Freeze stress of deciduous trees induces attacks by opportunistic ambrosia beetles

A broad host range and the utilization of living but weakened trees contribute, in part, to the invasion success of ambrosia beetles (Curculionidae: Scolytinae). The present study assessed the capability of freeze stress to induce attacks by ambrosia beetles. Freeze stress predisposed Cercis canadensis L., Cornus florida L., Malus pumila Mill. and Styrax japonicus Sieb. to attack under field conditions, although no attacks occurred on untreated trees. More attacks occurred on freeze‐stressed versus flood‐stressed M. pumila in Virginia but not for S. japonicus in Ohio. Attacks on flooded trees were skewed towards the base of the trunk, whereas attacks on freeze‐stressed trees mainly occurred around the upper regions of the trunk and into the branches. The predominant species recovered were Anisandrus maiche Stark and Xylosandrus germanus (Blandford) in Ohio, and Xylosandrus crassiusculus (Motschulsky) in Virginia. Ethanol emissions from trunks of S. japonicus were detected by solid phase microextraction‐gas chromatography–mass spectrometry at 1 day after imposing freeze stress, peaking 4 days after injury. Trees with an intolerance of freeze stress are predicted to be vulnerable to attack, especially when subjected to mild winter temperatures followed by late‐spring freezes. Freeze stress could thereby facilitate the destructiveness of exotic ambrosia beetles.


Introduction
Ambrosia beetles (Curculionidae: Scolytinae) in the tribe Xyleborini are among the most successful and destructive wood-boring insects worldwide (Hulcr & Stelinski, 2017). Thirty non-native species of Xyleborini are currently established in North America (Gomez et al., 2018) and the Xyleborini also represent half of the exotic Scolytinae introduced into Europe (Kirkendall & Faccoli, 2010). After dispersing from their natal galleries, adult females tunnel into host trees tissues, where they establish and cultivate gardens of their nutritional fungal symbiont (Hulcr & Stelinski, 2017). Adult females carry spores of their fungal symbiont within specialized structures called mycangia, which are then used to inoculate the tunnel walls of host tissues. The fungal symbionts are usually Ambrosiella and Raffaelea spp., although examples also exist of mutualism with Fusarium spp. (Kasson et al., 2013;Kostovcik et al., 2015;O'Donnell et al., 2015). Oviposition occurs after the fungal gardens are flourishing, with the larvae and adults consuming the fungal symbiont to properly develop and reproduce (French & Roeper, 1972;Biedermann & Taborsky, 2011). Xyleborini offspring are produced via haplodiploid reproduction and sibling mating subsequently occurs between the haploid males and diploid females. Their cryptic wood-boring behaviour, fungus farming lifestyle, haplodiploid reproduction and sibling mating likely contribute to the invasion success of ambrosia beetles (Smith & Hulcr, 2015;Gomez et al., 2018). A broad host tree range and association with weakened trees also likely promotes the establishment and proliferation of ambrosia beetles introduced into new habitats, with this relationship probably being facilitated by extreme climatic events producing an availability of weakened host resources (Ranger et al., 2013(Ranger et al., , 2015a.
Ambrosia beetles are often ranked among the most destructive insect pests of trees growing in ornamental nurseries because of their wood-boring behaviour and association with branch dieback and tree death (Oliver & Mannion, 2001;Fulcher et al., 2012;Ranger et al., 2016). They are also known to attack large stands of trees in forested systems (Grégoire et al., 2001;Henin & Versteirt, 2004;La Spina et al., 2013). In particular, Xylosandrus crassiusculus (Motschulsky) and Xylosandrus germanus (Blandford) comprise two of the most destructive ambrosia beetles of trees growing in ornamental nurseries and tree fruit orchards (Agnello et al., 2015;Ranger et al., 2016). Both X. crassiusculus and X. germanus are native to southeast Asia and have become established around the world, including parts of North America, South America, Africa, Europe and Oceania (Rassati et al., 2016;Gallego et al., 2017;Landi, 2017;Gomez et al., 2018;CABI, 2018a, b;Kavčič, 2018). Adult females of X. crassiusculus and X. germanus tunnel into the heartwood of host trees, where they cultivate fungal gardens of Ambrosiella roeperi T.C. Harr. & McNew and Ambrosiella grosmanniae C. Mayers, McNew & T.C. Harr., respectively (Harrington et al., 2014;Mayers et al., 2015). Although X. crassiusculus and X. germanus attack a broad range of 100-200 trees (Schedl, 1963, Weber & McPherson, 1983, both species exhibit a narrow preference for living but weakened trees emitting stress-induced ethanol (Ranger et al., 2013(Ranger et al., , 2015b. Ethanol acts as a chemical indicator of weakened trees (Kelsey, 2001) and also benefits the colonization success of ambrosia beetles by promoting the growth of their fungal gardens (Ranger et al., 2018). A variety of abiotic and biotic stressors can induce the production and emission of ethanol by trees, including flood and drought stress, frost injury, girdling, impaired root function, root and crown disturbance, pollutants, and pathogens (Kimmerer & Kozlowski, 1982;Kimmerer & MacDonald, 1987;Kelsey & Joseph, 2001;Kelsey et al., 2013;Ranger et al., 2015a). Previous studies have demonstrated that flood stress predisposes trees to attack, and that Xylosandrus spp. can distinguish among trees varying in the tolerance of flooding (Ott, 2007;Ranger et al., 2013Ranger et al., , 2015aReed et al., 2015). Based on this knowledge, Frank and Ranger (2016) developed a moisture management threshold to help growers avoid overwatering their trees and thereby minimize the risk of ambrosia beetle attacks. Characterizing the role of tree health on host selection and utilization by Xyleborini ambrosia beetles is therefore useful for developing management tactics.
In addition to the role of flood-stress, a growing number of studies suggest that freeze-stressed trees are also preferentially attacked by ambrosia beetles. For example, oak (Quercus rubra L.) trees growing in Germany that showed signs of frost damage but otherwise appeared healthy were attacked by X. germanus (Heidenreich, 1960;Gauss, 1960). Young sycamore (Acer pseudoplatanus L.) trees severely injured or killed by frost were also attacked by X. germanus (Heidenreich, 1960;Gauss, 1960). More recently, La Spina et al. (2013) documented that extreme frost events preceded attacks by X. germanus and other ambrosia beetles on large stands of European beech trees, Fagus sylvatica L., in Western Europe. Bark tissues of F. sylvatica experimentally-stressed using blocks of dry ice were also more attractive to X. germanus than untreated tissues (La Spina et al., 2013). Freeze stress also preceded attacks on thin-barked deciduous trees growing in ornamental nurseries in northeastern Ohio, U.S.A. (Ranger et al., 2015a).
Based on these previous studies, we hypothesized that deciduous trees subjected to freeze stress would be attacked by opportunistic ambrosia beetles targeting living but weakened trees, such as flood-stressed trees. To test this hypothesis, the present study aimed to: (i) assess the capability of freeze stress to induce attacks by ambrosia beetles; (ii) compare the preference behaviour of ambrosia beetles for freeze-stressed versus flood-stressed trees; and (iii) characterize the time course emission of ethanol from freeze-stressed trees.

Imposing freeze stress
Experiments were conducted in Ohio, U.S.A., and Virginia, U.S.A., because X. germanus and X. crassiusculus are the predominant exotic ambrosia beetles attacking trees in ornamental nurseries in these two states, respectively. Trees used in the Ohio experiments were 3-4 years old, 2.54 cm in diameter, approximately 1.2 m in height and were grown in 18.9-L pots containing a mixture of 90 : 10 pine bark and sphagnum peat moss, along with lime and Micromax Micronutrients (Scotts Co., Marysville, Ohio). Trees used in the Virginia experiments were approximately 3-4 years old, 2.54 cm in diameter, approximately 1.2 m in height and were grown in 28-L pots containing a mixture of aged pine bark and coarse sand (92 : 8), as well as dolomitic lime to stabilize pH. The media was top dressed with Osmocote Plus 15−9−12 (Scotts Co., Marysville, Ohio) slow-release fertilizer.
For the Ohio experiments, freeze stress was imposed by holding four container-grown Styrax japonicus Sieb. trees overnight (approximately 12 h) in a freezer (internal dimensions, length × width × height: 76.2 × 69.9 × 160.0 cm; model NSPF331WWW/0; Norlake Scientific, Hudson, Wisconsin) programmed to −10 ∘ C. For the Virginia experiments, freeze stress was imposed by holding container-grown S. japonicus, Cornus florida L. and Cercis canadensis L. trees overnight for three consecutive nights in a commercial cold storage facility maintained at −10 ∘ C. Trees were relocated outside the cold room during the day and returned in the evening. Malus pumila trees deployed in Virginia were only held overnight in a freezer. After their removal from the freezer, freeze-stressed trees were used for assessing ambrosia beetle host selection (Ohio and Virginia) and the induction of ethanol emissions (Ohio). The deciduous trees S. japonicus, C. florida and C. canadensis were selected for our experiments because previous studies have indicated that freeze stress predisposes them to attack by ambrosia beetles (Ranger et al. 2015b). Similarly, M. pumila was selected because of an intolerance of freeze stress (Cannell & Smith, 1986;Khanizadehi et al., 1989). We also incorporated these tree species into our experiments because X. germanus and X. crassiusculus are problematic in ornamental nurseries and tree fruit orchards.

Preference for freeze-stressed trees
Free-choice experiments were conducted under field conditions in Ohio and Virginia to assess the host selection behaviour of ambrosia beetles for freeze-stressed and untreated control trees.
Two separate experiments were conducted in Ohio. In both experiments, freeze-stressed S. japonicus trees (n = 4) were removed from a laboratory freezer after overnight exposure and paired with an untreated control tree (n = 4) in a deciduous woodlot in Wayne Co., Ohio (40 ∘ 45 ′ 40.85 ′′ N, 81 ∘ 51 ′ 14.71 ′′ W). Freeze-stressed and control trees were spaced 1.5 m apart with 4 m between adjacent pairs of trees. Trees were held under field conditions for 8 and 10 days after imposing freeze stress in Experiments 1 and 2 conducted in Ohio, respectively. Attacks on trees were counted at 1, 4, 7, and 8 days after imposing freeze stress and field deployment during Experiment 1. Similarly, attacks were counted at 1, 4, 7, 8, and 10 days during Experiment 2. On the last day of field deployment, the trunks were cut at the soil line and transferred to a walk-in cooler stored at 5 ∘ C. Pruning shears were next used to dissect the tunnels/galleries and recover the ambrosia beetle foundresses. Scolytinae specimens were stored in 70% ethanol and identified to species. Experiment 1 was conducted in Ohio from 10 July 2014 to 18 July 2014 and Experiment 2 was conducted from 29 July 2014 to 11 August 2014.
Two separate field experiments were also conducted in Virginia. In Experiment 1, freeze-stressed S. japonicus (n = 6) trees were freeze-stressed for three consecutive nights and then paired with a non-injured tree (n = 6) along the edge of a deciduous woodlot in Virginia Beach, Virginia, U.S.A. (36 ∘ 53 ′ 16 ′′ N, 76 ∘ 10 ′ 22 ′′ W). Freeze-stressed and untreated control trees were spaced 3 m apart with 6 m between adjacent pairs of trees. Trees were held under field conditions for 10 days after imposing freeze stress in Experiment 1. Attacks on trees were counted at 2, 4, 8, and 10 days after field deployment during Experiment 1. In Experiment 2, freeze-stressed C. canadensis (n = 6), C. florida (n = 6) and S. japonicus (n = 6) trees were removed from a freezer and arranged in randomized complete blocks with untreated trees of each species (n = 6 per species) along the edge of the aforementioned woodlot. A distance of 3 m was maintained between adjacent trees within a block and 6 m between adjacent blocks. Attacks on trees were counted at 4, 6, 12 and 14 days after field deployment during Experiment 2. On the last day of field deployment, the trunks were cut at the soil line, transferred to a walk-in cooler and stored at 5 ∘ C. Pruning shears were used to recover the ambrosia beetle foundresses, which were stored in 70% ethanol and identified to species. Experiment 1 was conducted in Virginia from 2 June 2014 to 12 June 2014 and Experiment 2 was conducted from 26 June 2014 to 10 July 2014.

Comparison between freeze-stressed and flooded trees
Because previous studies have demonstrated Xylosandrus spp. attack flood-stressed S. japonicus and other tree species (Ranger et al., 2013(Ranger et al., , 2015b, we conducted additional experiments to compare their preference behaviour when given a choice between freeze-stressed versus flood-stressed S. japonicus and M. pumila Mill trees. Container grown trees were freeze-stressed in accordance with the previously described methods for Ohio and Virginia, except that trees in Virginia were exposed to freeze stress for one night instead of three. Flood stress was imposed using a pot-in-pot system as described by Ranger et al. (2013). Freeze-stressed trees were deployed under field conditions the same day that they were removed from the freezer, whereas flood stress was imposed the same day trees were deployed under field conditions in Ohio and Virginia.
In Ohio, freeze-stressed (n = 12), flood-stressed (n = 12) and untreated control (n = 12) S. japonicus trees were deployed in a randomized complete block design within the aforementioned deciduous woodlot in Ohio. A distance of 4 m was maintained between adjacent trees within a block and 8 m between adjacent blocks. Because only four container-grown S. japonicus trees could fit in the aforementioned freezer used in Ohio, three groups of four freeze-stressed trees were deployed along with flood-stressed and untreated control trees on three separate occasions. Trees were deployed on 1 June 2018 for the first group, on 5 June 2018 for the second group and on 15 June 2018 for the third group. Attacks per tree were counted every 2-3 days. Trees in the first and second groups were cut at the base on 14 June 2018, and trees in the third group were cut at the base on 3 July 2018. Cut trunks were immediately transferred to a walk-in cooler stored at 5 ∘ C.
To compare the host-selection behaviour of ambrosia beetles attacking the freeze-stressed versus flood-stressed trees, we measured the vertical distribution of attacks on the experimentally-stressed trees deployed in Ohio and Virginia. After returning the cut trunks to the laboratory, a ruler was used to measure the distance (cm) from the base of the trunk to each attack, along with noting whether the attacks occurred to the trunk or branches. Ambrosia beetle specimens dissected from the attacked trunks were identified to species. The presence of eggs, larvae and pupae within each gallery was also quantified.
In Virginia, freeze-stressed (n = 8), flood-stressed (n = 8) and untreated control (n = 8) M. pumila trees were used. Trees were deployed in a randomized complete block design within a deciduous woodlot in Virginia Beach, Virginia (36 ∘ 52 ′ 3 ′′ N, 76 ∘ 17 ′ 12 ′′ W). A distance of 3 m was maintained between adjacent paired trees, and 6 m between adjacent pairs. Attacks per tree were counted every 2-3 days when trees were deployed from 18 April 2018 to 9 May 2018. The distance of attacks from the base of the trunk were measured (cm) as described above. Ambrosia beetle specimens recovered from the attacked trees were identified to species.

Ethanol emissions
Time course emissions of ethanol from the bark of freeze-stressed and uninjured S. japonicus trees were examined using solid phase microextraction-gas chromatographymass spectrometry (SPME-GC-MS). Styrax japonicus trees were freeze-stressed in accordance with the method described above for the Ohio location, and then immediately transferred to a greenhouse on the campus of the Ohio Agricultural Research and Development Center. Sampling chambers ( Fig. 1) were Chamber assembled using a metal frame and Teflon sheet for passively sampling volatile emissions from trunks of Styrax japonicus using solid phase microextraction (SPME)-gas chromatography-mass spectrometry. (A) Chambers were assembled by first wrapping two or three layers of Teflon tape around the trunk (vertical white scale bar = 20 cm). (B) A metal support frame was then secured to the trunk using an additional two or three layers of Teflon tape. (C) A syringe containing a retracted SPME fibre was then secured in parallel with the trunk using a zip tie. placed on the trunks of freeze-stressed and untreated S. japonicus trees to sample for ethanol emissions. Chambers were assembled by first wrapping two or three layers of Teflon tape around the trunk in a 4-5-cm band spaced approximately 21-22 cm apart (Fig. 1A). A metal support frame (diameter 6.5 cm; length 23 cm; volume 763.21 cm 3 ) was then secured to the trunk using an additional two or three layers of Teflon tape (Fig. 1B). A syringe containing a retracted SPME fibre (CAR/PDMS; 75 m coating; Sigma-Aldrich, St Louis, Missouri) was next secured in parallel with the trunk using a zip tie (Fig. 1C). The SPME fibre was positioned so the exposed fibre was in parallel with the trunk at a distance of 0.25-0.5 cm from the bark surface. A Teflon™ PFA sheet (length 60 cm, width 30 cm; Welch Fluorocarbon Inc., Dover, New Hampshire) was then placed around the chamber and the free ends were rolled together until snug against the metal frame (Fig. 1D, E). Plastic zip ties were tightened around the center and both ends of the chamber (Fig. 1F). Small gaps on either side of the SPME syringe were enclosed using a strip of activated charcoal filter (length 3 cm, width 1 cm; Hunter Fan Company, Tennessee). Because SPME is a passive sampling process, the chamber allowed for volatiles emitted from the bark epidermis (approximately 157 cm 2 ) to accumulate in the headspace of the chamber. Assembling a chamber required 4-5 min and could be accomplished by one person. After assembly, the SPME fibre was immediately exposed within the chamber for 3 h to passively collect volatiles accumulating in the headspace. After sampling, the SPME fibres were retracted into the syringe holder and removed from the chamber. The syringe holders were then disassembled and the SPME fibre assembly was sealed in a capped glass test tube and stored at −40 ∘ C until analysis. Ethanol was sampled from freeze-stressed (n = 3) and untreated control (n = 3) trees on day 0, before imposing freeze stress on the same day, and then at 1, 4, 8, 11 and 19 days after imposing freeze stress.
For analysis by GC-MS, SPME fibres were thermally desorbed for 2 min at 250 ∘ C in the injection port of an Agilent 7890B gas chromatograph (Agilent Technologies, Palo Alto, California) operated in splitless mode from 0-2 min and then a 1 : 20 split for the remainder of the run. A DB-5MS column (0.25 m × 30 m × 0.25 m; cross-linked/surface bonded 5% phenyl, 95% methylpolysiloxane; Agilent J&W, Santa Clara, California) was used for analysis: 40-60 ∘ C at 3 ∘ C/min and 60-230 ∘ C at 20 ∘ C/min. An Agilent 5977A mass spectrometer was operated in electron impact mode with a scan range of 40-415 amu. The external standard method was used for determining concentrations of ethanol present within the headspace of the chambers (Bartelt, 1997;Romeo, 2009). Serial dilutions containing known amounts of ethanol were applied using a pipette to filter paper discs (diameter 2.5 cm; qualtitative P5, medium porosity; Fisher Scientific, Fair Lawn, New Jersey) placed inside the chamber and sampled using SPME as described above. A glass rod (diameter 2.54 cm) was used in replacement of a tree stem within the sampling chamber. Peak areas associated with the ethanol standards were measured and a standard concentration curve was used to determine concentrations of ethanol associated with the trunk tissues. Fibres were conditioned before sampling by exposure within the injection port for 20 min at 250 ∘ C. In all cases prior to sampling, Teflon sheets and steel chambers were heated in an oven for 2 h at 37 ∘ C.

Statistical analysis
For the experiments concerning the preference for freezestressed trees, an unpaired t-test ( = 0.05) was used to compare the number of attacks per tree on freeze-stressed versus untreated control S. japonicus trees as part of both experiments conducted in Ohio (SAS Institute, 2001). An unpaired t-test was also used to compare attacks to trunks versus branches on freeze-stressed S. japonicus trees deployed in Ohio, along with the number of X. germanus and Anisandrus maiche specimens recovered from the trunks versus branches. Similarly, an unpaired t-test was used to compare the number of attacks on freeze-stressed versus untreated control S. japonicus trees deployed as part of the first experiment conducted in Virginia. A one-way analysis of variance (anova) and Tukey's honestly significant difference (HSD) was used to compare the number of attacks on freeze-stressed and uninjured C. canadensis, C. florida and S. japonicus trees deployed as part of the second experiment conducted in Virginia. An unpaired t-test was used to compare the number of X. crassiusculus versus X. germanus specimens recovered from freeze-stressed C. canadensis as part of Experiment 1 and C. canadensis, C. florida and S. japonicus trees deployed as part of Experiment 2 in Virginia.
For the experiments concerning the comparison between freeze-stressed and flooded trees, an unpaired t-test ( = 0.05) was used to compare the number of attacks on freeze-stressed versus flood-stressed S. japonicus trees deployed in Ohio, and also Virginia. A correlation analysis was used to assess the relationship between freeze stress and flood stress on the vertical distribution of attacks (SAS Institute, 2001). Furthermore, the Kolmogorov-Smirnov two-sample test was used to compare the vertical distribution of ambrosia beetle attacks on freeze-stressed versus flood-stressed trees (SAS Institute, 2001). An unpaired t-test was used to separately compare the number of A. maiche Stark versus X. germanus adults, eggs and larvae recovered from freeze-stressed and flood-stressed S. japonicus trees deployed in Ohio. According to the species of ambrosia beetles, an unpaired t-test was also used to compare the number A. maiche or X. germanus adults, eggs and larvae recovered from freeze-stressed versus flood-stressed trees deployed in Ohio. A one-way anova and Tukey's HSD were used to separately compare the number of ambrosia beetle adults recovered from freeze-stressed and flood-stressed M. pumila trees deployed in Virginia. According  to the species of ambrosia beetles, an unpaired t-test was also used to compare the number adults, eggs and larvae recovered from freeze-stressed versus flood-stressed trees deployed in Virginia. Repeated measures anova was used to compare the emission of ethanol from freeze-stressed S. japonicus trees over time (SAS Institute, 2001). Count data were log(x + 1) transformed prior to analysis but untransformed data are presented.

Preference for freeze-stressed trees
Ambrosia beetles preferentially attacked freeze-stressed S. japonicus trees over untreated control trees during two experiments conducted in Ohio ( Fig. 2A, B). In Experiment 1, ambrosia beetle attacks on S. japonicus were detected 1 day after imposing freeze stress ( Fig. 2A). By day 8, a mean ± SE of 22.5 ± 9.6 attacks per tree was recorded on freeze-stressed S. japonicus, whereas no attacks occurred on the untreated control trees. There was no difference in the mean number of attacks that occurred to the trunks versus branches of freeze-stressed S. japonicus (Table 1). In Experiment 2, ambrosia beetle attacks on S. japonicus were also initiated 1 day after imposing freeze stress (Fig. 2B). A mean of 37.3 ± 17.7 attacks per tree occurred by day 10, whereas no attacks occurred on the untreated control S. japonicus trees. No difference was detected in the number of attacks that occurred to the trunks versus branches of freeze-stressed S. japonicus trees (Table 1).
During the two free-choice experiments conducted in Virginia ambrosia beetles preferentially attacked freeze-stressed deciduous trees deployed under field conditions (Fig. 2C, D). In Experiment 1, a mean ± SE of 31.5 ± 13.6 attacks per tree occurred by day 10 to freeze-stressed S. japonicus trees, whereas no attacks occurred on the untreated control trees (Fig. 2C). In Experiment 2, a mean of 58.2 ± 18.1 attacks per S. japonicus tree, 37.7 ± 5.6 attacks per C. florida tree and 31.2 ± 5.6 attacks per C. canadensis tree were recorded at 14 days after imposing freeze stress (Fig. 2D). There was no difference in cumulative attacks among freeze-stressed C. canadensis, C. florida and S. japonicus trees, although there was a significant difference compared with the 0.0 ± 0.0 attacks on the untreated control trees (Fig. 2D).
In total, 54 X. germanus specimens were recovered from the freeze-stressed S. japonicus trees attacked during Experiment 1 in Ohio; no other Scolytinae species were recovered (Table 1). A mean ± SE of 13.5 ± 6.4 X. germanus were recovered per freeze-stressed S. japonicus. There was no difference in the number of X. germanus specimens recovered from trunks versus branches during Experiment 1 (Table 1). In total, 86 X. germanus and five Anisandrus maiche Stark were recovered from the freeze-stressed S. japonicus trees deployed during Experiment 2 in Ohio (Table 1). A mean of 21.5 ± 9.5 X. germanus specimens was recovered per freeze-stressed S. japonicus compared with 1.3 ± 0.5 A. maiche per tree (t = 2.57; d.f. = 6; P = 0.043). There was no difference in the number of X. germanus or A. maiche specimens recovered from trunks versus branches during Experiment 2 (Table 1).
In total, 162 X. crassiusculus and 2 X. germanus were recovered from freeze-stressed S. japonicus trees deployed during Experiment 1 in Virginia (Table 2). Similarly, 133 X. crassiusculus and 1 X. germanus were recovered from freeze-stressed S. japonicus deployed during Experiment 2 (Table 2). In total, 112 and 40 X. crassiusculus specimens were recovered from freeze-stressed C. florida and C. canadensis trees deployed during Experiment 2 in Virginia, respectively (Table 2). No X. germanus were recovered from freeze-stressed C. florida or C. canadensis trees.

Comparison between freeze-stressed and flooded trees
Ambrosia beetle host-selection and preference behaviour were compared when given a choice between freeze-stressed versus flood-stressed trees deployed in Ohio and Virginia. Under free-choice conditions, ambrosia beetles attacked both freeze-stressed and flood-stressed trees deployed in Ohio and Virginia. In Ohio, there was no difference in the mean ± SE of 17.1 ± 6.2 attacks per freeze-stressed S. japonicus tree compared with the 10.3 ± 2.3 attacks per flood-stressed tree (Fig. 3A). In Virginia, a mean of 154.6 ± 10.9 attacks occurred per freeze-stressed M. pumila tree, which was higher than the 28.8 ± 6.7 attacks per flood-stressed tree (Fig. 3B).
Because ambrosia beetles were observed to attack the branches of S. japonicus during the previously described experiments concerning the preference for freeze-stressed trees, we compared the vertical distribution of attacks on freeze-stressed versus flood-stressed trees. A significant difference was detected in the vertical distribution of ambrosia beetle attacks on freeze-stressed versus flood-stressed S. japonicus trees deployed in Ohio (Kolmogorov-Smirnov two-sample test; P < 0.0001) (Fig. 4A, B). Attacks on flood-stressed S. japonicus trees were skewed towards the lower portion of the trunk, whereas attacks on freeze-stressed S. japonicus mainly occurred around the middle to upper regions of the trunk and extended into the branches (Fig. 4A, B).
A significant difference was also detected in the vertical distribution of ambrosia beetle attacks on freeze-stressed versus flood-stressed M. pumila trees deployed in Virginia (Kolmogorov-Smirnov two-sample test; P < 0.0001) (Fig. 4C,  D). Attacks on flood-stressed M. pumila trees were skewed towards the lower portion of the trunk, whereas attacks on freeze-stressed M. pumila trees mainly occurred around the middle to upper regions of the trunk and extended into the branches (Fig. 4C, D).

Stress-induced ethanol emissions
Time course emissions of ethanol from the trunks of freeze-stressed and untreated control S. japonicus were  characterized using SPME-GC-MS (Figs 1 and 5). A significant repeated measures within-treatment effect was detected in the emission of ethanol from the freeze-stressed trees (F = 22.49; d.f. = 5, 10; P = 0.0087). Ethanol emissions were not detected on day 0 before inducing freeze stress, although a mean ± SE of 4.31 ± 2.36 ng of ethanol/cm 2 /h was detected 1 day after freeze stress (Fig. 5) (Fig. 5). Overall, ethanol emissions were significantly higher on day 4 after imposing freeze stress compared with day 1 and days 11 and 19 (Fig. 5). No ethanol emissions were detected from the bark of untreated control S. japonicus (data not included).

Discussion
Low temperature stress is one of the most important factors affecting the production of horticultural crops (Ashworth, 1992). Freeze injury is generally the result of a rapid drop in temperature below freezing after warm or mild weather during autumn, winter or spring months (Schoeneweiss, 1975;Burke et al., 1976). This scenario can occur at relatively small scales affecting a few individual trees (Motzkin et al., 2002;Ranger et al., 2015a) or at a regional scale affecting large stands of trees Freeze b 108.6 ± 8.9 a 3.9 ± 0.7 b 1.8 ± 0.5 bc 1.5 ± 0.5 cd 0.5 ± 0.3 cd 18.8 ± 8.1 a 2.5 ±  (La Spina et al., 2013;Kosiba et al., 2018;. Trees injured by freezing can then be predisposed to attack by secondary wood-boring insects (Solomon, 1995;Moraal & Hilszczanski, 2000;Vansteenkiste et al., 2004). The present study provides one of the few examples of ambrosia beetles attacking experimentally freeze-stressed deciduous trees, in addition to demonstrating the emission of stress-induced ethanol from trunks of injured trees. Most attacks were by non-native Xyleborini species that rapidly attacked the freeze-stressed trees. Targeting trees in the early stages of physiological stress might contribute to the invasion success of Xyleborini ambrosia beetles that utilize living but weakened trees (Kühnholz et al., 2001;Ranger et al., 2015a). Freeze-stressed trees were rapidly attacked during the present study, with attacks often beginning within 1-2 days after injury. Ambrosia beetles have also been observed to begin attacking flood-stressed trees within 1-2 days after flooding (Ranger et al., 2013(Ranger et al., , 2015b. Xylosandrus germanus was one of the predominant species excavated from freeze-stressed trees deployed in Ohio, whereas X. crassiusculus was the predominant species recovered in Virginia. The present study also documents the recovery of A. maiche from stressed trees deployed in Ohio. Notably, A. maiche was one of the predominant species recovered along with X. germanus when beetles were allowed to freely choose between freeze-stressed and flood-stressed trees. After first being reported from North America in 2009, A. maiche now appears to be established in eight states within the Midwestern and Eastern U.S.A. (Rabaglia et al., 2009;Atkinson, 2018). Successful colonization of freeze-stressed and flood-stressed trees, as made evident by the presence of eggs and larvae within galleries, was documented for A. maiche and X. germanus during the present study. These results provide additional insight into the utilization of trees in the early stages of physiological stress by certain Xyleborini ambrosia beetles.
By measuring the vertical distribution of attacks, ambrosia beetles were demonstrated to preferentially attack the lower trunk region on flood-stressed trees, whereas attacks on freeze-stressed trees mainly occurred around the middle and upper trunk regions and branches. Heidenreich (1960) also observed that trunks and branches of freeze-stressed trees were attacked by X. germanus, whereas trunks are mainly attacked on flood-stressed trees . Understanding how physiological stressors influence the vertical distribution of ambrosia beetle attacks could aid tree care professionals in determining what stressors potentially predisposed the trees to attack, thereby helping to predict and mitigate the risk of future attacks. Although not confirmed, we hypothesize the difference in distribution of attacks on freeze-stressed versus flood-stressed trees corresponds to the presence of ethanol within these tissues. Kelsey et al. (2013) demonstrated that more than four-fold ambrosia beetle attacks occurred above ethanol-infused sapwood tissues compared with the opposite side of the same log that did not any receive ethanol treatment. Gara et al. (1993) used a stainless steel chamber to detect ethanol emissions from the bark of lodgepole pine [Pinus contorta murrayana (Grev. and Balf) Engelm]. trees infected with heartwood decay fungi. The wire frame and Teflon-base chambers used as part of the present study allowed for the confirmation and measurement of ethanol emissions from the bark of freeze-stressed trees.
When cellular oxygen levels decline and are depleted, anaerobic respiration results in pyruvic acid being converted into acetaldehyde, which is then reduced to ethanol (Fall, 1999). Alcoholic fermentation typically occurs in the roots of plants or trees in response to flooding (Fall, 1999). Ethanol and acetaldehyde are not only then secreted by the roots, but also enter the transpiration stream and are subsequently emitted from the bark epidermis and the leaves (Kimmerer & Kozlowski, 1982;Mac-Donald & Kimmerer, 1993;Ranger et al., 2013Ranger et al., , 2015b). By contrast, an inability to uptake O 2 by cells within trunks injured by freeze stress would also lead to the production and emission of ethanol from these tissues (Kimmerer & Kozlowski, 1982;Forney et al., 2000;Obenland et al., 2003).
As noted previously, the presence of ethanol within host tree tissues benefits fungiculture by X. germanus and other ambrosia beetles by promoting the growth of the fungal symbiont and suppressing the growth of other fungal competitors (Ranger et al., 2018). Because foundress ambrosia beetles do not begin laying eggs until the fungal gardens are established, the presence of ethanol within the host tree tissues indirectly benefits the colonization success of these opportunistic ambrosia beetles. Healthy trees baited with an ethanol lure were not included as a positive control as part of the present study. However, Ranger et al. (2018) previously demonstrated that healthy C. canadensis and C. florida trees baited with ethanol lures were attacked but no fungal gardens or offspring were produced in the absence of in vivo ethanol within the host tree tissues. Still, additional studies are warranted to determine whether other compounds besides ethanol within the host tree tissues are beneficial to ambrosia beetle fungiculture and their colonization success.
A warming climate with a continued risk of freeze stress could result in increased attacks by Xyleborini ambrosia beetles, particularly for intolerant trees growing in ornamental nurseries and tree fruit orchards. Unusually warm periods followed by wide-spread freeze stress of trees have recently been reported for the U.S.A. and Europe (Ningre & Colin, 2007;Gu et al., 2008;Augspurger, 2009;Hufkens et al., 2012;Timlin et al., 2012;Kosiba et al., 2018;). An increase in global temperatures is anticipated to promote earlier plant development, although an increased risk of spring freeze injury has also been predicted even if the frequency and severity of late spring frosts remains the same (Hänninen, 1991;Arora, 2018;. Wild trees growing within their natural distribution ranges rarely experience severe frost injury , although trees growing in ornamental nurseries and tree fruit orchards will likely be vulnerable to attack. Trees being grown outside their normal distribution or the loss of freeze tolerance traits resulting from selective breeding programmes would contribute to the risk of injury and subsequent ambrosia beetle attacks. For example, a mild winter that occurred in northeastern Ohio in 2011-2012 resulted in trees budding up to 1 month earlier than normal (Timlin et al., 2012;NOAA, 2012). Vulnerable root and vegetative tissues were then exposed to six below freezing events that occurred in the region in April 2012. Ambrosia beetle attacks on trees growing in ornamental nurseries were subsequently documented in May 2012 after the spring flight of overwintering adults had commenced (Ranger et al., 2015a). Thin-barked deciduous trees with known intolerance of freeze stress were attacked, namely Acer palmatum Thunb., C. canadensis, Cladrastis kentukea (Dum.-Cours.) Rudd, S. japonicus and Zelkova serrata (Thunb.) Makino (Cappiello & Littlefield, 1994;Boerner et al., 1988;Brand, 2002;Reed, 2005;Ranger et al., 2015a). In support of these field observations, the present study demonstrated that experimentally-stressed C. canadensis, C. florida, M. pumila and S. japonicus were attacked by non-native ambrosia beetles. La Spina et al. (2013) previously reported that experimentally freeze-stressed tissues of F. sylvatica also attracted and induced attacks by X. germanus and other ambrosia beetles.
In summary, in the present study, we have demonstrated that freeze stress predisposes thin-barked deciduous trees to attack by opportunistic ambrosia beetles, especially X. crassiusculus and X. germanus. Both of these species are strongly attracted to trees emitting stress-induced ethanol, and freeze stress of S. japonicus was demonstrated to induce the emission of ethanol. The results of the present study indicate that trees intolerant of freeze stress growing in ornamental nurseries and tree fruit orchards could be at risk of attack by ambrosia beetles, especially when freeze stress is preceded by a mild winter. Additional studies are therefore warranted to characterize the freeze stress intolerance of trees commonly planted in ornamental nurseries and tree fruit orchards that are attacked by ambrosia beetles. The incidence of freeze stress and ambrosia beetle attacks on trees within nurseries and orchards should also be characterized and modelled to assist in developing management strategies.