Survival of artificial cohorts of larvae of the imported cabbageworm, Pieris rapae (L.), was assessed in a collard (Brassica oleracea L., var. acephala) patch on an organic vegetable farm in western Massachusetts. We recorded a significant drop in survival of larvae between the 4th and 5th instar due to parasitism by the parasitoid Cotesia rubecula (Marshall), which was introduced to New England from China in 1988. This was a large change from the survivorship pattern of P. rapae observed in the same area in 1985–1986, in which there was a significant drop in survival between 5th instars and pupae due to Cotesia glomerata (L.), the dominant parasitoid of P. rapae in New England at the time. In our study, parasitism by C. rubecula averaged 62.4 ± 8.6%, while no parasitism by C. glomerata was observed. The introduction of C. rubecula is now very likely to be benefitting organic vegetable producers in western Massachusetts, killing more larvae and at an earlier life stage than the previous parasitoid, reducing the potential for damage to cole crops.
The imported cabbageworm, Pieris rapae (L.) (Lepidoptera: Pieridae), is an invasive pest of Brassica crops in North America, Australia, and New Zealand (Jones et al. 1980). In the United States, 2 biocontrol agents have been released against P. rapae larvae during different historical periods, to reduce feeding damage to an acceptable level to growers.
The gregarious parasitoid Cotesia glomerata (L.) (Hymenoptera: Braconidae) was released in 1884 near Washington, D.C. (Clausen 1978). Cotesia glomerata parasitizes first and second instars of P. rapae, and kills P. rapae larvae at the end of the fifth instar, after most P. rapae larval feeding has occurred, so that parasitism usually does little to reduce within generation P. rapae damage. Indeed larvae parasitized by C. glomerata generally consume more food than unparasitized larvae during their development (Rahman 1970), although this varies with parasitoid brood size (Harvey 2000).
The second introduced larval parasitoid was Cotesia rubecula (Marshall) (Hymenoptera: Braconidae), a solitary, host-specific parasitoid. Introductions of this species occurred from 1960 to 1992 (Puttler et al. 1970; Williamson 1971 & 1972; Parker & Pinnell 1972; McDonald & Kok 1992; Lee & Heimpel 2005; Wold-Burkness et al. 2005). In 1988, releases of a Chinese strain of this species were made at 17 different locations in New England (Van Driesche & Nunn 2002). Like C. glomerata, C. rubecula attacks first and second instars of P. rapae, but kills fourth instars, before most feeding has occurred (Le Masurier & Waage 1993). Total feeding per larva is greatly reduced since 85% of feeding occurs in the last instar (Parker & Pinnell 1973) and final weight of C. rubecula-parasitized larvae is only 5–10% of that of mature healthy 5th instars (Harvey et al. 1999).
Furthermore, from an intergenerational perspective, while direct comparisons of rates of parasitism by these 2 Cotesia species are few, in a recent regional survey in the northern United States and Canada (Herlihy et al. 2012), average rates of parasitism were higher (47 ± 3% [95% CI, n = 1041]) for C. rubecula than C. glomerata (25 ± 3% [95% CI, n = 641]), across 32 sites, with sites being dropped for each species if it was not present. Recoveries of these species in this survey were nearly all disjunct and C. rubecula has largely replaced C. glomerata in the United States above latitude N 38° 48′, indicating that it is the superior competitor in that region (Herlihy et al. 2012). Such displacement, not the focus of this study, likely results from both intrinsic superiority of C. rubecula larvae (which are mandibulate and kill the non-mandibulate C. glomerata larvae in cases of co-parasitism) (Laing & Corrigan 1987) and behavioral traits that enable C. rubecula wasps to more efficiently find and exploit P. rapae (Cortesero et al. 1997). High rates of parasitism of P. rapae larvae in organic vegetation farms in western Massachusetts are now the norm, averaging 75% in both spring (Van Driesche 2008) and fall (Van Driesche unpub.) (Fig. 1).
Our study goal was to quantify the effects of C. rubecula on survivorship of P. rapae larval cohorts, using cage exclusion methods, and to compare the current survivorship pattern to that observed in western Massachusetts in the mid-1980s (Van Driesche 1988), when C. glomerata was the only important larval parasitoid of P. rapae. We predicted that (1) C. rubecula would be the dominant parasitoid because it has largely displaced C. glomerata in the study region (Van Driesche 2008; Herlihy et al. 2012), (2) rates of parasitism would be high, given the generally high levels of this species in earlier state surveys, and (3) survival of P. rapae larvae would decrease sharply from the fourth to fifth instar.
Fig. 1.
Distribution of parasitism rates by Cotesia spp. among organic vegetable farms in western Massachusetts in spring 2007 (data redrawn from Van Driesche (2008) and survey data of fall 2009, where the horizontal line is level of parasitism at site for cohort study in 2011. (Site numbers do not correspond between surveys.)
MATERIALS AND METHODS
Massachusetts Cotesia rubecula Parasitism Surveys
A survey Cotesia parasitism of P. rapae larvae in organic vegetable farms or private gardens in western Massachusetts was conducted in fall of 2009, using the same protocols as a survey in the same region in spring of 2007 (Van Driesche 2008). In brief, 19 organic vegetable farms or gardens in the four western counties of Massachusetts were visited, and the cole crops present (cabbage, collards, broccoli, Brussels sprouts) were examined. The survey was conducted from 10 Sept to 22 Oct of 2009, and all P. rapae larval stages, pupae, emerged pupae, parasitoid cocoons (live or emerged) on plants were collected, until at least 30 individuals had been located. These were then taken to the laboratory and dissected. Parasitism by C. rubecula or C. glomerata was determined based on detection and recognition of parasitoid eggs or larvae in dissected host larvae, together with the number of parasitoid cocoons (live or emerged) found during the field search. Parasitized larvae plus parasitoid cocoons (live or emerged) formed the parasitized group, which was divided by the count of all host or parasitoid stages collected to calculate the rate of parasitism (all other details as in Van Driesche [2008]). For C. glomerata cocoons, one group of cocoons was counted as one, since it represents one parasitized host.
Study Location
In June of 2011, 130 collard seedlings were planted at an organic vegetable farm (Intervale Farm in Westhampton, Massachusetts, 42°17.41′N, 72°46.19′W) to form an experimental plot where the interaction of C. rubecula and P. rapae larvae could be studied. Seedlings were planted in 6 rows of about 22 plants each, at 50 cm intervals. The land used for the patch had not been planted previously with cole crops, and was used the previous year for corn. In 2011, the area around the patch was planted to cole crops, corn, beans, and tomatoes. Neither the study patch nor the surrounding area were treated with pesticides during 2011.
Plants
Before each of 4 experiments, groups of 70 large collard plants (c. 30 cm tall with 7–10 leaves) in 15 cm dia pots were purchased from Harvest Farm in Whately, Massachusetts, where they had been grown in greenhouses without pesticides. After purchase, plants were held for about 1 week in the laboratory and watered daily until used in the experiment. Two days before starting an experiment, plants were taken to the field plot to acclimatize to outdoor conditions. Potted plants were placed within the existing rows between soil-grown collard plants; there were 15 potted plants in each of the plot's 4 rows (which were used for the main treatments) and 10 potted plants in a fifth row (with larvae needed to determine the accuracy of visual estimation of parasitism rates). Pots were set into evenly spaced shallow holes dug in the ground to help conserve moisture and prevent pots from blowing over, and were watered every other day during the experiment.
Insect Rearing, Handling, and Inoculation of Experimental Plants
Pieris rapae larvae used came from a laboratory colony initiated with butterflies collected at local sites near Amherst, MA, in 2010 and reared on collards. Butterflies were provided with sugar solution (“Instant Hummingbird Nectar,” Perky Pet Woodstream Corp. Lititz, PA) and allowed to oviposit on collard seedlings. After egg hatch, larvae were reared on seedling collards until just beyond the neonate stage (e.g., having taken their first meal and turned from yellow to green due to food in their guts). Young 1st instars were then taken, along with plants, to the field site where larvae were placed on the experimental plants using a paintbrush dipped in nectar water (sugar induces plant acceptance by neonates). Five 1st instars were placed on each potted plant.
Experimental Design
We tested the effects of 4 levels of natural enemy exclusion on survival of our cohorts of P. rapae larvae: (1) no exclusion (uncaged plants without Tanglefoot [Contech Enterprises Inc. Victoria, BC, Canada] on pots), (2) full exclusion (plants inside net sleeves and with Tanglefoot on pots), and (3, 4) 2 levels of partial exclusion, being either with sleeves and no Tanglefoot, or the reverse. There were 15 plants (replicates) for each level of exclusion and each plant was considered an experimental unit. Treatments were placed within their own rows, with all 15 plants arranged in one row, but immediately next to (< 0.5 m) a similar row of plants for treatments 2, 3, and 4. The maximum distance between pots of different treatments was 1.5 m. The experiment was repeated 4 times (Jul, Aug, Sept and Oct) to determine if important seasonal variation existed, as suggested by the slightly higher levels of C. glomerata parasitism in a fall survey vs. spring survey in Massachusetts (Van Driesche unpub. [Fall] vs. Van Driesche 2008 [Spring]). The same plot was used for all 4 runs of the experiment, but in each run, treatments were assigned to different rows.
In 2 of the 4 treatments, Tanglefoot was applied to pots to prevent walking predators from moving up the pots from the soil onto the plants. Tanglefoot was painted once in a band around the outsides of each pot in this treatment at the beginning of the experiment. For 2 of the 4 treatments, organdy sleeves were placed over the entire plant to exclude flying predators and parasitoids. Sleeves were held slightly away from the foliage of the potted plants by 4 stakes driven into the ground around each pot; sleeves were fitted over the stakes, dropping to the mid-pot level (about 5 cm above soil level). The bottom edges of the sleeves were left loose and were not secured to pots in order to facilitate sleeve removal, which was needed for frequent counting of larvae and watering of plants.
During insect counts, parasitism was estimated visually (based on shift in caterpillar color from green to whiter tones and visible shapes of parasitoids larvae in caterpillars) as soon as parasitism could be detected; however, these putatively parasitized larvae were left on plants for continued observation. To determine the relationship between visual estimates of parasitism and the real rates of parasitism (as would have been seen in dissection if larvae were removed and killed), in the fifth row we placed ten additional potted plants without organdy bags or Tanglefoot. Larvae in this group were placed on plants at the same time and in the same manner as for the cohorts of the main experiment. When larvae in the cohorts reached the fourth instar, we conducted a visual estimate of parasitism in the same manner as for the main experiment. We then checked our visual estimate by dissecting these extra larvae in the laboratory, using the same characters for species recognition as described in Herlihy et al. (2012). The comparison of parasitism rates of these larvae via field visual assessments and their subsequent dissection in the laboratory allowed us to measure the accuracy of our visual assessments of parasitism in the main experiment.
Data Collection
We counted surviving P. rapae larvae on all potted plants at a series of time points from first instars to pupae. Larval survival was checked visually in the field by examining potted plants leaf by leaf, recording all living larvae (by instar) and any parasitoid cocoons. For treatments in which plants were covered with organdy sleeves, these were lifted up to examine plants and then replaced. If any wild P. rapae eggs were found on unsleeved experimental plants, they were removed. During the first 2 runs (7–20 Jul and 4–16 Aug), plants were examined every second day, providing 6 observations per run. During the last 2 runs (3–24 Sept and 27 Sept to 17 Oct), the plants were initially examined every second day, but larvae developed more slowly due to lower temperatures and therefore after the first 2 observations the period between observations was changed to every third day, providing 7 observations in run 3 and 8 in run 4.
When the surviving larvae of a cohort reached the fourth instar, each larva was assessed visually in the field and classified as parasitized or not, by either C. rubecula or C. glomerata. (Parasitism by this stage was visible due to a change in body color and parasitoid species was recognized by larval size and appearance, changes which were familiar to us from rearing both parasitoids in the laboratory).
Life Table Construction
Life tables were constructed for cohorts of larvae from treatment groups within each monthly run of the experiment, but only 4 life tables (one per treatment, pooled over months) are presented. We used the marginal rate equation mB = dB/(1-mA), where mB is the marginal attack rate of the parasitoid, dB is the death rate from parasitism, and mA is the marginal attack rate for predation (here, disappearance), to calculate the underlying rates of parasitoid attack from the observed rates of parasitism and disappearance for fourth instars in life tables (the only stage with 2 observed mortality factors) (Elkinton et al. 1992).
Statistical Analyses
All statistical analyses were conducted using the program R (version 2.13.1). For P. rapae stage survival, analyses were based on insect counts (of particular stage) per plant, while for parasitism, data were proportions (of fourth instars) parasitized. Parasitism data were arcsine transformed in order to meet the assumption of normality. Number of survivors for selected P. rapae life stage and rates of parasitism were compared across run dates (month). Rates of parasitism were compared using a t-test for treatments with and without sleeves. Survival data were analyzed with an ANOVA, followed by a Tukey HSD test.
RESULTS
Rates of C. rubecula Parasitism in Regional Surveys
The rate of C. rubecula parasitism observed in the 2011 cohort survival experiment study (59–60% apparent parasitism [Table 1]) was typical of those seen in surveys under similar conditions in western Massachusetts in spring of 2007 and fall of 2009 (Figs. 1A and 1B), which together covered 39 locations. In the spring 2007 survey, 16 of 20 sites had C. rubecula parasitism rates equal or higher than the 2011 cohort study site, and in the fall 2009 survey, 17 of 19 sites did so. This indicates that results from our detailed cohort study are representative of results across western Massachusetts.
TABLE 1.
NUMBER OF LARVAE REACHING DESIGNATED LIFE STAGES AND RATES OF COTESIA RUBECULA PARASITISM FOR PIERIS RAPAE LARVAL COHORTS SUBJECTED TO DIFFERENT LEVELS OF NATURAL ENEMY EXCLUSION, IN AN EXPERIMENTAL COLLARD FIELD IN WESTAMPTON, MASSACHUSETTS (POOLED OVER FOUR EXPERIMENTAL MONTHS [JUL–OCT, 2011]).
Survival of P. rapae Cohorts
A 3-factor (month, sleeves, and Tanglefoot) factorial ANOVA showed that month of exposure had no significant effect on survival to fifth instar (F = 0.17; df = 3; P = 0.92) or survival to pupa (F = 0.05; df = 3; P = 0.83). Therefore, in further analyses, data were pooled across months to explore differences among the 4 levels of natural enemy exclusion.
In a 2-factor (sleeves and Tanglefoot) factorial ANOVA, differences in the survival of experimental cohorts to the pupal stage with sleeves (presence vs. absence) and Tanglefoot (presence vs. absence) as factors were highly significant (ANOVA, F = 63.2; df = 3; P < 0.0001). Survival of cohorts to the pupal stage with and without Tanglefoot did not differ significantly (for exposed [not sleeved]: t = 0.21, df = 6; P = 0.84; for sleeved: t = 2.43, df = 6; P = 0.051). Survival to the pupal stage in cohorts without organdy sleeves averaged 3.0 ± 2.9% with Tanglefoot and 3.7 ± 2.7% without Tanglefoot (pooled over all experimental runs). Survival to the pupal stage for cohorts on sleeved pots averaged 36.3 ± 6.6% with tanglefoot and 27.3 ± 3.4% without Tanglefoot (Table 1).
Because C. rubecula kills its hosts in the fourth instar, we also compared rates of survival to the fifth instar for treatments in all 4 runs. Differences in the survival of experimental cohorts to the pupal stage with sleeves (presence vs. absence) and tangle foot (presence vs. absence) as factors were highly significant (F = 53.4; df = 3; P < 0.0001). Survival to the fifth instar in cohorts without organdy sleeves averaged 5.3 ± 2.9% with Tanglefoot and 5.0 ± 1.3% without Tanglefoot (pooled over all experimental runs). Survival to the fifth instar for cohorts on sleeved pots averaged 43.0 ± 8.7% with Tanglefoot and 33.4 ± 5.2% without Tanglefoot (Table 1).
The survival of larvae in cohorts (Fig. 2) in all treatments decreased at a steady rate after an initial sharp drop between the first and second instar (which was seasonally more obvious in July and August), which was likely due to a combination of predation and failure of young larvae to establish and feed. There was another sharp drop in the number of larvae (considering visually parasitized larvae as “dead” for this comparison) in the unsleeved treatments during the fourth instar, due to the effect of parasitism by C. rubecula. The fourth instar was the first point at which parasitism could be scored visually and was the instar in which larvae parasitized by C. rubecula died.
Fig. 2.
The number of surviving P. rapae larvae (from first instar to pupa) in artificial cohorts established on potted collards in 4 successive months in 2011 on an organic vegetable farm in Westhampton, Massachusetts, USA. All cohorts began with 75 larvae on day zero. “Healthy” is defined as larvae that are alive, present, and not visibly parasitized (early stage parasitism is not detectable in living larvae). Numbers in parentheses on the x-axes are the dominant instar in the cohort on that sample date.
Parasitism of P. rapae by C. rubecula
Comparison of visually determined rates of parasitism to rates seen in larval dissection for 48 fourth instars placed in the experimental plot for this purpose (on the same dates as the cohort larvae in the main experiment) revealed that our estimates of the rates of parasitism by visual determination was 94% accurate. For these larvae, rates of parasitism based on visual inspection (65%) versus dissection of the same larvae (71%) were not significantly different based on an unpaired t-test (t = 0.81; df = 4; P = 0.47). The few cases of error in assessing parasitism visually were larvae that appeared unparasitized in the field but found to be parasitized upon dissection. There were no cases of larvae being classified as parasitized in the field by visual inspection that upon dissection were found not to be parasitized. Consequently, our reported parasitism rates are conservative.
A 2-factor (month, treatment) factorial ANOVA found that there was no effect of month on parasitism rates (F = 0.0805; df = 3; P = 0.97), and therefore data on parasitism were pooled by month to examine the effects of treatments on rates of parasitism. For treatments lacking organdy sleeves the rate of parasitism by C. rubecula was high (62.4 ± 8.6 %) (Fig. 3), in contrast to treatments protected by organdy sleeves, in which very little parasitism occurred (4.2 ± 4.0%) (ANOVA: F = 66.73; df = 3; P < 0.001). As with the survival data, the only significant difference in parasitism rates was between sleeved and exposed plants, with no effect of the presence or absence of tanglefoot on pots (Table 1).
Fig. 3.
Percentages of Pieris rapae larvae in experimental cohorts on potted collards that were parasitized by Cotesia rubecula during their larval life. Parasitism (± 95 % C.I.) is shown for 4 levels of natural enemy exclusion (with and without sleeve, and/or tanglefoot on pots), in 4 successive months in 2011 at an organic vegetable farm in Westhampton, Massachusetts.
Cohort Life Tables and Marginal Rates of Parasitoid Attack
Life tables for all treatments (4) by all runs (4) were constructed (16 tables) and examined. However, since run date (month of experiment) had no significant effect on any survival or parasitism parameters examined, tables were collapsed across run dates within treatments and the resultant 4 summary life tables, one per treatment, are presented (Table 2).
In a 3-way ANOVA of the effect of treatment, instar (first through fourth only) and run date on survival, we found no significant 3-way interaction (F = 0.75; df = 27; P = 0.65), but there was a significant effect of run date on survival during the early instar stages in all treatment types (F = 2.8, df = 3, P = 0.045). For this analysis, parasitized fourth instars were considered alive. Specifically, there was a significant difference between survival of all instars (first through fourth only) in all treatment types between the July and October experimental runs, but not between survival of instars in all treatment types of any other months based on a Tukey HSD test.
Marginal rate analysis (Table 1) revealed that parasitoid attack rates summed over both unsleeved treatment types (with and without Tanglefoot) were on average 13 percentage points higher (59.5 vs. 72.5%) than the apparent parasitism by C. rubecula as directly observed in field sampling.
DISCUSSION
Because it is often not feasible to monitor a biological control agent for many years after its release, an alternative is to measure the effects of the biological control agent after a lapse of some years, after it has had an opportunity to become fully effective. Monitoring cohorts of individuals with known and standardized age structures and starting points can provide especially useful data. In this case, use of cohorts of P. rapae larvae on potted collard plants placed on an organic farm, with and without natural enemy exclusion, allowed us to quantify losses by instar, giving a more precise measure of the survival patterns of P. rapae life stages in the field over time. Monitoring cohorts allowed us to determine that there is a relatively large drop in survival (72.5%) between the fourth and fifth instars due to parasitism by C. rubecula (see Table 2, average marginal rates of attack of 4th instars by C. rubecula for two uncaged cohorts).
TABLE 2.
LIFE TABLES FOR COHORTS OF PIERIS RAPAE LARVAE ON POTTED COLLARDS DEPLOYED IN A COLLARD PATCH AT AN ORGANIC VEGETABLE FARM, INTERVALE FARM, IN WESTHAMPTON, MA, SUMMED OVER 4 MONTHS (JUL, AUG, SEP, OCT) IN 2011, WITH 4 LEVELS OF NATURAL ENEMY EXCLUSION (+/-TANGLE FOOT ON POTS CROSSED WITH +/-SLEEVES OVER PLANTS).
continued
While tanglefoot had no or perhaps a marginally significant effect on survival of cohorts, sleeves were successful at reducing parasitism: survival of cohorts to pupae was much higher for cohorts with sleeves (31.8 ± 6.9%) than without (3.3 ± 2.7%). This suggests that parasitism rather than predation by ground predators was responsible for most P. rapae mortality in late instars (fourths), while, in contrast, predation by ground-dwelling predators was likely responsible for most P. rapae mortality in early instars (first through third), apart from failure of transplanted larvae to settle and feed. However, since this study was designed to estimate the effects of parasitism, our insights into predation are limited.
Our estimates of parasitism are likely to be underestimates for 2 reasons. First, parasitism in the field, measured visually, was determined to be only 94% accurate and was always in the direction of an underestimate. Secondly, parasitism rates in the cohort data were influenced by the problem of simultaneous mortality, since some larvae disappeared steadily in all exposed treatments, likely due to predation. As a correction for this issue, we calculated the underlying marginal rates of parasitism, which revealed underlying parasitoid attack rates that averaged 13 percentage points higher (59.5% vs 72.5%) than apparent mortality due to parasitism (Table 2) (Elkinton et al. 1992).
Our results are consistent with a study conducted in New Zealand that assessed the rates of parasitism of P. rapae by both C. rubecula and C. glomerata (Cameron & Walker 2002). It is clear that in western Massachusetts C. rubecula is now the dominant parasitoid of P. rapae and has displaced C. glomerata, as it did at many of the New Zealand sites studied by Cameron & Walker (2002) and in parts of North America (Biever 1992; Van Driesche 2008; Herlihy et al. 2012). In Massachusetts, C. rubecula attacked P. rapae at a higher rate in our 2011 cohort study than did C. glomerata in a previous study in 1985 and 1986 (Van Driesche 1988), in the same area before the introduction of C. rubecula. In general our results were similar to those found by Cameron & Walker (2002) in P. rapae populations with and without C. rubecula.
As expected, in comparison to the earlier study (Van Driesche 1988), the survival pattern of P. rapae has changed significantly since the introduction of C. rubecula. In the 1980s study, survival of P. rapae declined steadily throughout all life stages in the first generation each year. During the second, third, and fourth host generations, C. glomerata (the dominant parasitoid) caused approximately a 50% decrease in survival between the fifth instar and pupal stage. Since the displacement of C. glomerata by C. rubecula, the decrease in survival of P. rapae is now both steeper (72.5%) and earlier, occurring between the fourth and fifth instars. This is an important result for organic vegetable growers because earlier mortality of P. rapae caterpillars reduces damage per larva.
Cotesia glomerata in western Massachusetts is now a less common species, but one that remains a component of the P. rapae/collards agroecosystem (Van Driesche 2008). Partial displacement of C. glomerata from P. rapae means its density on the landscape as a whole is also likely to be lower, which should help reduce its parasitism rates to native pierids such as Pieris oleracea Harris (Herlihy 2012).
ACKNOWLEDGEMENTS
The authors thank Rick Tracy and Maureen Dempsey of Intervale Farm for use of their property The authors also thank David Wagner of the University of Connecticut and Elizabeth Jakob of the University of Massachusetts for their guidance. This material is based upon work supported by the National Institute of Food and Agriculture, U.S. Department of Agriculture, the Massachusetts Agricultural Experiment Station and the Department of Plant, Soil and Insect Sciences under Project number MAS00957 (Publication number 3485).
