The contribution of foliar endophytes to quantitative resistance to Melampsora rust


  • Anil K. H. Raghavendra,

    1. Department of Forest, Rangeland and Fire Sciences, University of Idaho, Moscow, ID, USA
    Search for more papers by this author
  • George Newcombe

    Corresponding author
    1. Center for Research on Invasive Species and Small Populations, University of Idaho, Moscow, ID, USA
    • Department of Forest, Rangeland and Fire Sciences, University of Idaho, Moscow, ID, USA
    Search for more papers by this author

Author for correspondence:

George Newcombe

Tel: +1 208 885 5289



  • Foliar endophytes of Populus do not induce the hypersensitive response associated with major genes for resistance to Melampsora leaf rust. But they could contribute to the quantitative resistance that represents a second line of defense. Quantitative resistance is thought to be determined by suites of minor genes in both host and pathogen that are influenced by the abiotic environment. Here, we determined the relative importance to quantitative resistance of foliar endophytes, one element of the biotic environment.
  • Leaves of six host genotypes differing in genetic resistance to Melampsora × columbiana were inoculated first with one of four foliar endophytes (Stachybotrys sp., Trichoderma atroviride, Ulocladium atrum or Truncatella angustata), and then with Melampsora.
  • These endophytes greatly reduced rust severity within inoculated leaves (i.e. local effects), but they had no systemic effect on rust of leaves not inoculated with endophytes. Differences among endophytes and their controls explained 54% of the total variation in quantitative resistance (i.e. rust severity); the six host/pathogen genotypes explained just 5%. In terms of magnitude of effect on rust severity, Stachybotrys, Trichoderma, Ulocladium and Truncatella were ranked in this order on all host/pathogen genotypes.
  • Endophytes may contribute significantly to quantitative resistance to Melampsora in leaves of Populus.


The discovery more than a century ago that plants possess major genes for resistance to pathogens (Biffen, 1905) has been of immense importance in biology and agriculture. A few years before Biffen's discovery, researchers in France had demonstrated that plants could be immunized against virulent pathogens by previous inoculation with nonpathogenic symbionts (Beauverie, 1901; Ray, 1901). The latter phenomenon has been variously known as acquired physiological immunity (Chester, 1933), cross-protection and induced resistance (Sequeira, 1984), and the inducers include symbionts such as mycorrhizal fungi and root endophytes (Van Wees et al., 2008). It is now clear that resistance induced by non-pathogenic symbionts can result in expression of genetic resistance in host plants (Wang et al., 2005). Thus, the separate discoveries at the turn of the twentieth century have, in a sense, now been integrated.

However, that integration is far from complete. The genetic resistance that is induced by symbionts is likely to be quantitative because major genes are typically induced only by specific pathogens (Jones & Dangl, 2006). Yet, both ecologists (Price et al., 2004) and plant breeders (Clair, 2010) still regard quantitative resistance as the product of minor genetic interactions between host and pathogen, influenced primarily by the abiotic environment. Foliar endophytes have largely been ignored, both theoretically and practically.

This is an important shortcoming because endophytic symbionts are known to enhance disease resistance in some manner (Arnold et al., 2003; Ganley et al., 2008), and foliar endophytes are also a common and diverse part of the biotic environment of plants. Endophytes might induce resistance locally or systemically (Bailey et al., 2006), but they might also directly interact with a pathogen. The extent of induction and of direct interaction could each vary with endophytes, and this inference raises many questions given the diversity of endophytes (Ganley et al., 2004; Arnold, 2007), particularly in leaves.

Of the three, nonclavicipitaceous classes of endophytes in plants, it is the shoot-restricted, largely foliar endophytes of ‘class 3’ that are the most diverse and least understood ecologically (Rodriguez et al., 2009). Although some foliar endophytes are known to enhance resistance in some plants, their specific contribution to quantitative resistance has not been determined.

In Populus affected by Melampsora rust, quantitative (Lefevre et al., 1998; Dowkiw et al., 2003) and major-gene resistance (Newcombe et al., 1996, 2001) can be distinguished by phenotype. Observation readily indicates that endophytes do not induce hypersensitive phenotypes typical of major genes in this system. As major genes may provide complete resistance it is necessary to pair host genotypes with virulent pathotypes of Melampsora in order to observe quantitative resistance (Flor, 1971; Van der Plank, 1984). This is the first condition for the determination of the contribution of endophytes to quantitative resistance.

The second condition is to ensure that both independent variables (i.e. endophytes on the one hand and combined host/pathogen genotypes on the other) actually vary. This can be achieved easily for endophytes by choosing distinct taxa from among the diverse foliar endophytes of Populus. However, combined host/pathogen genotypes are more challenging; distinct poplar genotypes need to be specifically distinguished in terms of the combined resistance and virulence genes of both host and pathogen, respectively. Such combinations have been termed interorganismal genotypes, in the sense of Loegering (1978). As defeated major genes can contribute to quantitative resistance (Dowkiw & Bastien, 2007), these genes should be included in the determination of these combined genotypes.

In the case of virulent Melampsora on Populus, quantitative resistance is known to vary continuously and it is commonly quantified using estimates of uredinial density (UD) and latent period (LP) (Newcombe, 1998; Newcombe et al., 2001). However, UD is more commonly employed as a more direct measure of quantitative resistance than LP. Endophytes are commonly isolated from both rusted and non-rusted poplar leaves.

Here, we varied endophytes associated with poplar leaf rust simply by isolating four distinct taxa. We then showed that six poplar genotypes were distinguishable when inoculated with Melampsora in terms of the combined expression of resistance and virulence. A total of 30 experimental combinations were thus available: four endophytes (plus their control) and six host/pathogen genotypes. We were then able to determine the extent to which these two variables (i.e. endophyte and host/pathogen genotype) explain quantitative variation in resistance to virulent Melampsora rust, the most important foliar pathogen of Populus (Newcombe, 1996). Finally, although it can be difficult to distinguish between the mechanism of local induction and that of direct interaction (Van Loon et al., 1998), and we did not attempt that here, we did determine whether endophyte effects were local or systemic. Local effects were contrasted with systemic effects that were expected to develop in younger leaves after the endophyte inoculations.

Materials and Methods

Plant material and culture

Of the six genotypes (clones) used in this study, the first four were Populus × generosa, or Populus trichocarpa × Populus deltoides hybrids (Eckenwalder, 2001): Genotype 1 (University of Washington, or UW, # 181-92-3246), genotype 2 (UW# 15-29), genotype 3 (UW# 20-88-183) and genotype 4 (family unknown). The other two were genotype 5 (OP 367, a Populus × canadensis, or P. deltoides × Populus nigra, hybrid) and genotype 6 (a P. trichocarpa genotype from Lapwai Creek, Idaho). One-year old, dormant cuttings of 30 cm in length were rooted in 5 × 6 inch pots containing autoclaved ‘Sunshine’ potting mix (Sun Gro Horticulture Inc., Bellevue, WA, USA) and grown in the glasshouses at the University of Idaho, USA. Cuttings for each genotype were obtained from a single tree to minimize variation among shoots within a genotype and the possibility of confounding between environmental condition and treatment. Plants were watered as required and fertilized at weekly intervals with 200 ppm of nitrogen (N) (15:16:17, Peter's Peat Lite Special; Scotts-Sierra Horticultural Products Co., Marysville, OH, USA) to the completion of the experiments. Glasshouse conditions were 16 h light : 8 h darkness, with temperatures ranging from 24 to 27°C. Insecticidal soap sprays were used to control mites which might otherwise have affected our experiments.

Isolation, culture and identification of foliar endophytes

Fungal endophytes were isolated from Melampsora-infected, P. trichocarpa leaves sampled from trees along the lower Clearwater River of Idaho, USA. These leaves were brought to the laboratory and surface-sterilized by rinsing them in 1% sodium hypochlorite (NaOCl) solution for 2 min followed by two rinses in SDW (sterile distilled water) for 1 min each. Surface-sterilized leaves were then cut to fit Petri plates containing potato dextrose agar (PDA) medium supplemented with streptomycin (0.05 mg l−1). The efficacy of the surface-sterilization protocol was confirmed by imprinting randomly selected surface-sterilized leaves on PDA in separate Petri dishes. Four isolates, designated PT1, PT2, PT3 and PT4, belonging to the genera Stachybotrys, Trichoderma, Ulocladium and Truncatella, were randomly chosen for this study from a pool of several morphotypes of each taxon. These isolates were further investigated via micromorphology and sequencing of the ITS region of their nuclear ribosomal DNA to determine species.

Genomic DNA of the four isolates was prepared using DNeasy Plant Mini Kit (Qiagen, Valencia, CA, USA) by following manufacturer's instructions. Universal fungal primers ITS1 and ITS4 were used to amplify the nuclear ribosomal region (White et al., 1990). Polymerase chain reactions were carried out in a total mixture volume of 25 μl containing a final concentration of 1.5 mM MgCl2, 0.5 mM dNTPs (Applied Biosystems, Foster City, CA, USA), 10 pmol of each primer, 1.5 U AmpliTaq® DNA Polymerase (Applied Biosystems), and 30-36 ng of genomic DNA. Samples were incubated in a thermal cycler (MJ Research, Waltham, MA, USA) at 94°C for 2 min for an initial denaturation. This was followed by 14 cycles of touchdown PCR as follows: 94°C for 30 s, 58°C for 30 s with a decrease of 0.8°C per cycle, 72°C for 30 s, and then by 25 cycles of 94°C for 30 s, 45°C for 30 s and 72°C for 45 s. Reactions were incubated at 10°C for 5 min after a final extension at 72°C for 3 min. The PCR products were purified using ExoSap-IT (USB, Cleveland, OH, USA) following manufacturer's instructions. Samples were sequenced using ITS1 and ITS4 primers and the sequencing reaction products were cleaned following differential precipitation with ethanol. Sequences were determined by an ABI 3130 xl automated sequencer (Applied Biosystems). A NCBI GenBank nucleotide blast search (unmodified options) was used to compare our sequences with those present in the GenBank. Isolate PT1 was also sequenced for the Ef-1α gene (translation elongation factor gene) for a better resolution of the species by Dr Gary J. Samuels (USDA-ARS, Beltsville, MD, USA). Similarly, the Alt a 1 gene of isolate PT4 was sequenced to resolve its species identity (Shipunov et al., 2009). All sequences were deposited in GenBank (Table 1).

Table 1. Species identity and GenBank accession number of sequences of the four endophytes used in this study
IsolateSpeciesGenBank Accession numbers
ITS (region)EF-1αa Alt a 1 b
  1. a

    Translation elongation factor gene was amplified to resolve species identity of isolate PT1.

  2. b

    Alternaria allergen gene was amplified to resolve the species identity of isolate PT4, see (Shipunov et al., 2009).

PT1 Trichoderma atroviride HM122044 JN580492  
PT2Stachybotrys sp. HM122045   
PT3 Truncatella angustata HM122046   
PT4 Ulocladium atrum EF589899   EF589841

Endophyte inoculum

Cultures ranging from 10- to 20-d-old were used to prepare spore suspensions. A spore suspension of each endophyte was prepared with many culture plates of each; these were then homogenized thoroughly with a Tissue Tearor (BioSpec Products, Bartlesville, OK, USA) and suspended in SDW. Spore concentrations were standardized to 7 × 10ml−1.

Combined host/pathogen genotypes

Virulent rust isolates were collected from the six genotypes in the field and maintained on ramets of the respective genotypes in the glasshouse. The abaxial surfaces of misted, LPI (leaf plastochron index)-4 and LPI-5 leaves were inoculated by evenly spreading urediniospores using a moist cotton swab (Newcombe, 1998). Inoculated plants were kept moist overnight under a plastic tent to enhance infection. Each of the six genotypes was inoculated with each of the six rust isolates and laid out as a completely randomized design. The infection type was scored for the 36 genotype–rust combinations (Newcombe et al., 2001) by visually examining hundreds of uredinia per combination. There was, however, no replication at the plant level in that only one ramet of each plant genotype was employed. Resistance was categorized as follows: R-0, no visible flecking or uredinia; R-1, few tiny uredinia with some necrotic flecking; R-2, small uredinia with only necrotic flecking; R-3, small- to medium-sized uredinia with chlorotic as well as necrotic flecking. The susceptible infection type (S) was recorded when medium to large uredinia were not accompanied by any chlorosis or necrosis. Genotypes with the same expressed resistance patterns were further distinguished in terms of quantitative resistance.

Leaf disk experiment (local-leaf level)

Leaf disks, 6 cm in diameter, were punched from detached LPI-4 leaves from each of the genotypes. For each genotype 50 leaf disks (10 for each endophyte and 10 for the control) were collected and placed abaxial surface up on filter papers moistened with 100 ppm of gibberellic acid in 10 × 8 cm diameter Petri plates. Given six genotypes, a total of 300 disks were used in this study. A spore suspension of each endophyte was sprayed on to these disks using a sterile hand-sprayer and care was taken to make sure that droplets were spread evenly on each of the disks. Control disks were sprayed with SDW. Each Petri plate was then sealed and incubated for 24 h in a growth chamber to enhance infection by the endophytes.

After 24 h the seal was removed from each Petri plate and disks were then inoculated with their respective rust isolates by slightly misting disks with SDW and spreading urediniospores evenly using a moist cotton swab. Petri plates were sealed again and incubated in a growth chamber using a completely randomized design. The growth chamber was maintained at 21°C with 16 h day illumination at an intensity of c. 70 μmol m−2 s1 for 14 d after which the experiment was terminated. Leaf disks were observed daily for 14 d after rust inoculation for the appearance of uredinia. The LP was calculated as time taken in days for the appearance of first uredinia from the day of rust inoculation. The UD equaled number of uredinia cm−2, recorded as counts for each leaf disk 14 d after rust inoculation.

Whole-plant experiment (local-leaf level)

Twenty-five ramets of rooted cuttings or plants from each genotype were randomly allocated to each endophyte treatment and control, which allowed for five replicates per endophyte and per control. Inoculations were made with endophyte spore suspensions using a sterile hand-sprayer. Care was taken to ensure that the abaxial surfaces of LPI-4 and LPI-5 leaves were sprayed evenly with spore suspensions. Plants were then transferred to a plastic tent already misted with water to maintain relative humidity. All plants sprayed with a particular endophyte were grouped together to avoid any cross-contamination with other endophytes during the inoculation; controls, sprayed with SDW, were also separate under the tent. After 24 h plants were removed from the tent and were inoculated with their respective virulent rust isolates to produce six genotype-rust combinations. Rust inoculum was swabbed on to the abaxial, misted surfaces of the LPI-4 and LPI-5 leaves. Controls were inoculated first to avoid any contamination with endophytes. Plants were again transferred to the plastic tent overnight to ensure successful rust infection. A completely randomized block design with five blocks (each block comprised a comprehensive set of replicates, i.e. 30 plants) was maintained throughout the experiment except during the inoculations. The LP was recorded on an individual plant basis, while the UD (uredinia cm−2) was recorded for each of the LPI-4 and LPI-5 leaves by counting uredinia in five different areas (20 × 15 mm; two on each side of the midrib and one at the leaf tip). To determine the reproducibility of our results this experiment was repeated a second time using the same protocol.

Confirmation of colonization by endophytes

We also determined successful infection and colonization of leaves by endophytes with additional plants inoculated with endophytes. Three plants for each of the four endophytes and the control were used; cuttings of genotype 2 were not available but the other five genotypes were. The LPI-4 and LPI-5 leaves from these plants were inoculated as before. A total of six leaves for each endophyte plus control from each of the five genotypes were observed for the presence of endophytes after surface sterilization. Surface-sterilized leaves were incubated in Ziploc (SC Johnson, Racine WI, USA) bags with sterile paper towels moistened with SDW at room temperature. These leaves were observed daily and fungi growing out of the leaves were identified such that the percentage of leaves with successful inoculants reisolation could be calculated.

Whole-plant experiment (systemic resistance)

Only genotypes 1, 3 and 4 and endophytes PT1 and PT2 were used for this experiment. After 2 wk of growth from single buds, plants were sprayed with endophytic spore suspensions as described previously. Four weeks after spraying plants with endophytes, the newly emerged LPI-4 and LPI-5 leaves were inoculated with rust inoculum to produce three genotype–rust combinations. It took 4 wk for the plants to produce LPI-4 and LPI-5 leaves that had not been exposed to endophyte inoculants. We again used a complete randomized block design consisting of four blocks; each block consisted of nine plants (3 genotypes × (2 endophytes + 1 control)) and a total of 36 plants were used in this experiment. The UD was recorded as described earlier.

Statistical analyses

For the leaf-disk (local-leaf level) experiment, the Cox proportional hazards regression was used to test the effect of endophyte, genotype and endophyte × genotype interaction on LP as it allowed analysis of right-censored leaf disks that failed to produce uredinia during the observation period. Model selection was not performed as we were interested in testing the effect of all three variables on survival time (LP). A random model analysis was carried out to determine the effects of genotype, endophyte and genotype × endophyte interaction on UD. All terms were considered as random, as genotypes, along with their respective virulent rust isolates, were randomly selected, as were endophytes (Model 1). Further, to evaluate differences among endophytes, a separate, mixed model analysis was performed with endophyte as a fixed effect while treating other terms as random (Model 1.1).

For the whole-plant experiment (local-leaf level), both LP and UD data were combined from two independent experiments (the original and the repeat) for analyses. However, separate analyses were also performed for both experiments (see the Supporting Information, Tables S1–S3). As disease was observed on every endophyte-inoculated and control plant, we used a mixed model to analyse the LP data with the following terms: genotype, endophyte, experiment, genotype × endophyte, experiment × genotype, experiment × endophyte, experiment × genotype × endophyte and blocks nested within experiment. Experiment and blocks within experiment were treated as fixed while other terms were random (Model 2). Further, to evaluate differences in LP among endophytes, we performed a separate, mixed model analysis with experiment, blocks within experiment endophyte and their interaction as fixed while treating other terms as random (Model 2.1).

A mixed model analysis was performed to analyse the UD data with the following terms: genotype, endophyte, experiment, blocks nested within experiment, genotype × endophyte, experiment × genotype, experiment × endophyte, experiment × genotype ×endophyte and genotype × endophyte × block interaction, nested within an experiment. Experiment and blocks within experiment were treated as fixed; other terms were random (Model 3). Further, to evaluate the differences among endophytes in UD, we performed a separate, mixed model analysis with experiment, blocks within experiment, endophyte and endophyte × experiment interaction as fixed while treating other terms as random (Model 3.1). To determine quantitative differences in host resistance among controls, we performed a separate analysis on control data with experiment, blocks within experiment, genotype and genotype × experiment terms as fixed effects (Model 3.2).

For the whole-plant experiment (systemic resistance), a mixed model analysis was performed with block as a fixed effect, while genotype, endophyte and genotype × endophyte interaction were treated as random (Model 4).

The Cox proportional hazard model was applied using PROC PHREG and the significance of the model was determined by the likelihood ratio test. The significance of the main and interaction variables were determined by the Wald χ2 statistics with their associated P-values. Comparisons of LP among endophytes were performed using the CONTRAST statement. Models 2 and 2.1 (i.e. LP data from whole-plant local-leaf level experiment) were analysed using PROC GLM and PROC MIXED. Appropriate F-tests for random variables were generated using RANDOM statement and TEST option in PROC GLM. Variance components were estimated from PROC MIXED with the restriction maximum likelihood (REML) option. The denominator degrees of freedom for fixed effects were determined using the Satterthwaite approximation. Mean comparisons followed Tukey's (< 0.05) after adjustment for multiple comparisons. The LP data were log-transformed before analyses to satisfy normality and homogeneity of variance assumptions. Models 1, 1.1, 3, 3.1, 3.2 and 4 were analysed as generalized linear mixed models (GLMM) using the PROC GLIMMIX procedure with the residual pseudo-likelihood method (RSPL). As UD data were based on counts, a Poisson distribution and a log link function were used. The Satterthwaite approximation was used to determine denominator degrees of freedom for fixed effects and mean comparisons followed Tukey's (< 0.05) after adjustment for multiple comparisons using LSMEANS statement. Significance of random variables was tested using the COVTEST statement. The dispersion parameter, or the residual variance, in these models was fixed at 1.0. All analyses were performed in SAS 9.2 (SAS Institute Inc., Cary, NC, USA). Original untransformed means (± SE) are presented in Tables and Figures.


Identification of endophytes

The identities of the four endophytes are presented in Table 1. Isolate PT1 was identified as Trichoderma atroviride based on its culture morphology, micromorphology and its ITS and Ef-1α sequences. On the basis of both ITS sequence and micromorphology, PT2 belonged to the genus Stachybotrys. Its nearest affinities were Stachybotrys cholorohalonata (GenBank: JN 942888) and Stachybotrys chartarum (HQ 2862164) with 461/464 and 462/464 sequence identity, respectively. However, its micromorphology did not match either of these species and it thus may represent an undescribed species. PT3 had only 98% sequence similarity (462 bp of 466 bp) to the ITS sequence of Truncatella angustata represented in GenBank (DQ 093715); morphologically it was indistinguishable from the type culture of T. angustata. Isolate PT4 was identified as Ulocladium atrum based on ITS and Alt a 1 gene similarities (Shipunov et al., 2009). Hereafter, isolates PT1, PT2, PT3 and PT4 are simply referred to genus (i.e. Trichoderma, Stachybotrys, Truncatella and Ulocladium, respectively).

Combined, host/pathogen genotypes

Four host genotypes were distinguished via infection type (Table 2). But, genotypes 1 and 5, and genotypes 3 and 6 were indistinguishable. Genotypes 1 and 5 displayed fully susceptible infection types to all six isolates of Melampsora. Genotypes 3 and 6 displayed the same pattern of infection types (i.e. R/R/S/R/R/S) to the six isolates of Melampsora. These two pairs could, however, be distinguished in terms of quantitative resistance, as reported below.

Table 2. Variation in combined, host/pathogen genotypes as revealed by major-gene interaction phenotypesa among the six Populus genotypes and corresponding virulent isolates of Melampsora in a fully factorial, inoculation experiment
Rust IsolatebGenotypes
Genotype 1Genotype 2Genotype 3Genotype 4Genotype 5Genotype 6
  1. Genotypes 1 and 5 and genotypes 3 and 6 were indistinguishable in terms of major-gene interactions, but they were further distinguished quantitatively (Fig. 2). Genotype 2 and genotype 4 were each distinguished by unique genes for resistance to isolates 6 and 3, respectively.

  2. a

    Infection type was recorded as either a resistant (R) or susceptible (S) interaction phenotype (Newcombe et al., 2001).

  3. b

    Rust isolates 1–6 were collected and employed as virulent isolates of genotypes 1–6 in all other inoculation experiments.


Four pathotypes of Melampsora × columbiana were also distinguished (Table 2). The isolates from genotypes 2 and 5 belonged to the same pathotype in that their infection types across the six genotypes were: S/S/R/S/S/R (Table 2). Similarly, the isolates from genotypes 1 and 4 shared a second pathotype: S/R/R/S/S/R. The isolates from genotypes 3 and 6 were unique third and fourth pathotypes. With genotype distinctions, reported later in the quantitative resistance among controls, we were able to conclude that the six poplar genotypes, inoculated with virulent isolates of Melampsora, were six distinct host/pathogen genotypes.

Leaf-disk experiment (local-leaf level) Latent period (LP)

The LP was not recorded from nearly 10% of leaf disks (i.e. 28 out of 300 leaf disks) as they failed to produce uredinia during the observational period of 14 d. Although the Cox proportional hazard analysis showed that the model with all three variables (i.e. endophyte, genotype and endophyte × genotype interaction) affected the survival time (LP) of Melampsora rust (likelihood ratio test: χ2 = 159.63; df = 29; < 0.001), only endophyte (χ2 = 43.05; df = 4; < 0.001) and host/pathogen genotypes (χ2 = 36.05; df = 5; < 0.001) were significant. Their interaction was not (χ2 = 28.44; df = 20; = 0.10). Overall, all four endophytes lengthened the time before eruption of uredinia (i.e. longer LPs or ‘survival times’) as the median LP was 8 d compared with the control with a median LP of 7 d. Leaves inoculated beforehand with Trichoderma (= 0.017), Stachybotrys (= 0.019) and Ulocladium (= 0.02) significantly slowed rust development compared with the control. Those inoculated with Truncatella were only marginally significant (= 0.062). The four endophytes did not differ from one another in slowing rust development.

Leaf-disk experiment (local-leaf level) Uredinial density (UD)

Analysis (Model 1) indicated that the UD was significantly influenced by endophytes, host/pathogen genotypes and their interaction (Table 3). Endophytes (including controls) accounted for 29% of the observed variation in UD in this assay. Host/pathogen genotypes and ‘host/pathogen genotype × endophyte’ interaction explained an additional 20% of the variation (Table 3). Analysis with endophytes as a fixed effect (Model 1.1) indicated that the four endophytes significantly reduced UD compared with controls, and there was roughly five times less disease in endophyte-treated leaf disks (Fig. 1a). However, the four endophytes did not differ significantly from one another.

Table 3. Summary of generalized linear mixed model (GLMM) analysis for the uredinial density in the leaf-disk (Model 1) and the whole-plant experiments (local-leaf level (Model 3) and systemic resistance (Model 4)
Source of variationUredinial density (Uredinia cm−2)
df F-valueχ2 P-valueVariance componentsVariation (%)
  1. Endophytes explained most of the variation in leaf-disk and whole-plant experiments (local-leaf level), but had no effect in the whole-plant (systemic resistance) experiment.

  2. a

    Significance test for random effects was conducted using covtest statement in Proc Glimmix.

  3. b

    Host/pathogen genotype (clone) represents the combined resistance and virulence genes of host and pathogen, respectively.

  4. c

    Endophytes were inoculated on LPI-4 leaves in leaf-disk experiment and LPI-4 and LPI-5 leaves of each plant in whole-plant (local-leaf level) experiments 24 h before rust inoculation. In the whole-plant (systemic resistance) experiment plants were inoculated with endophytes 4 wk before inoculating younger (E-) leaves with rust.

  5. d

    Data from two independent experiments (the original and the repeat) were combined for analyses.

Leaf-disk experiment (local-leaf level)
Random effecta
Host/pathogen genotypeb 115.55 < 0.001 0.33117.12
Endophytec 122.93 < 0.001 0.55428.63
Host/pathogen genotype × endophyte11.440.2300.0502.58
Error 1.000
Whole-plant experiment (local-leaf level)d
Fixed effect
Experiment1, 4.67.76 0.042
Blocks (experiment)8, 5901.39 0.197
Random effect
Host/pathogen genotype17.13 0.0038 0.1355.1
Endophyte110.07 0.0008 1.44353.58
Host/pathogen genotype × endophyte13.87 0.0245 0.0240.90
Experiment × Host/pathogen genotype115.85 < 0.001 0.0200.74
Experiment × endophyte14.35 0.0189 0.0632.34
Experiment × host/pathogen genotype × endophyte15.48 0.0096 0.0070.30
Host/pathogen genotype × endophyte × blocks (experiment)16.53 0.0053 0.0010.0003
Error 1.000
Whole-plant experiment (systemic resistance)
Fixed effect
Blocks3, 10.16 0.915
Random effect
Host/pathogen genotype10.010.907
Host/pathogen genotype × endophyte10.010.917
Figure 1.

Reduction in mean uredinial density (±SE) owing to endophytes in (a) leaf-disk and (b) whole-plant experiment (local-leaf level). Means with the same letter are not significant different (Tukey's multiple comparison test; < 0.05).

Whole-plant experiment (local-leaf level) Latent period

Analysis of variance of the combined data (Model 2) showed that 69% of the variation in LP was explained, overall, by our model. The LP was significantly affected by experiment, endophytes and combined host/pathogen genotypes (Experiment: F1, 4.6 = 8.63, = 0.04; endophyte: F4, 11.4 = 16.41, < 0.001. Host/pathogen genotype: F5, 12.46 = 11.36, < 0.001). Differences among host/pathogen genotypes and endophytes explained 31.25% and 25.0% of the variation, respectively. The ‘host/pathogen genotype × endophyte’ interaction was significant (F20,20 = 2.92, = 0.01) suggesting that the effects of endophytes on LP varied among the six host/pathogen genotypes (Table 4). However, this interaction explained only 6.25% of the observed variation. The LP remained unaffected by the following two- and three-way interactions: experiment × host/pathogen genotype, experiment × endophyte, experiment × host/pathogen genotype × endophyte and blocks nested within experiment. Analysis with endophytes as a fixed effect (Model 2.1) indicated that, as in the leaf-disk experiment, LPs were significantly shorter (i.e. faster eruption of uredinia) in the controls than in endophyte-inoculated plants (Table 4).

Table 4. Differences in mean latent period (± SE) of endophyte-inoculated and control plants in the whole-plant experiment, local-leaf level
Endophytea Latent period (d)
Host/pathogen genotype 1Host/pathogen genotype 2Host/pathogen genotype 3Host/pathogen genotype 4Host/pathogen genotype 5Host/pathogen genotype 6
  1. Endophytes varied in their ability to delay rust across the six host/pathogen Populus genotypes. Means with the same letter are not significantly different (Tukey's multiple comparison test; < 0.05).

  2. a

    Endophyte was analysed as a fixed effect.

  3. b

    Each value is a mean of 10 observations as latent period was measured on an individual-plant basis.

Stachybotrys 7.8b (± 0.21)ab 7.2 (± 0.09)b 8.3 (± 0.11)a 9.0 (± 0.11)ab 8.5 (± 0.11)a 7.6 (± 0.12)a
Trichoderma 8.2 (± 0.14)a 8.1 (± 0.16)a 8.2 (± 0.09)a 8.5 (± 0.12)bc 9.0 (± 0.11)a 7.6 (± 0.26)a
Truncatella 8.2 (± 0.09)a 8.0 (± 0.18)a 8.5 (± 0.12)a 9.5 (± 0.12)a 8.6 (± 0.15)a 7.3 (± 0.28)ab
Ulocladium 7.9 (± 0.16)ab 7.2 (± 0.09)b 8.1 (± 0.12)a 8.5 (± 0.12)bc 8.5 (± 0.03)a 8.0 (± 0.11)a
Control7.2 (± 0.17)b 6.2 (± 0.09)c 6.8 (± 0.09)b 8.3 (± 0.15)c 7.3 (± 0.03)b 6.5 (± 0.03)b

Whole-plant experiment (local-leaf level) Uredinial density

Analysis of the combined data from two experiments (Model 3) showed that seven of the eight variables significantly affected UD (Table 3). Variation among endophyte-inoculated and control leaves explained 54% of the observed variation in UD. Conversely, differences among host/pathogen genotypes accounted for only 5.0% of the observed variation. The ‘host/pathogen genotype × endophyte’ interaction, although significant (χ2 = 3.87, = 0.03), explained only 1.0% of the observed variation in UD. Experiment and its interactions (i.e. experiment × host/pathogen genotype, experiment × endophyte, host/pathogen genotype × endophyte × experiment and blocks × host/pathogen genotype × endophyte within experiment) were significant (Table 3), but blocks within experiment had no significant effect on UD. No interaction explained > 3.0% of the observed variation in UD. Analysis with endophytes as a fixed effect (Model 3.1) indicated that all four endophytes significantly reduced disease severity compared with controls: 15.9 times (Stachybotrys), 10.6 times (Trichoderma), 3.4 times (Ulocladium) and 0.64 times (Truncatella) (Fig. 1b). The four endophytes were ranked constantly in that same order across all host/pathogen genotypes as measured by magnitude of reduction in uredinial density (Table 5). Further, these patterns were observed even when the two experiments (the original and the repeat) were analysed separately (Tables S2, S3).

Table 5. Mean uredinial density (± SE) of endophyte-inoculated and control leaves across six host/pathogen Populus genotypes in the whole-plant experiment, local-leaf level
EndophyteUredinial density (Uredinia cm−2)
Host/pathogen genotype 1Host/pathogen genotype 2Host/pathogen genotype 3Host/pathogen genotype 4Host/pathogen genotype 5Host/pathogen genotype 6
Mean (± SE)Magnitude of rust reductionMean (± SE)Magnitude of rust reductionMean (± SE)Magnitude of rust reductionMean (± SE)Magnitude of rust reductionMean (± SE)Magnitude of rust reductionMean (± SE)Magnitude of rust reduction
  1. There was constant ranking in rust disease reduction by endophytes across six host/pathogen genotypes. Means with the same superscript letter are not significantly different (Tukey's multiple comparison test; < 0.05).

Stachybotrys 1.28 (± 0.24)d14.40.97 (± 0.17)d14.31.66 (± 0.22)d11.50.63 (± 0.12)d23.80.55 (± 0.10)d19.80.46 (± 0.08)d24.8
Trichoderma 2.28 (± 0.32)d7.71.15 (± 0.20)d11.92.08 (± 0.25)d9.01.08 (± 0.16)d13.50.91 (± 0.13)d11.40.59 (± 0.10)d19.0
Ulocladium 5.69 (± 0.71)c2.54.52 (± 0.36)c2.37.15 (± 0.47)c1.93.21 (± 0.27)c3.92.93 (± 0.25)c2.91.42 (± 0.15)c7.4
Truncatella 10.93 (± 0.72)b0.810.76 (± 0.55)b0.411.13 (± 0.55)b0.99.52 (± 0.47)b0.67.89 (± 0.40)b0.54.68 (± 0.27)b1.5
Control19.69 (± 0.92)a14.84 (± 0.64)a 20.81 (± 0.72)a15.63 (± 0.60)a11.45 (± 0.48)a11.85 (± 0.43)a

The UD (Model 3.2) varied significantly among controls of host/pathogen genotypes (Fig. 2). The six host/pathogen genotypes represented three levels of uredinial density or three levels of quantitative host resistance: host/pathogen genotypes1 and 3 (means of 19.8 and 20.9 uredinia cm−2, respectively) represented the highest level of resistance; host/pathogen genotypes 2 and 4 (means of 15.0 and 16.1 uredinia cm−2, respectively) were at an intermediate level; and host/pathogen genotypes 5 and 6 (means of 11.6 and 12.0 uredinia cm−2, respectively) represented the lowest level of resistance. Further, the two pairs of host/pathogen genotypes (i.e. 1 and 5, and 3 and 6), which were indistinguishable in terms of their pattern of major-gene interactions (Table 2), were easily distinguished in terms of quantitative resistance (Fig. 2).

Figure 2.

Three levels of uredinial density among the six host/pathogen Populus genotypes. These levels distinguished the two pairs (host/pathogen genotypes 1 and 5, and host/pathogen genotypes 3 and 6) that were not distinguished in terms of major-gene resistance (Table 2). Values are means ± SE. Means with the same letter are not significant different (Tukey's multiple comparison test; < 0.05).

Confirmation of colonization by endophytes

The success of endophyte inoculations was confirmed by the reisolation of inoculants from surface-sterilized leaves. The control and endophyte-inoculated leaves did produce other fungi that were easily distinguished from the inoculants. None of the inoculants were recovered from any of the control leaves. Inoculants were not recovered from other treatments. The two endophytes with the strongest effects in reducing rust severity, Stachybotrys and Trichoderma, were recovered from all of the inoculated leaves of all five genotypes. However, reisolation of the endophytes with weaker effects, Ulocladium and Truncatella, were not recovered from 100% of the leaves. Ulocladium was recovered from 4/6 leaves in genotypes 3 and 5 (66%), and from 5/6 leaves in genotypes 1, 4 and 6 (83%). Truncatella was recovered from 2/6 leaves (33%) in genotype 6, but not from any of the other genotypes.

Whole-plant experiment (systemic resistance)

Analysis (Model 4) indicated that none of the variables (i.e. blocks, host/pathogen genotype, endophyte, host/pathogen genotype × endophyte) had any significant effect on uredinial density (UD) in younger leaves of the same stems in which older leaves had been inoculated with endophytes before rust inoculation. This can be summarized as follows: blocks: F3,1 = 0.16, = 0.92; endophyte: χ2 = 0.00, = 1.00; host/pathogen genotype: χ2 = 0.01, = 0.91; and host/pathogen genotype × endophyte: χ2 = 0.01, = 0.91.


Foliar endophytes were surprisingly important contributors to quantitative resistance to leaf rust in our experimental system. Four foliar endophytes explained 54% of the variation in quantitative resistance (UD) among six poplar genotypes varying in genetic resistance to virulent isolates of Melampsora × columbiana. The local, within-leaf effects were strong enough that further investigation is needed to determine if foliar endophytes are the second line of defense behind major genes for resistance to leaf rust.

Even though the foliar endophytes of this study did not induce resistance systemically, they may have induced it locally (Van Wees et al., 2008). A competing explanation for the local effects of endophytes is direct interaction with Melampsora. Either mechanism could be compatible with the observed constant ranking of the four endophytes (i.e. Stachybotrys, Trichoderma, Ulocladium and Truncatella, in that order in terms of magnitude of effect; Tables 5, S2, S3), regardless of combined host/pathogen genotypes. It is extremely improbable [1/(4!)5 c. 1 in 8 00 000] that such constant ranking is due to chance. For local induction of host resistance to be compatible with constant ranking, the four endophytes would have to vary in their capacity to induce yet be insensitive to the differences in combined host/pathogen genotypes employed in this study. By contrast, direct interaction between endophytes and Melampsora could explain constant endophyte ranking without having to also account for interactions with combined host/pathogen genotypes.

It was essential for these inferences that the poplar genotypes were specifically distinguishable in terms of genes for resistance. If we had simply selected different genotypes of poplar without testing their resistance genes, we would not have been sure that there was actually variation in resistance genes. Just as otherwise distinct people could have the same genes for blood type, otherwise distinct, poplar genotypes could have the same resistance genes. In terms of major genes, genotypes 1 and 5 were identical in our tests in that they lacked resistance to the four pathotypes. Genotypes 3 and 6 were also identical as RRSRRS patterns were found for both. However, in terms of quantitative, genetic resistance (Fig. 2) host/pathogen genotype 1 was distinguished from genotype 5 and host/pathogen genotype 3 from genotype 6. In other words, when both major-gene and quantitative resistance were considered, each of the six host/pathogen combinations was a distinct genotype. Yet none of this host/pathogen variation affected the ranking of the four endophytes.

Endophyte infection may enhance disease suppression whether the mechanism involves local induction of host resistance or direct interaction with the pathogen (Mejia et al., 2008; Lee et al., 2009). Stachybotrys and Trichoderma had the strongest effects on uredinial density and they infected to the greatest extent, as measured by reisolation. Ulocladium and Truncatella had weaker effects on uredinial density and more poorly developed abilities to infect, although each was able to infect their source host, P. trichocarpa (genotype 6), with the expected stronger effects associated with colonization. In general, the effects of endophytes were strongest in P. trichocarpa, although effects were almost as strong in the five hybrids (Figs S1, S2). Therefore, if anything, this study may have underestimated the magnitude of endophyte effects by coupling endophytes isolated from P. trichocarpa with five hybrid genotypes.

In this study, four, specific endophytes contributed to quantitative resistance in individual leaves of six combined poplar/rust genotypes. However, overall foliar endophytes must be highly variable in nature at the level of a tree. Each poplar tree is likely colonized by diverse endophytes that vary from leaf to leaf (Santamaria & Diez, 2005; Albrectsen et al., 2010) and that may also vary among host genotypes. Overall, foliar endophytes of Populus are likely to vary more widely in their abilities to reduce rust severity than the four of this study. In other words, endophytes with stronger effects than Stachybotrys will likely be isolated from Populus; similarly, endophytes with no significant effect, or even a rust-enhancing effect will also likely be found. Furthermore, young foliage is likely to be uncolonized by endophytes (Stone, 1987; Arnold et al., 2003). Collectively, these likelihoods make it probable that the foliage of a tree is a highly variable mosaic with respect to endophyte-mediated resistance. This is quite unlike the major-gene, minor-gene and systemically induced forms of defense that should not vary within the foliage of a tree.


We thank two anonymous reviewers for their help with this manuscript. Christopher J. Williams, William Price and Jack Brown helped with statistical analysis. Cort L. Anderson and Ed Ismaiel (USDA-ARS, Beltsville, MD, USA) helped with sequencing. Posy Busby and Brian Stanton provided presubmission reviews. Financial support for the research was provided by Boise Cascade Inc. Greenwood Resources and Boise Cascade Inc. provided the poplar genotypes.