Combined effects of drought and insect herbivory on oak leaf endophytic communities and performance

Abstract

Leaf-associated microbiota are considered sensitive indicators of how plants respond to their environment. Yet, plants can be exposed to a wide range of biotic and abiotic stresses whose independent and combined effects on plant microbiota remain poorly understood, as do their cumulative consequences for plant performance. We conducted an experiment to expose saplings of four oak species to drought and herbivory by a piercing–sucking insect. We assessed the effect of water deficit on herbivory, as well as the individual and combined effects of water deficit and herbivory on fungal endophytic communities. To evaluate the cumulative impacts of biotic and abiotic stresses on plant performance, and the potential of the fungal endophytic communities as an indicator of plant responses to stress, we measured photosynthetic capacity, growth, and the accumulation of non-structural carbohydrates in twigs at the end of the growing season. Water deficit had no effect on insect herbivory but induced a decrease in the diversity of the leaf-associated fungal community. In contrast, herbivory shifted the composition of fungal communities through species replacement or addition depending on oak species. Herbivory altered photosynthesis, whereas drought but not herbivory reduced oak growth. None of these stresses influenced the storage of non-structural carbohydrates. Leaf-associated fungi responded to both stresses. However, the underlying mechanisms appeared to differ and the association with oak performance was weak. Overall, our study suggests a certain degree of resilience in trees subjected to interacting stresses, while also highlighting vulnerabilities that may have implications for tree health over the long term.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s00442-026-05995-2.

Introduction

Forests are essential ecosystems (Reichstein and Carvalhais 2019) whose functioning and the services they deliver are increasingly threatened by several stresses (Hartmann et al. 2022). Whether abiotic (heat waves, fires, storms, drought) or biotic (herbivores, pathogens), stresses have a well documented impact on tree growth and mortality, as well as on forest dynamics (Allen et al. 2010; Flower and Gonzalez-Meler 2015). However, much less is known on the interactive effects of these stresses on trees, and even less on the mechanisms shaping their integrated response to them (Teshome et al. 2020).

This gap in knowledge is especially important to fill when it comes to exotic insects that can become emerging pests in forest ecosystems (Liebhold et al. 1995). One such example is the oak lace bug (OLB), Corythucha arcuata (Hemiptera: Tingidae), a newly introduced piercing sucking insect that is rapidly spreading in European oak forests, with a preference for deciduous oaks (Csóka et al. 2020). This mesophyll feeder causes leaf discoloration by absorbing chloroplastic cell contents, thereby reducing the tree’s carbon assimilation (Nikolić et al. 2018). However, little is known about differences in susceptibility among deciduous oak species, as the limited available evidence shows inconsistent patterns between laboratory assays and field observations (Marković et al. 2021). This biological invasion adds to the pressure of increasingly severe and frequent droughts, which also disrupt tree physiology and carbon assimilation (Bréda et al. 2006). Drought induces stomatal closure, which progressively reduces CO₂ assimilation (Reddy et al. 2004). By reducing photosynthetic activity and altering xylem development, drought inhibits growth of major European deciduous oak species as Quercus cerris, Q. petraea, Q. pubescens, and Q. robur (Dickson and Tomlinson 1996; DELİGÖZ and BAYAR 2018). As resources are reallocated towards the production and storage of soluble sugars, free amino acids, and secondary metabolites in cell vacuoles to maintain cell turgor (Nawaz et al. 2025), prolonged drought may progressively deplete non-structural carbohydrate (NSC) reserves, as reported for Q. pubescens and Q. robur (Nardini et al. 2015; Di Iorio et al. 2024). Together, these physiological adjustments alter leaf quality for herbivores, leading to contrasting effects on tree susceptibility depending on herbivore feeding guilds (Gely et al. 2020; Carvajal Acosta et al. 2023). Previous research offers conflicting grounds for predicting drought’s effect on oak vulnerability to the OLB: native populations avoid drought-stressed foliage (Connor 1988), while mesophyll feeders respond inconsistently to drought in general (Huberty and Denno 2004). Concerns are nonetheless growing that the OLB may act as a latent bioaggressor (Karanikola and Tsikas 2025). The synergistic effects of OLB-induced discoloration and drought could strongly compromise tree carbon assimilation thereby depleting growth and impair the survival of young saplings.

Plant responses to combined biotic and abiotic stresses are modulated by fungal endophyte communities; therefore, characterizing changes in these communities is key to deciphering how trees cope with stress interactions (Stone et al. 2018; Trivedi et al. 2020). Consistent with the depletion of carbon assimilation and changes in leaf chemical environment (González-Klenner et al. 2022; Lajoie and Dariel 2025), drought has been shown to affect leaf microbiota, with changes in leaf water status modifying the dominance of particular taxa and potentially favouring pathogenic ones (Poudel et al. 2021; Trivedi et al. 2022; Cambon et al. 2023). In turn, fungal pathogens may themselves impair plant hydraulic functioning, for example by reducing stomatal conductance or obstructing water transport, thereby further increasing drought susceptibility (Hajji et al. 2009; Oliva et al. 2014). Plant microbiota is also partly shaped by insect herbivores through the introduction of microorganisms into plant tissues during feeding, including via insect saliva, with examples reported for insects from different feeding guilds (Lòpez-Fernàndez et al. 2017; Vacher et al. 2021). Previous studies have reported changes in fungal community composition following herbivory, including the promotion of phytopathogenic microorganisms (Lieutier et al. 2009; Humphrey and Whiteman 2020). These modifications add up to herbivory-induced defences that may additionally impose physiological stress or redirect resources towards defence, potentially reducing the plant’s ability to resist or tolerate subsequent pathogen infection (Smets and Koskella 2020; Pirttilä et al. 2023).

In this study, we used a factorial design to test whether drought and herbivory have independent, additive, or synergistic effects on foliar fungal communities, physiology, and growth of oak saplings, while addressing species-specific differences in vulnerability to independent and co-occurring stresses. We exposed saplings from four oak species to contrasting watering treatments, in the presence or absence of a piercing-sucking herbivore. We characterized leaf endophytic fungal communities and measured leaf stomatal conductance and photosynthesis. We further assessed tree performance through annual growth and the ability to produce and store non-structural carbohydrate at the end of the growing season. We first hypothesized (1) that oak species differ in their susceptibility to OLB herbivory, predicting that OLB-induced leaf discoloration would differ among species. We further hypothesized (2) that water deficit would affect oak susceptibility to the insect herbivore, with water-deficit saplings exhibiting different levels of OLB-induced leaf discoloration than well-watered ones. Finally, we hypothesized (3) that herbivory and water deficit would interact through overlapping internal leaf changes and constraints on carbon assimilation. Based on this hypothesis, we predicted (3a) additive or synergistic reductions in fungal diversity and alterations in community composition, potentially resulting in a higher relative richness of putative pathogens when both stressors occur simultaneously. We further predicted (3b) reduced photosynthetic performance, altered non-structural carbohydrate (NSC) content, and ultimately impaired tree growth.

Materials and methods

Study system

The experiment was conducted in a greenhouse (Cestas, France) for 7 months in 2023. 50 oak saplings from four species (Quercus robur, Q. petraea, Q. pubescens, and Q. cerris) were transplanted into 2 L pots filled with a soil–peat mixture (growing medium N.F.U 44551, based on blond peat, black peat and white sand), totaling 200 individuals. During the budburst period (March–April), the saplings were watered daily and randomly distributed in the greenhouse to minimize biases associated with light variation. After 3 months, saplings were distributed into 50 tetrads, each consisting of one sapling per species. Each tetrad was placed in a rearing cage (50 × 50 × 180 cm) that was randomly assigned to one of six experimental conditions, corresponding to the factorial combination of a two-levels irrigation treatment and a three-levels herbivory treatment.

Irrigation. In 25 cages, each pot was watered to saturation five times a week (high irrigation). Based on a preliminary experiment (details in supplementary S1), 25 cages received approximately 50% of the daily transpiration per sapling, five times a week (low irrigation). In order to maintain a constant level of water stress throughout the experiment and to adjust water supply according to environmental conditions, two control plants per species were weighed weekly before irrigation.

Herbivory pressure. We purposefully used herbivory both as a factor — in order to test the interactive effect of drought and herbivory on leaf microbiota, NSC and oak sapling performance — and as a response variable — to assess differences in the vulnerability of the four oak species to the OLB under different water supply levels. Specifically, a three-level herbivory treatment was applied two weeks after the onset of the irrigation treatment. It consisted in the introduction of none, 50, or 100 individuals of C. arcuata per cage. These infestation levels do not represent fixed categories of moderate or high natural infestation, as OLB densities vary widely throughout the season, but intended to create a discolouration gradient to facilitate the interpretation of synergistic or antagonistic effects of OLB and drought on oak performance. The insects were collected from adult oak trees in the city of Bordeaux.

Each treatment combination was replicated eight times, resulting in 48 saplings per oak species. In total, 192 saplings were randomly selected from the 200 available for the experiment. These two regimes were maintained during 3 months, after which all saplings were watered abundantly three times a week during 2 months. A HOBO data logger was placed in the greenhouse near the plants to continuously record temperature and humidity. The average monthly temperature and relative humidity were 24.6 °C and 49.8%, respectively.

Discolored leaf area caused by the OLB

OLB damages were monitored after two weeks and five weeks after the onset of the herbivory by estimating the percentage of discolored leaf area. We photographed (high quality smartphone) 10 haphazardly selected leaves from each plant against a black background. The images were analyzed using Mesurim (version 3.4 software) to quantify the proportion of yellow pixels relative to the total leaf area. Yellow pixels were selected manually and then Mesurim extended the selection to the entire image. Discolored areas were averaged at the individual level. Because OLB causes characteristic leaf discolouration that is clearly visible to the naked eye and can hardly be confused with other symptoms and that we visually verified that there were no signs of OLB presence in the control cages; these measurements were therefore performed only in cages where OLB had been introduced (50 and 100 treatments, n = 128).

Tree diameter growth assessment

Tree growth was assessed by comparing the mean stem diameter measured before budburst (March 2023) and again at the end of the growing season (October 2023). At each date, two diameter measurements were taken with a caliper at a mark drawn 2 cm above the substrate level, along two perpendicular directions. The two orthogonal measurements were averaged, and the resulting values were used directly in the statistical analyses without further transformation.

Photosynthesis

Within five days after discoloration quantification, operational quantum yield of photosystem II (PhiPS2) and stomatal conductance were measured using a Li-600 porometer-fluorometer on five randomly selected leaves per individual, varying in orientation and height. Three measurements were taken at different points on the same leaf with a total of 15 measurements per plant, and always at the same position on the leaf, avoiding veins and OLB. Measurement series recorded during periods when the instrument was deemed non-operational—likely due to contamination by honeydew or other external factors—were excluded from the analysis (4,733 measurements, corresponding to 83% of the initial dataset). Additionally, values outside the normal mathematical range (negative or above 1 values) for phiPS2 and stomatal conductance were also removed, representing 0.08% and 3% of the dataset, respectively.

Stem carbohydrates

Five cm long segments of the apical shoot of each sapling were collected at the end of the experiment, dried (48 h at 45 °C) and ground. The concentrations of starch and non-structural carbohydrates (glucose, sucrose, fructose) were done using the anthrone method: concentrations of starch and non-structural carbohydrates were determined by spectrometry (Biochrom Libra S22 spectrophotometer) against a sugar standard range and values were converted in mg·g⁻¹ of dry matter, yielding two concentration values per oak.

DNA metabarcoding

To characterize fungal communities, three leaves per oak were sampled at the end of the experiment and surface-sterilized following Compant et al. (2021). Sterilized leaf pieces were ground with a genogrinder (SPEX SAMPLEPREP 2010 Geno/Grinder) with liquid nitrogen. Total DNA was extracted using a CTAB adapted extraction method (Carrell and Frank 2014). Around 60 mg of leaves were incubated (45 min at 65 °C) with 900 µL CTAB solution (2% cetyl trimethylammonium bromide, 1% w/v polyvinylpyrrolidone, 100 mM Tris-HCl, 1.4 M NaCl, 20 mM EDTA). Proteins were removed with a two-step addition of 900 µl of chloroform-isoamyl alcohol (24/1 vol), centrifuging (15 min at 20,000 g at 4 °C) and supernatants were kept. Nucleic acids were precipitated by adding 100 µL of cold NaCl (5 M) and 1 ml of cold isopropanol, and frozen at -20 °C (12 h). After centrifugation (20 min at 20,000 g at 4 °C), the pellets were washed by a two-step addition of 1 ml of cold 70% ethanol and 500 µl of cold 95% ethanol. The air-dried pellets were resuspended in 50 µL of sterile H20. DNA was quantified by fluorescence with Qubit (Thermo Fisher Scientific Qubit 1X dsDNA HS Q33231). The ITS2 region of the RNA gene was amplified with ITS86F (Turenne et al. 1999) forward and the ITS4 (White et al. 1990) reverse primers, as recommended by Beeck et al. (2014). A first Polymerase Chain Reaction (PCR) amplification was done in 20 µL (2.5 µL of DNA (5 ng/µL), 2.5 µL of each primer (3 µM), 4 ul of Taq polymerase (5X HOT Firepol with 1.5mM MgCl2) and 9.5 µL sterile H2O) as follows: 15 min at 95 °C, 30 cycles of 30 s at 95 °C, 30 s at 60 °C, and 30 s at 72 °C. The amplicons were subsequently purified using magnetic beads (remove primers and primer dimers) on a Hamilton Star robotic platform following the manufacturer’s instructions. A second PCR was performed to attach dual indices and Illumina sequencing adapters. The PCR reaction mix (10 µL) included 2 µL of purified amplicons, 2 µL of Illumina RTU Index Primer1 and Primer2 (i7 2,5 µM and i5 2,5 µM), 2 µL of Taq polymerase (5X HOT FIREpol, 1.5mM MgCl2) and 4 µL H2O. The thermocycling conditions were as follows: 15 min at 95 °C, 12 cycles of 30 s at 94 °C, 30 s at 58 °C, 30 s at 72 °C, and a final extension at 72 °C for 10 min. Amplicons were quantified and pooled in equimolar concentrations before sequencing with the NextSeq 2000 (2 × 150 bp) chemistry (Illumina, San Diego, CA, USA). Demultiplexed, adapter‐trimmed fastq files were generated using the Illumina Reporter (Illumina, San Diego, CA, USA). To control for potential contaminants and false-positive sequences caused by tag-switching events (Esling et al. 2015), the molecular experimental design comprised extractions controls, PCR negative and positive (mock community of fungal species, Fig. S1) that were also sequenced, as well as unused tag combinations. All steps were performed by the Genome Transcriptome Facility of Bordeaux (France, https://doi.org/10.15454/1.5572396583599417E12).

Bioinformatic

In total, 318 millions of reads were produced. Bioinformatic analysis was performed using the FROGS pipeline (version 4.1, Escudié et al. 2018). The pipeline first assembled raw forward and reverse reads for each sample into paired-end reads with a minimum overlap of 10 nucleotides and a maximum mismatch of 0.1 using the VSEARCH algorithm (Rognes et al. 2016). The primers were removed using Cutadapt (Martin 2011). Ambiguous nucleotides whose size did not match the expected amplicon size (between 100 and 590 bp for the rDNA ITS gene) were filtered out. The sequences were then dereplicated and clustered using SWARM (Mahé et al. 2014) with an aggregation distance of 1 and the fastidious option activated. Chimeras were removed using VSEARCH. A filtering step was applied to retain only clusters representing at least 0.001% of the reads in the dataset. The highly variable part of the ITS2 region was extracted using ITSx (Bengtsson-Palme et al. 2013). Taxonomic assignment was performed on the representative OTU sequence using the UNITE Fungi 8.3 database (Abarenkov et al. 2010) using the RDP Classifier (Wang et al. 2007), which is reported in the literature to be faster while achieving comparable or even better accuracy than BLASTn-based approaches (Altschul et al. 1990; Liu et al. 2012; Nilsson et al. 2019). Finally, we cured the dataset using the metabaR R package v1.0.0 (Zinger et al. 2021). Non-fungal OTUs, contaminants, tag-jumps, low-quality sequences (< 96% sequence identity), and samples with insufficient sequencing depth (threshold of 10,000 reads per PCR) were removed. In addition, two OTUs with extremely low abundance (below a relative abundance threshold of 0.001%) were excluded. The final dataset comprised 447 fungal OTUs across 180 samples for a total of 266 million reads. Finally, we used the FungalTraits database (Põlme et al. 2020) to assign OTUs to the pathotrophic guild when their corresponding genus was classified as pathotrophic in the primary lifestyle category or exhibited any plant pathogenic capacity in the database. This resulted in 161 pathotrophic OTUs, among 375 OTUs classified at least at the genus level, distributed across 180 samples. The sequencing data have been deposited in the European Nucleotide Archive under accession PRJEB110006.

Statistical analyses

We first analysed whether OLB-induced leaf discoloration increased with herbivory pressure under contrasting irrigation conditions using a Generalized Linear Mixed-effects Model (GLMM) with a Gaussian error distribution. Fixed effects included species identity, irrigation and herbivory pressure (two-level factors: 50 and 100 OLB individuals). Because OLB-induced leaf discoloration cannot occur in the absence of herbivory, insect-free control plants were omitted in this analysis, thus totalling n = 119 oaks. In all subsequent analyses, insect-free control plants were retained and their discoloration values were set to 0, reflecting the absence of insect-induced damage. The cage was included as a random effect to account for potential non-independence of plants within the same cage. We further evaluated the sensitivity of this model by estimating the minimum detectable effect size for the main fixed effects using a using the pwr.f2.test function from the R package pwr (Champely et al. 2020), assuming a significance level of α = 0.05 and a statistical power of 0.80. The analysis was based on the residual variance estimated from the GLMM and the corresponding degrees of freedom of each tested effect.

OLB damage increased over the course of the season. Discoloration merely concerned 3% of leaf area on average at the time of the first assessment but reached 47% three weeks later. We considered that it would have made no sense to test the effect of OLB on oak photosynthesis at the beginning because of the very low level of discoloration. We ran separate models for the first and second assessments. The data from the first assessment were used to confirm that the irrigation treatment influenced photosynthesis. The data from the second assessment were used to test whether the irrigation treatment influenced the response of oak photosynthesis to OLB damage.

Finally, we tested the effects of irrigation and OLB damage on oak sapling growth, photosynthetic performance, carbohydrate concentrations in twigs and leaf endophytic fungal communities using GLMMs as described above, with one adaptation: herbivory pressure was replaced by the percentage of leaf area discolored by OLB feeding measured during the second assessment. We used GLMMs with beta and gamma error distributions to analyse PhiPS2 and stomatal conductance, respectively (mean per plant calculated from the second assessment, n = 131). We used GLMMs with Gaussian error distributions to analyse growth (n = 182) and carbohydrate concentrations, with separate models for starch and non-structural carbohydrates (n = 191 and n = 189, respectively).

Models were fitted in R version 4.4.2 using the glmmTMB function from the glmmTMB package (Brooks et al. 2017). A manual model simplification procedure was applied, iteratively removing non-significant terms, starting with the highest-level interactions, until only significant effects remained, before estimating the final model parameters. All reported values are presented as mean ± 95% confidence interval (CI) unless otherwise specified. The variance explained by the models was then recorded with R² values (fixed effects alone R²m and both fixed and random effects combined R²c, respectively, Nakagawa and Schielzeth 2013). Pairwise comparisons (Tuckey’s test) were performed using the emmeans package (Lenth et al. 2025).

For the analysis of endophytic fungal community diversity and pathogen diversity, reads count were rarefied using rrarefy() in the vegan package (Oksanen et al. 2025) to standardize sampling effort among samples and reduce biases associated with unequal sequencing depth and strong differences in OTU relative abundances (Weiss et al. 2017). The rarefaction threshold was set to the 0.25 quantile of the observed sequencing depth distribution across samples, as determined from rarefaction curves (Supplementary Fig. 2). The hill taxa function from the HillR package (Chao et al. 2014) was used to compute Shannon diversity (Hill number q = 1), as proposed by Calderón-Sanou et al. (2020). Diversity metrics were analysed using GLMMs models with Gaussian error distributions (n = 174). The composition of communities was computed with Aitchison distance metric after a centered log ratio transformation followed by Euclidean distance calculation for data sparsity constraint and differences in sequencing depth (Martino et al. 2019). We performed a permutational analysis of variance (PERMANOVA) to test the effects of oak species identity, irrigation and OLB damage treatments and their interactions with Cage set as a blocking factor with adonis2() from the vegan package (Oksanen et al. 2025). To distinguish the mechanisms by which the communities change, i.e. species replacement (turnover) or species loss or gain (nestedness), we used the function beta.pair in the package betapart (Baselga and Orme 2012) to partition the beta diversity within each host species. The turnover refers to the replacement of an existing OTU with a new one ; and the nestedness refers to change in the number of OTUs, involving gain or loss of OTUs. The effect of OLB damage on each diversity component was tested with a PERMANOVA on the distance matrix generated by the beta.pair function. Finally, we searched for the presence of taxon that would be statistically associated with a specific treatment by performing an indicator species analysis (IndVal). IndVal was performed with the multipatt function from the indicspecies R package (Cáceres and Legendre 2009) as follows : read counts were filtered at 5% prevalence (232 OTUs remaining), and compared in the six groups (Irrigation × Herbivory), with statistical significance assessed using 999 permutations.

Results

Effects of water deficit and herbivory on OLB-induced leaf discoloration

The introduction of OLB into cages led to an average of 47% leaf area discoloration after five weeks, with no significant difference (χ² = 2.89, P = 0.089, df = 1) between cages in which we initially introduced 50 (mean ± 95% CI = 44.0 ± 4.9%) or 100 OLB individuals (51.5 ± 5.8%). We observed no leaf discoloration in control cages in which no OLB were introduced.

Oak species identity explained a significant proportion of the variation in leaf discoloration (χ² = 29.49, df = 3, P < 0.001; Fig. 1). Quercus petraea was the most susceptible species, with mean leaf discoloration reaching 61%, whereas values in Q. robur were approximately twofold lower. Quercus cerris and Q. pubescens had intermediate discoloration levels (47% and 50%, respectively). Neither the interaction with species nor the main effect of irrigation was significant (χ² = 1.68, df = 3, P = 0.64; χ² = 0.34, df = 1, P = 0.556, respectively). The model explained 19% of the variation in leaf discoloration (R²m = 0.19), with none of this variation attributable to the random effect (R²c = 0.19). The sensitivity analysis of the model indicated that effects of approximately 5.4% points of leaf discoloration could be detected with 80% power for factors with one degree of freedom (i.e., irrigation, herbivory pressure, and their interaction). Effects involving three degrees of freedom (i.e., species and interactions involving species) required differences of approximately 6.3% points to be detected with similar power. This suggests that any undetected effect of irrigation on herbivory-induced damage would likely be small and unlikely to translate into substantial changes in oak performance or leaf fungal communities.

Fig. 1.

Leaf discoloration (% chlorosis) as a function of oak species (Quercus cerris, Quercus petraea, Quercus pubescens, Quercus robur) under two irrigation treatments (high irrigation, blue; low irrigation, red) and two levels of herbivory pressure (50, triangles; 100, filled circles). Small semi-transparent symbols represent raw data (n = 119); large symbols and error bars represent raw means ± 95% CI. Different letters (a, b and c) indicate significant differences between oak species according to a post hoc Tukey test following ANOVA (χ² = 29.49, df = 3, P < 0.001; R²m = 0.19, R²c = 0.19)

Diversity and composition of fungal communities in leaves

The endophytic fungal community was dominated by Ascomycota (96.7%), followed by Basidiomycota (2.9%). Mucoromycota accounted for less than 0.1%, and undefined fungi OTUs made up the remaining 0.4% of the studied community (Fig. S2, S3). At the family level, Cladosporiaceae was the most abundant (50%), followed by Erysiphaceae (22.1%), Aspergillaceae (10.9%), and Aureobasidiaceae (5.4%). The five most abundant OTUs, in decreasing order, belonged to the genera Cladosporium (OTU2 and OTU1, with 66,023,878 and 55,778,211 reads, respectively), Erysiphe (OTU3, with 43,244,420 reads), Penicillium (OTU4, with 15,388,487 reads), and Aureobasidium (OTU5, with 14,560,695 reads).

Endophytic fungal diversity responded only to irrigation, with diversity significantly reduced, by approximately 23%, in low-irrigated saplings (χ² = 13.1, df = 1, P < 0.001; Fig. 2) and showing no significant relationship with leaf discoloration (χ² = 0.45, df = 1, P = 0.5) or species (χ² = 4.29, df = 3, P = 0.23). The model explained 20% of the variance, with nearly half of it accounted for by the fixed effects (R²m = 0.09, R²c = 0.20).

Fig. 2.

Endophytic fungi diversity (Shannon index, Hill q = 1) under two irrigation treatments (high irrigation, blue; low irrigation, red) and three levels of herbivory pressure (None, open circles; 50, triangles; 100, filled circles). Small semi-transparent symbols represent raw data (n = 174), whereas large symbols and error bars indicate means ± 95% CI. Different letters (a and b) above treatments indicate significant differences between irrigation treatments according to a post hoc Tukey test following ANOVA (χ² = 13.1, df = 1, P < 0.001; R²m = 0.09, R²c = 0.20)

Pathotrophic fungal diversity also differed among irrigation treatments (χ² = 19.65, df = 1, P < 0.001; Fig. S4), with a 21% loss in low-irrigated saplings. The model explained 10% of the variance, with fixed effects accounting for all of it (R²m = 0.10, R²c = 0.10). No significant effect of leaf discoloration was detected (χ² = 0.03, df = 1, P = 0.86).

The composition of endophytic fungal communities varied along the leaf discoloration gradient, with patterns depending on oak species (F = 1.51, df = 3, P = 0.001, R² = 0.03, Table 1). Whereas we observed a slight turn over of fungal communities in the leaves of Q. robur discoloured by OLB feeding (β total = 5%, Turnover = 7% ; Table 2), fungal communities associated with intact leaves of Q. petraea (β total = 11%, nestedness = 19%; Table 2) was a nested subset of the communities observed in discolored leaves. Irrigation also affected community composition (F = 1.23, df = 1, P = 0.019, R² = 0.007, Table 1) but not total beta diversity defined by beta.pair (Sorensen index: F = 1.17, df = 1, P = 1, R² = 0.007) or its turnover and nestedness components. For saplings subjected to a single stress, corresponding to the highest level of herbivory pressure, the indicator species analysis identified eight indicator OTUs, of which five were classified as pathotrophs (Table S2). Under combined stress conditions, only one indicator species was detected, the saprotrophic fungus Bjerkandera adusta. In contrast, saplings not exposed to any stress were characterized by two indicator OTUs that were not classified as pathotrophs, belonging respectively to the Clavulinaceae and Rozellomycota lineages (Table 3).

Table 1.

Summary of PERMANOVA (adonis2) results testing the effects of leaf discoloration, irrigation, species, and their two-way interactions on the community composition (euclidean distance matrix on robust centered log-ratio-transformed reads count). Significant effects (p < 0.05) are in bold

Predictors	Df	R²	F	p_value
Species	3	0.03	1.75	0.001
Irrigation	1	0.007	1.23	0.019
Leaf discoloration	1	0.014	2.54	0.117
Species x Irrigation	3	0.02	0.91	0.71
Species x Leaf discoloration	3	0.03	1.48	0.003
Irrigation x Leaf discoloration	1	0.006	1.05	0.127

Table 2.

Summary of PERMANOVA (adonis2) results testing the effect of leaf discoloration on the three beta diversity components (total, turnover and nestedness distance matrices generated by beta.pair) associated with Quercus species. Significant effects (p < 0.05) are in bold

Predictors	Component	Df	R²	F	p_value
Q. cerris x Leaf discoloration	β total (Sorensen)	1	0.03	1.49	0.1
Q. pubescens x Leaf discoloration	β total (Sorensen)	1	0.01	0.40	0.98
Q. petraea x Leaf discoloration	β total (Sorensen)	1	0.11	4.75	0.001
	Turnover	1	0.08	3.50	0.040
	Nestedness	1	0.19	9.51	0.002
Q. robur x Leaf discoloration	β total (Sorensen)	1	0.05	2.09	0.031
	Turnover	1	0.06	2.63	0.027
	Nestedness	1	0.03	1.30	0.3

Table 3.

Stem starch and total soluble sugar concentrations (mg.g⁻¹ dry matter) in oak species. Homogeneous groups obtained using Tukey’s multiple comparison test. Means sharing the same letter do not differ significantly at α = 0.05 (Tukey HSD test)

Predictors	Model-based mean	SE	group
	Starch (mg.g⁻¹ DM)
Q. cerris	3.022	0.102	a
Q. pubescens	3.043	0.106	a
Q. petraea	3.268	0.104	ab
Q. robur	3.32	0.104	b
	Total soluble sugars (mg.g⁻¹ DM)
Q. cerris	0.492	0.026	ab
Q. pubescens	0.54	0.026	b
Q. petraea	0.518	0.027	ab
Q. robur	0.467	0.026	a

Photosynthesis

During the first survey, stomatal conductance was lower under low irrigation, with the magnitude of reduction varying among oak species (χ² = 9.22, df = 3, P = 0.026). Specifically, reducing irrigation halved stomatal conductance in Q. cerris (from 0.187 to 0.093 mol H₂O·m⁻²·s⁻¹) and Q. petraea (from 0.132 to 0.068 mol H₂O·m⁻²·s⁻¹), and reduced it by 65% in Q. pubescens (from 0.186 to 0.066 mol H₂O·m⁻²·s⁻¹) and 70% in Q. robur (from 0.155 to 0.045 mol H₂O·m⁻²·s⁻¹). Species identity and irrigation accounted for 40% of total variance in stomatal conductance (marginal R² = 0.40, conditional R² = 0.64). Still during the first survey, the operational quantum yield of photosystem II (PhiPS2) decreased under low irrigation (χ² = 3.92, df = 1, P = 0.048) and also varied among species (χ² = 11.7, df = 3, P = 0.009). Values were significantly lower in Q. cerris than in Q. petraea and Q. robur. The model accounted for 96% of total variance in PhiPS2 (marginal R² = 0.23, conditional R² = 0.96).

During the second survey (with leaf discoloration in models) PhiPS2 differed significantly among oak species (χ² = 13.4, df = 3, P = 0.004) the highest values were measured in Q. pubescens, the lowest in Q. petraea. PhiPS2 decreased significantly with increasing leaf discoloration (χ² = 6.80, df = 1, P < 0.001; model estimate ± SE = -0.003 ± 0.001). Model predictions indicate that PhiPS2 decreased from 0.73 to 0.68 between 0% and 100% leaf discoloration (Fig. 3). The model explained 91% of the variance in PhiPS2 (R²c = 0.91), whereas 33% was attributable to fixed effects (R²m = 0.33). PhiS2 was not significantly influenced by irrigation (χ² = 0.49, P = 0.48, df = 1). Stomatal conductance differed significantly among oak species (χ² = 11.04, df = 3, P = 0.011). Specifically, Q. petraea exhibited the highest mean predicted stomatal conductance (0.25 mol H₂O·m⁻²·s⁻¹), which was significantly greater than that of Q. robur (0.19 mol H₂O·m⁻²·s⁻¹). The model accounted for 77% of the total variance in stomatal conductance, with the majority of this variation being explained by the random effect (marginal R² = 0.02, conditional R² = 0.77). Neither leaf discoloration (χ² = 0.06, df = 1, P = 0.80) nor irrigation (χ² = 0.48, df = 1, P = 0.49) had a significant effect.

Fig. 3.

Relationship between leaf discoloration (%) and Phi PS2 (mol H₂O·m⁻²·s⁻¹) in four oak species. The fitted curves and shaded areas represent the predicted relationship and its 95% confidence interval, respectively, based on model predictions. Points correspond to raw data (n = 131). Colours indicate species identity: Q. cerris (green), Q. petraea (blue), Q. pubescens (orange) and Q. robur (pink). Different letters (a, b) indicate significant differences among oak species according to a post hoc Tukey test following ANOVA (χ² = 13.4, df = 3, P = 0.004; R²m = 0.33, R²c = 0.91)

Tree growth

Stem diameter increment was significantly higher in highly irrigated trees with species-specific responses (irrigation × species: χ² = 13.71, df = 3, P = 0.003). Notably, the reduction in growth under low irrigation was only significant for Quercus robur (estimate ± SE = 1.11 ± 0.21, t₍₂₀₎ = 5.28, P < 0.001) and Q. pubescens (estimate ± SE = 0.55 ± 0.21, t₍₂₀₎ = 2.67, P = 0.014), corresponding to an approximate 13% growth reduction for both species, which were also those exhibiting the highest overall growth (Fig. S5). There was a tendency towards reduced growth in presence of OLB, but this effect was only marginally significant (χ² = 3.82, df = 1, P = 0.051). The model explained 78% of the variation in diameter growth (R²m = 0.75), with little of this variation attributable to the random effect (R²c = 0.78).

Stem carbohydrates

Starch concentration in twigs differed significantly among species, albeit moderately, after accounting for irrigation regime and leaf discoloration (species: χ² = 12.77, df = 3, P = 0.005). The model explained 34% of the variance, with fixed effects accounting for a relatively small proportion (R²m = 0.08, R²c = 0.34). Post hoc Tukey comparisons indicated that Q. robur exhibited approximately 9% higher starch concentrations than Q. cerris and Q. petraea (Table S3). Similarly, total soluble sugar concentration varied significantly among species (χ² = 8.09, df = 3, P = 0.044), but the overall explanatory power of the model was lower (R²c = 0.22), with fixed effects contributing only marginally (R²m = 0.04). Tukey’s post hoc test showed that Q. robur had around 14% lower total soluble sugar concentrations than Q. petraea (Table S3). No significant effect of leaf discoloration was detected for either factor (starch : χ² = 0.26, df = 1, P = 0.608 ; total soluble sugar : χ² = 1.65 df = 3, P = 0.199 ; detailed interaction results are provided in S2).

Discussion

We analysed the response of European oak species to a combination of biotic and abiotic stress, and found that these trees exhibit a certain degree of resilience. We identified drought as the primary driver of oak performance and endophytic fungi diversity. In contrast, we found little effect of insect herbivory on oak sapling performances, despite considerable levels of discoloration and clear effects of insect damage on photosynthesis and endophytic fungal composition. However, observed changes in the leaf microbiome can, in the long term, pose a significant threat to tree health.

Fungal endophytic communities are known to mitigate the stress in plants (Stone et al. 2018; Trivedi et al. 2020). As such, these communities can serve as sensitive indicators of both stress exposure and overall plant health (Poudel et al. 2021; Trivedi et al. 2022). Consistently with previous studies (Bechtold et al. 2021; Karasov et al. 2022), we found that both water deficit and insect herbivory influenced leaf endophytic fungal communities, but found indications that the mechanisms at play are different and vary among oak species. Specifically, reduced water supply decreased the diversity of leaf endophytic fungi but had only a minor effect on their composition, suggesting a consistent filtering effect on fungal species on oak trees. In contrast, herbivory induced species-specific shifts in the composition of leaf endophytic communities resulting in an addition of fungal species in leaves of Q. petraea discoloured by the OLB (i.e., nestedness), or in the replacement of some species by others (i.e., turn-over) in Q. robur. Insect–plant interactions involve microbiota exchange, which—depending on the insect and plant species—can disrupt plant-associated microbial communities (Lieutier et al. 2009; Chung et al. 2013; Humphrey and Whiteman 2020; Pirttilä et al. 2023). The changes observed here may result from two non-exclusive mechanisms: the introduction of fungal taxa vectored by the insect and the induction of plant defense signaling pathways having pleiotropic effects on leaf associated microbes (Karasov et al. 2022; Pirttilä et al. 2023). For instance, the Colorado potato beetle (Leptinotarsa decemlineata) transfers its pathogenic fungal endosymbionts to the host plant, ultimately enhancing the beetle’s performance (Chung et al. 2013). Similarly, aphids feeding on wild tomatoes induce shifts in the diversity in the microbiome of the rhizosphere resulting in a bottom-up control on insect performance via plant-induced defenses (Trivedi et al. 2020; French et al. 2021). Moreover, plant-induced defenses can create compositional changes including a decline in diversity and an enrichment of phytopathogenic microbes (Humphrey and Whiteman 2020). Consistently, we identified five pathotrophic OTUs exclusively in saplings heavily impacted by OLB, which may explain the observed microbial imbalance. Among them, Aureobasidium sp.— one of the most abundant OTU —is known to cause leaf and shoot damage, thereby reducing photosynthesis (Czachura and Piątek 2025). Similarly, Erysiphe spp., the 2nd most abundant OTU, is widely recognized as powdery mildew pathogens and detrimentally affects tree growth and moisture regulation (Bert et al. 2016). By favoring specific microbial taxa with pathotrophic habits— which could be interpreted as an early warning sign of dysbiosis (Arnault et al. 2023)—the foliar damages induced by OLB could ultimately reduce plant fitness (Khare et al. 2018; Vannier et al. 2019). Importantly, the absence of synergistic effects between drought and herbivory on foliar fungal communities suggests that distinct and independent mechanisms are at play.

Because OLB is expected to avoid drought-stressed foliage (Connor 1988), but mesophyll feeders respond inconsistently to drought in general (Huberty and Denno 2004), we hypothesized that oaks subjected to low irrigation would display different levels of leaf discoloration than well-watered ones. Although our treatment did not necessarily impose an equivalent level of physiological stress across species, it did induce a measurable response to water limitation in all four species, as confirmed by the significant decrease in stomatal conductance and photosynthesis, and by changes in leaf microbial communities. Despite this confirmed physiological response to water deficit (48–70% reduction in stomatal conductance depending on species) we did not detect a significant effect of irrigation on the level of leaf discoloration caused by OLB. Because realized stress intensity was intrinsically tied to species identity rather than manipulated independently, this result indicates that, at the species-specific stress level actually achieved, OLB damage did not depend on oak water status; it does not rule out the possibility that damage could differ under more severe water deficit. Irrigation could also have longer-term effects on OLB through changes in its reproduction rate and population dynamics (Gutbrodt et al. 2011; Asiyeh Notghi Moghadam et al. 2022); however, we measured the damage sustained rather than insect population dynamics, and therefore cannot fully rule out this hypothesis either. In contrast to the limited irrigation effect, we found a clear effect of oak species identity on OLB-induced discoloration. Although our experimental design does not allow us to disentangle the respective contributions of host acceptance and host suitability, these results show that, after one insect generation, Q. robur was the least damaged by OLB, whereas the three other species displayed comparatively higher levels of OLB-induced decoloration, with Q. petraea being the most affected (61%; Fig. 1). Interestingly, the contrasting responses of the four oak species suggest that drought response, susceptibility to OLB herbivory, and fungal community responses may be linked. Q. petraea, for instance, combined a relatively limited reduction in stomatal conductance (48%) under low irrigation with the highest OLB-induced discoloration (61%, see Fig. 1) and the strongest shift in fungal communities. In sharp contrast, Q. robur displayed the strongest stomatal response to water deficit (70% decrease), but relatively little discoloration (30%, see Fig. 1) and a more limited fungal response. Together, these results suggest a species-specific pattern of responses to herbivory vs. drought, whereby greater drought susceptibility is associated with greater tolerance to herbivory and better conservation of fungal endophytic communities. However, Q. pubescens and Q. cerris did not follow this pattern, indicating that the relationship between drought tolerance, herbivore susceptibility, and microbiome responses is likely to be species-dependent and requires further investigation to understand what mechanisms are in play. Because oak species are known to differ in their drought tolerance and physiological thresholds (Higgs and Wood 1995), further investigation should focus on characterizing the mechanisms underlying species-specific differences in the response of herbivory to water treatment, for instance by controlling stress intensity across species.

The activity of the photosystem II decreased (i.e. decrease in CO₂ assimilation) proportionally to discolouration resulting from OLB feeding, in line with previous results (Nikolić et al. 2018; Paulin et al. 2020), but stomatal conductance remained unaffected. Given that C. arcuata is a cell-content feeder, the decrease of the photosynthesis likely reflects the depletion of the photosynthetic equipment in the leaf blade cells. Overall, at the scale and time point of our experiment, these results suggest that OLB herbivory had a stronger effect on photosynthetic activity than the level of water deficit imposed, possibly because herbivore damage accumulates as photosynthetic tissues are progressively damaged, whereas the physiological effects of water deficit can be transient and partly reversible when water availability improves. Besides these effects on photosynthesis, irrigation only affected oak sapling growth, whereas neither OLB herbivory nor irrigation affected carbohydrate storage. However, we are conscious that additional insights would be gained by monitoring the dynamics of resource allocation to different compartments — including collar and roots— over the course of the season (Barbaroux and Breda 2002; Perkovich and Ward 2021).

Overall, our results provide little evidence that biotic and abiotic stresses combined to impair the performance of oak saplings under controlled conditions. Water deficit emerged as the primary driver of oak sapling growth but had no visible influence on the damage caused by the oak lace bug in the short term. In contrast, the direct effects of herbivory on growth were limited, despite considerable leaf discoloration accompanied by a marked reduction in photosynthetic capacity. Herbivory was also associated with shifts in the composition of leaf endophytic communities and the presence of pathotrophic fungi, such as Erysiphe spp. The observed vulnerabilities highlight the need for further studies to determine the long-term consequences of interactive stresses in trees, and to establish whether the observed shifts in fungal communities and apparent physiological–growth decoupling may reflect early signs of dysbiosis or, conversely, a degree of resilience or compensatory capacity in young oaks.

Supplementary Information

Below is the link to the electronic supplementary material.

Authors’ contributions

TB conducted the stem carbohydrate quantification and metabarcoding experiments, performed data analyses, and wrote the manuscript. BC contributed to experimental design and setup, field measurements, data analysis, and manuscript writing. HS contributed to data analysis and manuscript writing. RB contributed to experimental design and setup, data analysis, and manuscript revision. OB contributed to experimental measurements and stem carbohydrate quantification. JC contributed to experimental setup, measurements, and manuscript revision.

Funding

T. Barlier was funded by the French National Institute for Agronomy Research (INRAE) and French National Forest Office (ONF) under the Convention de co-financement de thèse INRAE-ONF 2024–2028. This work was supported by the French Department of Forest health (DSF) under Grant agreement C-2024-099.

Data availability

The data was deposited in entrepot.recherche.data.gouv.fr which can be accessed at https://doi.org/10.57745/9VEH2P. The sequencing data reported in this paper have been deposited in the European Nucleotide Archive under accession number PRJEB110006.

Code availability

The code can be accessed at the same link as datasets : https://doi.org/10.57745/9VEH2P.

Declarations

Conflicts of interest

Not applicable.

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.