Host Plant Use by the Invasive Halyomorpha halys (Stål) on Woody Ornamental Trees and Shrubs

The brown marmorated stink bug, Halyomorpha halys (Stål) (Hemiptera: Pentatomidae) is an invasive plant-feeding insect native to eastern Asia. This herbivore is highly polyphagous, feeding on and damaging diverse plants, including field crops, vegetables, tree fruits, and ornamentals. Woody ornamental plants provide early- and late-season resources for adults emerging from and returning to overwintering sites, as well as feeding and breeding sites for H. halys throughout the growing season. In this study, we quantify the use of diverse plants by H. halys in two commercial nurseries in Maryland, recording data on the abundance of egg masses, early and late instar nymphs, and adults over a three-year study period. Our specific goals were to provide a quantitative comparison of the use of diverse plant species and cultivated varieties, identify non-hosts that could be used to create landscapes refractory to H. halys, and determine whether the use of plants varied across life stages of H. halys or the taxonomic status of plants. We found broad use of diverse plants in this study, identifying 88 host plants used by all life stages of H. halys. We also highlight the 43 plant taxa that did not support any life stage of H. halys and are thus classified as non-hosts. Interestingly, some of these plants were congeners of highly-used plants, underscoring high intrageneric and intraspecific variation in the use of plants by this polyphagous herbivore. We discuss how the selective planting of non-hosts, especially gymnosperms, may aid in reducing the agricultural and nuisance pest status of this invasive insect.

H. halys causes direct damage to plants, importantly to many commercial crops, through its feeding activities [11][12][13][14][15][16][17][18] and can cause indirect damage through the transmission of plant diseases including Paulownia witches' broom [19]. During the growing season of 2010, populations of H. halys burgeoned in the United States, and multimillion dollar losses were recorded on orchard crops including apples and peaches; vegetables such as sweet corn, peppers, and tomatoes; row crops including field corn and soybeans; vineyards; small fruit; and ornamental plants grown in landscapes and nurseries [14,15,20]. In addition to crop damage, H. halys is a severe nuisance pest during fall, winter, and spring when adults aggregate on commercial buildings and homes, enter and overwinter in domiciles, and egress in spring [15,[21][22][23][24][25]. These behaviors generate public concern, media attention, and a general outcry for management solutions [15,23,24]. In response to this demand, pest control companies provide services including the treatment of buildings and landscape plants where H. halys aggregate in autumn prior to entering structures [25].
Halyomorpha halys is highly polyphagous in both its native and invaded ranges. An important review of the Asian literature revealed 106 hosts distributed in 45 families, ranging from herbaceous annual vegetable crops to forest trees [12]. A synthesis of host use [24] reported 51 host plants in 32 plant families in Europe. This list included European native and non-native plants, ranging from herbaceous perennials to woody trees and shrubs. In the invaded North American realm, studies conducted in several counties in eastern Pennsylvania recorded H. halys on 73 species of plants ranging from annual crops to landscape trees [21]. Trees and shrubs, many of which were not native to North America, dominated the list of plants upon which H. halys was noted as abundant or common [21]. The greatest numbers of H. halys were found on tree of heaven, catalpa, yellowwood, paulownia, cherry, walnut, and redbud growing in non-managed woodlands in North Carolina and Virginia [26]. Additional investigations of ornamental and cultivated hosts have demonstrated temporal and stage specific shifts in host use of H. halys over the course of growing seasons [11]. Because of the use of plants of non-Asian origin in both Europe and North America, it is apparent that the host plant list from Asia will be an incomplete list of the plants used by H. halys in its invaded range.
The use of resistant plant material is a mainstay of integrated pest management for agronomic crops [26,27] as well as for ornamental plants in landscapes [28][29][30]. Several of the aforementioned reviews of H. halys noted significant variation in patterns of host use among woody landscape plants. However, these reviews focused on plants on which H. halys was observed feeding or breeding; with the exception of the survey by Bakken et al. [31], little or no information was presented on the plants that were not used as hosts by H. halys. Furthermore, although several Asian studies and reviews [12,24] noted H. halys utilizing many species of gymnosperms, gymnosperms have been conspicuously lacking in host lists from North America [11,21]. The extent of intraspecific variation in host plant use is also not well understood. Whereas previous studies have noted significant variation in the abundance of H. halys among varieties of tree fruits [32][33][34], little is known about intraspecific variation in host use among ornamental trees. Such information would be invaluable for developing surveys and management plans, and for identifying resistant plant material.
In this study, we examined patterns of host plant use by H. halys in large, diverse, commercial production nurseries in Maryland. Our specific goals were to quantify variation in the use of plants across life stages of H. halys, expand the host plant list by examining a broad diversity of plants, especially those that had not yet been surveyed for H. halys, and identify plants not used by any life stage, particularly ovipositing females. We also included gymnosperms and intraspecific variation in the plants surveyed, in order to develop a robust understanding of the variation in host plant use of H. halys in its invaded range.
Our study is novel in seeking to identify ornamental woody plants that are not included in the feeding or breeding repertoire of H. halys. By incorporating plants not used by H. halys into landscapes, we hope to obviate breeding sites and places where stink bugs aggregate prior to entering homes. This should reduce the need for treating structures and plants with insecticides to kill this pest [25]. Moreover, by identifying ornamental plants refractory to this pest, commercial growers of landscape plants could enjoy a marketing advantage by producing and selling plants that reduce the likelihood of autumnal home invasions. Commercial nurseries in this study provided a rich source of familial, generic, specific, and varietal variation in which to explore patterns of host use by H. halys on angiosperms and gymnosperms grown for planting in residential landscapes.

Ethics Statement
No endangered or protected species were involved in the study. We obtained permission from individual private commercial tree nurseries for access and data collection. Following the protocols of Venugopal et al. [35] and Martinson et al. [18], 1-min visual counts of H. halys were conducted per plant, encompassing the foliage, flowers, fruits/seeds, and bark to a height of up to 3 m on six trees per row. Undergraduate and graduate student observers were trained by E.J.B. and M.J.R. to ensure uniformity and consistency in the field protocols for data collection. H. halys abundance (total per plant) was recorded separately for four life stages: egg masses, early instar nymphs (instars 1-3), late instar nymphs (instars 4 and 5), and adults. Each year, we conducted repeated surveys for H. halys on each tree from late May to early August: in 2011, each tree was surveyed four times; in 2012 and 2013, each tree was surveyed six times. Sale of some of the sampled trees during the study period, as well as tree mortality from heat stress, disease, and physical damage, resulted in variable numbers of surveys, or 'tree visits,' for some trees (see S1 Table for number of visits for each plant taxon).

Tree identification
We recorded the genus, species, and cultivar (cultivated variety) of each tree. Trees without cultivar names are straight species, which have not been selected for particular traits. We completed the tree identification using nursery records and confirmed it using existing literature [36] to ensure consistent usage of scientific and cultivar names, common names, and spellings. We also ensured accuracy in nomenclature and taxonomy of the tree species and cultivars (hereafter, plant taxa) by comparing our records with The Plant List [37].

Statistical analysis
Host use by different life stages. Definitions of 'host' are highly variable across studies, and the concept of host has been used in several contexts. With respect to H. halys, Oda et al. [38,39] (reported in [12]) described breeding plants (those on which eggs, nymphs, and adults were observed) and feeding plants (those on which only adults were observed feeding). In North America, Nielsen and Hamilton [11] classified plants as hosts based upon consecutive observations of nymphal stages across multiple years. Because plants supporting all life stages are suitable for adult oviposition as well as nymphal development, these plants represent reproductive hosts for H. halys [40,41]. Following this delineation, we classify plant taxa with records of each of the four life stages (egg masses, early nymphs, late nymphs, and adults) of H. halys as hosts. On the other hand, we classify plant taxa without any of the H. halys life stages as nonhosts. Plants with some but not all of the life stages are classified as partial hosts.
To test whether the use of plants was similar across all life stages of H. halys, we calculated the proportion of plant taxa used by each life stage. We used Person's Chi-squared tests for pairwise (each life stage) statistical comparisons of these proportions.
We further ranked the plants classified as hosts based on mean abundance of H. halys, calculated as the summed abundance of H. halys nymphs and adults divided by the total number of 1 min. surveys conducted for that taxon.
Use of angiosperms and gymnosperms. The use of angiosperms and gymnosperms by H. halys was analysed through generalized linear models (GLM) assuming a Quasi-Poisson error distribution and log link function [42]. GLMs were performed for each life stage with the abundance of H. halys as the response variable and taxonomy as the predictor, accounting for differences in tree visits across the cultivars (through 'offset' statement). Significant differences in the model estimated means were identified through Tukey's HSD comparisons (α = 0.05).
All statistical analyses were performed in the program R [43] and associated statistical packages. Nomenclature and other taxonomic details were verified using package 'Taxonstand' [44]. Tukey's HSD were performed with the package "multcomp" [45]. Coefficients estimated by the GLMs were extracted and plotted using "ggplot2" [46].

Host use by different life stages
Our surveys for H. halys encompassed 254 unique taxa of ornamental trees and shrubs. Over the three year study period, we conducted 52,717 one minute surveys, recording a total of 589 egg masses, 23,697 early instar nymphs, 3,525 late instar nymphs, and 10,925 adults. At least one life stage of H. halys was present on 211 (83%) of the surveyed plant taxa. Egg masses were present on 99 (39.0%), early nymphs on 176 (69.3%), late nymphs on 144 (56.7%), and adults on 198 (78.0%) taxa.
Host plant use differed among life stages, with a general increase in the proportion of available plant taxa utilized with increasing insect developmental stage. Specifically, adults were found on a higher proportion of plant taxa than were late nymphs and egg masses, and late nymphs were found on a higher proportion of plant taxa than were egg masses (Table 1). Early nymphs, on the other hand, used nearly as many plant taxa as adults and more than late nymphs. There were no cultivars on which only the egg masses, but no other stages, were recorded (see S1 Table). Despite differences in the proportion of plant taxa used by each life stage, broad patterns of overlap in plant use were also apparent across life stages (Table 1).
All life stages of H. halys were observed on 88 plant taxa, which we therefore classified as hosts. On the other hand, no life stages of H. halys were observed on 43 plant taxa, and these were classified as non-hosts. The remaining 123 cultivars were classified as partial hosts by virtue of the presence of at least one but not all life stages. of the numbers of each life stage found on each plant taxon over the course of the study is available in S1 Table.

Discussion
By surveying 254 species and cultivated varieties of trees and shrubs under the controlled conditions of production nurseries, we were able to quantify the abundance and host plant use of H. halys. We identified 88 host taxa, 43 non-host taxa, and 123 partial hosts supporting some but not all life stages of H. halys. In doing so, we discovered variation in H. halys host use among plants that may guide the development of landscapes refractory to this stink bug.
Our results agree with previous work in North America [11,21,31], Asia [12], and Europe [24] which observed one or more life stages of H. halys on a broad range of woody plants in managed and non-managed settings. Favored hosts found in this study match those of previous ones for several genera ( Table 2). Prunus is a genus that appears on our list of the 25 most utilized hosts and other lists of common hosts for H. halys [21,31]. Other genera found on our list of the most commonly used hosts that also appear on other host lists include Malus, Syringa, Acer, Cladrastis and Cercis [21,31] (see S2 Table for an explicit comparison between the present study and previous studies for plant genera that include host taxa). Lee et al. [12] noted the affinity of H. halys for hosts in the Fabaceae and Rosaceae. In addition to these families, Oleaceae, Sapindaceae, Rutaceae, Ulmaceae, Moraceae, Altingiaceae, and Malvaceae supported the greatest abundances of H. halys over three years of this study ( Table 2). Because of our broad sampling, we were also able to identify host plants in nine genera that, to the best of our knowledge, have not previously been listed as host plants for H. halys: Cercidiphyllum, Evodia, Gleditsia, Halesia, Nyssa, Oxydendrum, Parrotia, Pseudocydonia, and Styrax.
We also identified 43 plant taxa not used at all by H. halys in our study. Interestingly, these taxa included congeners of plants we identified as hosts, underscoring strong intraspecific and intrageneric variation in H. halys abundances and host plant use. For example, H. halys was never observed on several varieties of Acer palmatum while most cultivars of its congener Acer rubrum were heavily utilized. Several cultivars of Ginkgo biloba supported notable numbers of H. halys nymphs and adults, whereas Ginkgo biloba 'Saratoga' supported none.
In our study, the timing or absence of fruiting resources on particular taxa likely contributed to their status as non-hosts [18]. For example, early fruiting cultivars of shrubs like Hamamelis x intermedia or non-fruiting trees such as male Ginkgo biloba 'Saratoga' were devoid of all life stages of H. halys and classified as non-hosts. Flowering and fruiting of Hamamelis occurs in winter and early spring, well in advance of the arrival of H. halys into the nursery [18,35]. In addition, some trees do not fruit abundantly until they reach maturity [47]; this could potentially lead to an underestimation of the list of plants used by the highly mobile stages. However, 10 of the 28 non-host taxa for which we have phenology data (2 angiosperms and 8 gymnosperms; data available for Year 3 only) did bear reproductive structures. Thus, plant maturity and fruiting are not the only reasons that plants are on the non-host list. In interpreting our designation of taxa as non-hosts, we urge caution for several taxa for which relatively few observations were made. For example, the number of observations of Acer davidii was four over the entire course of the study; due to the small number of tree visits, the placement of this species as a non-host is subject to future modification (Table 3, S1 Table). By contrast, Acer davidii's congener Acer palmatum var. dissectum Inaba Shidare was observed 144 times over the three years of the study and its designation as a non-host is therefore well-supported (Table 3, S1 Table).
The patterns of intraspecific variation in host use by H. halys we report here mirrors those of other studies that demonstrate variation in host use among varieties of apples. Fujisawa [32] and Funayama [33] attributed intraspecific variation in patterns of host use to differences in fruiting times among cultivars of apples. Funayama [33] noted the importance of multiple hosts in the normal development of H. halys. Recent work by Martinson et al. [20] demonstrated the strong positive relationship between the presence of fruit and the abundance of H. halys adults on individual trees. The remarkable ability of H. halys to track quality resources in space and time has been detailed in several studies [12,14,[16][17][18]. This explains at least in part the pattern of broad host use in H. halys, as different species and cultivars will present resources of differing quality in the nurseries throughout the growing season.
Stage specific differences in patterns of host use in this study reflect those found in previous studies of H. halys on woody plants. Stage specific shifts in host use were noted as different instars of H. halys tracked resources on different woody hosts [11]. In non-managed settings in several locations in North Carolina and Virginia, the broadest range of hosts were used by adult H. halys, the fewest hosts were used as oviposition sites, and nymphs utilized many more hosts than ovipositing females, but slightly fewer hosts than adults [31].Similarly, we found egg masses on the fewest numbers of plant taxa, whereas highly mobile adults were found on the greatest number (Table 1). Early and late instar nymphs were found on intermediate numbers of hosts.
A noteworthy finding is the difference in the use of gymnosperms by H. halys as observed in our study to that in Asian literature. The use of gymnosperms by H. halys is well established in Asia [12,32,38,[48][49][50][51][52]. Specifically, some gymnosperms such as Japanese cedar serve as important hosts for overwintered adults early in the season [12,51]. However, our study confirms the use of several families, genera, species, and cultivars of gymnosperms as partial hosts for H. halys in North America (Table 3), but none of the gymnosperms studied were classified as hosts. Although gymnosperms are used as partial hosts by H. halys, it is noteworthy that gymnosperms supported far lower H. halys abundances than did angiosperms (Fig 1). Moreover, the list of non-hosts was dominated by gymnosperms. The Pinaceae and Cupressaceae in particular each contained several genera and species of non-hosts.
The practical implications of our study are that several species and cultivars of woody ornamental trees and shrubs presently in production do not appear to be utilized by any life stage of H. halys and by our definition, they are not hosts. By planting these varieties, landowners may enjoy lower abundances of H. halys in their landscapes, with the additional benefit of spawning fewer H. halys that would become nuisance pests as they enter homes and businesses later in autumn. Gymnosperms can help to diversify landscapes with trees that provide valuable ecosystem services including water infiltration, carbon sequestration, and as a buffer against invasive species [53]. Our findings provide evidence that gymnosperms provide a rich source of plant material refractory to H. halys for use in landscapes. We also believe that growers who produce these resistant varieties may enjoy a marketing advantage in states and countries within the invaded range of H. halys.
Supporting Information S1