Infestation of the gall midge Dasineura oleae provides first evidence of induced plant volatiles in olive leaves

Abstract In this study, we present the first characterization of herbivore-induced plant volatiles (HIPVs) released from infested olive leaves. The gall midge Dasineura oleae is a specific pest of Olea europaea and endemic of the Mediterranean Basin, an area in which severe outbreaks currently occurred. Little is known about the damage caused by the pest and the relationship with its host. Since gall formation and larval feeding activity may lead to the release of specific plant volatile compounds, we investigated the volatile profiles emitted from infested plants compared with healthy plants under both laboratory and field conditions. Additionally, the volatiles emitted from mechanically damaged plants were considered. A blend of 12 volatiles was emitted from olive trees infested by D. oleae. Of these, β-copaene, β-ocimene, cosmene, unknown 1 and unknown 3 were found to be exclusively emitted in infested plants. The emission of germacrene-D, (E,E)-α-farnesene, and (Z,E)-α-farnesene, α-copaene, (E)-4,8-dimethylnona-1,3,7-triene, (E)-β-guaiene and heptadecane significantly increased in infested trees. Linalool, β-copaen-4-α-ol, β-bourbonene, β-cubebene, β-elemene, β-copaene and δ-amorphene were found only in the field trial and showed differences depending on the level of infestation and the plant stage. (Z)-3-Hexenol, (E)-4-oxohen-2-enal, and 2-(2-butoxyethoxy)-ethanol, were exclusively emitted from the leaves after mechanical damage. In a field trial in Italy, we also demonstrated spring synchronization between adults of D. oleae and O. europaea trees. Analyses of morphoanatomical malformations of gall leaves showed that tissue alterations occur at the spongy parenchyma causing an increase of the leaf blade thickness. We speculate that tissue alterations may lead to HIPV release, in turn potentially attracting D. oleae natural enemies.


Introduction
The olive leaf gall midge, Dasineura oleae (Angelini, 1831) (Diptera: Cecidomyiidae), is an endemic monophagous species of Mediterranean countries hosted by olive trees (Olea europaea L.) (Dogănlar et al., 2011). In recent years, several outbreaks have been reported in the Mediterranean (Simoglou et al., 2012;Caselli et al., 2021;Picchi et al., 2021). In particular, the Tuscany region (Italy) has registered a dramatic increase in infestations throughout its area, with large outbreaks recorded in Grosseto Province (i.e., Gavorrano and Capalbio) and Massa-Carrara Province since 2013 (Tondini and Petacchi, 2019). During 2020, the outbreaks extended farther north, reaching the olive areas around Sarzana on the border between Liguria and Tuscany.
As with other cecidomyiids, D. oleae is characterized by a very short adult life span of approximately 12-72 h under laboratory conditions (unpublished data). Adults therefore need to find young olive leaves on which to lay their eggs in a relatively short time. The capacity of the female midge to find a suitable host in the appropriate phenological stage for larval development is fundamental to guarantee the best chances of survival of the progeny (Hall et al., 2012). In gall-forming insects, gall formation is generally caused by the trophic activity of the larva after hatching (Rohfritsch and Shorthouse, 1982). When galls occur on leaves, the photosynthetic capacity of the plant is reduced, and the quantity and quality of the yield could be compromised (Martinez et al., 1992;DeClerck-Floate and Price, 1994;Gonzales et al., 2005;Huang et al., 2014). Studies on the impact of D. oleae galls on olive leaf physiology have demonstrated that both net photosynthesis and stomatal conductance are negatively influenced by midge infestation (Caselli et al., 2021).
In this study, we characterized the headspace volatile compounds released by olive plants of cv Frantoio infested by D. oleae under both field and laboratory conditions, comparing these profiles to those of healthy and mechanically damaged plants. For the first time, we described induced volatiles released by infested O. europaea leaves. Collecting field data, we examined the synchronization between olive leaf gall midge attack and the phenology of O. europaea. Furthermore, we performed a morpho-anatomical analysis of the gall tissue to analyze leaf modifications due to D. oleae gall formation.

Laboratory experiments
One-year-old rooted cuttings of O. europaea cv Frantoio were purchased from a plant nursery supplier (SPO, Società Pesciatina Olivicola, Pescia, Pistoia, Italy) in March 2020. Eighteen uniform plants were grown in 1.4-liter plastic pots containing sphagnum peat:pumice (1:1, v:v) and trained on a single shoot. Plants were drip-irrigated every day. The rooted cuttings were singly placed in cylindrical PVC cages (diameter: 25 cm, length: 80 cm) equipped with transparent chiffon fabric (mesh size: 0.05 mm) to allow aeration. On 10 April 2020, nine rooted cuttings were infested by placing five shoots with D. oleae galls taken from an infested orchard (length: 8 nodes) into each cage. Before the application, each infested shoot was carefully inspected to avoid any insect pests other than D. oleae. Adults of D. oleae emerged 24-48 h after the infested shoots were placed into the cages, mating occurred, and females laid eggs on the cutting leaflets. One month after the infestation (May 2020), leaf galls were visible. The other nine rooted cuttings were maintained not infested and treated as a control group. Infested and control plants were kept separated in two different climatic chambers under controlled conditions (24 ± 1°C, 16:8 (L:D) photoperiod) until the end of the experiment to avoid priming effects.
In order to distinguish pest-specific volatiles due to an active damage from the emissions of damaged tissues alone, we also sampled the volatiles emitted by mechanical damage, another 12 O. europaea cv Frantoio rooted cuttings (1 year old) were purchased from the nursery and kept under the same environmental conditions of the other laboratory plants. After 2 weeks of adjustment phase, six plants received damage, while the other six plants were considered as a control group. The mechanical damage was performed by using a metal punch (mod. 785, Zenith s.r.l., Manerbio, Italy) to cut 0.55 cm 2 holes on the first ten fully expanded apical leaves (two holes per leaf) of the treatment group. The headspace volatiles were collected from mechanically damaged plants starting immediately after the damage.

Field experiments
Field investigations were done to characterize the volatiles released from D. oleae infested trees under field conditions. Three different olive orchards located in Castelnuovo Magra (La Spezia, Italy), an area in which a strong outbreak of D. oleae has been recorded, were chosen for these trials. For each orchard, six olive trees of cv Frantoio ranging from 30 to 50 years old were selected. Headspace volatiles were collected from one shoot per tree. The shoots analyzed were ca. 30 cm long with no fruits and were chosen at mid-canopy height by uniformity of leaf number and size through visual observation. Since the whole area surrounding the fields that were sampled registered outbreaks of D. oleae, it was not possible to find a field of trees that was not infested as a control group. Therefore, sampling fields were classified as highly infested (field H) when the rate of infested leaves was over 20% and gall density was over two galls per leaf or lightly infested (fields L1 and L2) when the infestation rate was below 20% and the gall density was below two galls per leaf, according to Tondini and Petacchi (2019).

VOC collection
VOCs were collected using closed loop stripping analysis (CLSA) for 3 h (Kunert et al., 2009). A distal shoot portion of approximately 30 cm was selected and enclosed within a plastic bag (BVOC-bag, Cuki® oven bag, Cofresco, Volpiano, Italy). Six empty plastic bags were sampled as negative controls. Air samples were collected using an adsorbent trap (glass tube, 6.5 × 0.55 × 0.26 cm 3 , loaded with 1.5 mg activated charcoal; CLSA filter LR-type; Brechbühler AG, Schlieren, Switzerland). The trap was fitted to a 12-V graphite vacuum pump (Fürgut, Tannheim,
Germany) using a short Teflon tube. The pump circulated air at a rate of ca. 1 liter min −1 within the VOC-bag. After collection, the volatiles were eluted from the CLSA filters with 100 μl of dichloromethane. The volatiles were analyzed by injecting 2 μl of each sample into a gas chromatography-mass spectrometer (GC/MS) (7890, Agilent Technologies, Santa Clara, USA) equipped with a mass selective detector-MS (5975C, Agilent Technologies). A GC/MS nonpolar HP-5 MS column (Agilent Technology, 30 m × 0.25 mm × 0.25 μm film thickness) was used for separation with a constant helium flow of 1.2 ml min −1 and at an average velocity of 39.723 cm s −1 . The oven temperature program was started at 50°C for 1.8 min and heated to 250°C at a rate of 7.3°C min −1 . The total run time was 34.19 min. The mass spectrometric detector was operated in scan mode (m/z 35-400 amu). Volatiles were initially identified by comparing their mass spectra with those found in two libraries: NIST 14 (National Institute of Standards and Technology, 2008) and Wiley 7 (John Wiley, NY, USA); a mixture of n-alkane standards (nC8-nC20, Sigma-Aldrich, St. Louis, Missouri, USA) was used to calculate linear retention indexes (LRIs) of the detected compounds (Van den Dool and Kratz, 1963). The identification of VOCs was based on the comparison of their retention index r with the retention indexes reported in the literature (PubChem, Nist, and Pherobase) (Babushok et al., 2011). The compounds were finally confirmed by comparing their retention times with those available from laboratory standards (Sigma-Aldrich, St. Louis, Missouri, USA). Compounds that were also detected in the negative control (VOC-bag) were considered contaminants. The amount of volatiles was reported as the 10 −4 total ion current (TIC) mean ± standard deviation (SD).

Number of leaves, galls, and measurement of leaf area per sampled shoot
At the end of the collection of volatiles for both field and laboratory experiments, each shoot enclosed within the plastic bags was cut and defoliated to count the number of leaves and galls through visual observation and to measure the area of all the sampled leaves. For the area measurements, an LI-3000C Portable Leaf Area Meter (LI-COR, Lincoln, Nebraska, USA) was used.
Monitoring of the D. oleae flight pattern and phenology of O. europaea To record the flight curve distribution of D. oleae, a field campaign was planned in an unirrigated olive orchard of ca. 1 ha, located at 60 m above sea level (a.s.l.) in Gavorrano (Grosseto, Italy) (42°54 ′ 28.30 ′′ N, 11°00 ′ 10.65 ′′ E), where pest infestations have been recorded in the last few years and in which weekly monitoring for the identification of the D. oleae developmental stage is ongoing. Standard transparent delta traps (Csalomon®, Budapest, Hungary) with a horizontal sticky base were used for these trials. Traps were nonattractive and odorless to avoid bias due to color and odor preference (Ranamukhaarachchi and Wickramarachchi, 2007). Twenty-five traps, placed 20 m apart from one another, were hung on trees at a height of 0.5 m above the ground, and D. oleae males and females captured per trap were recorded weekly during the entire month of April 2020. O. europaea phenology has been determined by matching the growing degree days (GDDs) of olive trees during the sampling period and phenological observations in the field, with the BBCH (Biologische Bundesanstalt, Bundessortenamt und Chemische Industrie) scale as the reference point for O. europaea (Cortés et al., 2002). The GDDs were calculated using the formula by Orlandi et al. (2005) considering a threshold temperature of 10°C for O. europaea (Marchi et al., 2012). The Hydrological Service of Tuscany Region website was used to obtain daily maximum and minimum temperature data (Regione Toscana, 2020). The Braccagni (Grosseto, Italy) meteorological station was chosen because of its proximity to the sampling field.

Characterization of D. oleae attack by gall morphoanatomical observations
For the anatomical observations of the gall structures at the time of the experiment, ten leaves for both the control and infested treatments were chosen. A portion of 0.5 cm 2 was cut from the central region of healthy leaves and from the gall of infested leaves and fixed in an FAA solution (45% ethyl alcohol, 5% glacial acetic acid, 10% formaldehyde 8:1:1, v/v/v). Later, leaf samples, once rinsed in water, were dehydrated in a graded ethanol series and embedded in Histoplast. The tissues were cross-sectioned (10 μm) with a Shandon microtome (Shandon Inc., Pittsburgh, Pennsylvania, USA), mounted on glass slides and stained with 0.01% toluidine blue (Sigma-Aldrich, St. Louis, Missouri, USA) for 12 min. Observations were performed under an optical microscope (Fluophot, Nikon, Shinjuku, Japan). Representative selected sections were photographed with a Leica EC3 digital camera equipped with a microscope.

Statistical analysis
The data were analyzed using R software (R Core Team, 2020). The leaf surface and the number of galls between groups in the laboratory and field trials were analyzed by analysis of variance (ANOVA). In the peak analysis, contaminants found in the solvent alone were excluded from the analysis. Peak area integrations (TIC values) were compared between treatments by ANOVA. The leaf surface and the number of galls were accounted for as controlling variables. Data were log-normalized to fit a normal distribution. In case of a statistically significant treatment or field effect, a Tukey post-hoc test was performed. Although the potted plants had no (control) or some (infested) galls, the field plants had different degrees of infestation. Thus, it was not possible to consider a control field, and the amount of galls was considered a controlling variable. In the laboratory trial, since no gall was present on the control plants, only the number of leaves and the leaf surfaces were controlled for. Principal component analysis (PCA) of the laboratory data was performed using the R package 'factoextra' (Kassambara and Mundt, 2020). The PCA calculated the combination of the VOC area data by extracting eigenvalues and eigenvectors of a correlation matrix and then highlighting principal components. A two-dimensional score plot was created to compare the volatile profiles from control and infested plants. All the data are reported as the mean ± SD. The figures were created by using the R packages 'ggplot2' and 'cowplot' (Wickham, 2016;Wilke, 2020).
PCA allowed us to determine the clustering of the volatile profiles collected from infested and healthy-potted saplings. The score plot ( fig. 1) reports the PCA results and shows the difference between infestation conditions. Although healthy plants are clustered together, infestation by D. oleae caused a remarkable deviation along the principal components. Since PC1 explains almost half of the total variability (45.4%) the power left to PC2 (12.9%) is only a few points away from PC3 (9.89%) and PC4 (7.12%), resulting in dimension 2 contributing a relatively low explanatory power.

VOC characterization of infested field plants
A total of 37 compounds were collected from the field trees. Among them, caryophyllene and germacrene D were the most abundant, without differences between the olive orchards and the relative infestation level, while β-copaene and unknown 1 were remarkably more abundant in the more highly infested field (field H) (table 1). Some VOCs such as linalool, β-bourbonene, β-cubebene, β-elemene, aromadendrene, α-muurolene, β-copaen-4 α-ol, α-bisabolol, and unknown 4 (table 1) were found only in the field trials, regardless of the infestation level. Many VOCs were present only in highly infested plants: linalool, DMNT, cosmene, α-copaene, (E,E)-α-farnesene, unknown 1, and heptadecane, thus confirming what was observed under laboratory conditions. Additionally, (Z )-3-hexenyl acetate, higher in the control-potted plants, was more abundant in the low infested fields (fields L1 and L2). Benzaldehyde, which was not different between the potted plants, was instead not detected in field H. However, some compounds such as methyl salicylate showed a reversed pattern that increased in the highly infested field. The compounds β-bourbonene, β-cubebene, β-elemene, β-copaene, (Z,E)-α-farnesene, and δ-amorphene differed between the fields regardless of the infestation level (table 1).

Number of leaves, galls, and measurement of leaf area
Laboratory conditions Leaf areas ranged between 179.59 ± 52.89 and 198.09 ± 58.13 cm 2 per plant in the control and infested groups, respectively. Statistical analysis showed no significant differences between the two groups (F 1,16 = 0.499, P = 0.4363) ( fig. 2a). The number of leaves per plant was not significantly different (F 1,16 = 0.674, P = 0.3296) between control (60.78 ± 10.83) and infested (71.56 ± 24.85) plants ( fig. 2a). The number of galls per shoot in the infested plants was 134.89 ± 111.82, while the number of galls per leaf was 1.73 ± 1.06. In the cuttings used for the volatiles released by mechanical damage, the amount of leaves per shoot was 50.60 ± 16.77 and 50.33 ± 12.81 (F 1,10 = 0.582, P = 0.658) for control and damaged plants, respectively, while the leaf surface was 156.24 ± 52.89 and 163.16 ± 45.83 cm 2 (F 1,10 = 0.637, P = 0.547), thus showing no significant difference between the groups.

Flight curve of D. oleae and O. europaea phenology
The flight curves of D. oleae males and females showed the same trend, with a greater presence of males than females during the whole period of sampling ( fig. 3). On 8 April, a peak of adult flight activity was evident for both sexes, as the traps recorded a total of 2066 males and 1004 females. On this date, several swarms of D. oleae were noticed around young olive tree shoots and females during oviposition. The minimum presence of D. oleae was observed on 22 April, the last date of sampling, with a total catch of 512 males and 190 females.
Through visual observation and referring to the BBCH scale for O. europaea, phenological stage number 9, corresponding to external leaflets opening with their extremity intercrossing, was reached on 8 April, at 55.9 GDDs. On 15 April, at 75.7 GDDs, phenological stage number 33 was identified ( fig. 3).  (1), NIST WebBook (2), or PubChem (3). The review of Babushok et al. (2011) has also been considered (4). When possible, the LRI was verified by a standard compound ( †). Sampling fields were classified as highly infested (H) or lightly infested (L1 and L2). The volatile compounds were collected by the CLSA technique for 3 h. The compounds were identified by mass spectrometry and confirmed by LRIs available from the literature or laboratory standards when available. The table also reports the P-value of the ANOVA test of the TIC amount. The number of samples where the compound was found is reported between brackets. Compounds in bold are present only in infested plants, while the gray background identifies the compounds varying between infestation conditions. Asterisks indicate statistical significance (*P < 0.05, **P < 0.01, ***P < 0.001).

Morphoanatomical observations
The healthy leaves of O. europaea cv Frantoio have a typical elongate-elliptical shape, with a smooth dark green adaxial surface and a lighter and less smooth abaxial surface ( fig. 4a). A longitudinal healthy leaf section is presented in fig. 4b, in which palisade parenchyma I and II (PPI and PPII), spongy parenchyma (SP), and peltate trichomes (PT) are clearly recognizable. D. oleae galls are roundish, a few millimeters long (3-6 mm), and they are irregularly present just on the adaxial leaf surface (fig. 4a). The section in fig.  4c shows a belt of cell proliferation just near the larval chamber at the SP. The larval chamber is clearly recognized in fig. 4d. The thickness of this leaf section (2 mm) is threefold greater than the thickness of the healthy leaf (0.35 mm) ( fig. 4b).

Discussion
Studies on olive leaf volatile profiles are currently scarce and based on hydrodistillate fractions, which are different from headspace emissions (Campeol et al., 2001;Flamini et al., 2003;Brahmi et al., 2012;Malheiro et al., 2016). In the present study, headspace collections of olive volatiles were recorded under both laboratory and field conditions. Control-potted plants emitted a blend of 26 VOCs, including the aldehydes nonanal and benzaldehyde, the esters methyl salicylate, 3-hexenyl butyrate and hexyl acetate, the ketone sulcatone, and the sesquiterpenes caryophyllene, (E, E)-α-farnesene and germacrene D, previously identified by Malheiro et al. (2015Malheiro et al. ( , 2016. Other VOCs are reported here for the first time, such as myrcene, (Z)-3-hexenyl acetate, 2-ethylhexanol, DMNT, carveol, α-copaene, tetradecane, humulene, bicyclosesquiphellandrene, (E)-β-guaiene, (Z,E)-α-farnesene, δ-amorphene, liguloxide, β-oplopenone, heptadecane, γ-cadinene, and unknown 2. This study, for the first time, provides evidence about HIPVs for olive trees due to insect attack to the leaves. Previous studies have only investigated fruit HIPV profile due to B. oleae larvae infestation (Alagna et al., 2016) and the behavioral responses of its parasitoid Psyttalia concolor to olive drupe HIPVs (Giunti et al., 2016a). Moreover, although gallers usually manage to suppress VOCs (Borges, 2018), D. oleae induces a change of the volatile profiles of attacked plants in the amount of emissions and in the released compounds. A blend of 12 VOCs is distinctive of olive leaves infested by D. oleae: DMNT, β-ocimene, α-copaene, β-copaene, cosmene, germacrene D, (E)-β-guaiene, (Z, E)-α-farnesene, (E,E)-α-farnesene, unknown 1, unknown 3, and heptadecane. Some were found to be specific only to infested plants in laboratory trials (β-ocimene, cosmene, β-copaene, unknown 3, and unknown 4), while others showed increased emission in infested plants both in the laboratory and in field collection (DMNT, α-copaene, (E)-β-guaiene, (Z,E)-α-farnesene, (E, E)-α-farnesene, liguloxide, and heptadecane). This herbivoreinduced emission observed in the study is in agreement with other studies that reported some of the same compounds in infested plants of other species (Suckling et al., 2012). DMNT and α-copaene were released in Medicago truncatula Gaertn.  Volatiles were collected from six potted plants after 20 holes of 3 mm were made on ten leaves. Volatile emissions were compared with those of control plants collected on the same day. The volatiles were collected by the CLSA technique for 3 h. a LRI = Linear retention index calculated in relation to n-alkanes. b LRI = Linear retention index already published in peer-reviewed journals and listed in PubChem. When possible, the LRI was verified by a standard compound ( †).
following the feeding activity of Spodoptera exigua (Hübner) (Arimura et al., 2008). In the study of Zeng et al. (2017), α-farnesene and ocimene were released after tea leaf (Camellia sinensis (L.) Kuntze) stimulation by jasmonic acid combined with mechanical damage, simulating aphid damage. (E, E)-α-Farnesene and β-ocimene were also emitted by cotton plants treated with methyl jasmonate (Rodriguez-Saona et al., 2001). The sesquiterpene (E,E)-α-farnesene and the monoterpene β-ocimene are already known as constituents of olive oils, and when infested by B. oleae, olive fruits increase the emission of these two compounds (Giunti et al., 2016b). In the present study, (E, E)-α-farnesene and β-ocimene were the most abundant molecules in the infested potted plants, thus confirming a key role of these molecules in the olive tree-insect interactions. (E)-β-Ocimene is reported as an HIPV from B. oleae exploited by virgin males of the braconid parasitoid P. concolor to boost their mate searching activity (Giunti et al., 2018). The monoterpene cosmene, identified in the present study, was also found as an HIPV in apple leaves infested by Pandemis heparana Denis & Schiffermüller (Giacomuzzi et al., 2016). Most of the other compounds are in agreement with other studies, as they were identified as VOCs of olive trees (Flamini et al., 2003;Baratella, 2011). The main VOCs collected from the field trials were caryophyllene and germacrene D, regardless of the infestation level between fields. Although caryophyllene is one of the main constituents of the olive leaf profile (Flamini et al., 2003), germacrene D might instead act as an HIPV since its emission increased in response to midge infestation in the laboratory collection. Moreover, statistical analysis revealed that germacrene D is also affected by the leaf surface of the sampled shoots both in laboratory and field trials. Two other VOCs, (Z,E)-α-farnesene and heptadecane, are commonly emitted in small amounts by apple leaves and olive leaves (Bengtsson et al., 2001;Dursun et al., 2017). In the present study, they were directly related to the leaf surface of the shoots sampled in the laboratory trials, and their emission increased in infested plants.
A substantial number of VOCs are usually emitted due to oxidation processes after plant tissue injury. Considering the volatile emission due to mechanical damage allowed us to notice the field H (n = 6) is highly infested, while fields L1 (n = 6) and L2 (n = 5) are infested at a lower level. Letters report statistical significance.
difference between mechanical wounding and insect activity, thus identifying the set of specific herbivore-induced volatiles. The analysis revealed that (Z )-3-hexenyl acetate was the main released compound after mechanical wounding. Indeed, this ester is known to be released from damaged leaves (Oluwafemi et al., 2011), usually together with its original alcohol (Z)-3-hexen-  1-ol (Kikuta et al., 2011;Sufang et al., 2013), which was also found in our collected VOCs. (Z )-3-Hexen-1-ol is found in a typical green leaf formed by lipoxygenase of linolenic acid of the cell membranes. Both 3-hexen-1-ol and (Z)-3-hexenyl acetate were indeed reported as the main VOCs collected by severed olive leaves in Malheiro et al. (2016) confirming their emission after cell membrane breakup. Among the 12 compounds previously reported as HIPVs by D. oleae, (E,E)-α-farnesene was also released following mechanical damage in our study, but the 20-fold rise of emission in the infested plants and the consistency of the data under both laboratory and field conditions confirmed the role of this sesquiterpene as HIPV (Giacomuzzi et al., 2016).
The laboratory trials allowed for the use of uninfested plants, thus revealing four molecules (β-ocimene, cosmene, β-copaene, and unknown 1) released only in the plants that were attacked and not detected in the control plants. However, although cosmene, β-copaene, and unknown 1 were also consistently higher in the plants of the highly infested field, β-ocimene was higher in the highly infested field and in one less-infested field. Some compounds were reduced in the infested plants in the laboratory trials. γ-Cadinene, although completely absent in infested potted plants, was present in the mature field plants but without a difference between infestation levels. (Z )-3-Hexenyl acetate and 2-ethylhexanol, which are more abundant in control-potted plant collections, were always present in field trials. However, the lack of uninfested field trees prevented comparison with the volatile profiles of controlled conditions and understanding whether the emission of these VOCs was actually reduced due to infestation. For this reason, the number of galls in the sampled branch was used to predict the amount of volatiles. Linalool, β-copaen-4 α-ol, and heptadecane were found to be dependent on the number of galls. Although linalool and heptadecane are known as common HIPVs (Moayeri et al., 2007;War et al., 2011), β-copaen-4 α-ol has so far been reported only as a sesquiterpene characteristic of many plant species (Dória et al., 2010;Magierowicz et al., 2020), but not as an herbivory-induced compound. Moreover, β-copaen-4 α-ol was detected only in the field collection and not in the laboratory collection, thus suggesting an emission more related to mature plants or other parameters. Linalool, which was also not observed in the laboratory trial, is characteristic of highly infested plants in the field, together with DMNT, cosmene, α-copaene, (E,E)-α-farnesene, unknown 1, and heptadecane, already observed in potted plants as HIPVs. The reason linalool appeared only in the field collections might be dependent on the age or the phenological stage of the plant, since the plants used in the laboratory trials were young saplings. For instance, in Rao et al. (2010) older leaves of palmarosa (Cymbopogon martinii (Roxburgh) Watson) contained more linalool than young leaves. Also in hop (Humulus lupulus L.), the linalool content increased with the age of the leaves (Matsui et al., 2012). Hence, β-copaen-4 α-ol and linalool might act as HIPVs only in mature plants. A possible effect due to plant age might also explain the difference in the compounds β-bourbonene, β-cubebene, β-elemene, β-copaene, and δ-amorphene between field L1 and fields L2 and H. These sesquiterpenoids were not present in the young potted saplings, thus appearing only in field plants, regardless of the infestation level. Sesquiterpenoids are often involved in communication between species since they can travel long distances via advective transport in gas (Huang and Osbourn, 2019). Other VOCs were detected only in the field collection, with no difference between fields (aromadendrene, α-muurolene, α-bisabolol, and unknown 4).
Although certain studies pointed out differences in volatile emissions between young and old leaves (Takabayashi et al., 1994;Rao et al., 2010;Matsui et al., 2012), only a few studies are available about these differences when comparing young (few-week-old saplings) and mature trees (30-year-old olive trees). In the study of Street et al. (1997), the emissions of Pinus pinea L. seemed to vary significantly from mature forest to young plantation, and age-related changes were found in the volatiles released by wounded phloem of Picea abies (L.) Karsten seedlings (Kännaste et al., 2013). Whether similar changes can occur in broad-leaf species such as O. europaea has not been reported, but it might provide an explanation for the emissions observed in the present study. A genotype effect cannot be excluded, but it seems to be unlikely, since olive trees are reproduced mainly by cutting, and the cultivar chosen for the study was the same for all the trials.
The relationship between gall-inducing species and host plants is a complex association in which insects redirect plant physiology and, consequentially, the growth of the attacked organs for the purpose of gaining advantages (e.g., nourishment and shelter) (Shorthouse et al., 2005). For this reason, the synchronization of gall-inducing insects with their host is fundamental, particularly for short-lived insects, such as cecidomyiids (Yukawa, 2000;Tondini and Petacchi, 2019). Furthermore, in a tritrophic interaction (host plant-pest-natural enemies) each component has its biological rhythm concerning the behavior and the physiology of all the members (Allemand et al., 1994;Fantinou et al., 1998;Turlings et al., 1998;Zhang et al., 2010). The efficiency of the tritrophic system is maintained by synchronizing the rhythms across the trophic levels that is influenced by biotic and abiotic factors (Zhang et al., 2010). For instance, the emission of HIPVs by plants to recruit natural enemies of herbivore is a typical phenomenon involved in a synchronized tritrophic system (Zhang et al., 2010). However, studies about rhythm synchronization in tritrophic interactions are currently scarce. This paper gives the first demonstration of the synchronization between D. oleae adults (both females and males) and the phenology of O. europaea. A greater number of D. oleae adults during the 2020 sampling campaign was observed on 8 April. This is in accordance with our previous data that recorded the highest amount of D. oleae exactly in the same period also during the previous years, although in those cases only females were recorded (Tondini and Petacchi, 2019). In the same flying period, the phenology of O. europaea reached the 9th stage of the BBCH scale, in which the external leaflets opened and their tips intercrossed (Cortés et al., 2002). The GDD on 8 April was 55.9. However, the GDD recorded in conjunction with the 9th stage of the BBCH scale usually ranged between 70 and -100 (Marchi et al., 2012). This phenomenon can be explained by the climatic conditions verified during the period between January and April 2020. The temperature during these months has never been too frigid, with a mean temperature of 9.9°C and a quite scarce rainfall, particularly in February (4.6 mm) (Regione Toscana, 2020). During March, rainfall was more abundant (57 mm) and concurrently with mild weather, abundant rainfall can be the cause of a vegetative restart of O. europaea, even if the GDD was not greater than 56 (Regione Toscana, 2020). During the third sampling date, on 15 April, researchers observed that the olive shoots reached 30% of the final length, classifying this phenological stage as the 33 rd stage of the BBCH scale (Cortés et al., 2002). During the whole period of sampling, D. oleae eggs were observed on olive leaflets, confirming the 490 Alice Caselli et al.
preference of this cecidomyiid for fresh leaves, unlike other gall midges colonizing hard leaves or other tissues, such as the yew gall midge, Taxomyia taxi (Inchbald) and the blueberry gall midge, Dasineura oxycoccana Johnson (Dernisky et al., 2005;Miller and Raman, 2019). The hypothesis that the D. oleae sex ratio presents a bias in favor of females in the proportion of 2:1 by Hallett and Heal (2001) is confuted in the current study, having observed exactly the opposite relationship. However, this phenomenon may result in a false estimation, since during direct sampling from the field several secondary factors are considered, such as the differential mortality of sexes due to food quality and/or overwintering diapause (Tabadkani et al., 2012).
Leaf galls induced by cecidomyiid trophic activity generally show alterations in cells and tissues, with a consequent modification of the leaf blade shape (Albert et al., 2013;de Alcantara Guimarães et al., 2013). A basic study on the leaf tissue modifications as a consequence of D. oleae feeding action has been recently reported by our group (Caselli et al., 2021). We observed that the first-stage larva forms an entrance hole on the adaxial olive leaf surface and then it develops in a gallery within the SP of the leaves. Spongy parenchymatic tissue is often modified after gall agent attack, similar to Clusia lanceolata Cambessèdes following Clusiamyia nitida Maia trophic action (de Alcantara Guimarães et al., 2013) and Copaifera langsdorffii Desfontaines following midrib gall development (Oliveira and Isaias, 2010). Additionally, the feeding activity of Dasineura mali on the upper apple leaves surface, causes characteristic distortions followed by a change in the foliage color (Lo et al., 2015). Near the larval chamber of D. oleae, the olive SP cells divide and become spherical with a small intercellular space, such as in Guarea macrophylla Vahl subsp. tuberculata Vellozo (Kraus et al., 1996). Here, it is demonstrated that the formation of the larval chamber by D. oleae induces the development of new cell layers in the SP, as described for other cecidomyiids by Albert et al. (2013) and de Alcantara Guimarães et al. (2013). We observed a spongy hyperplastic parenchyma, as described for Aspidosperma spruceanum Benth. ex Müll. Arg. by Formiga et al. (2011) and consequentially, the galled leaf portion is thicker than the unaffected leaf portion. Further studies may focus on the metabolic compounds formed in the parenchyma tissues during the D. oleae gall development. Furthermore, parenchymal alteration may have a role in inducing a plant defense response (Sousa et al., 2020). Among plant defense strategies, the emission of HIPVs plays an important role in the recruitment of natural enemies that may control the attack of herbivore insect pests (Gebreziher, 2020).

Conclusions
The analysis of the volatiles emitted by plants infested by galls of D. oleae provided a clear set of 12 insect-induced compounds, and the data were consistent between the laboratory and the field trials. Following the insect attack, the emission amount of some VOCs increased significantly. Other HIPVs were not present in control plants and appeared only after the infestation. The sampling technique did not cause any mechanical damage that might interfere with the plant volatile emissions. Some volatiles seemed to be related to the age of the plant rather than to the infestation status.
This study confirms a close relationship between D. oleae and its exclusive host, O. europaea, highlighting a spring synchronization of midge flight activity and the vegetative restart of the olive tree. As a consequence of the larval feeding action, the olive leaves show modifications of the leaf shape blade due to the structural alteration of the spongy parenchymatic tissue. Further studies are needed to investigate the relationship between leaf structure modifications and VOCs emitted by O. europaea, shedding light on the ecological role of volatiles. Forthcoming behavioral tests might investigate whether the change in the volatile profile of the plant elicits a response in adults of D. oleae and its natural enemies.