Spillover of a biological control agent (Chrysolina quadrigemina) onto native St. Johnswort (Hypericum punctatum)

Study organisms

The genus Hypericum contains nearly 500 species spread throughout temperate parts of the world (Nürk & Blattner, 2010). This study focused on two species that co-occur in old-fields in central New York: the invasive H. perforatum and its native congener, H. punctatum. Both are perennials. H. perforatum is native to Europe, western Asia, and northern Africa, but arrived in the United States no later than the 17th century (Josselyn, 1672). One well-studied defensive chemical in Hypericum species is hypericin (Avato & Guglielmi, 2004; Smelcerovic et al., 2007), a compound that has increased toxicity in the presence of visible light (i.e., phototoxicity), with an action spectrum peak between 540 and 610 nm (Fields, Arnason & Philogène, 1990). In addition to its effect on invertebrate herbivores, hypericin is also toxic for grazing livestock (Bourke, 2000), which prompted the introduction of Chrysolina quadrigemina as a biological control agent for H. perforatum in the 1940s (Andres, 1985). The native range of C. quadrigemina overlaps with that of H. perforatum in the Mediterranean region, where it consumes only H. perforatum and H. tomentosum and oviposits on H. perforatum (Wilson, 1943). Despite ingestion of phototoxic hypericin, adult Chrysolina quadrigemina expose themselves to sunlight by perching high up on Hypericum stems during the day (Fields, Arnason & Philogène, 1990). In laboratory experiments, however, the beetles have demonstrated negative phototropism (Wilson, 1943).

Location

We conducted our field experiments in former agricultural fields belonging to Cornell University, Ithaca NY. To determine the natural distribution of Hypericum, we collected data in 5 fields (Durland Bird Sanctuary, 42°26′19″N, 76°23′54″W; Freese Road, 42°28′2″N, 76°26′38″W; Neimi Road West, 42°30′1″N, 76°26′7″W; Neimi Road East, 42°30′1″N, 76°26′3″W; and Whipple North, 42°29′32.13″N, 76°25′40.88″W). We established the transect experiment and sun/shade experiment in fields surrounding an experimental field station (Dryden NY, 42°27′51″N, 76°26′40″W).

Feeding assays

To determine whether beetles preferred the native or exotic Hypericum species, we conducted feeding assays. In July 2011, we collected adult beetles from wild Hypericum plants, stored them in containers with a damp paper towel, and starved them for 24 h prior to feeding trials. We tracked whether the beetles were collected on H. punctatum (N = 17) or H. perforatum (N = 8). We collected as many beetles as possible for the feeding trials before they began their late-summer period of inactivity, and although we attempted to collect equal numbers of beetles from each plant species on three days over the course of one week, we did not manage to find as many beetles on H. perforatum during the second and third days of collection. For each trial (N = 25), we paired undamaged stems of equal length from greenhouse grown Hypericum plants, placing them in a water tube and inserting these into a soil-filled pot, 10 cm in diameter. We placed one beetle on the soil surface between the two stems, encased the pots in mesh bags for 2.5 h, and recorded which plant, if any, the beetle chose to eat by the end of the trial. All trials occurred within an outdoor shade tent. To determine whether the beetles preferred the native or exotic Hypericum, we analyzed feeding preference with a χ-square test. We excluded trials in which the beetles made no choice (N = 11), using only data in which the beetle fed on one of the two plants offered (N = 14). Beetles never fed on both species during a trial. We analyzed these data using the Wilson score interval because when sample sizes are small, the commonly used Wald interval performs poorly and the Clopper–Pearson interval gives an actual coverage probability much greater than the nominal confidence level (Agresti & Coull, 1998).

Natural surveys

We assessed the natural herbivory patterns and habitat distribution of the two Hypericum species by recording their occurrence along transects (N = 8) that began at the forest edge and continued 100 m into the field, or for fields less than 200 m wide, to the midpoint of the field (two transects measured 60 m and one 40 m). To avoid biasing our samples by light exposure, we collected data from transects with differing exposures (i.e., two northeast-facing fields, two southeast-facing fields, one south-facing field, and three west-facing fields). We measured any Hypericum that fell within 1 m of either side of the transect, recording its distance from the forest edge and total number of leaves (as a proxy for plant size). In addition, we quantified how many leaves were heavily damaged (>50% surface area) by herbivores. Leaves with less than 50% damage usually had slime trails indicative of slug herbivory (J Tingle, pers. obs., 2010) and we excluded them from analyses of beetle damage.

Because H. punctatum did not occur past 40 m from the field edge (see Results), we restricted our analyses of both species to the first 40 m. We used mixed effects binomial logistic regression in R (glmer; package lme4; Bates, Maechler & Bolker, 2012) to examine how the number of damaged versus undamaged leaves changed with distance from the forest edge. The model included distance from forest edge and species plus their interaction as predictors, and transect as a random intercept effect. Here and elsewhere, we calculated the variance of the mixed effect model with and without the random effect (package MuMIn; Bartón, 2015) as recommended by Nakagawa & Schielzeth (2013). We also examined how leaf number, our proxy for plant size, changed with distance from the forest edge, using the same predictors. Because H. perforatum and H. punctatum have very different leaf sizes, we first standardized total leaf number (i.e., total number of consumed and unconsumed leaves) by mean leaf number per species (i.e., x_i∕μ_x). This eliminated the main effect of species, but enabled us to better examine the interaction between species and distance.

Transect experiment

Given the beetle’s documented preference for open, sunny areas (Fields, Arnason & Philogène, 1990), we established an experiment to test whether distance from nearby forest edges affected the amount of beetle herbivory that the native and exotic Hypericum received.

In March 2011, we began cold stratification (4 °C, 3 weeks) of Hypericum seeds (Companion Plants, Inc. Athens OH). We sowed the seeds in commercial potting soil (Pro-mix “BX” with biofungicide, Premier) in 36-well trays in a greenhouse, thinning them to 1 individual per cell as needed. Plants received water ad libitum and weekly fertilizer (21-5-20 NPK, 150 ppm). We planted the Hypericum into the field on 7–9 June 2011 after acclimating them in an outdoor shade tent.

We planted both species of Hypericum along transects in a single northeast-facing forest edge (42°28′2″N, 76°26′38″W) where both species naturally occur. Each transect (N = 5 pairs) began at the forest edge (0 m) and ran 50 m into the field, with a Hypericum planted every 5 m (N = 11/transect). Transect pairs consisted of a native-only and an exotic-only transect separated by 1 m, and were approximately 20 m away from neighboring pairs.

In late July 2011, we assessed herbivore damage (number of leaves heavily damaged versus undamaged). In September 2011 we harvested aboveground tissue of each Hypericum and of all the other plants in a 0.25 m² plot around the Hypericum plant, dried the material in a drying oven (50 °C), and separately determined the mass of the Hypericum and surrounding vegetation to 0.1 g.

To investigate how distance from edge and species identity impacted herbivory in the transect experiment, we again used mixed effects binomial logistic regression. The model included distance from edge and species plus their interaction as predictors, with transect pair as a random intersect effect. We also looked at how naturally-occurring competitor biomass changed with distance from the edge using a mixed effects linear model (lmer in R), with distance as the main effect and transect pair as a random spatial block. Finally, we looked at how naturally-occurring competitor biomass predicted Hypericum biomass with a mixed effect linear model and transect pair again as a random effect.

Sun/shade experiment

To directly assess how herbivory impacted the growth of the native H. punctatum, we conducted a 2 × 2 full-factorial experiment that manipulated shade and protection from herbivores. We positioned 11 replicate blocks along forest edges. To prevent biases, we placed the replicate blocks along forest edges that differed in cardinal directions and ensured that each was separated from its nearest neighbor by at least 20 m. Each replicate block contained four 1 m² plots representing each of the four shade and caging combinations. We manipulated shade by locating half of the plots under the canopy at the field edge. We located paired sun plots 15 m away from the forest edge. Light measurements taken between 10:45 am and 12:15 pm on 26 July 2010 (Li-Cor, Model LI 250 Light Meter) showed that on average, shaded plots received 30% the amount of mid-day light that sunny plots did (t_1,10 = − 7.56, p < 0.0001). Within each shade or sun manipulation, we also protected half of the plants from herbivores by completely enclosing the cages with bridal veil. The unprotected cages had bridal veil on the top only. Protected plants received heavy damage (>50% of leaf area eaten) on 8% of their leaves, versus 16% of leaves in unprotected plants (F_1,76 = 2.6, p = 0.111). Thus, the herbivory manipulation did not completely exclude herbivores, but was effective at reducing herbivory. We planted H. punctatum on 21 June 2010 into each plot using the same propagation methods as those described in the transect experiment, except that we used locally collected H. punctatum seeds. We trimmed the resident vegetation in each plot to the ground and planted two small H. punctatum plants in opposite corners of the plot, approximately 1 m apart. Afterwards, we did not further manipulate the vegetation. We assessed herbivory and measured plant height on 28 June and three weeks later on 19 July 2010.

To examine how sun versus shade impacted patterns of herbivory, we used binomial logistic regression to assess the number of damaged versus undamaged leaves in the uncaged plots. To assess how sun and herbivore exposure impacted Hypericum growth rates, we used a factorial ANOVA with light (sun vs. shade) and protection (caged vs. uncaged) as main effects, plus their interaction, as well as replicate block as a random intersect. We calculated relative growth rate as: ${\displaystyle \frac{\Delta height}{{height}_{initial}}\times \frac{1}{days}}$ We then ln + 1 transformed the data to improve the normality of the residuals. If both of the original H. punctatum individuals survived, we took the average relative growth rate of the two. One uncaged, sunny plot and one caged, sunny plot experienced complete mortality and were excluded from the analysis.

Does the beetle preferentially exploit the native or exotic Hypericum species?

In the feeding assays, 11 beetles chose H. punctatum, three chose H. perforatum, and 11 made no choice. A Wilson score interval at 95% confidence for the 14 beetles that made a choice does not include 0.5, indicating beetle preference for H. punctatum (0.52–0.92), but the small sample size leaves us little power to determine the strength of this preference.

Amongst naturally-occurring Hypericum, we found that natives on average had heavy damage on 23% of leaves compared to only 13% of leaves on the exotic (Fig. 2A) and this trend was consistent with distance from the forest edge (Fig. 3A). In contrast, native H. punctatum planted along an experimental transect received less damage than H. perforatum (21% of leaves vs. 27%; Fig. 2B). However, distance from the forest edge strongly influenced how much damage each species received (Fig. 3B; species ×distance: χ² = 29.2, p < 0.0001), with the exotic receiving more damage towards the forest edge and the native receiving more damage away from the forest edge.

Is beetle herbivory concentrated in sunny areas?

Distance from forest edge shaped how much heavy damage each species received. In natural populations, the likelihood of heavy damage increased with distance from forest edge for both species, but more dramatically so for H. punctatum than for H. perforatum (Fig. 3A; species × distance: χ² = 21.3, p < 0.0001; $R_{\mathrm{marginal}}^2=0.08$; $R_{\mathrm{conditional}}^2=0.08$). In the transect experiment, the likelihood of damage decreased with distance from forest edge for H. punctatum, but increased for H. perforatum (Fig. 3B; species × distance: χ² = 29.2, p < 0.0001). However, the random transect effect explained more of the variation in damage ($R_{\mathrm{marginal}}^2=0.01$; $R_{\mathrm{conditional}}^2=0.19$). In the sun/shade experiment, H. punctatum in sunny and shady uncaged plots did not differ in the amount of damage they received (χ² = 1.1, p = 0.300), though small sample size may have hindered our ability to detect differences between treatments. Thus, we have limited evidence to suggest that C. quadrigemina feeding in our experimental fields was restricted to sunny areas.

Does beetle herbivory limit the natural distribution of H. punctatum?

Native H. punctatum grew closer to the forest edge than did the exotic H. perforatum (14.8 ± 1.6 m vs. 34.4 ± 5.9 m; mean ± 95% CI), and we found no H. punctatum past 40 m from the edge (Fig. 4). Yet, both species appeared to grow more vigorously in open, sunny areas. Although the size of Hypericum planted in the transect experiment did not differ with distance from the forest edge (distance: χ² = 1.4, p = 0.224; species × distance: χ² = 0.2, p = 0.640), H. punctatum in the sun/shade experiment had 48% higher growth rate in the sun than in the shade (χ² = 9.4, p = 0.002) and naturally occurring individuals of both Hypericum species tended to be larger towards the center of the fields (distance: χ² = 5.4, p = 0.019; species × distance: χ² = 2.9, p = 0.085). Thus, at least two lines of evidence suggest that H. punctatum grew larger or more quickly in open conditions compared to the shaded conditions where it naturally occurred.

Despite the frequency of attack by C. quadrigemina observed across experiments, we found no evidence that beetle herbivory limited Hypericum growth. Excluding beetles in the factorial manipulation of shade and herbivory had no effect on relative growth rate (χ² = 0.2, p = 0.632). There was also no interaction between beetle exclusion and sun/shade treatment (χ² < 0.1, p = 0.786). We also found no evidence that the competitive environment limited Hypericum distribution. The biomass of naturally-occurring competitors in the transect experiment did not change with distance (χ² < 0.1, p = 0.782) and there was no relationship between competitor biomass and Hypericum size (χ² = 0.9, p = 0.332).