Study system
Fuchsia parviflora (Onagraceae) is an endemic species from Mexico that inhabits ravines in pine-oak forests between 1550 and 2500 m asl, encompassing the states of Michoacán, Guerrero, Jalisco, Nayarit and Estado de México. It is a hummingbird and bee pollinated dioecious shrub of 1.5–4 m (Breedlove 1969; González et al. 2018), with smaller tubular red flowers in females (3.2–5.7 mm long) than in males (8.5–11.2 mm). Fruits are rounded berries with 14–20 small seeds 1.9–2.3 mm long. Opposite oblanceolated to ovate leaves (30–75 mm long) are pubescent in both surfaces (Breedlove 1969). The flowering season is from June to December, with a flowering peak in August (González et al. 2018). Their fructification period is from August to December, with a peak in fruit production during October (González et al. 2018), and their leaves are shed from January to March, suffering a total or partial lack of leaves in March.
Study Site And Field Procedures
Four populations of Fuchsia parviflora were located in the central part of Michoacán, Mexico within three municipalities which are part of the Pátzcuaro and Zirahuén lake basins: Pátzcuaro (Residuos Sólidos), Salvador Escalante (Agua Verde and Zirahuén) and Quiroga (Chupícuaro; SI. S1).
Twenty females and 20 males from 1–4 m high were tagged in each population. Herbivory intensity was evaluated as the proportion of leaves damaged per plant and as the proportion of leaf area removed by herbivores. We decided to estimate leaf herbivory using both methods because they offer complementary information, besides, even whether the proportion of removed area was higher for one gender, it might be lower at the individual plant level because there might be less leaves that have been damaged in total. To estimate the proportion of leaves damaged per plant, the total number of leaves and the total number of leaves with herbivory were quantified in each plant. We considered leaves with herbivory as those that presented some type of mastication (missing holes or fragments in the edge of the leaf) or skeleton (bites without touching the leaf rib). To estimate the total number of leaves in plants higher than two meters, the total number of leaves in a branch multiplied by the number of branches were quantified. Surveys were performed during the five phenology stages during 2017, approximately every two months: before flowering (June), during flowering peak (August) during fruit production peak (October), after flowering and fruit production (January), and during leaf shedding (March).
To determine the proportion of leaf area removed, 10 leaves per plant with herbivory were collected at random (from the top, intermediate and bottom part of each shrub) during each census and transported in a cooler to the laboratory. To obtain the proportion of area removed, the collected leaves were scanned to estimate their total area and the leaf area removed with the aid of Image J 1.x (Schneider et al. 2012). During 2018 herbivory was estimated with both methods only during October, as it was the month with higher herbivory during 2017 (Fig. 1).
Trichome density was estimated in young leaves only (2 months old aprox.), because in older leaves they are lost. To this end, a leaf from 10 males and 10 females was collected in each population and the number of trichomes were estimated in a 1 x 1 cm square, both in the adaxial and the abaxial leaf surfaces using a magnifying glass.
Cafeteria Experiment
In October 2018, 10 caterpillars from an unidentified butterfly but recognized as the main herbivore of F. parviflora and 10 generalist crickets (Pyrgomorphidae), that were also observed feeding at the leaves from the study species, were collected from Chupicuaro population. All caterpillars were of similar size, probably belonging to the same larval stage. Before the experiment, both caterpillars and crickets were starved for 12 hours. Then, caterpillars were placed in Petri dishes with humid filter paper to keep moisture, and crickets in plastic jars with little holes in the covers. Two leaves of approximately the same size, one from a female and one from a male plant, were placed at the extremes of each Petri dish or plastic jar. We used five male and five female plants for the experiment previously extracted from three populations (4 plants from R. Sólidos, 4 from Chupícuaro, and 2 from Agua Verde) and kept in pots inside a shade house. Leaves were cut with a scalpel and collected with forceps and gloves to avoid any contact that could have an effect on the herbivore’s preference. Caterpillars and crickets were placed in the center and observed until they chose one of the leaves. Once the insect’s election was observed, they were allowed to feed on the leaves to calculate the total area removed. Each caterpillar and cricket was offered leaves of a different female and male plant every day for 5 days.
Statistical Analyses
To determine differences between males and females, phenological stages and populations in the proportion of leaves damaged and the proportion of area removed per leaf during 2017, we conducted a generalized linear mixed model (GLMM) with binomial distribution and logit function. The model included: i) the proportion of leaves damaged and ii) the proportion of area removed as dependent variables. The effect of gender, phenological stage (census at which the data was taken) and population, as well as all their interactions were included as fixed factors, while plant nested within population was included as a random effect (Pinheiro and Bates 2000). In a second GLMM, we considered only data from October 2017 and 2018, as this month was the one with the highest herbivory intensity (see results). The model included the effect of the year, gender and population, as well as their interactions as fixed factors. We also included the plant nested within population as a random effect, while the proportion of leaves damaged and the proportion of leaf area removed were the dependent variables. In both models, we included population as fixed factor because we were interested to test differences among populations. For all response variables, we developed saturated models which included all main terms and their interaction. We then remove the non-significant terms and obtained the minimal adequate model (Crawley 2012) and finally, and performed multiple comparison tests.
To determine whether the number of trichomes differed among genders, a general linear model was performed considering gender, population and the side of the leaf (adaxial and abaxial surfaces) as fixed factors, as well as the number of trichomes as the dependent variable. The number of trichomes were square root transformed.
Finally, differences between genders in the number of leaves chose by each caterpillar and cricket and in the proportion of area removed were analyzed by means of a paired sample t-test (males vs females). All analyses were conducted in R v. 3.3.2 (R Development Core Team 2016), while for mixed models we used the lme4 package (Bates et al. 2015). Multiple comparisons were conducted with the multcomp package v1.4-16 (Hothorn et al. 2008).