Common garden experiments suggest terpene-mediated associations between phyllosphere microbes and Japanese cedar

Abstract

Plant–microbe interactions in the phyllosphere provide invaluable information on plant ecology, with implications for ecosystem functioning and plant–atmosphere feedbacks. The composition of phyllosphere microbes varies significantly depending on host lineages, geographic regions, and climatic conditions. However, the factors driving these variations in interactions with plants remain poorly understood. Biogenic volatile organic compounds (BVOCs) emitted by plants may be important in these interactions. Here, we quantified the composition of phyllosphere microbial communities and terpene emissions from leaves of Japanese cedar (Cryptomeria japonica) trees grown in two common gardens from cuttings collected from natural populations across Japan. Amplicon sequencing revealed that microbial communities differed significantly between gardens and among host populations. Analysis of BVOC profiles showed that the camphene and total terpene emission rates were associated with bacterial composition, whereas that of ent-kaurene was marginally linked to fungal composition. The relative abundances of certain fungal genera that include the species reported to cause disease in Japanese cedar, the emission rates of most monoterpenes and a sesquiterpene β-farnesene were correlated with the climatic conditions at the origin sites of the cedar trees. These findings highlight the intricate relationships between phyllosphere microbes and terpene emission from host trees and suggest the role of climatic factors in shaping these associations.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-16496-2.

Introduction

The aerial surface of plant leaves, called the phyllosphere, harbors a diverse array of microbes including bacteria and fungi¹. Associations of phyllosphere microbes with host plants have a considerable impact on plant physiology and ecology. Phyllosphere microbes can influence plant fitness by causing or suppressing disease^2–5 or by promoting plant growth and stress tolerance^6,7. Plant defense against pathogens and the counteradaptation by the latter further complicate plant–microbe interactions^7–10. Microbial infection may also affect plant–atmosphere feedbacks through microbial metabolization of gases emitted by plants^11–14. Studies of plant–microbe interactions can lead to insights into key issues in ecology, such as the evolution of plant traits and drivers of ecosystem functions that influence biosphere–atmosphere interactions.

Shifts in the composition of the phyllosphere microbial community are significant indicators of changes in plant–microbe interactions^5,15,16. This composition is associated with environmental factors including climate^17–19 season^17,20 land use^21,22 location^17,23,24 and geographic distance²¹ as well as by host-related factors such as species^{17,19,23,25,26} and traits^17,19,27. The association of host genotypes with microbial composition in the phyllosphere has been investigated using known characterized mutants or accessions^20,24,28. However, further investigation of genetic and phenotypic variation within wild plant species is necessary to elucidate the associations and coevolution of phyllosphere microbes with plants.

Biogenic volatile organic compounds (BVOCs) are plant-derived secondary metabolites that affect plant–microbe interactions^10,29. BVOCs affect plant resistance to microbes by directly inhibiting microbial growth by damaging microbial cell membranes^30,31 and inducing plant immune responses³². On the other hand, some microbes metabolize BVOCs stored in or emitted from plants and use them as a carbon source, suggesting their adaptation to plant resistance mechanisms^11,33–35. While many laboratory studies have suggested a relationship between pathogen-induced BVOC emission by plants and phyllosphere microbes^32,35,36 few have investigated the roles of constitutive BVOC emission by plants under natural conditions in plant–microbe associations^37,38. Significant inter- and intraspecific variations in both the composition of the phyllosphere microbial community and emission rates of BVOCs have been reported for wild plants^37,39,40. Elucidating the drivers and interrelationships of these variations is a challenge in understanding plant–microbe associations mediated by plant BVOC emission.

BVOC emission can also have a significant impact on atmospheric chemical processes and climate. Oxidation of emitted BVOCs can indirectly affect the radiative balance by increasing the lifetime of certain greenhouse gases due to a decrease in oxidative capacity in the atmosphere⁴¹. Secondary organic aerosols produced by BVOC oxidation act as cloud condensation nuclei and may affect regional climate⁴². Guenther et al.⁴³ estimated that terrestrial plants emit about 1 billion tons of BVOCs per year, of which about 92% is emitted by forest trees. These estimates are based on the model of the standardized rates of BVOC emission by plants. However, this model does not fully account for variations in the amounts and composition of emitted BVOCs among and within plant species or the influences of plant associations with microbes. Therefore, elucidating the relationship between plant–microbe associations and plant BVOC emissions is important for understanding the plant–aerosol–climate feedbacks.

Japanese cedar, Cryptomeria japonica (Cupressaceae), is one of the major silvicultural tree species in Japan and emits substantial amounts of terpenes, including monoterpenes (MTs), sesquiterpenes (SQTs), and diterpenes (DTs)^44–46. In comparison with MTs, SQTs and DTs have high molecular weights and are highly reactive in the atmosphere^47,48. Since Japanese cedar accounts for a large proportion of Japan’s total forest area (approximately 20% including plantations)⁴⁹ these terpenes may have a non-negligible impact on the surrounding atmospheric environment. A study based on a common garden experiment reported significant intraspecific variation in the composition of terpenes stored in and emitted from the leaves of Japanese cedars⁴⁶. Intraspecific genetic variation in Japanese cedar has been well studied, revealing a geographic genetic structure with four main clades: the Pacific Ocean side clade, two Sea of Japan side clades, and the Yakushima Island clade^50–52.

Intraspecific variation in physiological and morphological traits of Japanese cedar has been revealed by common garden experiments, suggesting adaptation to the environments at the origin sites^53–55. The reasons for differences in the quantity and composition of stored and emitted terpenes among Japanese cedar populations are still unclear. However, these differences may be influenced by the local climate and pathogen communities at the origin sites of the populations⁴⁶. The problem is that the pathogen community data used in that study were based on previously published information, and no quantitative data were obtained on the actual microbial community on the leaf surface of Japanese cedar. To clarify whether the intraspecific variations in terpene emissions observed in this species reflect the variations in the phyllosphere microbial community, quantitative investigations of both microbial communities and BVOCs emitted by the host are necessary.

Here, we investigated the phyllosphere microbial communities of Japanese cedar by amplicon sequencing using cedar trees that sourced from diverse wild populations across Japan, known for their genetic differentiation^50–52cultivated under controlled conditions in two common gardens (Supplementary Fig. S1). We also measured terpene emission rates in some of these trees from three populations in one common garden. Our objectives were to elucidate (1) the drivers of the variation in the phyllosphere microbial community composition among Japanese cedar populations, (2) the influence of local climates on cedar–microbe association, and (3) the relationship between terpene emission rates and these microbial compositions. Our study provides new insights into the mechanisms of plant–microbe associations mediated by host-emitted BVOCs.

Results

Composition of the phyllosphere microbial community of Japanese Cedar

Amplicon sequencing identified 2,933 bacterial amplicon sequence variants (ASVs) and 5,040 fungal ASVs across all samples from the phyllosphere microbial communities of Japanese cedar planted in the two common gardens. After rarefaction, the total number of bacterial ASVs was 2,871 and that of fungal ASVs was 4,865. The richness of bacterial ASVs (9,099 reads) was 89 ± 64 (mean ± SD) and that of fungal ASVs (17,427 reads) was 136 ± 121 per sample. In the bacterial communities, the most abundant phyla were Proteobacteria (58.9% of all reads), Acidobacteriota (6.8%), Myxococcota (5.9%), Verrucomicrobiota (4.6%), Bacteroidota (3.0%), Firmicutes (2.5%), and Actinobacteriota (1.7%) (Fig. 1a, b). The relative abundance of Myxococcota tended to be higher in the Kawatabi common garden, whereas that of Acidobacteriota was higher in the Tsukuba common garden. In the fungal communities, the most abundant classes were Dothideomycetes (32.4% of all reads), Eurotiomycetes (30.7%), Leotiomycetes (9.2%), Lecanoromycetes (7.6%), Sordariomycetes (2.6%), Cystobasidiomycetes (2.5%), Agaricomycetes (2.1%), and Tremellomycetes (1.8%) (Fig. 1c, d). The relative abundance of the classes Leotiomycetes and Lecanoromycetes tended to be higher in the Tsukuba common garden than in the Kawatabi common garden.

Fig. 1.

Compositions of the phyllosphere microbial communities of Japanese cedars planted in the two common gardens. Each bar represents the phyllosphere microbial community of each individual in (A, C) the Kawatabi and (B, D) the Tsukuba common garden. Bar arrangement is based on the value of the climatic PC1 (from high to low) at the origin site of each individual. (A, B) Relative abundances of the eight most abundant bacterial phyla. (C, D) Relative abundances of the eight most abundant fungal classes. The original population of each sample is indicated under the bar (see Supplementary Fig. S1 and Table S4 for the names and locations).

The composition of the bacterial and fungal communities was significantly influenced by common garden and host population (Fig. 2). In the bacterial community, 38.3% of the total variation was explained by common garden (PERMANOVA: R² = 0.130, p = 0.001), host population (R² = 0.124, p = 0.014), and their interaction (R² = 0.128, p = 0.009). In the fungal community, 40.1% of the total variation was explained by common garden (PERMANOVA: R² = 0.176, p = 0.001), host population (R² = 0.116, p = 0.028), and their interaction (R² = 0.115, p = 0.033).

Fig. 2.

Differences in the composition of the phyllosphere (A) bacterial and (B) fungal communities of Japanese cedar between the common gardens at Kawatabi (blue) and Tsukuba (orange). Differences were visualized using nonmetric multidimensional scaling (NMDS) based on Bray–Curtis dissimilarities. The density of color represents the climatic PC1 at the origin site of each tree. MDS1 indicates the first ordination axis generated by NMDS, and MDS2 indicates the second.

To investigate how climatic conditions at the origin sites of the host populations influence their effects on the composition of the microbial communities, we obtained the climatic data^56–59 at these locations. A principal component analysis (PCA) of these data revealed that 73.8% of the total variation was explained by the first two PCs (Supplementary Fig. S2), which were used for the subsequent analysis. Climatic PC1, which explained 50.57% of the total variation, corresponded to higher annual mean temperature and mean monthly potential evaporation and corresponded to lower standard deviation of the monthly mean temperatures and the sum of the monthly precipitation amount for the months with mean temperatures below 0 °C, suggesting that higher values of PC1 correspond to a warmer and less snowy climate. Climatic PC2, which explained 23.19% of the total variation, corresponded to higher mean monthly near-surface relative humidity and annual precipitation amount and a lower coefficient of variation of the monthly precipitation estimates, suggesting that higher values of PC2 correspond to higher amounts of precipitation and its stability.

The total explained variation in microbial community composition was reduced when climatic PC1 and PC2 at the host origin sites were used as explanatory variables instead of host population. The result of PERMANOVA testing the effects of climatic PCs and their interactions with common garden on the variation in bacterial community composition was as follows: R² = 0.014 and p = 0.044 for PC1, R² = 0.011 and p = 0.136 for the interaction between PC1 and common garden, R² = 0.009 and p = 0.335 for PC2, R² = 0.010 and p = 0.160 for the interaction between PC2 and common garden. For the total variation in fungal community composition, the result of PERMANOVA was R² = 0.010 and p = 0.126 for PC1, R² = 0.007 and p = 0.671 for the interaction between PC1 and common garden, R² = 0.009 and p = 0.277 for PC2, and R² = 0.008 and p = 0.346 for the interaction between PC2 and common garden. Of the detected fungal genera, 10 included the fungal species that have been reported to cause disease in Japanese cedar⁶⁰ (Supplementary Table S1). Most of these genera were predicted to be plant pathogens or saprotrophs based on the FungalTraits database⁶¹ suggesting functional associations with plants. The aggregated abundances of these fungi after rarefaction were negatively correlated with climatic PC1 at the host origin sites (Wald test: β = − 0.423 ± 0.158, z = − 2.675, p = 0.007; Fig. 3). However, the influence of common garden and climatic PC2 on these abundances was not statistically significant (both p > 0.1).

Fig. 3.

Relationship between the abundance of the fungal genera that include the species reported to cause disease in Japanese cedar after rarefaction and climatic PC1 at the origin sites of the host trees. The solid line represents fitted values from a negative binomial generalized linear model, where the aggregated abundance of all the genera was modeled as a function of climatic PC1. Shaded areas indicate the 95% confidence intervals of the model predictions. See Supplementary Table S1 for the list of these genera.

Relationships between terpene emissions, phyllosphere microbial community composition, and climate

Japanese cedar trees in the Kawatabi common garden emitted 12 MTs, 2 SQTs, and 2 DTs (Fig. 4a). β-Caryophyllene was excluded from the subsequent analysis since it was detected from only one tree (Supplementary Table S2). PCA showed that PC1, PC2 and PC3 collectively explained 71.7% of the total variation in the log-transformed basal emission rates of terpenes (Fig. 4b, Supplementary Fig. S3). Terpene PC1 negatively corresponded to higher basal emission rates of most MTs and was significantly positively correlated with climatic PC1 at the tree origin sites (Student’s t-test: β = 0.444 ± 0.194, t = 2.294, p = 0.031; Supplementary Fig. S4a, b). Terpene PC2 corresponded to the higher basal emission rates of some MTs, ent-kaurene and phyllocladene but was not significantly correlated with the climatic PCs (no explanatory variables were selected by the model selection; Supplementary Fig. S4c, d). Terpene PC3 corresponded to the higher basal emission rates of some MTs, β-farnesene, and phyllocladane and was significantly negatively correlated with climatic PC1 at the tree origin sites (Student’s t-test: climatic PC1, β = − 0.417 ± 0.172, t = − 2.430, p = 0.023; Supplementary Fig. S4e, f).

Fig. 4.

Terpene emission profiles of Japanese cedar in the common garden at Kawatabi. (A) Basal emission rates (ng gdw^− 1 h^− 1) of monoterpenes and methyl salicylate (green to yellow), sesquiterpenes (purple), and diterpenes (blue) emitted from each cedar individual originating from three populations. (B) Principal component analysis of log-transformed basal emission rates (ng gdw^− 1 h^− 1) of terpenes. The color of the vectors indicates the type of terpenes (green: MTs, purple: SQTs, blue: DTs). The first 2 axes are shown.

The basal emission rate of total terpene, as well as that of camphene specifically, was significantly correlated with bacterial community composition (total terpene, R² = 0.280, p = 0.023; camphene, R² = 0.472, p = 0.003; Fig. 5a). The basal emission rates of α-pinene, 3-carene and the sum of the basal emission rates of MTs was marginally correlated with bacterial composition (α-pinene, R² = 0.213, p = 0.061; 3-carene, R² = 0.189, p = 0.089; MTs, R² = 0.198, p = 0.082). The basal emission rate of ent-kaurene was marginally correlated with fungal community composition (R² = 0.214, p = 0.066; Fig. 5b). The higher basal emission rate of camphene and the higher basal emission rates of total terpene corresponded to higher abundances of the phyla Actinobacteriota, Firmicutes, Bacteroidota, and Cyanobacteria. The higher basal emission rate of ent-kaurene corresponded to higher abundances of the classes Taphrinomycetes, Tremellomycetes, and Agaricomycetes.

Fig. 5.

Vector fitting of log-transformed basal emission rates (ng gdw^− 1 h^− 1) of terpenes onto NMDS coordinates constructed from the compositions of the phyllosphere (A) bacterial and (B) fungal communities of Japanese cedar from three populations (blue: Ajigasawa, grey: Azouji, orange: Yakushima). (A) Bacterial phyla and (B) fungal classes whose relative abundances were significantly correlated with the NMDS coordinates are also shown. Only loading vectors of significantly correlated terpenes and microbial taxa are shown (solid: p < 0.05, dashed: p < 0.1). The color of the vectors and the letters indicates the type of variable (grey: microbial taxa, black: total terpene, green: MTs, blue: DTs).

Discussion

Using common garden experiments, we revealed that differences in the composition of both bacterial and fungal communities in the phyllosphere of Japanese cedar were dominated by differences between common gardens rather than by differences among host wild populations (Fig. 2). This result suggests that local environments or local microbial communities in common garden primarily determine the composition of microbial communities in the phyllosphere. This result is consistent with those of previous studies on grasses^20,24,62 and broad-leaved trees⁶³.

Despite the dominant effect of common garden, we also detected differences in microbial composition among wild populations of Japanese cedar (Figs. 2 and 3). Genetic variation among cedar populations may also influence the selection of phyllosphere microbes from the local microbial communities. The aggregated relative abundances of some fungal genera, which included the species reported to cause disease in Japanese cedar⁶⁰ (Supplementary Table S1), were negatively correlated with warm and less snowy climate at the host origin sites (Fig. 3). This suggests the influence of climate on coevolution of some phyllosphere microbes and Japanese cedar. Several studies have suggested a correlation between microbial community composition and the environment, including climate^19,64–68. Some disease-causing species of these genera have been reported in wild Japanese cedar populations, mainly in the northern region of Honshu Island⁴⁶ which is consistent with the results of this study. It is possible that the climate clines in the distribution pattern of these fungi may have influenced their coevolution with associating plants. Field-grown Japanese cedar trees from cold-climate populations have been suggested to store higher amounts of terpenes⁴⁶ and these fungi may be better adapted than other microbes to this defense trait. Most of these fungi were predicted to be functionally associated with plants as pathogens or saprotrophs (Supplementary Table S1). The higher abundance of these fungi in the phyllosphere of cedar trees from cold-climate populations may provide them with an ecological advantage in infecting the host or being involved in the early stage of litter decomposition⁶⁹.

Our data suggest that the basal emission rates of some terpenes in Japanese cedar may be related to the composition of the phyllosphere microbial community. The basal emission rates of total terpene were correlated with the composition of the bacterial community, as well as the basal emission rates of some MTs (Fig. 5a). These results suggest that the emission of terpenes, primarily MTs, collectively mediates the cedar–bacteria association. The emission of multiple terpenes is simultaneously induced by plant defense signaling pathways^10,12,70. Camphene, together with other MTs α-pinene and β-pinene, has been suggested to play a pivotal role in the induction of defense responses against bacterial pathogens in Arabidopsis thaliana³². Given the correlation between the basal emission rate of most MTs (Fig. 4), multiple MTs may control the bacterial communities on Japanese cedar.

While the basal emission rate of the DT ent-kaurene was marginally correlated with the composition of the fungal community, the overall results suggest only a weak association between terpene emission and the composition of the cedar fungal community. Nevertheless, the emission of ent-kaurene might still affect the composition of the fungal community. This aligns with our previous research⁴⁶ which suggested a correlation between the composition of the fungal pathogen community in each cedar population and the relative basal emission rates of the SQT α-farnesene and the DT ent-kaurene. Furthermore, the essential oil of Japanese cedar, which contains ent-kaurene as a major component, is known to have antimicrobial activity⁷¹.Therefore, the emission of ent-kaurene likely plays an ecological role in the cedar–fungi association.

The weak observed correlation could be due to several factors: the determined emission rates of ent-kaurene were undetectable in eight samples (Supplementary Table S2). Considering the value of the limit of detection (LOD) (0.55 ng, Supplementary Table S3), the basal emission rate of ent-kaurene in these samples might be smaller than in the others. Improving the sensitivity of the detection method will confirm this prediction. In addition, outliers in the ent-kaurene emission data may have influenced the correlation test. Increasing the sample size and the number of populations used may elucidate a more detailed association between ent-kaurene emission and the composition of the fungal community, considering the large variation in the basal emission rate of ent-kaurene among Japanese cedar populations⁴⁶.

Some microbes are able to regulate or use plant defenses^10,72. Firmicutes was one of the bacterial phyla whose relative abundance was higher on the hosts with a higher basal emission rate of total terpene (Fig. 5a). Some strains of Firmicutes induce immunity in tomato and are relatively abundant in plants with enhanced defense⁷³. The induction of plant immunity often leads to the emission of terpenes^12,32,70. However, the elevated level of terpene emission observed in this study may not be induced by bacteria, as the terpenes emitted by Japanese cedar were suggested to be released primarily from storage pools⁷⁴. Nevertheless, the possibility remains that Firmicutes may tolerate cedar defenses mediated by enhanced terpene emission. The class Taphrinomycetes, whose relative abundance was positively related to the basal emission rate of ent-kaurene (Fig. 5b), include the genus Taphrina. All known Taphrina species are plant pathogens⁷⁵ and have various functional genes required for plant infection by these fungi, such as genes encoding enzymes involved in plant hormone biosynthesis or degradation of plant cell walls⁷⁶. The positive correlations between the relative abundances of these microbes and the emissions of some terpenes may indicate microbial adaptation to the phyllosphere environment of Japanese cedar. However, our study did not test the causality of the BVOC–microbe relationships. Terpene application to the phyllosphere of cedar trees that emit a small amount of terpene (e.g., trees from the Yakushima population) will verify the influence of terpene emission on the shifts in microbial composition. Inoculation assays using cultured microbial strains will reveal the microbial abilities to tolerate the phyllosphere environment of Japanese cedar. Terpene application assays on cultured microbes will clarify the functionalities of both terpenes and microbes in their relationships. Further investigation is needed to identify the mechanisms behind the BVOC-mediated association of phyllosphere microbes with Japanese cedar.

We demonstrated that common garden was the primary factor influencing microbial composition in the phyllosphere of Japanese cedar (Fig. 2). Multiple factors may contribute to the observed differences in microbial composition between the two gardens. Geographic distance may affect the community assembly of local microbes, particularly those with limited dispersal capabilities^21,67. Fungi have larger spores⁷⁷ and are more limited in dispersal than bacteria^67,78,79 which may explain the greater effect of common garden on the fungal composition in this study. The two gardens differed in their local environments since they are located at different latitudes, and the surrounding land use is also very different: the Kawatabi common garden is surrounded by a forest, whereas the Tsukuba common garden is surrounded by farmland and an urban area. These differences in the local environment, such as climate^17,18 and land use^21,80 may also have influenced the microbial composition in the two gardens.

Although only three populations were used, our study suggested a negative correlation between warm, humid, and less snowy climates at the origin sites of the populations and the basal emission rates of most MTs and β-farnesene (Supplementary Fig. S4). Hiura et al.⁴⁶ suggested that the amounts of stored, but not emitted, MTs and the compositional ratio of emitted β-farnesene in Japanese cedar are negatively correlated with warm and less snowy climates, suggesting the existence of climate clines. The partial pressure calculated from the stored amounts of MTs in Japanese cedar was suggested to be positively correlated with their emission rates from the same leaf⁷⁴. Therefore, the similar pattern of variations among Japanese cedar populations in stored and emitted MTs may be attributed to the emission of MTs from storage pools in leaves. However, it is possible that the three populations used in this study may not fully represent the total variation in the climate at the origin sites of the cedar natural populations. Future studies should include additional populations, particularly those on the Pacific Ocean side, to understand the impact of local climates on the development of the BVOC–microbe association. We expect that consideration of both abiotic and biotic environmental conditions will further our understanding of the mechanisms of BVOC emission in plants in natural environments.

Climate change may influence the interpretation of our results. Considering the faster rate of temperature increase at high altitudes⁸¹ and latitudes⁸² differences in today’s annual mean temperature among the origin sites of the cedar populations may be smaller than those in the data we used. Using recent climatic data may result in a more unclear climatic trend in both microbial abundance and BVOC emissions. Nevertheless, the relative warmth of the origin sites of the cedar populations to each other would remain largely unchanged, and similar trends to those observed in our results would be detected if we used more recent climatic data.

In conclusion, our findings revealed differences in the phyllosphere microbial community compositions of Japanese cedar both between common gardens and among wild populations, as well as the correlation between the compositions and basal emission rates of some terpenes by host trees. The results also suggest that the local climate at the origin sites of the Japanese cedar populations is a significant factor influencing variations in the phyllosphere microbial communities and terpene emission rates in this species. This study is among the few to demonstrate in outdoor settings the relationship between the emission patterns of BVOCs and both abiotic and biotic environments in trees with substantial BVOC emissions. Further investigation is required to elucidate the mechanisms of plant–microbe associations mediated by BVOC emission and the impact of biotic associations on plant–atmosphere feedbacks.

Methods

Sampling sites and collection of phyllosphere Microbiome

We collected leaves at a height of about 1.5 m from cedar trees (about 5–6 m tall) on 2 June 2023 in the common garden at Kawatabi Field Science Center, Tohoku University (38.78°N, 140.73°E) and on 26 June 2023 in the common garden in the Tsukuba Experimental Forest, University of Tsukuba Mountain Science Center (36.11°N, 140.10°E). Wild populations of Japanese cedar are widely distributed in Japan, ranging from Yakushima Island, the southern limit of their distribution, to the northern limit of Ajigasawa (Aomori Prefecture)⁸³. Trees from 14 wild populations were grown from cuttings in the two common gardens (Supplementary Fig. S1 and Table S4). Leaves were collected from 3 to 10 individuals per population (n = 108: Tsukuba = 50, Kawatabi = 58). Using scissors surface sterilized with 99% ethanol, leaves (ca. 5 g per tree) were collected in plastic bags, placed in cool bags with ice packs, brought back to the laboratory within 1 day of collection, and stored at 4 °C for one to two days.

Extraction of microbial DNA and amplicon sequencing

DNA of phyllosphere microbes was extracted using the protocol of Tian et al.⁸⁴ with modifications. Collected leaves were soaked in 40 mL of buffer (0.1 M potassium phosphate buffer with 0.1% Tween-80, pH 7.0) and microbes were separated from the leaf surface by sonication for 10 min in a Bransonic ultrasonic bath (Branson Ultrasonics, Danbury, CT, USA). Microbes were precipitated by centrifugation (10,000 ×g, 15 min) in a high-speed refrigerated microcentrifuge (Tomy, Nerima, Tokyo, Japan), suspended in 100 µL of PBS and stored at − 20 °C. DNA was extracted with a Fast DNA Spin Kit for Soil (MP Biomedicals, Irvine, CA, USA) following the manufacturer’s protocol.

Bacterial and fungal DNA was amplified using Ex Taq Hot Start Version (Takara Bio, Kusatsu, Shiga, Japan). The V4 region of bacterial 16 S rRNA gene was amplified with the 515 F (5′-GTGYCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACNVGGGTWTCTAAT-3′) primer set⁸⁵. The PCR mixture (20 µL) contained 1× Ex Taq buffer, 0.2 mM dNTP mixture, 0.05 U Ex Taq HS (Takara Bio, Kusatsu, Shiga, Japan), 0.5 µM forward and reverse primers, 1 µM chloroplast and mitochondria peptide–nucleic acid clumps (to minimize the amplification of chloroplast and mitochondrial 16 S rRNA genes from host plants⁸⁶, and 1 µL of the template. PCR was performed as follows: 45 s at 95 °C, 28 cycles of 15 s at 95 °C, 10 s at 78 °C, 30 s at 50 °C, 30 s at 72 °C, and 5 min at 72 °C. The ITS2 region of fungal DNA was amplified with the fITS9 (5′-GAACACAGCGAAATGTGA-3′) and ITS4ngsUni (5′-CCTSCSCTTANTDATATGC-3′) primer set^87,88. The PCR mixture was as described above except that peptide–nucleic acid clumps were omitted. PCR was performed as above except that the cycles at 78 °C were omitted. The second PCR was then performed to add adapter sequences to amplicons. The PCR mixture (20 µL) contained 1× Ex Taq buffer, 0.2 mM dNTP Mixture, 0.05 U Ex Taq HS, 0.5 µM forward and reverse adapters, and 14 µL amplicons from the first PCR. PCR was performed as follows: 45 s at 95 °C, 8 cycles of 15 s at 95 °C, 30 s at 50 °C, 30 s at 72 °C, and 5 min at 72 °C. Amplicons were purified using NucleoMag NGS Clean-up and Size Select magnetic beads (Macherey-Nagel, Dueren, Germany) following the manufacturer’s protocol. DNA concentration of each sample was measured using Qubit dsDNA Quantification Assay Kits (Thermo Fisher Scientific, Waltham, MA, USA), and an approximately equimolar pool of amplicons was constructed. This pool was commercially sequenced using the Illumina MiSeq platform with the MiSeq Reagent Kit v3 for 2 × 300 bp paired-end reads in a single run (Macrogen, Tokyo, Japan).

Amplicon sequence processing, ASV classification, and taxonomic assignment

Amplicon sequencing yielded 15,561,772 total raw reads. Sequencing data were analyzed in QIIME 2 v2023.9⁸⁹. Primer sequences of the V4 regions of the 16 S rRNA gene and ITS2 region were removed from the 5′ ends of both forward and reverse reads, resulting in a total of 8,467,176 bacterial and 6,799,366 fungal trimmed reads. Quality filtering, chimera removal, and inference of ASVs were performed using the DADA2 pipeline⁹⁰. Bacterial ASVs were inferred from paired-end sequences and fungal ASVs were inferred from single-end sequences, yielding 7,073,527 bacterial and 6,252,014 fungal denoised reads. Taxonomy was assigned to each bacterial ASV using a naïve Bayes classifier^91,92 based on the SILVA database⁹³. The ASVs assigned to chloroplasts or mitochondria and ASVs for which no taxa were assigned at the kingdom level were excluded. Taxonomy was assigned to each fungal ASV using BLAST +⁹⁴ based on the UNITE database⁹⁵. The ASVs assigned to taxa other than fungi at the kingdom level and ASVs that were not assigned to any taxa at the kingdom level were excluded. A total of 3,465,682 filtered bacterial reads (32,090 ± 13,829 reads per sample, mean ± SD) and 5,547,104 filtered fungal reads (51,362 ± 12946 reads per sample, mean ± SD) were used in subsequent analyses. The total number of bacterial ASVs was 2,933 and that of fungal ASVs was 5,040 after the filtering.

Collection of terpenes emitted by Japanese Cedar

We collected BVOCs emitted by cedar trees grown in the common garden at Kawatabi Field Center in June 2023; the trees originated from three populations: Ajigasawa (n = 9), Azouji (n = 10), and Yakushima (n = 6). The dynamic branch enclosure system⁹⁶ was used for BVOC collection according to Hiura et al.⁴⁶ except that ambient air was used as the purge gas instead of compressed air from gas cylinders. Briefly, intact branches were enclosed in a Teflon bag with a purge air inlet that supplied ambient air, from which volatile organic compounds and ozone had been removed by charcoal filters. Two air samples (2 L each) per enclosure were collected simultaneously in adsorbent tubes: one for MT measurements and the other one for SQTs and DTs. We followed Hiura et al.⁴⁶ for the analysis of the collected terpenes. Briefly, MT samples were analyzed using a custom-built thermal desorption unit coupled to a gas chromatograph (GC) equipped with a mass selective detector (MSD) and a flame ionization detector (FID) (Agilent 6890/5973, Agilent, Santa Clara, US). SQT and DT samples were analyzed using a GC/MSD (Agilent 6890/5973, Agilent, Santa Clara, US) following solvent extraction and concentration. The limit of detection (LOD) for each compound was determined using a weighted linear regression for the calibration curve. The regression was weighted by the reciprocal of the square of the concentration following the method described by Gu et al.⁹⁷. The LOD values are listed in Supplementary Table S3.

To standardize emission rates, we applied the G93 model⁴³ and converted the measured emission rates to basal emission rates, which were defined as the emission rates at a standard leaf temperature of 30 °C, using the values of the constant β (MTs, 0.17; SQTs, 0.20; DTs, 0.21) for models of the rates of terpene emission by Japanese cedar reported previously^44,98. For statistical analysis, the basal emission rates were common logarithm-transformed after adding 1 (to transform < LOD values to 0) to avoid zero or negative values. Data of emitted terpenes are shown in Supplementary Table S2.

Acquisition and analysis of climatic data at the origin sites of the wild Japanese cedar populations

The climate data at the origin sites of the 14 wild populations of Japanese cedar for 1981–2010 were obtained from the CHELSA V2.1 database^56–59. The data cover part of the period during which the source trees of the cedar clones were growing at the origin sites of the cedar populations. Therefore, we assumed that the data reflect differences in their growth environments. The following data were obtained: annual mean temperature (°C, bio1), standard deviation of the monthly mean temperatures (10^− 2 °C, bio4), annual precipitation amount (kg m^− 2 year^− 1, bio12), coefficient of variation of the monthly precipitation estimates (kg m^− 2, bio15), mean monthly near-surface relative humidity (%, hurs_mean), mean monthly potential evaporation (kg m^− 2 month^− 1, pet_penman_mean), monthly mean temperatures (°C, tas01–12), and monthly precipitation amount (kg m^− 2 month^− 1, pr01–12). The sum of the monthly precipitation for the months with mean temperatures below 0 °C was calculated for each location and used as an index of snowfall (kg m^− 2, snow). This index and the six average climatic variables other than the monthly means were summarized into PCs by PCA using the prcomp function in R v4.3.3⁹⁹ (Supplementary Fig. S2).

Statistical analysis

Data analysis and visualization were performed using the tidyverse¹⁰⁰ phyloseq¹⁰¹ vegan¹⁰² MASS¹⁰³ and ggplot2¹⁰⁴ packages in R v4.3.3⁹⁹. Using the vegan¹⁰² package, the ASV table was rarefied to the minimum reads per sample to ensure equal sequencing depth (bacteria: 9,099 reads, fungi: 17,427 reads). After rarefaction, the total number of bacterial ASVs was 2,871 and that of fungal ASVs was 4,865. The ASV richness of bacteria and fungi was calculated using the specnumber function in the vegan¹⁰² package. Bray–Curtis dissimilarities among the microbial communities were calculated to measure their beta diversity using the vegan¹⁰² package. Nonmetric multidimensional scaling (NMDS) based on Bray–Curtis dissimilarities was performed to visualize beta diversity (Fig. 2). A permutational multivariate analysis of variance (PERMANOVA) with the Bray–Curtis dissimilarities as the response variable was performed to test for differences in microbial community composition between common gardens and among populations, which were included together with their interactions as explanatory variables in the test. PERMANOVA was also performed using the climatic PC1 and PC2 at the origin sites of host individuals as explanatory variables instead of their populations to test for correlation between microbial community composition and climate at the origin sites of host individuals. The list of fungal species that have been reported to cause disease in Japanese cedar throughout Japan was obtained from Kobayashi⁶⁰. Fungal ASVs of the same genus as these species in the list were selected as the fungal genera that may be associated with Japanese cedar. The functional groups of these fungal genera were also predicted using the FungalTraits database⁶¹ to validate their functionality in association with plants. To explore the drivers of the total relative abundances of these fungi, a negative binomial generalized linear model was fitted with the aggregated abundances of these fungi after rarefaction as the response variable using the glm.nb function with a log link function in the MASS¹⁰³ package (Fig. 3). Common garden and the climatic PC1 and PC2 at the origin sites of host individuals were included in the model as explanatory variables. The significance of the explanatory variables was tested by the Wald test.

To analyze the terpene emission profiles of Japanese cedar, we performed PCA of the basal emission rates of terpenes after log transformation using the prcomp function (Fig. 4b). To investigate the relationship between the terpene emission profile and the local climate at the origin sites of the cedar trees, we fitted linear regression models with the terpene PCs as the response variable and the climatic PCs at the host origin sites as the explanatory variables, resulting in 3 models (Supplementary Fig. S4). We performed variable selection on the 3 models using the stepAIC function in the MASS¹⁰³ package to address the multicollinearity between the climatic PCs for the 3 populations included in the analysis. The significance of the explanatory variables in each model was tested by Student’s t-test.

The following procedure was used to analyze the relationship between the basal terpene emission rate and phyllosphere microbial community composition (Fig. 5). First, to avoid a large distortion of the entire NMDS plot due to the microbial community outliers, we detected them using CLOUD analysis¹⁰⁵ a non-parametric test that is based on a distance matrix between microbial communities and detects microbial samples whose mean distance to neighbors is significantly different from those of other samples. Outliers in the bacterial and fungal communities were detected using the mean Bray–Curtis dissimilarity to five neighbors. The only detected outlier (the fungal community of one tree from the Azouji population) was excluded from the subsequent analysis. Next, NMDS based on the Bray–Curtis dissimilarities among the microbial communities was performed only for those individuals from which emitted terpenes were collected. Correlation was then tested between the NMDS coordinates constructed from the phyllosphere microbial community compositions and the basal emission rates of terpenes by vector fitting in the vegan¹⁰² package. The basal emission rate of each terpene, their total sum, or the sum for each category (i.e., MT, SQT or DT) was used for the vector fitting after log transformation. Correlation was also tested between the abundances of bacterial phyla or fungal classes and the NMDS coordinates constructed from the community compositions by vector fitting.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

S.I. performed statistical analysis and wrote the first draft; T.I.K and T.H. generated hypotheses and designed research; S.I. and T.I.K. collected and performed experiments with microbial samples; Y.O. sampled terpenes emitted from the Japanese cedars; Y.O. and T.S. analyzed the collected terpenes; Y.S. and Y.T. designed and maintained the common gardens. All authors participated in manuscript preparation.

Data availability

The datasets generated and analysed during the current study are available in the DDBJ Sequence Read Archive under the accession number PRJDB20421. All R scripts used for the data analysis are available on GitHub (https://github.com/Satoyoshi-ISHIZAKI/CedarPhyllosphere_SciRepo.git).

Declarations

Competing interests

The authors declare no competing interests.