Preimaginal Stages of the Emerald Ash Borer, Agrilus planipennis Fairmaire (Coleoptera: Buprestidae): An Invasive Pest on Ash Trees (Fraxinus)

This study provides the most detailed description of the immature stages of Agrilus planipennis Fairmaire to date and illustrates suites of larval characters useful in distinguishing among Agrilus Curtis species and instars. Immature stages of eight species of Agrilus were examined and imaged using light and scanning electron microscopy. For A. planipennis all preimaginal stages (egg, instars I-IV, prepupa and pupa) were described. A combination of 14 character states were identified that serve to identify larvae of A. planipennis. Our results support the segregation of Agrilus larvae into two informal assemblages based on characters of the mouthparts, prothorax, and abdomen: the A. viridis and A. ater assemblages, with A. planipennis being more similar to the former. Additional evidence is provided in favor of excluding A. planipennis from the subgenus Uragrilus.


Introduction
The emerald ash borer (EAB), Agrilus planipennis Fairmaire (Coleoptera: Buprestidae), is a metallic wood-boring beetle indigenous to eastern Asia, including China (Beijing, Hebei, Heilongjiang, Inner Mongolia/Nei Mongol, Jilin, Liaoning, Shandong, Sichuan, Tianjin, and Xinjiang); Taiwan; Japan; Korea; Mongolia; and Russian Far East [1, 2,3]. In addition, Jendek and Grebennikov [4] state that A. planipennis occurs in Laos. In China, A. planipennis typically causes only minor damage to native tree species, generally attacking weakened or dying Asian ash (Fraxinus spp., Lamiales: Oleaceae) such as Fraxinus chinensis Roxb., F. mandshurica Rupr., and F. rhychophylla Hance [2,3]. However, A. planipennis readily infests and kills both stressed and healthy North American ash species including F. americana L., F. pennsylvanica Marshall, and F. velutina Torr. when planted in China [5,6] and has become one of the most serious invasive insect pests killing tens of millions of healthy ash trees in Eastern North America since its discovery in 2002 [1,7,8,9] and in Moscow, Russia [10]. It has been estimated that between the years 2009-2019, 17 million landscape ash trees in urban areas across 25 states will require treatment, removal and replacement at a cost of approximately $10.7 billion [11]. The large-scale mortality now occurring to native ash in forested and urban settings in North America will undoubtedly change urban landscapes and impact forest system processes, including threatening many other insect taxa with close evolutionary and ecological ties to ash [12]. Besides ash trees, A. planipennis was reported to feed more rarely on Juglans mandshurica Maximowicz, Pterocarya rhoifolia Siebold & Zuccarini (Fagales: Juglandaceae) and Ulmus davidiana Planchon (Rosales: Ulmaceae) in Asia [1,13]. In Europe, there is great concern that A. planipennis will spread westward from Moscow and threaten European ash species such as F. angustifolia Vahl, F. excelsior L., and F. ornus L. [10,13,14,15].
The higher levels of resistance demonstrated by Asian ash species to A. planipennis as compared with European and North American ash species is likely related to the fact that Asian ash species co-evolved with A. planipennis, while those in Europe and North America did not [16]. The evolutionary arms-race [17] between the wood-boring A. planipennis and its native Asian ash hosts has allowed Asian ashes to develop a suite of physical and phytochemical defenses that protect the trees against A. planipennis infestation except during times of environmental stress such as drought [18]. However, the non-Asian ash species lack these resistance mechanisms and thus are easily infested by A. planipennis even when healthy. A similar situation occurs in the case of Agrilus anxius Gory (bronze birch borer), a North American birch (Betula spp., Fagales: Betulaceae)-infesting species, that is usually only capable of infesting stressed North American birch, but can easily infest and kill European and Asian birch when planted in North America [19].
An effort is currently underway to identify relatives of A. planipennis that may pose a risk to North American woody plants if accidentally introduced [20]. Increased knowledge of larval morphology, along with a sound understanding of basic biology and ecology should help elucidate key evolutionary adaptations that allow some Agrilus Curtis species to become highly invasive when introduced to new environments. Of particular interest are adaptations that contribute to the ability of A. planipennis to effectively attack and kill healthy ash trees and undermine their defenses.
Volkovitsh & Hawkeswood [24] segregated Agrilus larvae into two informal groups or assemblages based on 1) presence or absence of microsetal areas along the anterior margin of the labrum and 2) of distinct zones of microspinulae concentrated on the internal surface of the maxillae (Figure 1) (Figures 1, 2, 3), fringe of microspinulae between maxillary stipes and base of maxillary palpus, and microspinulae concentrated subapically on the mala and internal surface of the stipes and cardo (Figure 1b) [24]. Species included by Volkovitsh & Hawkeswood [24] (Figures 3f, 3j, 4f, 5d, 5g). These species have a dense microsetal/microspinulated area on the anterior margin of the labrum, the epipharynx (Figures 3b, 3d-3g), and the internal surface of the maxillae, more than species in the A. viridis assemblage.
Despite its economic importance as an invasive species, all preimaginal stages, which include egg, instars I-IV, prepupa (nonfeeding terminal phase of instar IV), and pupa, of A. planipennis remain superficially described. Generalized descriptions of the larvae have been included in various biological or ecological studies [2,3,27,28]. Moreover, detailed descriptions of Agrilus larvae have been reported for only a small number of species in the genus [24,29,30,31,32,33,34,35]. Generalized information on biology and morphology of mostly Palearctic [34,36,37,38] and North America Agrilus species [39,40] is more common. With more than 2,750 species [21] recognized in the genus Agrilus, it remains a monumental task to amass descriptions and life history data for all Agrilus species worldwide. This study presents the first detailed description of the egg, larval instars I-IV, prepupa, and pupa of A. planipennis and compares the larvae to 7 Agrilus species to determine its affinity.

Description
We present a detailed morphological description for instar IV of Agrilus planipennis, along with egg, instars I-III, prepupa, and pupa. Since overall morphology is very similar between all instars, we describe only important distinguishing characteristics for stages I-III and prepupa for brevity. See 'Discussion' for more elaboration.
External lateral surface porous-like; dorsally smooth and shiny with streaked gelatinous appearance, resembling plastic film from above; symmetric depressions medially on lateral margins of dorsum. Glue-like substance on venter of egg that helps it adhere to bark surface.

Discussion
Our results provide additional evidence [1, 4,42] in favor of excluding A. planipennis from the subgenus Uragrilus. The larvae of 3 additional species currently classified in Uragrilus were examined for this study: A. anxius, A. ater, and the type species of the subgenus, A. guerini. These 3 species fall within the A. ater assemblage sensu Volkovitsh & Hawkeswood [24] because they have a pubescent, laterally expanded labrum and maxillae with a dense covering of microspinulae on the internal surface. In fact, larvae of A. anxius resemble Palaearctic species of Uragrilus, but additional phylogenetic and comparative studies are required to confirm relatedness. Larvae of A. planipennis instead share more features with A. politus and other species in the A. viridis assemblage sensu Volkovitsh & Hawkeswood [24]. These species have a glabrous labrum and microspinulae concentrated subapically on  the mala and internal surface of the stipes and cardo. This suggests that based on larval characters, A. planipennis does not belong in the subgenus Uragrilus as proposed by Alexeev [22] based on adult characters and its subgeneric position is unclear. Based on adult features, A. planipennis is considered to be more closely related to species in the A. cyaneoniger group [1,4], but the immature stages of species in this group remain unknown.
Larval characters useful in species discrimination include: 1) overall shape of abdominal segments; 2) pigmentation of pronotal and prosternal grooves [30]; 3) shape of either groove (entire or bifurcated); 4) presence or absence of glabrous space surrounding either groove; 5) structure of terminal processes [30], including the number, shape, and size of the excretory ducts (invagination of the inner surface of the urogomphi sensu Petrice et al. [41]), and presence/absence of ledges, particularly in latter instars; 6) extent of pilosity and shape of anterior margin of labrum (glabrous or pubescent and margin shape); 7) setation of labial prementum, which includes the relative length of apical setae on the corner sclerites; distance between bases of apical setae to posterior border of microsetal area (Alexeev ratio, Figure 1b); shape of posterior border of microsetal area (i.e., arcuate, zigzag, truncate, etc.) and of entire setal labial area; 8) sclerotization, shape of apical teeth of the mandible, and size of penicillum (Figures 1c, 3i, 7h); 9) extent of pilosity, proportions, and shape of apical antennal segment; and 10) size, shape, and number of spiracular trabeculae. The characters and their states are described below: 1. The overall shape of each abdominal segment, more pronounced in posterior segments, of A. planipennis is trapezoidal or bell-shaped, having the posterolateral angles produced laterad (Figures 6a, 10a, 14) (less in the prepupa, Figure 8a), differing from other known Agrilus larvae which have individual subquadrate abdominal segments (Figure 5h). The function of trapezoidal abdominal segments remains unknown. 2-4. The pronotal groove of A. planipennis is posteriorly bifurcate (Figures 6f, 6g, 10d-10e) and lacks a smooth space or border surrounding the groove. A similar pronotal groove is found in other species such as A. biguttatus (A. ater assemblage) (Figure 5d), therefore a posteriorly bifurcated pronotal groove is not unique to A. planipennis. Alternatively, the pronotal groove in other species may be entire as in A. politus and A. anxius, or also bordered by a glabrous area as in A. australasiae (Figure 5a), and to a lesser extent in A. guerini (Figures 5b-5c). The prosternal groove is entire in the species examined, including A. planipennis (Figure 10b). Some species may have a short posterior bifurcation (e.g., A. anxius). A smooth area may border the prosternal groove as in A. guerini (Figure 5b), A. anxius, and A. australasiae (Figure 5a), but absent in A. planipennis (Figures 10b-10e) and A. biguttatus (Figure 5d). The extent of the smooth area and posterior bifurcation may differ among species. 5. The terminal processes of A. planipennis are long, cylindrical and narrow and surrounded by few setae. With each subsequent instar the terminal processes become longer and the number of subdivisions or ledges increases. Instar I has 2 excretory ducts, older instars, including the prepupa, have 3 excretory ducts on each terminal process: apically, medially, and basally. As the larva matures (beginning with instar III), ledges or subdivisions begin to appear along the mesal (internal) margin of the terminal process  and the excretory ducts become deeper and more defined (Figures 11e, 11h, 11i, 12d-12f). The medial and basal excretory ducts do not extend laterad or posterad, but are limited to the internal margin (Figure 11h). Other species may have either the medial excretory duct greatly extending laterad with the basal excretory duct confined to the internal margin, for example as in A. subcinctus Gory [41], or both excretory ducts greatly extending laterad as in A. anxius, A. biguttatus (Figure 5g), and A. politus. Whether the excretory ducts greatly extend laterally in A. guerini remains unclear as we only examined a slide mounted larval preparation of this species. However, superficially, the terminal processes of this species resemble the terminal processes present in instar III of A. planipennis. In A. guerini, the medial and basal excretory ducts are more pronounced and extend slightly laterad; furthermore, the entire process is not cylindrical but laterally compressed. All species examined, except A. planipennis, have shorter and stouter terminal processes with the apical excretory duct being moderately wide and lack the numerous subdivisions or ledges present in instars III, IV and prepupa of EAB. 6. Variations on the shape and pilosity of the labrum are highly informative also in delimiting assemblages above the species level (e.g., A. viridis and A. ater assemblages). In addition to the presence or absence of pilosity on the anterior margin of the labrum between the A. viridis and A. ater assemblages, the overall shape of the labrum also differs between these assemblages. Species in the A. ater assemblage have slight lateral expansions directly beyond the apex of the palatine sclerites, making the anterolateral margin of the labrum, for species in the A. ater assemblage, subapically produced (Figures 3b, 3d, 3e), while for species in the A. viridis assemblage it is uniformly rounded (Figures 3a, 3c). The shape of the anteclypeus differs slightly among species (Figures 3a-3g), however, no specific pattern was apparent for these assemblages. 7. The labium is very useful in distinguishing among Agrilus species and features of this structure were used extensively by Alexeev [30,32] in his keys and descriptions of larvae of Palearctic Agrilus (Figures 1b, 4a-4g, 8b-8e). Agrilus planipennis has a sinuate, almost zigzag posterior contour of the microsetal area and the space between the anterior margin of the labrum and the posterior border of the microsetal area is equal to approximately 1/3 of the distance from the anterior margin to the bases of the apical setae (Figures 1b, 4a, 8b-8e). This ''Alexeev ratio'' [29] varies among species and can be defined as the distance between the anterior margin and posterior border of the microsetal area over (/) the distance between the anterior margin and the bases of the apical setae of the corner sclerites of the prementum (Figure 1b). A species-assemblage-level character found on the labiomaxillary complex is either the presence of microspinulae concentrated subapically on the mala and internal surface of the stipes and cardo (i.e., A. viridis assemblage) (Figures 4c, 4g) or a dense covering of microspinulae on the internal surface of the maxillae (i.e., A. ater assemblage) (Figures 4d-4f).   Figure 7h) or a completely smooth margin. The shape of the mandibles of A. planipennis and A. politus is very similar, being deltoid, while for A. australasiae the mandibles are quadrate to subquadrate. The penicillum in A. planipennis and A. politus is large, a characteristic typical of borers feeding on hard wood. However, the structure of the apex, cutting edge, and the shape of mandibles appears related to the density of the larval food [43]; being adaptive characters and not necessarily indicative of phylogenetic relationship. 9. Spiracles of A. planipennis are more circular and complete (thoracic spiracles more ''closed'' than abdominal spiracles) than in A. australasiae [24]. 10. The last segment of the antenna in A. planipennis is quadrate, while in A. australasiae it is deltoid. However, all sensory structures are present in both species with minor differences in position and size of microspinulae, located laterally and smaller in A. planipennis (Figure 7l) and apically and larger in A. australasiae [24]. Instars I and II of A. planipennis lack the fringe of microspinulae around the apex of antennal segment 1.
Agrilus planipennis larvae are recognized by the following combination of character states, including the first 6 states, which are unique among the species examined: 1, trapezoidal abdominal segments; 2, segment 10 setation sparse; 3, narrow, cylindrical terminal processes; 4, with numerous ledges appearing after instar II; 5, zigzag posterior contour of the microsetal area on prementum; 6, space between the anterior margin of prementum and posterior border of microsetal area is equal to approximately 1/3 of the distance from the anterior margin to the bases of the apical setae; 7, terminal processes with 2-3 excretory ducts; 8, smooth area between microdenticles and pronotal and prosternal grooves lacking; 9, pronotal groove posteriorly bifurcating; 10, prosternal groove entire; 11, labrum glabrous with margin not produced anterolaterally; 12, microspinulae concentrated subapically on the mala and internal surface of the stipes and cardo; 13, mandibles deltoid with well-defined apical teeth and large penicillum; 14, antennal segment 2 quadrate.

Differences among instars
Minor differences exist between instars of A. planipennis [27], including the degree of pigmentation of sclerotized structures such as mandibles, as well as setation and relative size. The developmental stages can be distinguished by the number of excretory ducts making up the terminal processes (2 in instar I and 3 in instars II, III, IV+prepupa) and the presence (instars III, IV+prepupa) or absence (instars I+II) of ledges. Among instars I-IV, the ventral antero-lateral setae of the labrum do not increase in size, therefore the relative size of the setae decreases with each instar (Figures 2a-2h). Differences also exist in the thoracic and abdominal compression [compression of the prepupa, being much shorter than instar IV (Figures 14a, 14b)] and the subsequent curling of the prepupa, becoming J-shaped, which is a major behavioral difference. The shape of the microspinulae differs among instars I+II and III+IV+prepupa, having comb-like  microspinulae in the former assemblage (Figures 12a-12c) and differentiated (single) microspinulae in the latter (Figure 9f).
Variation in the size and shape of the following structures has been used to determine the number of larval instars and duration of stadia for A. planipennis and other Agrilus larvae: terminal processes (frequently referred to as urogomphi), prothoracic plate, body width and length, and epistome width/length ratio (erroneously referred to as peristome [27]).

Conclusion
This study upholds the segregation of Agrilus larvae into two assemblages based mainly on differences in the mouthparts, the A. viridis and the A. ater assemblages as proposed by Volkovitsh & Hawkeswood [24]. Based on features of the larvae, retention of A. planipennis in the subgenus Uragrilus, which includes also A. ater and A. guerini as suggested by Alexeev [22,29], is dubious and substantiates recent studies [4] suggesting A. planipennis to be most closely related to species in the A. cyaneoniger group based on characters of the adult. However, that hypothesis could not be explicitly addressed in this study since immature stages of those species remain unknown.
While A. planipennis shares a similarly shaped posteriorly bifurcated pronotal groove with A. biguttatus, they differ in key characters, mainly the mouthparts and terminal processes. Even though larvae of A. planipennis are more similar to those in the A. viridis assemblage than to those in the A. ater assemblage (where species of Uragrilus cluster), adult characters do not support the placement of A. planipennis in the subgenus Agrilus where A. viridis and A. politus are currently classified based on adult characters. As such, given the limited knowledge of immatures in the genus (described for approximately 50 species) and pending a comprehensive phylogenetic analysis, this arrangement of classifying larvae into two major assemblages is for utilitarian purposes and not necessarily a reflection of evolutionary history.
Accurate identification of all life stages is essential to detect and successfully control and contain the spread of invasive forest pests like A. planipennis. Sets of characters herein described and illustrated will form the basis for future studies aimed at understanding the phylogeny of Agrilus. Understanding the evolutionary history of a group of organisms allows scientists not only to make predictions about potential invasive species with similar evolutionary histories and adaptations, but also helps scientists determine ways to manage invasive pests.
Homology of terminal processes with urogomphi is unwarranted. Urogomphi are derivates of the 9 th abdominal segment [45,46] while terminal processes are located on the 10 th segment. Terminal structures present in Agrilus should be termed terminal processes and are continuous with the 10 th segment. In some Buprestidae species, these terminal processes are present only in neonate larvae and lost in the mature larvae (Buprestis Linnaeus) [47] or they are present in all the larval instars and lost only in the prepupa (Anocisseis Bellamy) [48]. In Aphanisticini and Ethonion Kubáň there is a pair of lightly sclerotized tubercles instead of processes on the 10 th segment [24]. We consider terminal processes to be secondary ectodermal structures of the 10 th segment. Functionally, terminal processes serve to aid in the compression of excrements and as a support during larval movement within the galleries, as such, forming a morpho-functional complex with shortened VIII-X abdominal segments [29].
Abbreviations (codens) for institutions and collections used in the text follow Evenhuis [49]: NMNH-National Museum of Natural History, Washington, DC, USA.

Slide preparation
To study larval structures, microslides were prepared following the method used by Alexeev [29] using Fohr-Berlese media that acts also as a clearing agent to decompose soft tissue. Two slides were prepared per larval specimen: 1) mouthparts and 2) larval integument.
1. Mouthparts were separated from the head capsule along the posterior margin of the hypostome-pleurostome-epistome complex ( = peristome; all apical sclerotized structures of the head) using dissecting microscissors (Fine Science Tools, Foster City, CA, USA). Cutting into sclerites was avoided during dissection. Once the mouthparts were separated from the head, the mandibles were ''popped out'' with a pin or sharp forceps by gently exerting lateral (external) pressure on the inner subapex of the mandibles. Once both mandibles were extracted, the peristome complex was separated by inserting a pin between the hypostome-epistome suture (pleurostome). Both antennae and labrum were retained with the epistome. Any remaining external tissue was then removed from all sclerotized parts. The Fohr-Berlese media was placed in the center of a clean slide in the shape of a cross. The dissected mouthpart sections were arranged with the external surface upwards and along the y-axis of the cross, starting with the pair of mandibles, then the epistome and last the hypostome+pleurostome. Four minute pieces of firm paper were placed at the corners of the cross to prevent damage of the mandibles by pressure. A glass cover slip, previously rinsed in alcohol and dried, was slowly lowered over the Fohr-Berlese preparation slide from the margin of the liquid to avoid creating bubbles in the medium. Additional Fohr-Berlese media was placed along the sides of the cover slip to fill any gaps. The medium was drawn under the cover slip. 2. After the mouthparts were separated, the larval body was cut along a pleural line from the thorax to approximately the 9 th abdominal segment. The head capsule remained intact. The body was placed into 10% KOH aqueous solution and boiled until soft tissues were dissolved and the integument became completely transparent (approximately 5-10 minutes). The transparent integument was rinsed three times in water. Fohr-Berlese media was placed on a cleaned slide and then the integument was positioned with the external surfaces of the dorsum and the venter facing upward. This step took several minutes because the integument often became twisted during the rinsing process. Working over a black background was found to be helpful. After the integument was completely extended, the cover slip was placed on the slide from the margin of the liquid and very gently and slowly lowered with forceps or a pin to avoid bubbles. The slides were continuously maintained in a horizontal position and then placed for a few hours in an oven at approximately 30uC.

Imaging
The following equipment was used for observation and imaging: Larval body: a Leica (Wetzlar, Germany) MZ 9.5 dissecting microscope with a Leica DFC290 mounted camera. Instar IV slides: a Leica DME light microscope with a Panasonic (Secaucus, NJ, USA) Super Dynamic WV-GP460 analogous camera. Instar I: a Leica DM 5000 B polarizing microscope with a Leica DFC320 mounted camera. Instars II, III, prepupa, pupal body: a Zeiss (Oberkochen, Germany) Discovery.v20 stereomicroscope and an AxioCam HRc; mouthparts of instars II, III, prepupa: a compound microscope Leitz DIAPLAN with an AxioCam HRc. Scanning electron micrographs (SEM) were taken with a Philips XL-30 ESEM with LaB6 electron source.
Ideally, A. planipennis larvae would be compared to the larvae of species in the A. cyaneoniger species-group, which are hypothesized to be the closest relatives of A. planipennis based on adult features [1,4]. However, immature stages of species in this group remain unknown (i.e., A. agnatus Kerremans, A. auristernum Obenberger, A. bifoveolatus Kerremans, A. cyaneoniger Saunders, A. lafertei Kerremans, A. lubopetri Jendek, A. qinling Jendek). For this reason, A. australasiae, described in detail by Volkovitsh & Hawkeswood [24] and 6 other distantly related species, were used for comparison. Collection of samples from Bath, MI, USA was approved by the landowner, Mr John Valo. No specific permits were required for the collection of samples from the Forest on the Vorskla River field station in Russia in 1971, it was not privately owned or protected, and the field study did not involve endangered or protected species.