Summary
Flowering plants often exhibit declining investment in gametes, seeds, and floral organs among sequentially blooming flowers within inflorescences. However, few experimental studies have determined the relative importance of proximate underlying causes such as architectural constraints or resource competition among flowers, and fewer still have identified likely selective factors that favor sequential variation in allocation. By combining observational and experimental studies in natural populations of Aquilegia canadensis, we showed that ovule and pollen production declined with flower sequence by 9% and 18%, respectively, due to architectural effects. However, a much stronger decline in fruit set (27%) and seeds per fruit (58%) was mediated entirely by competition for resources among flowers. Reducing resources by defoliation greatly reduced fruit set (−37%) and seeds per fruit (−61%) and increased the severity of the decline in fruit set. Experimentally alleviating resource preemption by first flowers greatly increased the seed production of later flowers (+39%) to the extent that fourth flowers on manipulated plants set as many seeds as first flowers on control plants. In contrast, aspects of floral morphology including sepal size, nectar spur length, and herkogamy showed little or no sequential variation and were relatively insensitive to experimental manipulation of resources. We also investigated whether the positional decline in allocation was adaptive. Disproportionate investment in early flowers was not associated with a seasonal decline in important environmental factors (e.g., light, water, soil nutrients) or cross‐pollination. Marker‐gene assays also failed to show that early flowers consistently produced a greater proportion of high‐quality outcrossed progeny. All flowers had a high probability of being eaten by herbivores (mean = 70%). However, the risk of herbivory was lower for early flowers (64%) than later flowers (79%). Herbivory may, therefore, select for higher allocation to early flowers as well as flexible investment mediated by interflower competition because it allows the reallocation of resources to later flowers when early flowers are lost to herbivores. This adaptive hypothesis is further supported by comparative data from closely related A. caerulea, which differs from A. canadensis in the nature of herbivory as well as the pattern and mechanism of sequential variation in reproductive allocation.
Methodology
Study species and population
Aquilegia canadensis L. (Ranunculaceae) is a shortlived, spring-flowering perennial common throughout eastern and central North America (Whittemore 1997). Each plant produces a basal rosette of 1–5 ternate leaves, which is likely the main site of photosynthesis. The main storage organ is a thumb-sized rootstock. Mature plants produce one or rarely two panicles of large flowers. First flowers are not located closer to the basal leaves and roots than later flowers. The long primary shoot of the inflorescence terminates in a single flower (the 1° flower). Relatively short secondary shoots arise from axillary meristems subtended by small cauline leaves on the primary shoot and are also terminated by individual flowers (2° flowers). Even shorter 3° shoots arise from axillary meristems on 2° shoots, and so on. A typical inflorescence consists of one 1° flower plus two or three 2°, and more rarely some 3° flowers. Primary flowers almost always open first, followed by 2° and 3° flowers. Although flower and inflorescence development has not been studied in fine detail, there are no indications of floral preformation. Whole inflorescences and individual flowers develop from extremely small meristems during the current flowering season. Moreover, all floral organs (pistil, stamens, spurs, sepals) increase in size throughout floral development until the flowers fall apart (Griffin et al. 2000). All this suggests that there is the potential for short-term dynamic allocation of resources within the inflorescence, and that perturbations of inflorescence and flower development may alter the sequential pattern of resource allocation.
Variation in reproductive allocation across the flowering sequence
We quantified the decline in ovules, pollen, fruit, and seed across the flowering sequence under field conditions. For ovules and pollen, we randomly selected 40 plants with ≥4 buds across all subpopulations in 1997; collected every flower produced just before anthesis; counted carpels, anthers, and ovules; and estimated pollen production by completely grinding the anthers in 0.5 mL 3:1 glycerol:lactic acid and counting pollen in eight replicate 0.009 mL subsamples using a haemocytometer. Because flowers were collected before pollination, variation in gamete production would be due to architectural effects only (sensu Diggle 1997). To quantify fruit set, seeds per fruit, and the mass of individual seeds for naturally pollinated flowers, we randomly selected 22 plants in 1997 and 36 in 1999, each with ≥ 4 flowers, collected all resulting fruits, and counted and weighed (to 0.01 mg, 1999 only) the seeds in each. Because natural pollination supplies sufficient pollen for maximum fruit set and seed production (Appendix C), sequential variation in fertility could arise from a combination of architectural effects and resource competition among flowers (Diggle 1997). The naturally pollinated plants from 1999 were the controls for the experimental manipulations described in the following paragraphs.