Black poplar (P. nigra L.) trees were grown from hardwood cuttings (photosynthesis experiment and isoprene experiment) or softwood cuttings (transcriptome experiment) obtained from different tree genotypes growing in a common garden in Isserstedt, Germany (50°57′28.5″N 11°31′17.4″E). One genotype was used for the transcriptome experiment as well as for the photosynthesis and isoprene measurements, whereas two different genotypes were additionally used for photosynthesis measurements. The cuttings were potted in 2 l- pots, grown in the greenhouse (18/20°C, night/day, relative humidity 60%, natural light with 9–14 h photoperiod, supplement light for 12 h, SON-T Agro; Philips, Andover, MA, United States) and transferred to an environmental chamber [18/20°C, night/day; relative humidity 60%; photoperiod 16 h, MT 400 (Eye, Uxbridge, United Kingdom)] 2 days before the onset of the experiment. All three experiments (photosynthesis experiment, isoprene experiment and transcriptome experiment) were conducted separately at different time points but under the same temperature and humidity conditions. All trees were used ca. 4 months after potting and had reached a height of about 0.5 m. Trees used for photosynthesis and isoprene measurement did not show any noticeable shoot growth at time of experiment.

Uredospores of the biotrophic poplar leaf rust fungus (M. larici-populina Kleb.) were obtained from naturally infected black poplar trees growing in the above-mentioned common garden. The identity of the fungus was verified by using specific primers for the internal transcribed spacer region of M. larici-populina as described in Eberl et al. (2018). The pathogen was amplified by infecting 1-year-old trees, and after 2–3 weeks uredospores were harvested with a scalpel and a brush. Spores were stored at -20°C until the start of the experiment either dried over silica overnight (photosynthesis experiment and transcriptome experiment) or used within 2 months after harvesting (isoprene experiment). Plants were inoculated with the fungus by spraying a mixture of water and spores (dry: 1 mg ml^-1 and fresh: 1.5 mg ml^-1) on the abaxial side of each leaf (approximately 1 ml per leaf) and covering each tree with a polyethylene terephthalate (PET) bag (Bratschlauch, Toppits, Minden, Germany), which was kept closed for 1 day to ensure sufficient humidity for spore germination. Control groups received the same treatments but were sprayed with water only. First sporangia in the rust-infected trees were visible on the abaxial side of the leaves at 7 dpi (days post-infection) (Supplementary Figure S1), which matches the time course of infection in the literature (Hacquard et al., 2011). Under natural conditions individual poplar leaves are usually exposed to many cycles of rust infection. We decided to investigate only a single cycle of infection to better determine which stage of infection influences plant processes.

Photosynthetic parameters were measured on the second mature leaf (counting from the apex) of six trees from both groups (“control” and “rust-infected”; n = 6) at six different time points: 1 day before rust infection (-1 dpi), 4 h post-infection (hpi), 1, 7, and 10 dpi. Due to unexpectedly high CO₂-levels in the air supply (Supplementary Figure S3) at one measurement time (4 dpi), we excluded this data point from further analysis. All measurements were conducted at the same time of the day (9 am–12 pm) except for 4 hpi (1 pm–4 pm). The leaf was put into a custom-made single leaf chamber [chamber: polyoxymethylene, lid: poly(methyl methacrylate), openings sealed with sponge rubber; for picture see Supplementary Figure S2], which was connected to a LI-6400XT Portable Photosynthesis System (LI-COR, Lincoln, NE, United States). For each treatment a separate chamber was used to avoid contamination by fungal spores of the control leaves. The air supply for the LI-6400XT was filtered through active charcoal and humidified to 12%. An LED lamp (850 PAR at leaf position; 5 W, warm white and cool white; Roschwege GmbH, Greifenstein, Germany; for spectrum see Supplementary Figure S4) was placed over the leaf chamber as light source. The two different treatments were measured alternatingly to avoid temporal effects for one of the groups. The leaf was allowed to equilibrate to the LED light for 10 min before it was connected to the LI-6400XT, and photosynthetic parameters (photosynthetic rate, stomatal conductance and intercellular CO₂) were measured for 10 min. The photosynthetic rate and stomatal conductance were normalized to the leaf area, which was determined with Photoshop CS5 (Adobe, San Jose, CA, United States) from a picture taken at the end of the experiment with a reference field of known size. As the trees were not actively growing during the experiment, the leaf size was assumed to be the same throughout the whole experiment. After the last measurement, the leaf was flash-frozen in liquid nitrogen for phytohormone and sugar analysis.

Phytohormones and sugars were analyzed from 10 mg freeze-dried, ground leaf material of the photosynthesis experiment (n = 6). Phytohormone analysis was carried out on an LC/MS/MS system as previously described (Eberl et al., 2018). Data were processed using ANALYST 1.5.2 (AB Sciex, Framingham, MA, United States) and hormones were quantified relative to the peak area of their corresponding standard (D₄-salicylic acid and D₆-abscisic acid; Santa Cruz Biotechnology, Dallas, TX, United States). For sugar analysis, extracts were diluted 1:10 with water prior to analysis on an Agilent 1200 HPLC system (Agilent, Santa Clara, CA, United States) coupled to an API 3200 tandem mass spectrometer (AB Sciex). The analytes were separated on an hydrophobic interaction liquid chromatography (HILIC)-column (apHera NH₂ Polymer; 15 × 4.6 mm, 5 μm; Supelco, Bellefonte, PA, United States) with a water/acetonitrile gradient (flow, 1.0 ml min^-1), for more details see Madsen et al. (2015). The data were processed using ANALYST 1.5.2 (AB Sciex) and the compounds were quantified using an external standard curve. For this, a mixture of glucose, fructose and sucrose (Sigma-Aldrich, St. Louis, MO, United States) was analyzed at six different concentrations ranging from 20 to 1.25 μg ml^-1.

Isoprene emission from rust-infected and control trees was measured from the second mature leaf from the apex in six trees of each treatment (n = 6). The same conditions, setup and time points (from 1 to 10 dpi) as for the photosynthesis experiment were used, but the material of the single leaf chamber was changed to aluminum (Supplementary Figure S2) to avoid volatile contamination, while the lid was still made from poly(methyl methacrylate). The ambient temperature of 20°C used in our experiments is lower than the temperature applied in comparable studies on isoprene emission (e.g., Brilli et al., 2007) but better simulates the conditions that trees of this poplar species and the pathogen experience in the field. Furthermore, the relative humidity was reduced to 6% in order to avoid excessive transpiration inside the chamber (due to a larger leaf size compared to the photosynthesis experiment). Isoprene emission was analyzed with a proton transfer reaction mass spectrometer (PTR-MS; Ionicon Analytik, Innsbruck, Austria). A detailed description of the PTR-MS can be found in Lindinger and Jordan (1998). Before starting the analysis the PTR-MS was calibrated by using the Gas Calibration Unit, a system generating clean air mixed with precise flows of an isoprene gas standard (Ionicon Analytik). The capillary line of the PTR-MS was connected with the outflow of the leaf chamber. The proton transfer reactions occurred in the reaction chamber (drift tube) between the primary ion H₃O⁺ coming from the ion source, and the isoprene in the sampled air. In the drift tube the pressure was in the range of 2.3 mbar and the E/N ratio (electric field/particle density) was 137 Td (1 Td = 10^-17 V cm²). Isoprene was monitored at the mass signal 69 (m/z). The raw count-rate signal intensity () of the isoprene was normalized (ncps) to the cps sum of the primary ion and water cluster, and to the drift tube pressure. Each leaf was given a 10 min equilibration period in the leaf chamber under LED light. Then the leaf chamber was connected to the PTR-MS to monitor isoprene emission for 15 min. This time was enough to reach steady-state conditions of isoprene emission. The average of cps during the steady-state period was used to calculate the emission rate, from which the background (empty leaf chamber, Supplementary Figure S2) was subtracted. Isoprene emission was normalized to the leaf area, which was determined by a picture taken at the beginning of the experiment. As the trees were not actively growing during the experiment, the leaf size was assumed to be the same throughout the whole experiment. Leaf area was calculated as described in “Photosynthesis measurements.” After the last measurement (10 dpi), the second mature leaf was flash-frozen immediately in liquid nitrogen for metabolite analysis.

Leaves sampled from the isoprene experiment (n = 6) were ground in liquid nitrogen and then lyophilized. The MEP pathway metabolites (see Figure 5 legend) were extracted twice with a 250 μl solution of 50% acetonitrile containing 10 mM ammonium acetate, pH 9.0, using 5 mg dry tissue. After vortexing and micro-centrifugation, 200 μl of the supernatant from both extracts were combined, transferred into a new 1.5 ml tube and dried under a stream of nitrogen gas at 40°C. The residue was dissolved in 100 μl of 10 mM ammonium acetate, pH 9.0, and, after vortexing, 100 μl of chloroform was added. The upper aqueous phase, separated by centrifugation, was transferred into a new tube and diluted with 1 volume of acetonitrile. After centrifugation for 5 min to remove any precipitate, the supernatant was transferred to an HPLC vial. MEP pathway metabolites were analyzed on an Agilent 1260 Infinity HPLC system (Agilent) connected to an API 5000 triple quadrupole mass spectrometer (AB Sciex). A 5 μl portion of the extract was injected and the metabolites were separated on a HILIC XBridge Amide column (150 × 2.1 mm, 3.5 μm; Waters, Milford, MA, United States) with a HILIC guard column containing the same sorbent (3.5 μm, 10 × 2.1 mm) and a SSITM high pressure pre-column filter (Sigma-Aldrich) using two solvents: 20 mM ammonium bicarbonate adjusted to pH 10.5 with ammonium hydroxide (solvent A) and 80% acetonitrile containing 20 mM ammonium bicarbonate, pH 10.5 (solvent B). The solvent gradient profile started with 100% of solvent B which decreased to 60% in the first 15 min, followed by an isocratic gradient with solvent B. Separation was performed at 25°C with a flow rate of 500 μl min^-1. The mass spectrometer operated in negative ionization mode with ion spray voltage -4500 eV, turbo gas temperature 700°C and nebulizer gas 70 psi. The MEP pathway metabolites were analyzed using the MRM conditions described by Wright et al. (2014). The metabolite concentrations were calculated by using external standard curves, and were normalized to the [¹³C]-labeled internal standards of each intermediate (González-Cabanelas et al., 2016) added to the extract after the first extraction step.

Leaves sampled from the isoprene experiment at 10 dpi were ground in liquid nitrogen and 50 mg of fresh tissue was extracted in light-protected tubes with 1 ml of acetone by shaking for 6 h at 4°C in the dark. After centrifugation for 5 min at 2350 g at 4°C, 800 μl of the extract was transferred into a new light-protected tube and 200 μl of water was added. After spinning the samples for 1 min at 5000 rpm at 4°C, they were transferred to brown glass vials for analysis on an Agilent 1100 Series HPLC with UV/VIS diode array detector. The detector was set at 445 nm for the detection of carotenoids and at 650 nm for the chlorophylls. The pigments were separated on a Supelcosil column LC-18 (7.5 cm × 4.6 mm × 3 μm; Sigma-Aldrich) using an acetone (solvent A)/1 mM NaHCO₃ (in water, solvent B) gradient with a flow rate of 1.5 ml min^-1. The initial mobile phase consisted of 65/35% (v/v) solvent A/solvent B. Then, solvent A was linearly increased to 90% within 12 min and to 100% over 8 min. 100% solvent A was kept for 2 min and then decreased to 65% again within 3 min. Quantification was done using external standard curves. Authentic standards of the chlorophylls and β-carotene (Santa Cruz Biotechnology) were analyzed in a range from 0.1 to 0.00625 mg ml^-1. Lutein, neoxanthin, and violaxanthin were assumed to have the same response factor as β-carotene.

To investigate transcriptional changes in black poplar leaves upon rust infection, eight trees grown from green cuttings were selected and half of them were inoculated with M. larici-populina uredospores. Leaves from the trees of both treatments (“control” and “rust-infected”; n = 4) were harvested 8 dpi and flash-frozen in liquid nitrogen. RNA was isolated with the InviTrap Spin Plant Mini Kit (Stratec Biomedical AG, Birkenfeld, Germany), including a DNase digestion (DNase kit, Qiagen, Hilden, Germany). RNA concentration and quality was analyzed with a NanoDrop 2000c spectrophotometer (Peqlab Biotechnologie GmbH, Erlangen, Germany) and the RNA 6000 Nano Kit on a Bioanalyzer (Agilent). Sequencing was done at the Max Planck-Genome-Centre (Köln, Germany) on a HiSeq 2500 (Illumina, San Diego, CA, United States) with 9 Mio reads per sample. Quality control measures, including the filtering of high-quality reads based on fastq file scores, the removal of reads containing primer/adapter sequences, and trimming of the read length, were carried out using CLC Genomics Workbench v9.1³. The same software was used for de novo transcriptome assembly, combining two replicates of each RNA-Seq treatment group, and selecting the presumed optimal consensus transcriptome as previously described (Vogel et al., 2014). The final de novo reference transcriptome assembly (backbone) of P. nigra contained 81,580 contigs (sets of overlapping sequence segments that together represent a continuous region of the original RNA). Minimum contig size was 250 bp with an N50 contig size of 1320 bp. The transcriptome was annotated using BLAST, Gene Ontology (GO) and InterPro terms (InterProScan, EBI), enzyme classification (EC) codes, and metabolic pathways (Kyoto Encyclopedia of Genes and Genomes, KEGG) as implemented in BLAST2GO v4.1⁴. Based on the BLAST hits, the contigs were designated as being of either plant or fungal (i.e., M. larici-populina) origin. To assess transcriptome completeness, we performed a BUSCO⁵ (Benchmarking Universal Single-Copy Orthologs) analysis by comparing our assembled (plant-derived only) transcript set against a set of highly conserved single-copy orthologs. This was accomplished using the BUSCO v3 pipeline (Waterhouse et al., 2017) compared to the predefined set of 303 Eukaryota single-copy orthologs from the OrthoDB v9.1 database. Our assembled transcriptome was determined to be 87.8% complete and only 3.6% of the BUSCO genes were missing. Digital gene expression analysis was carried out using CLC Genomics Workbench v9.1 to generate BAM (mapping) files, and QSeq Software (DNAStar Inc., Madison, WI, United States) was then used to estimate expression levels. The log₂ (RPKM) values (normalized mapped read values; geometric means of the biological replicate samples) were subsequently used to calculate fold-change values. To identify differentially expressed genes, we used the Student’s t-test (as implemented in Qseq) corrected for multiple testing using the Benjamini–Hochberg procedure to check the false discovery rate (FDR). A gene was considered significantly differentially expressed if the FDR-corrected p-value was less than 0.05. Fisher’s exact test was used as part of BLAST2GO to identify the overrepresentation of GO terms among lists of differentially expressed genes between treatment groups. The GO-enriched bar charts were simplified to display only the most specific GO terms by removing parent terms representing existing child terms using the function “Reduce to most specific terms” in BLAST2GO. A GO term was considered significantly enriched if the p-value corrected by FDR control was less than 0.05.

All data were tested for statistical assumptions, i.e., normal distribution and homogeneity of variances. Whenever necessary, the data were log-transformed (salicylic acid, DMADP + IDP). For the photosynthetic parameters and the isoprene emission a two-way repeated measures ANOVA was performed using “time” as a within-subject factor and “rust infection” as a between-subject factor. For end-point measurements (phytohormones, sugars, MEP pathway metabolites and carotenoids) an independent student’s t-test was performed for each compound. Correlations between photosynthetic parameters and phytohormones were tested using bivariate Pearson’s r correlation. All statistical analyses were performed with SPSS 17.0 (SPSS, Chicago, IL, United States).

In order to investigate the influence of fungal infection on photosynthesis in black poplar, we measured photosynthetic parameters at various times during the development of rust infection in the leaves.

The photosynthetic assimilation rate in uninfected control trees was generally stable at 9–12 μmol CO₂ m^-2 s^-1 during the time course of measurements (Figure 1A). The photosynthetic rate in rust-infected poplar trees, however, dropped by approximately 50% within the first 4 hpi from 12.1 ± 1.32 to 6.1 ± 0.86 μmol m^-2 s^-1. This reduction persisted throughout the later time points. The factor “rust infection” strongly influenced the photosynthetic rate, as also shown statistically [repeated measures ANOVA: F_(1,10) = 15.524, P = 0.003]. However, there was also a significant interaction of time and rust infection [repeated measures ANOVA: F_(4,40) = 8.891, P < 0.001]. The patterns of stomatal conductance were similar to those observed for photosynthetic assimilation rate. Control plants without rust infection showed a slight increase in stomatal conductance throughout the experiment from 0.09 ± 0.003 mol m^-2 s^-1 at the first measurement to 0.13 ± 0.012 mol m^-2 s^-1 during the last three measurements (Figure 1B). Rust fungus infection decreased stomatal conductance in poplar trees by more than 50% as early as 4 hpi, when it declined from 0.11 ± 0.015 to 0.06 ± 0.005 mol s^-1 m^-2. As the infection progressed, the stomatal conductance of rust-infected trees remained at around 50% of that measured from uninfected controls. The effect of “rust infection” was statistically highly significant [repeated measures ANOVA: F_(1,10) = 19.815, P = 0.001], as was the interaction of time and rust infection [repeated measures ANOVA: F_{(4, 40)} = 10.665, P < 0.001]. However, the intercellular CO₂ levels did not differ between the different time points, nor between the two treatments, i.e., rust-infected and control black poplar leaves (Supplementary Figure S5).

After observing the reduction in photosynthetic assimilation rate in rust-infected poplar leaves, we analyzed glucose, fructose and sucrose levels in leaves harvested at 10 dpi. For these sugars, no significant changes in concentration were observed after infection of leaves by the rust fungus (Table 1).

Our results show that photosynthetic activity is downregulated in poplar leaves right after the onset of infection with the rust fungus, and remains lower over the course of infection without affecting soluble sugars.

To analyze phytohormones that might be involved in regulating anti-pathogen defense and photosynthesis, leaves were collected at 10 dpi.

Salicylic acid increased fivefold in rust-infected black poplar leaves compared to uninfected controls [331 ± 45 ng g^-1 DW in controls; 1580 ± 390 ng g^-1 DW in rust-infected trees, Figure 2A], which is highly significant [Student’s t-test, t₍₁₀₎ = -4.687, P = 0.001]. ABA, a phytohormone regulating stomatal opening, among other processes, also increased in rust-infected black poplar leaves, reaching a level twice as high as in the control leaves (36.2 ± 6.10 ng g^-1 DW in controls; 72.1 ± 11.34 ng g^-1 DW in rust-infected trees, Figure 2B). Additionally, ABA levels in rust-infected leaves correlated negatively with stomatal conductance and photosynthetic rate measured at 10 dpi (Table 2). In control leaves, however, no such correlation could be observed.

Taken together, we could show that rust infection increased SA as well as ABA in poplar leaves, whereas the latter shows a negative relation to photosynthetic parameters.

To evaluate the effect of pathogen infection on isoprene emission, we monitored leaves of infected and uninfected young poplar trees using a PTR-MS at different time points before and during infection with the rust fungus (M. larici-populina).

The isoprene emission from uninfected controls showed slight variations throughout the time course of the experiment, ranging from 2.7 to 3.4 nmol m^-2 s^-1. Infection with the rust did not significantly change emission of isoprene compared to control trees or compared to the levels measured before the pathogen inoculation (Figure 3). The fluctuations over time were similar to those in uninfected controls, leading to a significant statistical effect of “time” [repeated measures ANOVA: F_(4,40) = 1.475, P = 0.031]. However, there was no significant effect of “infection” or the interaction “time × infection” on isoprene emission. This shows that rust infection does not affect the emission of isoprene from black poplar trees.

We analyzed intermediates of the MEP pathway and levels of the photosynthetic pigments, chlorophylls and carotenoids. Like isoprene, these are also produced from DMADP and IDP originating from the MEP pathway. The analyzed leaves were sampled together with any fungal spores and mycelium present in infected leaves. The MEP pathway intermediates, DXP, MEP, CDP-ME, and MEcDP, were present at similar levels in rust-infected and control leaves (Figures 4A–D). However, the amount of DMADP and IDP (DMADP + IDP), the final products of both the MEP and MVA pathways (not separable in our LC-MS analysis), were significantly higher in rust-infected leaves [Student’s t-test, t₍₁₀₎ = -3.503, P = 0.006] (Figure 4E). The concentration of DMADP + IDP in rust-infected leaves was more than double that of control leaves (4.1 ± 0.46 nmol g^-1 FW in controls; 9.3 ± 1.60 nmol g^-1 FW in rust-infected trees). On the other hand, many of the major isoprenoids that are known to be produced from MEP pathway-derived C₅ units, including the carotenoids, lutein, neoxanthin, violaxanthin, and chlorophylls a and b, did not change after rust infection (Table 3). β-Carotene, however, increased by more than 50% in leaves from rust-infected poplars compared to uninfected controls (0.59 ± 0.072 mg g^-1 FW in controls; 0.93 ± 0.090 mg g^-1 FW in rust-infected; Student’s t-test, t₍₁₀₎ = -3.100, P = 0.011). The orange-colored uredospores of the rust fungus M. larici-populina were also analyzed separately to determine if these contained any carotenoids, and β-carotene, but none of the other poplar carotenoids, was found (Supplementary Table S1). The concentration of β-carotene was more than threefold higher in fungal spores than in uninfected poplar leaves.

DMADP and IDP, the C₅ building blocks of all isoprenoids, can be produced by either the plastidic MEP pathway or the cytosolic MVA pathway (Rodríguez-Concepción and Boronat, 2015). To determine whether rust infection could influence either pathway, we carried out transcriptome analysis of uninfected and rust-infected black poplar trees. No differential expression of genes involved in the MEP pathway was observed (Figure 5). However, genes encoding biosynthetic enzymes of the MVA pathway increased in expression in rust-infected leaves compared to uninfected controls (Figure 5). Two enzymes catalyzing early steps of the MVA pathway, acetoacetyl-coenzyme A thiolase and 3-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) synthase, had especially higher RPKM values (reads per kilobase of transcript per million mapped reads) in leaves infected with the rust fungus. The transcriptome analysis also identified MVA pathway transcripts of fungal origin, likely from M. larici-populina. Contigs annotated as fungal HMG-CoA synthase and all downstream MVA pathway enzymes were detected in the transcriptome of rust-infected leaves (Supplementary Table S2), but not in control leaves.

Thus rust infection did not affect the levels of intermediates of the MEP pathway. However, the amounts of the isoprenoid building blocks DMADP + IDP increased, as a result of up-regulation of the MVA pathway. In addition, there was an increase in β-carotene, which likely arose from the isoprenoid biosynthesis by the rust fungus itself.

Despite the central importance of photosynthesis in supplying plants with carbon, energy, and reducing equivalents, we observed an immediate and sustained decrease in photosynthetic activity in rust-infected black poplar leaves of nearly 50%, from 4 hpi to 10 dpi (Figure 1). Infection of poplar with Melampsora leaf rust was earlier reported to be associated with a marked reduction in photosynthesis (Zhang et al., 2010, 2016; McKown et al., 2014; Jiang et al., 2016; Gortari et al., 2018), but the detailed temporal dynamics of this reduction had not been studied. In willow, a similar pattern of decreased net photosynthetic rate was observed after rust infection, but stomatal conductance changed only at late time points (Toome et al., 2010). A negative impact of rust on transcripts of photosynthetically relevant genes in poplar was also reported at 6 dpi (summarized in Major et al., 2010). Interestingly, we did not observe any changes in the intercellular CO₂ concentration (C_i) after rust infection (Supplementary Figure S5). A decreased C_i would be expected when the assimilation rate decreases after stomatal closure due to the lowered availability of CO₂. The lack of decrease of C_i in our study might be explained by temporal dynamics, i.e., a fast reduction and subsequent equilibration of C_i before the first measurement. Alternatively, a signal – most likely phytohormonal – might have regulated stomatal conductance and assimilation rate simultaneously, so that reduced uptake and reduced consumption of CO₂ would balance out without any net effect on the C_i. A decline in stomatal conductance and assimilation rate with stable C_i has been observed before after herbivory (Meza-Canales et al., 2017). Also, the influence of biotic stress-related phytohormones on photosynthesis has been shown in other systems (Popova et al., 1988; Tang et al., 2017). A transcriptional analysis conducted at early time points of infection would help to better understand the physiological mechanisms leading to the rapid decline in photosynthesis.

In addition to the decrease in photosynthesis we observed elevated levels of SA and ABA (Figure 2) in poplar leaves upon rust infection. SA is known to play a central role in plant defense against biotrophic organisms such as rusts or mildew, and induces a hypersensitive response, the expression of pathogenesis-related genes and other responses (Derksen et al., 2013). ABA, on the other hand, is primarily known to mediate responses to abiotic stresses such as drought (Tuteja, 2007) and to control stomatal closure (Acharya and Assmann, 2009). Stomata are natural openings through which the rust fungus can enter the intercellular spaces of the leaf within the first 6 h after inoculation (Hacquard et al., 2011). Yet rapid closure of stomata might prevent the pathogen from entering, as was observed in tomato leaves infected by Pseudomonas syringae (Melotto et al., 2006). A similar mechanism might occur in poplar leaves and would explain the fast decrease in stomatal conductance and the increased levels of ABA observed in infected tissue. Phytohormone analysis of earlier time points and exogenous application of ABA and SA will help to disentangle the signaling networks between these two hormones and control of photosynthesis. Considering the fast response of photosynthetic parameters to rust infection and the changes in phytohormone content, we infer an active control of stomatal closure by the plant. Previous work suggested that a mechanical disturbance by fungal hyphae could also be involved in reducing stomatal conductance (Jiang et al., 2016), perhaps especially at later time points.

Although rust infection triggered a drastic decrease in the photosynthetic assimilation rate of poplar (Figure 1), soluble sugar content did not decrease in rust-infected compared to uninfected control leaves (Table 1). Given the rapid production of sucrose and hexose sugars as photosynthetic assimilates, it is surprising that their levels were not affected by the decline in photosynthesis. Soluble sugars can also originate from breakdown of storage carbohydrates or transport from other tissues. Since biotrophic pathogens utilize hexose sugars from their hosts (Voegele and Mendgen, 2011), it is assumed that infected tissues become carbon sinks even when photosynthetically active in order to satisfy the increased demand for carbon (Berger et al., 2007). This would require the mobilization of carbohydrates from other parts of plant. Such mobilization might account for the decline in wood production in poplar observed on rust infection, which causes significant economic losses in infected plantations (Frey et al., 2005; Wan et al., 2013). Another explanation for the maintenance of constant sugar levels in rust-infected leaves with a concurrent decrease in assimilation rate could be the reduced export of sugars. Further work is needed to elucidate the mechanisms responsible for maintaining sugar levels under these conditions.

Isoprene is emitted in large amounts by poplar and other tree species, and it is assumed to protect leaves from heat stress or reactive oxygen species (Sharkey et al., 2008). However, the role of isoprene under biotic stress is poorly studied. We observed no influence of rust infection on isoprene emission from black poplar (Figure 3). Similar patterns of stable isoprene emission were observed under drought stress conditions (Pegoraro et al., 2004; Brilli et al., 2007), which also trigger ABA-mediated stomatal closure. On the other hand, studies on the effects of insect herbivory or mechanical wounding have reported variable outcomes, showing either an increased (Brilli et al., 2011), stable (Müller et al., 2015), decreased (Brilli et al., 2009; Jardine et al., 2013; Jiang et al., 2018), or time-dependent (Loreto and Sharkey, 1993; Loreto et al., 2006; Portillo-Estrada et al., 2015) isoprene emission patterns after stress application.

In contrast to our results, another study investigating the influence of rust infection on poplar trees observed lower isoprene emissions from infected compared to uninfected trees (Jiang et al., 2016) possibly due to the more severe level of infection. Although in our study rust significantly decreased photosynthetic parameters and altered hormone levels, our infected leaves did not have necrotic lesions (Supplementary Figure S6). Such necrotic lesions, however, were present on the leaves used by Jiang et al. (2016), and may have induced the reduction in isoprene emission. Necrosis leads to premature death of cells and hence could reduce the area of living tissue with which the plant synthesizes isoprene. The stability of isoprene emission under various stress conditions suggests that this compound is of vital importance to the physiology of isoprene-emitting plants. This importance might be due to a direct effect of isoprene, for example, by reducing oxidative stress, or an indirect effect by maintaining flux through the MEP pathway (Logan et al., 2000), which provides essential compounds for plant metabolism. After many years of research, the physiological role of isoprene is still to a large extent unknown.

Considering the tight metabolic connections between photosynthesis and the MEP pathway, we expected to observe lower MEP pathway activity after rust infection due to the reduction in photosynthesis. However, the transcription of genes encoding biosynthetic enzymes of the MEP pathway did not change after rust infection (Figure 5). In addition, the stable levels of MEP pathway intermediates (Figures 4A–D) suggest a constant metabolic flux through the pathway despite fungal infection. Consistent with this, the levels of the chlorophylls and most carotenoids, the main non-volatile products of the MEP pathway, did not change after rust infection (Table 3). Constant levels of the chlorophylls in leaves after rust infection were also recently reported in another poplar species (Gortari et al., 2018). However, the levels of DMADP and IDP (quantified together in our LC-MS analysis) increased after infection (Figure 4E). DMADP and IDP can be produced by both the plastidic MEP pathway and the cytosolic MVA pathway (Hemmerlin et al., 2012). Since the cells in the leaf were disrupted for chemical analysis, the DMADP and IDP present in both cellular compartments were analyzed simultaneously. An increased transcript abundance of genes involved in the early steps of the MVA pathway (Figure 5) suggests the increased DMADP + IDP levels observed in infected tissue might be derived from this pathway. This explanation is also supported by the recently reported increased emission of sesquiterpenes upon rust infection of poplar (Eberl et al., 2018). Alternatively, the increased DMADP + IDP levels could have resulted from the fungal metabolism in rust-infected tissue. Future research using ¹³C-labeled CO₂ (Ghirardo et al., 2014) or glucose (Hemmerlin et al., 2012) could elucidate the biosynthetic origin of the increased DMADP + IDP.

In addition to higher DMADP + IDP-levels we also found higher amounts of β-carotene in rust-infected leaves compared to controls (Table 3). When rust spores were analyzed separately for carotenoids, high concentrations of β-carotene but no other carotenoids were found (Supplementary Table S1). We then analyzed the transcriptome of infected poplar leaves for Melampsora-specific genes involved in isoprenoid biosynthesis and found transcripts for genes encoding all steps of the MVA pathway (Supplementary Table S2). In contrast to plants where carotenoids are produced via the MEP pathway, their biosynthesis in fungi occurs via the MVA pathway (Disch and Rohmer, 1998). Taken together, it is likely that M larici-populina produces β-carotene in its hyphae and spores. The fungus M. larici-populina in the infected tissue is therefore the most probable cause of the increased β-carotene and DMADP + IDP levels. The fungus might use this pigment in spores to attract spore dispersers (Cano et al., 2013) or in hyphae to scavenge oxygen radicals (Davoli and Weber, 2002) that are produced by plants as defenses against infection (Ferreira et al., 2006).