ‘labelled orthostichy’) were further subdivided by node. For each plant part or group of seeds, two subsamples of dried material were weighed and combusted at 900 °C for 2 min in a biological oxidizer (model 4000; Harvey Instrument Corp., Hillsdale, NJ). The CO2 liberated from each sample was collected into 10 ml of 14C cocktail (Harvey Instrument Corp.) and transferred into scintillation vials, which were counted for 5 min each in a scintillation counter (LS230; Beckman, Fullerton, CA) to measure sample activity [disintegrations per minute (d.p.m.)].

For each plant part or group of seeds, the mean specific activity (d.p.m. per mg dry weight) was calculated, which reflects the ability to draw assimilate from the labelled leaf, independent of the part’s size. Total activity of roots and stems was calculated as the product of specific activity and total weight of each part. Mean total activity per seed was calculated as total activity of the subsample divided by the number of seeds in the subsample.


All analyses were performed using the SAS statistical package ( SAS Institute Inc. 1985). A series of t-tests was used to determine whether the origin of the maternal plant affected offspring morphology (height, basal stem diameter or number of primary branches). As only one of these comparisons indicated a significant difference within a day-length treatment due to maternal latitude (the mean basal stem diameter was greater in northern- than southern-derived plants grown under northern day-length conditions, P = 0.014), all plants within a treatment were pooled. To detect a block effect due to position along the greenhouse bench, final plant height and primary branch number were regressed against block position within each daylight treatment (PROC REG). No effect of position was found for the southern-treatment plants, and only height was affected significantly for the northern-treatment plants (P = 0.004). Pairwise comparisons (PROC GLM) indicated that only the endmost block differed in mean height from the other blocks in the northern treatment, and excluding that block from the regression eliminated the effect (P = 0.37). Block was therefore not included as a covariate in any of the analyses.

Differences between day-length treatments in morphological traits at harvest (height, basal stem diameter, number of primary and secondary branches, stem and root weight) were examined using t-tests. Multivariate repeated-measures analyses (PROC GLM) ( von Ende 1993) were used to compare the patterns of change in these traits during development under the two day-length conditions. The spatial distribution of seed production among branches was characterized by calculating the mean position of nodes on each plant, weighted by the number of seeds produced at each node. This mean centre of seed production was compared between day-length treatments using a t-test.

For inflorescences produced on primary branches, the number of flowers and seeds per inflorescence varied with node position, and nodes were correlated within individuals. To account for the non-independence of nodes within an individual, flower and seed numbers were compared between treatments using a multivariate repeated-measures analysis (PROC GLM) ( von Ende 1993). It is not valid to compare nodes directly between treatments because they occupy different relative positions. For example, plants in the two treatments differed in total number of nodes produced and in the position where the greatest number of flowers was observed (see the Results). Nodes on each plant were therefore converted to ‘relative nodes’ (ranging from 0 to 1) by dividing node position by the total number of nodes on that plant. Relative nodes were then partitioned into five groups: relative nodes between 0 and 0.20 fell into the bottom fifth, relative nodes between 0.21 and 0.40 into the next fifth, etc. The number of flowers or seeds per inflorescence was then averaged within each fifth, and these values were used as dependent variables in the analysis. This method allowed comparisons between equivalent portions of the plants while preserving much of the spatial pattern of flower and seed production. Differences in other reproductive traits (percentage seed set, seed weight) were examined using t-tests. A multivariate repeated-measures analysis was used to compare the temporal pattern of leaf senescence between the day-length treatments.

anovas were used to test the effects of labelling and day-length treatments on the specific activity and mean total activity per seed of seeds produced at the labelled node and at all the other nodes within the labelled orthostichy. Before analysing the activity of seeds produced at the other nodes, their mean activity was divided by the mean activity at the labelled node to minimize the effect of any between-plant variation in total amount of activity assimilated (due to leaf age, light intensity, stress, etc.). This relative activity was then averaged across unlabelled nodes for each plant to generate the mean relative activity of unlabelled nodes (within the labelled orthostichy) for each plant. Specific activity and total activity of roots and stems were analysed with an anova.

Sectoriality, or the degree to which assimilates were confined to the labelled orthostichy, was compared across treatments by a repeated-measures anova in which orthostichy was the repeated factor. A repeated-measures analysis accounts for the non-independence of the orthostichies within an individual. Differences between developmental stages in specific activity of seeds in each orthostichy were analysed with an anova, and comparisons of least squares means between stages were adjusted using the sequential Bonferroni technique ( Rice 1989). The number of seeds and the total weight of seeds produced in each orthostichy was compared within each day-length treatment using an anova to detect any systematic difference among orthostichies in seed production. No differences were found (seed number, P = 0.97; seed weight, P = 0.52); therefore the appropriate null hypothesis (indicating no sectoriality) was that carbohydrates would be distributed uniformly across orthostichies.


Growth, architecture and reproduction

Photoperiodic regime had a strong effect on flowering time, plant size and architecture. All plants grown under the southern day-length treatment began flowering at week 9, almost 3 weeks earlier than the northern-treatment plants. The shape of the growth curves differed between treatments, as indicated by significant week-by-day-length treatment interactions for all five traits (P < 0.0001). t-tests comparing treatments by week indicated that growth in the two groups was essentially identical until the southern-treatment plants flowered ( Fig. 2). The groups then began to diverge as the southern-treatment plants stopped growing stouter and made no more vegetative nodes ( Fig. 2a,b). Southern-treatment plants showed an early height advantage (week 5, t-test, P = 0.0001) that disappeared not long after the onset of flowering once internode elongation in these plants was complete. The northern-treatment plants continued to produce vegetative nodes and grow in height until they also began to flower ( Fig. 2c). In both treatments, few secondary branches elongated until after they and all the primary branches had begun flowering, perhaps because flowering at primary branch apices releases apical dominance in Perilla ( Woolhouse 1983). Nevertheless, no additional secondary branches could have been initiated after flowering, and once all existing branches had elongated increases in branch number levelled off ( Fig. 2d). By the end of the experiment, southern-treatment plants had fewer nodes and thinner stems, and were significantly shorter and more sparsely branched than northern-treatment plants (P < 0.0001 for each trait). Because the number of inflorescences was equal to the total number of branches on a plant plus one (the terminal inflorescence), the southern-treatment plants made significantly fewer inflorescences than the northern-treatment plants. Total root weight did not differ between treatments (P = 0.98) but northern-treatment plants had significantly heavier stems (P = 0.0001).

Figure 2.

Mean plant size and architecture by week for each day-length treatment: (a) diameter of the stem (mm) at the base of the plant; (b) number of nodes on the main axis; (c) plant height (cm); and (d) the number of primary (solid lines) and secondary (dashed lines) branches. Solid symbols indicate the period during and after flowering. Squares indicate northern-treatment (N) plants; triangles indicate southern-treatment (S) plants, The shape of growth curves differed between day-length treatments for all five traits (repeated-measures manova: P < 0.0001 for all traits; BSD, F = 20.8, d.f. = 10; nodes, F = 136.5, d.f. = 10; height, F = 45.1, d.f. = 11; primary branches, F = 139.5, d.f. = 10; secondary branches, F = 83.3, d.f. = 5). Arrows indicate the week that treatment means first differed statistically (t-tests, significance levels adjusted for multiple comparisons). Most traits were not measured weekly during senescence when plants were fragile, but secondary branches continued to elongate after flowering and were counted weekly until harvest.

Day-length treatments also differed in the distribution of seeds among inflorescences. First, day-length treatment affected the vertical distribution of flowers and seeds among branches ( Fig. 3). Both flower and seed production were centred just above node 7 (7.10 ± 0.76 and 7.18 ± 0.23, respectively) for the southern-treatment plants, and just above node 9 (9.25 ± 0.35 and 9.28 ± 0.33) for northern-treatment plants. The difference between day-length treatments was significant in both cases (t-test, P < 0.0001). Secondly, the number of flowers and seeds produced per inflorescence on the primary branches was greater in the southern- than the northern-treatment plants ( Fig. 3, Fig. 4b,c and Table 1). The number of seeds matured per flower on primary branches did not differ between treatments, but seeds per flower on secondary branches was higher in the southern-treatment plants ( Fig. 4c). The larger number of seeds per inflorescence produced by southern-treatment plants partially compensated for a smaller number of inflorescences, to the extent that total seed number per plant did not differ statistically between treatments ( Fig. 4d). Southern-treatment plants still made fewer seeds, however, and given the small sample size and large individual variation this difference (up to 14%) may still be biologically important, even if it is not statistically significant.

Figure 3.

Mean (± SE) number of seeds produced per primary branch at each node. The y-axis represents node position, counting from the base to the tip of the plant. Southern-treatment plants made more flowers (P = 0.02, data not shown) and more seeds (P = 0.01) per inflorescence than northern-treatment plants (repeated-measures manova).

Figure 4.

Means (± SE) of reproductive traits for the two day-length treatments. Significant differences between treatments are indicated by stars: ***P = 0.0003, **P = 0.001, *P < 0.1. Bars representing primary and secondary branches were stacked when the sum of their values was the total for the plant; bars are separate for non-additive traits.

Table 1.  Summary of the results from the multivariate repeated-measures analysis of variance comparing day-length treatments for the number of flowers and seeds per inflorescence. Inflorescences on each plant were divided into five zones distributed along the main axis
Flowers per inflorescence
Day-length treatment16.100.0222
Zone × day-length treatment45.620.0041
Seeds per inflorescence
Day-length treatment17.140.0143
Zone × day-length treatment42.360.0922

The greater number of seeds per inflorescence was not offset by a decrease in seed weight; in fact, the trend was towards heavier seeds in the southern-treatment plants ( Fig. 4e). Mass per inflorescence was therefore higher in southern-treatment plants and, as a result, the total amount of biomass in seeds (including both primary and secondary branches) did not differ between the treatments ( Fig. 4f). Again, statistical power was low, but the small variation in total seed biomass (less than 3%) suggests nearly complete compensation by seed number and mass for the large difference in inflorescence number.

Flowering date also significantly affected the onset and rate of leaf senescence. Plants in the southern day-length treatment began losing their leaves most rapidly during week 15, 3 weeks earlier than the period of most rapid leaf loss in the northern-treatment plants ( Fig. 5). When the rate of leaf loss was standardized to flowering date, leaf loss was most rapid during the sixth week after flowering in both treatments. Nevertheless, the two groups did differ in abscission rate, as indicated by a significant interaction between day-length treatment and week in the repeated-measures analysis of variance ( Table 2). Once abscission began, plants in the northern treatment lost their leaves more rapidly than the southern-treatment plants.

Figure 5.

Mean percentage leaves abscised per plant by week for each day-length treatment. Relative week (weeks since the onset of flowering) is represented by the x-axis. The period of most rapid leaf loss is also labelled by absolute week (weeks since the experiment began) next to the appropriate points on each curve.

Table 2.  Summary of the results from the multivariate repeated-measures analysis of variance comparing day-length treatments for the percentage leaves abscised by week
Day-length treatment19.620.0030
Week × day-length treatment83.590.002

Carbohydrate translocation

In both day-length treatments, the amount of radioactivity per seed was highest in seeds produced on the axillary branches of the labelled leaf (the ‘labelled branch’) ( Fig. 6), as was specific activity of seeds (data not shown). For these branches, both specific activity and activity per seed varied significantly with the labelling-by-day-length treatment interaction ( Table 3): for both leaf positions labelled during the vegetative stage, the northern-treatment plants had more activity per seed, but when leaf 6 was labelled during the second reproductive stage, the southern-treatment plants had the higher activity per seed in the labelled branches ( Fig. 6).

Figure 6.

Mean (± SE) activity (d.p.m.) per seed by node position within the labelled orthostichy for each day-length and labelling treatment. Note log scale. The developmental stage at the time of labelling is given to the right, and the position of the labelled leaf is given above each graph and on the x-axis. anova was used to compare the activity of seeds at the labelled nodes only; activities at other nodes are shown for comparison. Significant differences between day-length treatments in the activity of seeds in the labelled branch are indicated by stars (Bonferroni-adjusted comparisons of least-squares means): ***P = 0.0007, **P = 0.0024, *P = 0.012.

Table 3. anova summary table for activity at the labelled nodes and for activity at the unlabelled nodes, relative to the labelled node, within the labelled orthostichy. Activities were log-transformed prior to the analysis
Labelled nodesUnlabelled, relative to labelled nodes
Activity per seed
Full model1110.360.0001112.340.025
Day-length treatment10.010.91811.000.324
Labelling treatment517.150.000152.030.095
Day-length × labelling treatment54.580.00252.110.084
Specific activity
Full model1113.310.0001113.070.005
Day-length treatment11.350.25210.670.418
Labelling treatment520.060.000152.540.044
Day-length × labelling treatment57.800.000153.180.017

There was some evidence of an effect of day-length treatment on the specific activity of seeds produced on the other (non-labelled) branches within the orthostichy of the labelled leaf ( Fig. 7 and Table 3). The specific activity of seeds on these branches, relative to seeds on the labelled branch, was significantly higher in the northern- than the southern-treatment plants when leaf 6 was labelled during the second reproductive stage (P = 0.008). There was, however, no difference between day-length treatments in specific activity at any of the other developmental stages ( Fig. 7), nor at any stage was there an effect of day-length on total activity per seed in unlabelled branches relative to the labelled branch ( Table 3).

Figure 7.

Mean (± SE) specific activity (d.p.m. per mg) of seeds at each node within the labelled orthostichy, relative to the specific activity of seeds at the labelled node. When the relative specific activity was averaged over all non-labelled nodes within a plant and compared between day-length treatments, the only significant difference was for plants labelled at leaf 6 during the second reproductive stage (Bonferonni-adjusted comparisons of least squares means; P = 0.008).

The activity of roots and stems was somewhat affected by day-length treatment, as indicated by the significant and nearly significant labelling-by-day-length treatment interactions ( Table 4 and Fig. 8). The only significant differences between day-length treatments within a labelling treatment were for the first reproductive stage when leaf 9 was labelled. At this stage, the northern-treatment plants had higher specific activity in both stems (P = 0.0075) and roots (P = 0.0068) but there was no difference in total activity. Developmental stage had a more consistent effect on stem and root activity, which dropped between the first and second reproductive stages. This was true for both total activity and specific activity in northern- and southern-treatment plants labelled at leaf 6 and northern-treatment plants labelled at leaf 9.

Table 4. anova summary table for stem- and root-specific activity. Activities were square-root transformed prior to the analysis
Full model119.770.0001
Day-length treatment11.620.209
Labelling treatment519.030.0001
Day-length × labelling treatment52.370.054
Full model1122.530.0001
Day-length treatment12.160.148
Labelling treatment546.020.0001
Day-length × labelling treatment53.000.020
Figure 8.

Mean (± SE) specific activity (d.p.m. per mg) of roots and stems for each labelling and day-length treatment. Note log scale. Differences between day-length treatments are indicated by stars (t-tests adjusted for multiple comparisons): **0.005 < P < 0.01.

Carbohydrate translocation was highly sectorial at all developmental stages, but the degree of sectoriality varied with both day-length and labelling treatment ( Fig. 9 and Table 5). When plants were labelled at leaf 6 during the vegetative stage, the northern-treatment plants had significantly more activity in their labelled orthostichies, and thus were more sectorial, than the southern-treatment plants. This pattern had reversed by the second reproductive stage, when the southern-treatment plants were significantly more sectorial. Sectoriality increased over time in the southern-treatment plants at leaves 6 and 9, and in the northern-treatment plants when leaf 9 was labelled. In northern-treatment plants labelled at leaf 6, however, sectoriality was high at the vegetative stage and remained constant over time ( Fig. 9).

Figure 9.

Mean (± SE) specific activity (d.p.m. per mg) of seeds within an orthostichy for each labelling and day-length treatment. Note log scale. Developmental stage at the time of labelling is given to the right, and the position of the labelled leaf is given above each graph. Differences between day-length treatments in the activity of the labelled orthostichy are indicated by stars within the graphs (repeated-measures manova): ***0.0001 < P < 0.0005. Differences within a day-length treatment between developmental stages were as follows. Southern-treatment leaf 6: vegetative vs. first reproductive P < 0.001, first vs. second reproductive P = 0.0001; northern-treatment leaf 6: NS; southern-treatment leaf 9: first vs. second reproductive P = 0.0003; northern-treatment leaf 9: first vs. second reproductive P = 0.005.

Table 5. anova summary table for sectoriality of resource movement by day-length and labelling treatment. Activities were log-transformed prior to the analysis
Orthostichy × day-length treatment24.140.022
Orthostichy × labelling treatment109.930.0001
Orthostichy × day-length × labelling treatment102.050.037


Growth, architecture and reproduction

In Perilla, the effect of photoperiod on architecture is important because of the constraints that architecture places on reproduction. Architecture limits meristem number and, in this species, meristem number is a resource essential for flower production. This experiment has demonstrated that photoperiod affects architecture simply by determining flowering time: plant growth curves did not differ between day-length treatments until the southern-treatment plants began to flower ( Fig. 2). At flowering, apical meristems were converted into inflorescences, which precluded further vegetative development and thereby fixed plant architecture.

Because the southern-treatment plants flowered earlier, they had fewer branch apices available for flowering than the plants in the northern treatment ( Fig. 2d). Consequently, architectural constraints were more strongly limiting for the southern-treatment plants. Despite limitations on inflorescence number, these plants suffered only a moderate (approximately 14%) reduction in seed number compared with the northern-treatment plants, and total seed mass did not differ between treatments ( Fig. 4d,f). Plasticity at the level of the inflorescence allowed southern-treatment plants to compensate for fewer inflorescences by increasing the number of flowers and seeds produced per inflorescence ( Fig. 3 and Fig. 4b).

Although seed production is theoretically not limited in either treatment (given indeterminate inflorescences), total seed number per plant actually remained within a fairly narrow range ( Fig. 4d), indicating that seed production was in fact limited by one or more resources. Plants in both treatments made a similar total number of seeds, and therefore the number of seeds per inflorescence was somehow less limited in the southern-treatment plants. One possible explanation is that mineral nutrients set the upper limit on seed production. Plants in both treatments presumably had equal access to nutrients because they had the same root mass (t-test, P = 0.98) and received the same amount of fertilizer every week. If mineral nutrients limited seed production, then perhaps all the plants were able to produce roughly the same mass of seeds ( Fig. 4d,f) and the southern-treatment plants simply packaged them into fewer but larger inflorescences ( Fig. 3).

If, instead, carbon limited seed production, then a higher photosynthetic rate might explain how southern-treatment plants can support more seeds per node (and thus per leaf) than the northern-treatment plants. Because the southern-treatment plants flowered nearly 3 weeks earlier than the northern-treatment plants, the leaves present during seed fill were, on average, younger. In Perilla, photosynthetic rate reaches its peak at full leaf expansion and declines steadily with leaf age ( Batt & Woolhouse 1975), and the leaves on the southern-treatment plants were therefore probably capable of assimilating more carbon during seed fill than leaves in the later-flowering group.

Another mechanism by which plants can enhance total carbon gain is by retaining actively photosynthetic leaves for a longer time. Some evidence for this was found among southern-treatment plants. Although leaf drop in this treatment occurred 3 weeks earlier than in the northern treatment, it began at the same whole-plant developmental stage (late in seed development) ( Fig. 5). Thus, in Perilla, leaf loss is not simply a function of leaf age but instead appears to be triggered by seed maturation, as in soybean ( Noodén 1984). Moreover, senescence proceeded more slowly in southern-treatment plants. The mechanism behind the treatment difference in senescence rate is not clear, but it may have allowed southern-treatment plants access to relatively more carbon to support seed fill.

Carbohydrate translocation

Just as the distribution of seeds among branches differed between day-length treatments, so did the distribution of carbohydrates. This result is consistent with other work showing that architecture influences carbohydrate translocation patterns ( Chapman et al. 1992 ; Botella et al. 1993 ). The stage at which plants were labelled and the location of the source leaf also significantly affected patterns of carbohydrate translocation, as in a previous greenhouse labelling experiment ( Preston 1998b).

Some of the translocation patterns were similar between the day-length treatments. For example, in both treatments, allocation to roots gradually tapered off throughout development ( Fig. 8). This pattern is consistent with the observation that flowering in Perilla is accompanied by a decline in root growth ( Woolhouse 1983). Both day-length treatments also showed a drop in stem activity between the first and second reproductive stages ( Fig. 8), which was similar to the pattern observed in the earlier labelling experiment ( Preston 1998b). In that experiment, plants harvested early (1 week after labelling) still showed high levels of stem activity, which suggested that assimilates were remobilized between labelling and the final harvest, perhaps to support seed fill.

The greatest difference between the treatments was in the activity per seed at labelled nodes ( Fig. 6). In plants labelled at the vegetative stage, activity per seed was significantly higher in the northern- than the southern-treatment plants, and activity per seed was negatively associated with seed number. In plants labelled at the second reproductive stage, this pattern was reversed. These contrasting patterns may be explained by a difference between labelling treatments in the origin of the labelled assimilates (i.e. current vs. remobilized) allocated to seeds. Plants labelled while vegetative had not even begun to flower when they were labelled, and therefore any activity recovered in seeds must have been remobilized from elsewhere in the plant. In contrast, plants labelled at the second reproductive stage were maturing seeds at the same time that they took in labelled carbon, thus developing seeds had access to labelled current assimilate.

In the plants labelled while vegetative, assimilate was remobilized and then distributed among seeds developing at the labelled node. There were fewer seeds per node in the northern- than in the southern-treatment plants (significant for node 6, t-test, P = 0.0006), suggesting that the activity per seed was greater under the northern treatment because these plants had fewer seeds over which to spread a similar amount of remobilized carbon. In plants labelled during the second reproductive stage, seeds were importing carbon at the time plants were labelled, and results from these plants are consistent with conventional expectations that current assimilates flow from a source leaf into the nearest ( Wardlaw 1990) most active sink ( Wardlaw 1990; Minchin et al. 1993 ) that has a direct vascular connection to the source leaf ( Watson & Casper 1984; Wardlaw 1990). For example, seeds at the labelled node maintained their priority over seeds at more distant nodes ( Fig. 6) and seeds in other orthostichies ( Fig. 9). Furthermore, the amount of activity translocated to each seed at the labelled node increased between the first reproductive stage and the second, when the demand for carbon at the labelled node was greatest. This was most dramatic in the southern-treatment plants labelled at leaf 6, where the activity of seeds at the labelled node increased not only in absolute terms ( Fig. 6) but also relative to other seeds developing in the labelled orthostichy ( Fig. 7).

The size of the increase in activity per seed between the first and second reproductive stages appeared to be related to the number of seeds maturing on each inflorescence. The southern-treatment plants, which made more seeds per inflorescence than the northern-treatment plants, showed the more dramatic increase between reproductive stages. This pattern was echoed within the northern day-length treatment: the greatest increase in activity per seed occurred at node 9, the node with the most flowers per inflorescence. It is important to note that activity per seed increased, not just total activity of all seeds at a node, because it suggests that the number of seeds produced at a node influenced the amount of labelled assimilate available to each seed at that node.

These results show clearly that differences between the day-length treatments in morphology and reproduction were correlated with differences in carbon allocation, but factors other than assimilate demand may also have influenced the observed patterns. In addition to the activity of tissues that import assimilate (‘sinks’), such patterns typically depend on the amount of assimilate supplied by the source leaf and on the conducting tissue between them ( Minchin & Thorpe 1993; Minchin et al. 1993 ). The plants in the two day-length treatments differed in all three of these features. First, of the two treatment groups, southern-treatment plants made more seeds per inflorescence, which suggests that the assimilate demand at each node was also greater. Secondly, the southern-treatment plants had, on average, younger leaves at flowering and thus a potentially greater supply of assimilate at each source leaf than the northern-treatment plants. Finally, it is possible that the plants differed in vascular capacity as well. In Perilla, phloem development is inhibited at flowering and sieve tubes become partially blocked by callose ( Woolhouse 1983). As the northern-treatment plants had more time to develop phloem, they may have had more functional phloem cells during seed fill than the southern-treatment plants. It is therefore probable that variation in each of these components contributed to treatment differences in translocation patterns.

Nevertheless, the results from this experiment point to a prominent role for sink demand in carbon movement, whereas the roles of pathway and source appear to be less important. For example, differences in phloem capacity should have been manifest by the first reproductive stage, but translocation patterns were most similar at this stage. Furthermore, at every whole-plant developmental stage, the average age of leaves on the southern-treatment plants was less than the average for the northern-treatment plants, yet activity in the southern-treatment seeds was higher only during the second reproductive stage. In contrast, variation in the carbon demand of seeds, stems and roots over development was accompanied by parallel shifts in the distribution of activity among these tissues. Therefore, the interpretation of the data focuses on differences in patterns of seed production between the two treatments, even though carbon demand was only one factor influencing translocation patterns.

Ecological consequences of architecture

In this experiment, the photoperiodic environment strongly affected plant architecture, which in turn affected both the distribution and packaging of seed production and the allocation of carbon within the plant body. Whereas architectural effects on reproduction are obviously important to plant fitness, the ecological significance of the observed translocation patterns are not as clear. In a previous labelling experiment ( Preston 1998b), leaves varied in the amount of assimilate they supplied to branches at other nodes. In those plants, the consequences of leaf loss for the carbon budget of the rest of the plant would depend on the particular export patterns of the lost leaf. In this experiment, the day-length treatments did not differ in the absolute amount of assimilate that was exported beyond the labelled node (P = 0.95). Rather, they differed only in the amount of assimilate exported to other nodes relative to the amount retained at the labelled node. Thus, it is not clear that damage to a given leaf would have a differential effect on the whole-plant carbon economies of northern- and southern-treatment plants. It seems more likely that the distribution of seeds among branches would be ecologically important. For example, losing a source leaf to disease or herbivory would remove the primary source for a greater proportion of the seeds from a southern- than a northern-treatment plant.

In contrast, the relationship between architecture and reproduction obviously had potential fitness effects. Under some conditions, plants are able to overcome architectural constraints on reproduction, as was observed in this experiment. Compensation is not always possible, however. For example, there was no evidence that southern-grown plants in a garden experiment ( Preston 1998a) produced a greater number of flowers per inflorescence than the northern-grown plants. Together, these two experiments demonstrate that environmental effects on a whole-plant trait (architecture) can impose a constraint on another trait (inflorescence number). Under the right conditions, however, plasticity at other morphological levels (flowers per inflorescence and seeds per flower) can compensate for this constraint.


I thank David Ackerly, R. Lanier Anderson, Niels Anten, Carl Schlichting, Maxine Watson, and an anonymous referee for helpful comments on the manuscript, and D. Ackerly and Harold Mooney for providing space and a supportive atmosphere in which to write. I am grateful to the Indiana University greenhouse staff for care of the plants. Partial support for this research came from the Indiana Academy of Sciences, Sigma Xi, and the Indiana University Department of Biology.

Received 18 May 1998revision accepted 28 January 1999