Plants might allocate chemical defenses unequally within attractive units of flowers including petals, sepals, and bracts because of variations in the probability of florivory. Based on the optimal defense theory, which predicts that plants allocate higher chemical defenses to tissues with higher probabilities of herbivore attack, we predicted that distal parts and sepals would have higher chemical defense allocations than proximal parts and petals. To test this prediction, we compared total phenolics and condensed tannins concentrations as well as presence of florivory within attractive units of ten angiosperm species. In agreement with the prediction, the overall results showed that the distal parts had higher total phenolics and condensed tannins than the proximal parts. On the other hand, contrary to the prediction, petals and sepals showed no tissue-specific variations. Florivory was more severe on the distal parts than the proximal parts, although statistical support for the variation was slightly weak, while the variations in presence of florivory between petals and sepals differed between the distal and proximal parts. These results may support the prediction of the optimal defense theory because distal parts of attractive units had higher presence of florivory and concentration of chemical defenses.

Studied species

We selected ten perennial species (Table 1) whose flowers are mono-colored or have a uniform pattern along their attractive units, based on visual inspection. We selected these species to obtain sufficient samples for chemical measurements and to minimize the effects of color pattern on distributions of chemical compounds along attractive units. One exceptional species (Houttuynia cordata) having no perianth but white showy petaloid bracts was also used. For species belonging to Asteraceae (Aster ageratoides, Aster microcephalus and Helianthus tuberosus), we used only ray flowers because dissecting small tubular flowers was impractical. Aquilegia buergeriana has two floral phenotypes (purple and yellow) and we measured the chemical defenses and florivory separately for each phenotype.

Study sites

We collected samples from twelve locations in Miyagi, Iwate, and Aomori prefectures in northern Japan (Table 1). Aobayama, Aramaki, and Kawauchi are suburbs of Sendai city, Miyagi Prefecture, where we sampled flowers growing in natural vegetation facing paved or unpaved roads. Kagitoriyama and Izumigatake are mountainous regions in Miyagi Prefecture, where we collected samples growing in deciduous-coniferous mixed forests. Koeji and Nakanose are also suburbs of Sendai city, where we collected samples growing in natural vegetation along the Hirose river. Tsuta, Tashiro, and Hakkoda are mountainous regions in Aomori Prefecture, where we sampled flowers growing in damp vegetation in Tsuta, in dry vegetation in Tashiro, and in natural vegetation in the Botanical Garden of Tohoku University in Hakkoda. Hayasaka and Sodeyama are plateau grassland areas in Iwate Prefecture, where we sampled flowers growing in natural vegetation.

Sample collection

We sampled non-damaged flowers without apparent color changes caused by aging for chemical measurement during June to September 2016. For the chemical trait measurements, a minimum amount of dry mass was required. Although measuring the chemical traits for each individual would be ideal, for many species it was not possible to obtain sufficiently large samples from single individuals. Therefore, to keep some natural variation in the chemical traits among individuals and simultaneously fulfill the minimum requirement for the chemical measurements, we made sample groups consisting of several individuals. Because the number of flowers needed for one measurement depended on the flower size, we collected a suitable number of flowers for each group for each species (Table 1). In each location, we randomly selected patches of several individuals so that we could treat flowers collected from one patch as one sample group. We collected flowers for each sample group from individuals growing close to each other. All procedures after collection of flowers were applied to each sample group so that flowers from different groups were never mixed together. We cut flowers at their pedicels or peduncles and returned them to the laboratory as soon as possible.

In the laboratory, we separated attractive units, i.e., petal, sepal, and bract samples into several parts using scissors or razors to compare the differences in chemical concentrations among them. Because of the diversity of floral morphology, we couldn’t apply a single segmentation method to all species. Therefore, we applied simple anatomical segmentations for most species but applied specific segmentation methods for some species. The list of segmentation methods of attractive units of flowers for each species and assignment of categories [i.e., position (distal or proximal) and part (petal or sepal)] for each segmented attractive unit is shown in Table 2.

For species belonging to Asteraceae (A. ageratoides, A. microcephalus, H. tuberosus) and H. cordata, we divided each petal and bract into the proximal and distal parts at the center. For the species having colored sepals that attract pollinators (L. auratum, and A. buergeriana), we divided proximal and distal parts of petals and sepals separately by applying the segmentation rule above. Spurs of A. buergeriana were separated for measurement of chemical traits and treated as proximal parts of petals.

Because H. sieboldii has corollas consisting of a narrow tubular section, an expanding section, and a lobe section, we divided them into these three parts for measurement of chemical traits. For statistical analyses, we treated the former two as the proximal parts and the lattermost as distal parts.

Aconitum japonicum has zygomorphic flowers consisting of petals with spurs and three types of decorative sepals (lower sepals, lateral sepals, and a helmet) covering petals. Hence, we separated them into 1) petals with spurs, 2) lower sepals, 3) lateral sepals, and 4) helmet. The three types of sepals were also separated into distal and proximal parts by the same rule noted above. Because the petals including spurs of A. japonicum were quite small, the need to obtain sufficient sample for the chemical measurements precluded separating them into distal and proximal parts. Thus, the chemical traits were measured for seven groups for this species.

Cardiocrinum cordatum sets slightly warped actinomorphic flowers horizontally. Each flower has three petals and sepals. The upper side of each flower has two sepals and one petal, while the lower side has one sepal and two petals. Therefore, we measured the chemical traits for the upper and lower petals and sepals separately. Also, because flowers of the species seemed to consist of a narrow tubular section, an expanding section, and a lobe-like section similar to H. sieboldii, we separated the petals and sepals into these three parts. For the statistical analyses, we treated the tubular and expanding sections as proximal and the lobe-like section as distal.

Flowers of H. fulva have inner and outer petals, so we separated the attractive parts of each flower into inner petals, outer petals, and sepals. Each of these three attractive parts was further divided into distal and proximal parts.

During sample dissection, we removed any pollen attached to floral samples by brushing gently. For the species having spurs (A. japonicum and A. buergeriana), we also removed nectar from them. We dried the separated samples with silica gel and kept them at room temperature in the silica gel until they were used for the chemical measurements.

Chemical trait measurement

Before the chemical measurements, we ground dried samples of a segmented group into powder using Micro SmashTM MS-100R (TOMY DIGITAL BIOLOGY CO., LTD., Tokyo, Japan). Following the Folin-Ciocalteu method (Julkunen-Tiitto, 1985), we measured total phenolics concentration using tannic acid (Wako Pure Chemical Industries, Ltd., Osaka, Japan) as the standard. We followed the butanol-hydrochloric acid-iron assay to measure condensed tannin concentration using commercially available cyanidin chloride as the standard (Makkar et al., 1999). The concentrations of total phenolics and condensed tannins were calculated from the absorbance values obtained using a Multiskan GO spectrophotometer (Thermo Fisher Scientific, Vantaa, Finland). To control for possible effects of differences in conditions among the measurement batches (i.e., sets of samples measured simultaneously), we measured all flower parts from the same sample group in one measurement batch and measured only one sample group in one batch for each species so that the fluctuation among measurement batches similarly affects all flower parts. In addition, our preliminary measurement showed that measurements with different sample weights resulted in slightly different measured values even for the same sample (Fig. 1). To reduce the effects of sample weight variation, we equalized the sample weight of each part to 15 mg, the standard weight of a powdered sample needed for the measurements, when possible. However, because some separated parts of some sample groups lacked the required weight, we equalized the sample weight of each part to the minimal weight among the parts in the same sample group. Because the dry weights were too light to equalize, we omitted the distal part of H. sieboldii in three sample groups. Also, because the weights of distal and proximal samples of lateral sepals in eight sample groups of A. japonicum were less than 5 mg, the chemical traits were measured using mixed samples of distal and proximal parts to avoid inaccurate measurement due to the measurement limit of the spectrometer.

Measurements for presence of florivory

It was impractical to precisely measure the amount of florivory for each part of attractive units for species in Asteraceae because they have many small ray flowers and in Ranunculaceae because they have complex floral structures. To apply the same method to all species and all parts of attractive units, we measured florivory as presence/absence. During the sample collections for the chemical measurements, we measured presence of florivory for randomly selected flowers. We recorded signs (i.e., missing parts on petals, sepals, and bracts) of florivory damage for proximal and distal parts of petals/sepals/bracts of each sampled flower in a binary manner. After the measurements, non-damaged flowers without apparent color changes due to aging were included in the chemical measurement samples.

References

Julkunen-Tiitto R. 1985. Phenolic constituents in the leaves of northern willows - methods for the analysis of certain phenolics. Journal of Agricultural and Food Chemistry 33: 213-217.
Makkar HPS, Gamble G, Becker K. 1999. Limitation of the butanol-hydrochloric acid-iron assay for bound condensed tannins. Food Chemistry 66: 129-133.

Figure legend

Fig. 1: Relationship between measured sample weight and concentration of total phenolics in Lilium auratum shown by the preliminary measurements. Symbols connected by a dashed line represent same sample measured with different weights. The figure file is included in the methods directory of the dataset.

Tables

Table 1: Detailed information of the species and sampling sites. Numbers of flowers and sample groups for each sampling site are also shown.

† Because petals of Aconitum japonicum are too small for separate measurements of chemical traits for distal and proximal parts, chemical traits of whole part of petals were measured for the species.

Table 2: List of sample segmentation methods for each species. Assigned categories (Sepal or Petal and Distal or Proximal) for the floral parts are also shown.

Please see the readme.md or readme.html files in the dataset for the details of the dataset.