American Journal of Botany 101(3): 467–478. 2014.
TREE GENOTYPE AND GENETICALLY BASED GROWTH TRAITS STRUCTURE TWIG ENDOPHYTE COMMUNITIES1
LOUIS J. LAMIT2,5, MATTHEW K. LAU2, CHRISTOPHER M. STHULTZ3, STUART C. WOOLEY4, THOMAS G. WHITHAM2, AND CATHERINE A. GEHRING2 2Department
of Biological Sciences and Merriam-Powell Center for Environmental Research, Northern Arizona University, Flagstaff, Arizona 86011-5640 USA; 3Math, Science & Technology Department, University of Minnesota Crookston, Crookston, Minnesota 56716 USA; and 4Department of Biological Sciences, California State University Stanislaus, Turlock, California 95382 USA • Premise of the study: Fungal endophytes asymptomatically inhabit plant tissues where they have mutualistic, parasitic, or commensal relationships with their hosts. Although plant–fungal interactions at the genotype scale have broad ecological and evolutionary implications, the sensitivity of endophytes in woody tissues to differences among plant genotypes is poorly understood. We hypothesize that (1) endophyte communities in Populus angustifolia (Salicaceae) twigs vary among tree genotypes, (2) endophyte variation is linked to quantitative tree traits, and (3) tree genotype influences interspecific fungal interactions. • Methods: Endophytes were isolated from twigs of replicated P. angustifolia genotypes in a common garden and characterized with PCR-RFLP and DNA sequencing. Twig length and diameter, aboveground tree biomass, and condensed tannins were also quantified. • Key results: (1) Aspects of fungal community structure, including composition and total isolation frequency (i.e., abundance), varied among genotypes. (2) Aboveground biomass and twig diameter were positively associated with isolation frequency and covaried with composition, whereas twig length and condensed tannin concentration were not significantly correlated to endophytes. (3) Fungal co-occurrence patterns suggested negative species interactions, but the presence of significant co-occurrences was genotype dependent. • Conclusions: The species is often assumed to be the most important ecological unit; however, these results indicate that genetically based trait variation within a species can influence an important community of associated organisms. Given the dominance of plants as primary producers and the ubiquity of endophytes, the effect of host genetic variation on endophytes has fundamental implications for our understanding of terrestrial ecosystems. Key words: aboveground biomass; condensed tannins; co-occurrence analysis; endophyte; Populus angustifolia; Salicaceae; tree genotype.
Plants are typically viewed as discrete entities; often overlooked are the numerous plant-associated fungi that can have strong influences on plant growth, pest resistance, and stress tolerance (Rodriguez et al., 2009; Hoeksema et al., 2010). Of growing interest are Type 3 endophytes (Rodriguez et al., 2009; subsequently referred to simply as endophytes). These fungi
live within apparently healthy aboveground tissues of all plants surveyed, are horizontally transmitted, can form diverse communities with members representing distantly related fungal lineages, and are believed to represent an important component of the millions of undescribed fungal species (Arnold, 2007; Blackwell, 2011). Endophytes obtain nutrients and energy from plants and, in some cases, modulate host traits such as drought tolerance, photosynthetic efficiency, growth, and pathogen and herbivore resistance (Arnold, 2007; Rodriguez et al., 2009). Given the intimate link between endophytes and their hosts, genotypic differences within host species may have important effects on the structure of endophyte communities, which may feed back to influence plant growth and fitness. Furthermore, because endophytes are ubiquitous within plants, the primary source of energy fixation for terrestrial ecosystems, understanding the relationship between intraspecific host plant variation and endophytic fungi is fundamental to our understanding of the functioning of terrestrial ecosystems. Historically, research on endophytes largely focused on characterizing community shifts among host plant species, tissue types, and tissue ages, or across spatial and temporal scales (e.g., Carroll and Carroll, 1978; Faeth and Hammon, 1996; Collado et al., 1999; Sahashi et al., 1999; Arnold et al., 2001, 2003; Santamaria and Diez, 2005; U’Ren et al., 2012; Zimmerman and Vitousek, 2012; de Souza Sebastianes et al., 2013). These studies show that endophytes can exhibit various levels of host
1 Manuscript received 28 January 2014; revision accepted 4 February 2014. The authors thank R. J. Deckert, A. E. Arnold, T. Wojtowicz, and members of the Gehring Laboratory and Cottonwood Ecology Group for technical assistance and/or comments on this manuscript and the Ogden Nature Center for support of our common gardens and field housing. Special thanks to C. M. Pace for drawing Fig. 1. Chemical analyses were performed in R. L. Lindroth’s laboratory at the University of Wisconsin, Madison. Research was supported by the National Science Foundation Frontiers in Integrative Biological Research grant DEB-0425908 to the Cottonwood Ecology Group. Additional support to L. J. Lamit was provided by a National Science Foundation Integrative Graduate Education and Research Traineeship, an Achievement Rewards for College Scientists scholarship, and a Mycological Society of America graduate fellowship. 5 Author for correspondence (e-mail: [email protected]); present address: Michigan Technological University, School of Forest Resources and Environmental Science, 1400 Townsend Dr., Houghton, MI 49931-1295 USA
doi:10.3732/ajb.1400034
American Journal of Botany 101(3): 467–478, 2014; http://www.amjbot.org/ © 2014 Botanical Society of America
467
468
[Vol. 101
AMERICAN JOURNAL OF BOTANY
and tissue specificity. In some cases, endophytes show fidelity to specific host species even across large temporal scales (e.g., Carroll and Carroll, 1978). In contrast, communities frequently vary across large and small spatial scales, even when the same host species is sampled (e.g., Arnold et al., 2001; U’Ren et al., 2012; Zimmerman and Vitousek, 2012). Importantly, Elamo et al. (1999) provided evidence that abundances of leaf endophyte species vary among maternal families of mountain birch (Betula pubescens subsp. czerepanovii (Orlova) Hämet-Ahti) in two spatially separated common gardens, located in contrasting environments. This finding points to the potential importance of host plant genetics in shaping the distributions of endophytes. Endophyte sensitivity to host genetics has important evolutionary implications. Selection, founder effects, and genetic drift may influence heritable plant traits important to endophytes, leading to shifts in endophyte communities in response to host evolution. The influence of endophytes, either positive or negative, may also be a selective force on host plants (Elamo et al., 1999; Miglia et al., 2007), altering the relative fitness of plant genotypes. Furthermore, phenotypic differences among host genotypes may alter endophyte interactions with one another, potentially altering their population dynamics and evolution. Although host–endophyte evolutionary dynamics have been discussed for more than three decades (e.g., Carroll, 1988; Todd, 1988; Elamo et al., 1999; Miglia et al., 2007; Arnold et al., 2009), the role of host genotype in shaping host–endophyte interactions and endophyte–endophyte interactions needs to be more fully elucidated. A small number of studies show that endophytes can be sensitive to intraspecific variation in plant genetics (e.g., Todd, 1988; Elamo et al., 1999; Pan et al, 2008; but see Miglia et al., 2007; Korkama-Rajala et al., 2008). However, there are at least three significant gaps in our knowledge that warrant further investigation of plant genetic effects on endophytes. First, the majority of endophyte research with woody plants focuses on fungi inhabiting leaves and needles. This is despite the apparent ubiquity of fungi in woody tissues (e.g., Chapela, 1989; Sieber et al., 1991; Hutchison, 1999; Sahashi et al., 1999; Santamaria and Diez, 2005; de Souza Sebastianes et al., 2013), data suggesting that endophyte communities can differ between woody tissue and leaves (e.g., Sieber et al., 1991; Sahashi et al., 1999; Santamaria and Diez, 2005; de Souza Sebastianes et al., 2013; but see discussion in Arnold, 2007), and the likelihood that some of these fungi are economically important latent pathogens (Christensen, 1940; Chapela, 1989; Desprez-Loustau et al., 2006). Second, there is little understanding of the genetically variable plant traits to which endophytes may be linked. These plant traits may include phytochemistry (Bailey et al., 2005) and productivity (Todd, 1988). Identification of these traits is important if we are to understand the ecological and evolutionary forces that shape plant–endophyte interactions and will help further the development of a predictive-mechanistic framework for understanding endophyte distributions. Third, it is unclear whether differences in plant genotypes modulate interspecific interactions among endophytes. Work with species interactions using other fungi demonstrates that interactions can play important roles in structuring communities (e.g., Koide et al., 2005; Fukami et al., 2010), and limited research shows that endophytes co-occur in patterns indicative of interspecific interactions (Pan and May, 2009). Modulation of endophyte species interactions by tree genotype has implications for endophyte community structure and function, which may cause genotype-specific feedbacks to host trees.
In the present study, we examine the fungal communities associated with Populus angustifolia James (Salicaceae; narrowleaf cottonwood) twigs to understand how endophytes respond to genotypic differences in their host. First, we hypothesize that endophyte community structure (i.e., isolation frequency [a proxy measure for abundance], richness, diversity, and composition) will vary among tree genotypes in a common garden. This approach is particularly strong for addressing this hypothesis because a common garden contains replicated genotypes of the same age, planted in a random order over a relatively homogenous area, thus allowing for the control of confounding environmental and age influences that occur in naturally established field sites. Second, we hypothesize that endophytes will show relationships with quantitative tree traits such as growth (including twig length, twig diameter, and aboveground biomass) and condensed tannins, which are known to be genetically based in cottonwoods (e.g., Bailey et al., 2005; Lojewski et al., 2009). The plant vigor hypothesis (Price, 1991) suggests that herbivores intimately associated with host tissues (e.g., gallers and stem borers) preferentially attack vigorously growing plants because they contain more resources (e.g., carbohydrates, nutrients). Because of their close association with plant tissues, endophytes may also favor vigorously growing plants, leading to a variety of potential community patterns in relation to plant productivity. We predict that endophyte isolation frequency will be positively associated with increases in tree growth, as has been found for arthropods (Price, 1991; Stone et al., 2010) and that endophyte species composition will covary with variation in growth among trees. In contrast, we predict that endophyte isolation frequency will be negatively related to the concentration of condensed tannins (CTs) in twigs due to the potential antifungal properties of CTs (Kraus et al., 2003; Bailey et al., 2005; Holeski et al., 2009), and that CTs will likely affect other aspects of endophyte community structure. Third, we hypothesize that endophytes will exhibit cooccurrence patterns indicative of interspecific interactions, as commonly seen in other microbial communities (HornerDevine et al., 2007). Importantly, co-occurrence patterns will differ among tree genotypes because each genotype represents a unique habitat defined by its genetically based traits. Identifying relationships among tree genotype, tree traits, and twig endophytes is a key step in understanding the hidden evolutionary and ecological dynamics that govern the distribution of these ubiquitous plant symbionts. MATERIALS AND METHODS Study system—Populus angustifolia is a mid to upper elevation (~1300– 2500 m a.s.l. in the region of the present study) foundation tree species in riparian habitats of interior western North America. Sampling efforts were focused on replicated genotypes of P. angustifolia growing in a common garden established in 1991. The garden is located in Ogden, Utah (latitude = 41.248146, longitude = −111.999830, elevation = 1302 m a.s.l.) and was planted with clonally replicated Populus genotypes from cuttings collected along the length of the nearby Weber River. The garden is contained within an approximately 1.2 ha area, with trees planted 4–7 m apart. Trees were planted in random order, and by chance, members of the subset of trees used in this study ranged from neighbors (within 4–7 m) to across the length of the garden (~230 m) from each other. Spatial analyses indicated that trees in close proximity were not more likely to have similar endophyte communities relative to trees far away from each other (Appendix S1, see Supplemental Data with the online version of this article). Each tree genotype was characterized with 35 codominant restriction fragment length polymorphism (RFLP) markers to verify that all were genetically distinct (Martinsen et al., 2001; M. Zinkgraf et al., Northern Arizona
March 2014]
LAMIT ET AL.—TREE GENETICS AND TWIG ENDOTYPES
University, unpublished data). A hybrid index generated from the RFLP markers indicated that all but one genotype used in this study were P. angustifolia genotypes; genotype 18, from the 2008 data set, was a backcross hybrid with P. fremontii S. Watson (Lamit et al., 2011; M. Zinkgraf et al., Northern Arizona University, unpublished data). Sampling methods—Fungal endophyte communities were examined from multiple tree genotypes on two dates, approximately 2 yr apart. The goal with the two samples was not to make direct comparisons between the 2 yr. The first sample (collected mid-June 2006, during the growing season) was collected to assess the response of endophytes in twigs to differences among tree genotypes and to examine potential links between genetically based tree traits and endophytes. The second sample (collected in late April 2008, during leaf flush) was collected to complement but not repeat the first to (1) verify that twig endophyte communities differ among tree genotypes, in a mostly independent set of trees (only one tree genotype overlapped between the two samples, genotype 1023), (2) sample more 1-cm twig segments from a tree, covering greater shoot age, (3) let fungal isolates emerge for a longer time, and (4) identify isolates with higher resolution molecular methods. The two endophyte samples were collected as follows (see Fig. 1). In midJune 2006, 15 twigs were selected from a minimum of three locations around each tree (10 genotypes, 3–9 trees per genotype, 49 trees total), although only 13 or 14 twigs could be obtained from a very small number of trees (Table 1). A single 1-cm-long segment was taken from the middle of the 3-yr-old growth section (formed in 2003) from each twig for fungal isolations (726 total segments; Table 1). In late April 2008, two twigs were cut from the north and south sides from each tree (5 genotypes, 2–5 trees per genotype, 17 trees total), and the entire 2- and 3-yr-old growth sections (formed in 2005 and 2006) were cut into 1-cm segments, pooled per age class and 32 segments per tree (16 per age class, 544 segments total) were selected at random for isolations (Table 1). Fewer twigs were taken from each tree in the 2008 sample to reduce damage to trees caused by clipping large numbers of twigs. Twigs for each sample year were selected from the trees at a height between 2 and 5.5 m from the ground. The following isolation and culturing procedures were used for both sampling dates. Twig segments were rinsed with tap water to remove adhering debris, and then surface sterilized by immersion for 5 min in 70% ethanol, followed by 5 min in 50% Clorox bleach, then three successive rinses in sterile reverse osmosis water (Bailey et al., 2005). During each step, twig segments were vigorously agitated. Twig segments were individually placed on Petri dishes of 2% potato dextrose agar (PDA), and left for one (2006 sample) or two (2008 sample) months. Petri dishes were examined weekly for fungal growth. Colonies of fungi were sorted into morphological categories (morphotypes) based on mycelia color, texture, and growth characteristics at the center and edges of the isolate, hyphal height above the medium, hyphal depth into the
469
medium, characteristics of reproductive structures, growth-rate, and color changes in the PDA. Endophyte isolation frequency was expressed as the percentage of twig segments from a single tree that grew fungi. Representatives of each morphotype were photographed, and a subset from the 2006 sample was archived as live vouchers. Molecular methods were used to identify operational taxonomic units (OTUs) that approximate fungal species. A piece of mycelium (~2 mm diameter) was removed from each colony, and genomic DNA was extracted using DNeasy 96-well Plant Kits (Qiagen, Valencia, CA, USA). PCR amplification of the internal transcribed spacer (ITS) region of the ribosomal DNA (including the 5.8S gene), located between the 18S and 28S genes, was accomplished using the primers ITS1F and ITS4 (White et al., 1990; Gardes and Bruns, 1993). Amplified PCR product was visualized on an agarose gel. For the 2006 samples, the PCR product from each isolate was digested with the restriction enzymes MboI and HinfI and unique banding patterns were treated as OTUs. PCR-RFLP profiles were obtained from 154 isolates, including at least 20% of the isolates of each colony morphotype. An Applied Biosystems 3730 Genetics Analyzer (Foster City, California, USA) at the University of Arizona Genetics Core facility was used to obtain sequences of the ITS region from 25 isolates, representing 1 to 5 individuals of all but one OTU (Ogden_32), a singleton with a unique morphology and PCR-RFLP profile that did not provide a good quality sequence. For the 2008 sample, restriction enzyme digestion was not conducted. Instead, the ITS region of a random selection of 10% or more of the fungal colonies representing each unique morphotype (except OTU Ogden_33, a singleton with a unique morphology that would not amplify) was sequenced using an Applied Biosystems 3730 Genetics Analyzer at Northern Arizona University’s Environmental Genetics and Genomics Laboratory (102 sequences total). For both sample dates, the forward and reverse sequences for isolates were edited and combined into a consensus sequence using the program BioEdit (Hall, 1999). From the 2008 sample, 28 of the isolates had good quality sequences in only one direction, and these forward or reverse sequences were also used. The program cd-hit-est (Huang et al., 2010) was used to group sequences at the 95% similarity level, which has been demonstrated as an appropriate similarity cutoff for conservative delimitation of species boundaries of endophytic Ascomycota (U’Ren et al., 2009). All sequences were subject to BLASTn searches in the NCBI DNA sequence database (http://blast.ncbi.nlm.nih.gov) to obtain tentative taxonomic information. Representative sequences from each OTU were deposited in the NCBI DNA sequence database under accession numbers JX978230 to JX978260. Concurrent with endophyte sampling in 2006, twigs were collected to measure CT concentrations. Twigs were selected at a height between 2 and 5.5 m from the ground, the same as for endophyte sampling. From each tree, the first ~20 cm behind the bud scar for the current year was clipped from a single twig and flash-frozen (Fig. 1). This segment encompassed the same age as the tissue
Fig. 1. The locations of the endophyte samples (2006 and 2008) and twig-level plant traits (2006 only) taken from Populus angustifolia twigs. Segments are labeled with the year in which they grew.
470 TABLE 1.
[Vol. 101
AMERICAN JOURNAL OF BOTANY Endophyte variables from each genotypes, for each sample date. Twig segments
RFLPs or sequencesa
Isolates
Tree genotype by year
Trees per genotype
Total
Mean/tree ±1 SD
Total
Mean/tree ±1 SD
2006 10 1005 1008 1012 1017 1023 996 H-10 T-15 WC-5 Total:
4 4 5 4 9 4 9 3 3 4 49
59 60 74 58 135 59 132 45 45 59 726b
14.8 ± 0.5 15.0 ± 0.0 14.0 ± 0.4 14.0 ± 1.0 15.0 ± 0.0 14.8 ± 0.5 14.7 ± 0.5 15.0 ± 0.0 15.0 ± 0.0 14.8 ± 0.5
18 16 25 21 57 30 57 9 20 8 261
4.5 ± 2.6 4.0 ± 2.0 5.0 ± 2.9 5.3 ± 1.0 6.3 ± 2.6 7.5 ± 1.0 6.3 ± 2.9 3.0 ± 1.0 6.7 ± 1.2 2.0 ± 1.8
12 12 15 10 35 15 32 6 13 4 154
2008 1000 1023 11 18 999 Total:
2 4 5 3 3 17
64 128 160 96 96 544
32.0 ± 0.0 32.0 ± 0.0 32.0 ± 0.0 32.0 ± 0.0 32.0 ± 0.0
24 89 125 26 37 301
12.0 ± 5.7 22.3 ± 4.3 25.0 ± 4.9 8.7 ± 3.2 12.3 ± 2.5
14 28 35 11 14 102
Isolation frequency, Mean/tree ±1 SD
OTU richness, Mean/tree ±1 SD
Simpson’s index, Mean/tree ±1 SD
3.0 ± 2.6 3.0 ± 2.2 3.0 ± 2.7 2.5 ± 0.6 3.9 ± 1.5 3.8 ± 1.9 3.6 ± 2.2 2.0 ± 1.0 4.3 ± 2.1 1.0 ± 1.2
30.5 ± 17.5 26.7 ± 13.3 33.9 ± 19.7 36.5 ± 8.4 42.2 ± 17.3 51.0 ± 7.5 43.2 ± 19.7 20.0 ± 6.7 44.4 ± 7.7 13.8 ± 12.9
3.0 ± 1.6 3.0 ± 0.8 3.0 ± 1.2 3.3 ± 0.5 3.1 ± 0.9 3.0 ± 0.8 3.4 ± 1.1 2.0 ± 1.0 3.3 ± 1.2 1.5 ± 12.9
0.51 ± 0.3 0.62 ± 0.1 0.50 ± 0.3 0.64 ± 0.1 0.53 ± 0.2 0.57 ± 0.1 0.60 ± 0.2 0.38 ± 0.3 0.59 ± 0.2 0.51 ± 0.4
7.0 ± 0.0 7.0 ± 2.4 7.0 ± 2.0 3.7 ± 0.6 4.7 ± 0.6
37.5 ± 17.7 69.5 ± 13.6 78.1 ± 15.3 27.1 ± 10.0 38.5 ± 7.9
6.5 ± 0.7 5.8 ± 0.5 6.8 ± 1.3 5.0 ± 0.0 6.0 ± 0.0
0.78 ± 0.0 0.63 ± 0.1 0.73 ± 0.1 0.74 ± 0.1 0.75 ± 0.1
Total Mean/tree ±1 SD
a
The total and average/tree number of RFLPs are reported for 2006 and ITS sequences for 2008. Nine segments were not included in the final 2006 sample because a small number of trees did not yield 15 branches. Note: ITU = internal transcribed spacer, OTU = operational taxonomic units; RFLP = restriction fragment length polymorphism.
b
from which endophytes were isolated (3-yr-old), as well as younger and older tissue. Condensed tannin concentrations in different woody tissues (e.g., twigs and bark from the trunk) of P. angustifolia tend to be similar when compared within an individual tree (L. J. Lamit et al., unpublished data), so the CT concentrations of slightly older and younger tissue should be representative of the twig material that was sampled for endophytes. Samples were freeze-dried and ground in a Wiley mill to pass through a 40-mesh screen. The concentration of soluble CTs was quantified with the acid butanol method of Porter et al. (1986) using purified CTs from P. angustifolia as the standard (Hagerman and Butler, 1994). Two scales of tree growth traits were measured in conjunction with the 2006 endophyte sample to examine potential links between twig endophytes and tree vigor. First, growth traits of individual twigs were measured to represent tree growth in the immediate location of the endophyte community. To assess the possibility that shoot growth during the growing season of community establishment (2003) had a lasting influence on the endophyte community in the 3-yr-old twig segments from which endophytes were isolated, shoot growth from 2003 was measured as the distance between the bud scars created during budbreak in 2003 and 2004 (Fig. 1). These measurements therefore reflect shoot growth during 2003. Additionally, diameter at the middle of the same twig segment was measured (Fig. 1). As twig diameter continues to expand each year, it captures the width expansion from the time of shoot formation in 2003 to the time of diameter measurement (April 2007). Twig width measurements also reflect the differences in sizes of twig segments cultured for endophytes because although the length of the segment was standardized (i.e., 1 cm), the width still varied. On most trees, 15 twigs were measured and averaged. However, fewer twigs were measured on a small number of trees due to lack of available twigs at the standardized sampling height (2–5.5 m). Second, total aboveground tree biomass was measured, which is an indicator of tree size and an integrative metric of a tree’s performance during the course of its life. Total aboveground tree biomass was estimated for each tree using the Populus biomass equation of Fischer et al. (2006), from diameter at breast height (DBH) measured in fall 2007. Because all trees were the same age and grew in the same environment, differences in growth traits among tree genotypes can be attributed to tree genetics. Statistical analyses—The effect of tree genotype on univariate variables was examined using the program R 2.8.1 (R Foundation for Statistical Computing, Vienna, Austria) with the packages lme4, vegan, and RLRsim. Linear models for total isolation frequency, OTU richness, Simpson’s diversity index
and all tree traits (2006 data set only; CTs, twig length, twig diameter, total aboveground biomass) were fit with restricted maximum likelihood (REML; Shaw, 1987), treating genotype as a random effect. After rank transformation, the same methods were used to test for variation among tree genotypes of the two most common endophyte OTUs, whose top BLAST hits closely matched Cytospora chrysosperma (Pers.) Fr. (OTU = Ogden_08) and Phoma Sacc. (OTU = Ogden_22). These OTUs are of special interest because they represented >50% of the isolates recovered in both sampling dates, and species belonging to Cytospora and Phoma are frequently isolated as endophytes from Populus spp. but are pathogenic in some circumstances (Christensen, 1940; Chapela, 1989; de Gruyter and Scheer, 1998; Hutchison, 1999; Santamaria and Diez, 2005; Desprez-Loustau et al., 2006). For each variable, a restricted likelihood ratio test (Crainiceanu and Ruppert, 2004) was used to examine the effect of genotype. Broad-sense heritability (the proportion of total phenotypic variance due to differences among tree genotypes: H2) was calculated from variance components estimated with REML (Shaw, 1987). The effect of tree genotype on community composition (i.e., the individual abundances of all OTUs in an observation treated together as a multivariate variable) was tested using distance-based permutational multivariate analysis of variance (PERMANOVA; Anderson, 2001), coupled with nonmetric multidimensional scaling (NMDS) and indicator species analysis (see McCune and Grace, 2002). A PERMANOVA was performed in the program Primer 6.1.15 with the PERMANOVA+ 1.0.5 add-on (PRIMER-E, Plymouth, UK), while ordinations and indicator species analyses were run in the program PC-ORD 5 (MjM Software, Gleneden Beach, Oregon, USA). Bray–Curtis dissimilarity was used for both NMDS ordinations and PERMANOVA. First, PERMANOVA was run with each data set using the matrix of raw abundances (i.e., the number of isolates) for each OTU. In a second set of PERMANOVA analyses, the abundances of each OTU were relativized by their OTU’s maximum abundance observed in a data set (McCune and Grace, 2002; hereafter referred to as relativized by OTU maximum). This relativization puts the abundances for all OTUs on the same relative scale, thereby reducing the likelihood that compositional differences are driven by a small number of OTUs isolated in high frequencies. PERMANOVA was conducted with genotype as a random effect. Variance components were used to estimate broad-sense community heritability as the variance attributable to genotype divided by the total variance, which is analogous to the method used for univariate variables. To further understand the magnitude of endophyte community variability among tree genotypes, we also report the standard deviation of the effect of genotype,
March 2014]
LAMIT ET AL.—TREE GENETICS AND TWIG ENDOTYPES
calculated as the square root of the variance component due to tree genotype. Variance components from PERMANOVA are expressed in terms of squared units of dissimilarity, Bray–Curtis in this case, which is scaled as a percentage in PRIMER. The square root of the genotype variance component can therefore be interpreted as the average percentage community dissimilarity among genotypes (see details in Anderson et al., 2008). To graphically represent composition, NMDS ordinations were performed with data relativized by OTU maximum values. Unique groupings of genotypes were observed in the NMDS ordinations; therefore, indicator species analysis was applied to the OTU matrices to identify which OTUs were driving the clustering of genotypes. The 2006 endophyte sample was used to test our hypothesis that the four quantitative tree traits (CTs, twig length, twig diameter, aboveground biomass) were linked to variation in the endophyte community. These analyses were focused on two endophyte community traits, total isolation frequency, and OTU composition (using data relativized by OTU maximum values), which showed the most consistent effects of tree genotype. First, bivariate relationships between each tree trait and each endophyte variable were examined using individual tree phenotype values, and mean values for each genotype. Pearson’s correlation was used for tests with total endophyte isolation frequency, and distance-based linear model regression (DISTLM) was used for endophyte composition. To obtain a statistic from DISTLM that was comparable to the r-value from Pearson’s correlations, the R2 value was square-root-transformed (Lamit et al., 2011). To understand how the four tree traits related to each other, we ran Pearson’s correlation analyses between each pair of traits using individual tree phenotype and genotype mean values. Next, the effect of each tree trait on endophytes was examined while controlling for the effects of the other tree traits and tree genotype. With these analyses, we hypothesized that if a tree trait remains significant while holding genotype and the other traits constant, the significant trait may be linked to the underlying mechanism by which tree genotype influences endophytes. Total endophyte isolation frequency was examined with a REML-based linear model with tree genotype as a random effect and all tree traits as covariates. A restricted likelihood ratio test was used to test the effect of tree genotype, while Wald χ2 tests were used to test covariates. Equivalent models were examined for OTU composition using PERMANOVA with Type III sums of squares. These analyses were performed with the packages lme4, RLRsim and car in R 2.8.1 and Primer 6.1.15 with PERMANOVA+ 1.0.5. PERMANOVA and DISTLM were implemented with Bray–Curtis dissimilarity. Total aboveground tree biomass was natural-log-transformed in all analyses to linearize its relationships with response variables. Patterns of endophyte OTU co-occurrences, which may be indicative of species interactions, were examined with the C-score (Stone and Roberts, 1990). The C-score is computed as the average number of checkerboard units (i.e., one OTU occurs where another does not) among all OTU pairs in the community. This metric is robust to Type I error (Gotelli, 2000) and has been used previously to infer patterns of species interactions among endophytes (Pan and May, 2009). To test the significance of the observed C-score, we used a null modeling approach in which presence–absence matrices of OTUs were randomized with an algorithm that maintained observation totals and the relative number of isolates for OTUs (i.e., fixed-fixed algorithm; Gotelli, 2000). This algorithm was chosen because relative and total abundances were expected to differ among OTUs and observations, respectively. A total of 10 000 randomizations were run for each analysis. A single P-value for each analysis was calculated as the number of simulated C-scores equal to or more extreme than the observed C-score for each community. Because our hypothesis about co-occurrences was nondirectional, we used an alpha-level of 0.025 to test each C-score. A standardized effect size (SES), the difference between the observed C-score and the mean of the simulated C-scores, was generated for all analyses to facilitate comparisons among analyses. Higher SES values are indicative of negative relationships, as this results from OTUs tending not to co-occur (i.e., more checkerboard units). SES values less than zero are indicative of positive relationships that result in aggregation of species (i.e., positive co-occurrences). Three co-occurrence analyses were conducted separately for both endophyte sample dates: (1) using all trees pooled together, with the communities from each tree treated as replicates (generating a single C-score and SES value), (2) using the mean values of species abundances for each genotype in a single analysis (also generating a single C-score and SES value), and (3) individual analyses for each genotype with only communities from replicates of a single genotype at a time (generating a C-score and SES value for each genotype). Genotype 1000 (2008 sample) was excluded from analyses because it only had two replicates. Co-occurrence analyses were conducted in R using the commsimulator function in the bipartite package.
471 RESULTS
The endophytic fungal community and its response to tree genotype— A total of 33 fungal OTUs were recovered (Fig. 2; Appendix S2). Fungi emerged from 261 of the 726 twig segments in 2006 (~36%). These fungi grouped into 12 unique PCRRFLP types, seven of which were singletons (Fig. 2A; Appendix S2). Several isolates in the 2006 data set were difficult to separate morphologically or with PCR-RFLP; BLAST results suggested these represented several members of the genus Alternaria, and they were lumped into one OTU group for statistical analyses of the 2006 data set (OTU = AltComp2006 in Fig. 2A). Of the 576 twig segments plated in 2008, 301 grew endophytes (~52%), which represented 27 OTUs including 13 singletons (Fig. 2B; Appendix S2). All top BLAST matches were fungi from the division Ascomycota, and many were taxa sampled from plant tissues, including Populus spp. (Appendix S2). Consistent with our hypothesis that tree genotype influences endophyte community structure, some endophyte variables exhibited significant variation among tree genotypes in the 2006 data set (Table 2; Figs. 2A, 3A, B). Total endophyte isolation frequency differed among tree genotypes, with broad-sense heritability estimates indicating that 22% of the total variation in isolation frequency was due to tree genotype (Table 2; Fig. 3A). In contrast, species richness and Simpson’s diversity showed no evidence of a response to genotype (Table 2). Community composition exhibited a significant nonrandom pattern among tree genotypes, but because total variation of endophyte composition was high (i.e., variance due to genotype plus residual variance), broad-sense heritability estimates suggested that tree genotype accounted for 9% of the total variation in composition (using raw or relativized data; Table 2; Fig. 3B). However, in terms of Bray– Curtis dissimilarity, the compositions of endophyte communities were 13.7–15.2% dissimilar among genotypes, on average (square root of genotype variance component from PERMANOVA for the raw community matrix = 13.7, and for the matrix relativized by OTU maximum values = 15.2). Between the two distinct groupings of communities in tree genotypes observed in the NMDS ordination (Fig. 3B), indicator species analysis revealed that OTU Ogden_06 was an indicator of group 1 (genotypes WC-5, 10, H-10, 1008, 1005; indicator value = 39.5, P = 0.050), while Ogden_12 (indicator value = 51.1, P = 0.002) and Ogden_08 (indicator value = 67.7, P < 0.001) were associated with genotype group 2 (genotypes 1012, 1017, 996, 1023, T-15). Of the two individual OTUs who were examined individually, Ogden_08 exhibited significant variation among tree genotypes (RLR = 2.501, df = 1, P = 0.043, H2 = 0.210) while Ogden_22 did not (RLR = 0.000, df = 1, P = 1.00, H2 = 0.000). Communities isolated for the 2008 data set were also sensitive to tree genotype (Table 2; Figs. 2B, 3C, 3D), lending further support to our hypothesis that tree genotype influences endophyte community structure. Broad-sense heritability estimates indicated that tree genotype explained 72% of the variation in total isolation frequency and had a marginal effect on OTU richness, while Simpson’s diversity showed no evidence of variation among tree genotypes (Table 2; Fig. 3C). Community composition exhibited a significant nonrandom pattern among tree genotypes, and broadsense heritability estimates suggested that tree genotype accounted for 24% and 11% of the total variation in composition (using raw and relativized data, respectively; Table 2; Fig. 3D). In terms of Bray–Curtis dissimilarity, the compositions of endophyte communities were 17.4–22.8% dissimilar among genotypes, on average (square root of genotype variance component from PERMANOVA
472
AMERICAN JOURNAL OF BOTANY
[Vol. 101
Fig. 2. Relative isolation frequencies for endophytes from the (A) 2006 and (B) 2008 data sets, based on mean values of each operational taxonomic unit (OTU) for each genotype. Rare OTUs are present 1–3 times in a year’s entire data set.
for the raw community matrix = 22.8, and for the matrix relativized by OTU maximum values = 17.4). Among the three clusters of genotypes observed in the NMDS ordination (see Fig. 3D), indicator species analysis revealed that OTU Ogden_03 was a marginally significant indicator of group 1 (genotype 18; indicator value = 48.5, P = 0.078), Ogden_12 (indicator value = 56.5, P = 0.032) and Ogden_22 (indicator value = 74.1, P = 0.002) were associated with genotype group 2 (genotypes 999, 11, 1023), while group 3 (genotype 1000) had no significant indicators. Of the two individual OTUs who were examined individually, Ogden_08 exhibited significant variation among tree genotypes (RLR = 6.281, df = 1, P = 0.006, H2 = 0.590) while Ogden_22 showed marginally significant variation among genotypes (RLR = 2.080, df = 1, P = 0.064, H2 = 0.386). Relationships between tree traits and twig endophyte communities— All plant traits varied significantly among tree genotypes (Table 2; condensed tannins: mean = 2.11% dry mass, SD = 0.52% dry mass; aboveground tree biomass: mean = 198
711.1 g, SD = 169 968.5 g; 2003 twig length: mean = 4.83 cm, SD = 3.01 cm; 2003 twig diameter: mean = 0.33 cm, SD = 0.066 cm). Tree genotype explained 21–48% of the variance in the tree traits, demonstrating that these traits have a genetic basis (Table 2). Aboveground tree biomass exhibited significant positive phenotypic and genotypic correlations with twig diameter, while twig diameter exhibited a positive phenotypic correlation with twig length of the year of establishment and a similar (but nonsignificant) trend with twig length using genotype mean values (Appendix S3). Results suggest that variation in aboveground tree biomass and twig diameter are associated with variation in the endophyte community (Tables 3 and 4; Fig. 4). Correlation analyses show that total aboveground biomass and twig diameter were positively associated with total endophyte isolation frequency using tree phenotype data, but only aboveground biomass showed a significant positive relationship with isolation frequency using genotype mean values (Table 3). Condensed tannin concentrations and twig length were not significantly
March 2014]
LAMIT ET AL.—TREE GENETICS AND TWIG ENDOTYPES
TABLE 2.
Effect of tree genotype on endophyte community variables for each sample date and on tree traits for the 2006 data set. RLR or Pseudo-F
Response variable 2006 Endophyte isolation frequency Endophyte OTU richness Endophyte Simpson’s diversity index Endophyte OTU composition Endophyte OTU composition (relativized by OTU maximum) Condensed tannin concentration (% dry mass) Aboveground tree biomass (g) Twig length: year of community establishment, 2003 (cm) Twig diameter: year of community establishment, 2003 (cm) 2008 Endophyte isolation frequency Endophyte OTU richness Endophyte Simpson’s diversity index Endophyte OTU composition Endophyte OTU composition (relativized by OTU maximum)
P
H2
2.747 0.000 0.000 1.453 1.500
0.040 0.435 1.000 0.051 0.024
0.22 0.00 0.00 0.09 0.09
12.487
TREE GENOTYPE AND GENETICALLY BASED GROWTH TRAITS STRUCTURE TWIG ENDOPHYTE COMMUNITIES1
LOUIS J. LAMIT2,5, MATTHEW K. LAU2, CHRISTOPHER M. STHULTZ3, STUART C. WOOLEY4, THOMAS G. WHITHAM2, AND CATHERINE A. GEHRING2 2Department
of Biological Sciences and Merriam-Powell Center for Environmental Research, Northern Arizona University, Flagstaff, Arizona 86011-5640 USA; 3Math, Science & Technology Department, University of Minnesota Crookston, Crookston, Minnesota 56716 USA; and 4Department of Biological Sciences, California State University Stanislaus, Turlock, California 95382 USA • Premise of the study: Fungal endophytes asymptomatically inhabit plant tissues where they have mutualistic, parasitic, or commensal relationships with their hosts. Although plant–fungal interactions at the genotype scale have broad ecological and evolutionary implications, the sensitivity of endophytes in woody tissues to differences among plant genotypes is poorly understood. We hypothesize that (1) endophyte communities in Populus angustifolia (Salicaceae) twigs vary among tree genotypes, (2) endophyte variation is linked to quantitative tree traits, and (3) tree genotype influences interspecific fungal interactions. • Methods: Endophytes were isolated from twigs of replicated P. angustifolia genotypes in a common garden and characterized with PCR-RFLP and DNA sequencing. Twig length and diameter, aboveground tree biomass, and condensed tannins were also quantified. • Key results: (1) Aspects of fungal community structure, including composition and total isolation frequency (i.e., abundance), varied among genotypes. (2) Aboveground biomass and twig diameter were positively associated with isolation frequency and covaried with composition, whereas twig length and condensed tannin concentration were not significantly correlated to endophytes. (3) Fungal co-occurrence patterns suggested negative species interactions, but the presence of significant co-occurrences was genotype dependent. • Conclusions: The species is often assumed to be the most important ecological unit; however, these results indicate that genetically based trait variation within a species can influence an important community of associated organisms. Given the dominance of plants as primary producers and the ubiquity of endophytes, the effect of host genetic variation on endophytes has fundamental implications for our understanding of terrestrial ecosystems. Key words: aboveground biomass; condensed tannins; co-occurrence analysis; endophyte; Populus angustifolia; Salicaceae; tree genotype.
Plants are typically viewed as discrete entities; often overlooked are the numerous plant-associated fungi that can have strong influences on plant growth, pest resistance, and stress tolerance (Rodriguez et al., 2009; Hoeksema et al., 2010). Of growing interest are Type 3 endophytes (Rodriguez et al., 2009; subsequently referred to simply as endophytes). These fungi
live within apparently healthy aboveground tissues of all plants surveyed, are horizontally transmitted, can form diverse communities with members representing distantly related fungal lineages, and are believed to represent an important component of the millions of undescribed fungal species (Arnold, 2007; Blackwell, 2011). Endophytes obtain nutrients and energy from plants and, in some cases, modulate host traits such as drought tolerance, photosynthetic efficiency, growth, and pathogen and herbivore resistance (Arnold, 2007; Rodriguez et al., 2009). Given the intimate link between endophytes and their hosts, genotypic differences within host species may have important effects on the structure of endophyte communities, which may feed back to influence plant growth and fitness. Furthermore, because endophytes are ubiquitous within plants, the primary source of energy fixation for terrestrial ecosystems, understanding the relationship between intraspecific host plant variation and endophytic fungi is fundamental to our understanding of the functioning of terrestrial ecosystems. Historically, research on endophytes largely focused on characterizing community shifts among host plant species, tissue types, and tissue ages, or across spatial and temporal scales (e.g., Carroll and Carroll, 1978; Faeth and Hammon, 1996; Collado et al., 1999; Sahashi et al., 1999; Arnold et al., 2001, 2003; Santamaria and Diez, 2005; U’Ren et al., 2012; Zimmerman and Vitousek, 2012; de Souza Sebastianes et al., 2013). These studies show that endophytes can exhibit various levels of host
1 Manuscript received 28 January 2014; revision accepted 4 February 2014. The authors thank R. J. Deckert, A. E. Arnold, T. Wojtowicz, and members of the Gehring Laboratory and Cottonwood Ecology Group for technical assistance and/or comments on this manuscript and the Ogden Nature Center for support of our common gardens and field housing. Special thanks to C. M. Pace for drawing Fig. 1. Chemical analyses were performed in R. L. Lindroth’s laboratory at the University of Wisconsin, Madison. Research was supported by the National Science Foundation Frontiers in Integrative Biological Research grant DEB-0425908 to the Cottonwood Ecology Group. Additional support to L. J. Lamit was provided by a National Science Foundation Integrative Graduate Education and Research Traineeship, an Achievement Rewards for College Scientists scholarship, and a Mycological Society of America graduate fellowship. 5 Author for correspondence (e-mail: [email protected]); present address: Michigan Technological University, School of Forest Resources and Environmental Science, 1400 Townsend Dr., Houghton, MI 49931-1295 USA
doi:10.3732/ajb.1400034
American Journal of Botany 101(3): 467–478, 2014; http://www.amjbot.org/ © 2014 Botanical Society of America
467
468
[Vol. 101
AMERICAN JOURNAL OF BOTANY
and tissue specificity. In some cases, endophytes show fidelity to specific host species even across large temporal scales (e.g., Carroll and Carroll, 1978). In contrast, communities frequently vary across large and small spatial scales, even when the same host species is sampled (e.g., Arnold et al., 2001; U’Ren et al., 2012; Zimmerman and Vitousek, 2012). Importantly, Elamo et al. (1999) provided evidence that abundances of leaf endophyte species vary among maternal families of mountain birch (Betula pubescens subsp. czerepanovii (Orlova) Hämet-Ahti) in two spatially separated common gardens, located in contrasting environments. This finding points to the potential importance of host plant genetics in shaping the distributions of endophytes. Endophyte sensitivity to host genetics has important evolutionary implications. Selection, founder effects, and genetic drift may influence heritable plant traits important to endophytes, leading to shifts in endophyte communities in response to host evolution. The influence of endophytes, either positive or negative, may also be a selective force on host plants (Elamo et al., 1999; Miglia et al., 2007), altering the relative fitness of plant genotypes. Furthermore, phenotypic differences among host genotypes may alter endophyte interactions with one another, potentially altering their population dynamics and evolution. Although host–endophyte evolutionary dynamics have been discussed for more than three decades (e.g., Carroll, 1988; Todd, 1988; Elamo et al., 1999; Miglia et al., 2007; Arnold et al., 2009), the role of host genotype in shaping host–endophyte interactions and endophyte–endophyte interactions needs to be more fully elucidated. A small number of studies show that endophytes can be sensitive to intraspecific variation in plant genetics (e.g., Todd, 1988; Elamo et al., 1999; Pan et al, 2008; but see Miglia et al., 2007; Korkama-Rajala et al., 2008). However, there are at least three significant gaps in our knowledge that warrant further investigation of plant genetic effects on endophytes. First, the majority of endophyte research with woody plants focuses on fungi inhabiting leaves and needles. This is despite the apparent ubiquity of fungi in woody tissues (e.g., Chapela, 1989; Sieber et al., 1991; Hutchison, 1999; Sahashi et al., 1999; Santamaria and Diez, 2005; de Souza Sebastianes et al., 2013), data suggesting that endophyte communities can differ between woody tissue and leaves (e.g., Sieber et al., 1991; Sahashi et al., 1999; Santamaria and Diez, 2005; de Souza Sebastianes et al., 2013; but see discussion in Arnold, 2007), and the likelihood that some of these fungi are economically important latent pathogens (Christensen, 1940; Chapela, 1989; Desprez-Loustau et al., 2006). Second, there is little understanding of the genetically variable plant traits to which endophytes may be linked. These plant traits may include phytochemistry (Bailey et al., 2005) and productivity (Todd, 1988). Identification of these traits is important if we are to understand the ecological and evolutionary forces that shape plant–endophyte interactions and will help further the development of a predictive-mechanistic framework for understanding endophyte distributions. Third, it is unclear whether differences in plant genotypes modulate interspecific interactions among endophytes. Work with species interactions using other fungi demonstrates that interactions can play important roles in structuring communities (e.g., Koide et al., 2005; Fukami et al., 2010), and limited research shows that endophytes co-occur in patterns indicative of interspecific interactions (Pan and May, 2009). Modulation of endophyte species interactions by tree genotype has implications for endophyte community structure and function, which may cause genotype-specific feedbacks to host trees.
In the present study, we examine the fungal communities associated with Populus angustifolia James (Salicaceae; narrowleaf cottonwood) twigs to understand how endophytes respond to genotypic differences in their host. First, we hypothesize that endophyte community structure (i.e., isolation frequency [a proxy measure for abundance], richness, diversity, and composition) will vary among tree genotypes in a common garden. This approach is particularly strong for addressing this hypothesis because a common garden contains replicated genotypes of the same age, planted in a random order over a relatively homogenous area, thus allowing for the control of confounding environmental and age influences that occur in naturally established field sites. Second, we hypothesize that endophytes will show relationships with quantitative tree traits such as growth (including twig length, twig diameter, and aboveground biomass) and condensed tannins, which are known to be genetically based in cottonwoods (e.g., Bailey et al., 2005; Lojewski et al., 2009). The plant vigor hypothesis (Price, 1991) suggests that herbivores intimately associated with host tissues (e.g., gallers and stem borers) preferentially attack vigorously growing plants because they contain more resources (e.g., carbohydrates, nutrients). Because of their close association with plant tissues, endophytes may also favor vigorously growing plants, leading to a variety of potential community patterns in relation to plant productivity. We predict that endophyte isolation frequency will be positively associated with increases in tree growth, as has been found for arthropods (Price, 1991; Stone et al., 2010) and that endophyte species composition will covary with variation in growth among trees. In contrast, we predict that endophyte isolation frequency will be negatively related to the concentration of condensed tannins (CTs) in twigs due to the potential antifungal properties of CTs (Kraus et al., 2003; Bailey et al., 2005; Holeski et al., 2009), and that CTs will likely affect other aspects of endophyte community structure. Third, we hypothesize that endophytes will exhibit cooccurrence patterns indicative of interspecific interactions, as commonly seen in other microbial communities (HornerDevine et al., 2007). Importantly, co-occurrence patterns will differ among tree genotypes because each genotype represents a unique habitat defined by its genetically based traits. Identifying relationships among tree genotype, tree traits, and twig endophytes is a key step in understanding the hidden evolutionary and ecological dynamics that govern the distribution of these ubiquitous plant symbionts. MATERIALS AND METHODS Study system—Populus angustifolia is a mid to upper elevation (~1300– 2500 m a.s.l. in the region of the present study) foundation tree species in riparian habitats of interior western North America. Sampling efforts were focused on replicated genotypes of P. angustifolia growing in a common garden established in 1991. The garden is located in Ogden, Utah (latitude = 41.248146, longitude = −111.999830, elevation = 1302 m a.s.l.) and was planted with clonally replicated Populus genotypes from cuttings collected along the length of the nearby Weber River. The garden is contained within an approximately 1.2 ha area, with trees planted 4–7 m apart. Trees were planted in random order, and by chance, members of the subset of trees used in this study ranged from neighbors (within 4–7 m) to across the length of the garden (~230 m) from each other. Spatial analyses indicated that trees in close proximity were not more likely to have similar endophyte communities relative to trees far away from each other (Appendix S1, see Supplemental Data with the online version of this article). Each tree genotype was characterized with 35 codominant restriction fragment length polymorphism (RFLP) markers to verify that all were genetically distinct (Martinsen et al., 2001; M. Zinkgraf et al., Northern Arizona
March 2014]
LAMIT ET AL.—TREE GENETICS AND TWIG ENDOTYPES
University, unpublished data). A hybrid index generated from the RFLP markers indicated that all but one genotype used in this study were P. angustifolia genotypes; genotype 18, from the 2008 data set, was a backcross hybrid with P. fremontii S. Watson (Lamit et al., 2011; M. Zinkgraf et al., Northern Arizona University, unpublished data). Sampling methods—Fungal endophyte communities were examined from multiple tree genotypes on two dates, approximately 2 yr apart. The goal with the two samples was not to make direct comparisons between the 2 yr. The first sample (collected mid-June 2006, during the growing season) was collected to assess the response of endophytes in twigs to differences among tree genotypes and to examine potential links between genetically based tree traits and endophytes. The second sample (collected in late April 2008, during leaf flush) was collected to complement but not repeat the first to (1) verify that twig endophyte communities differ among tree genotypes, in a mostly independent set of trees (only one tree genotype overlapped between the two samples, genotype 1023), (2) sample more 1-cm twig segments from a tree, covering greater shoot age, (3) let fungal isolates emerge for a longer time, and (4) identify isolates with higher resolution molecular methods. The two endophyte samples were collected as follows (see Fig. 1). In midJune 2006, 15 twigs were selected from a minimum of three locations around each tree (10 genotypes, 3–9 trees per genotype, 49 trees total), although only 13 or 14 twigs could be obtained from a very small number of trees (Table 1). A single 1-cm-long segment was taken from the middle of the 3-yr-old growth section (formed in 2003) from each twig for fungal isolations (726 total segments; Table 1). In late April 2008, two twigs were cut from the north and south sides from each tree (5 genotypes, 2–5 trees per genotype, 17 trees total), and the entire 2- and 3-yr-old growth sections (formed in 2005 and 2006) were cut into 1-cm segments, pooled per age class and 32 segments per tree (16 per age class, 544 segments total) were selected at random for isolations (Table 1). Fewer twigs were taken from each tree in the 2008 sample to reduce damage to trees caused by clipping large numbers of twigs. Twigs for each sample year were selected from the trees at a height between 2 and 5.5 m from the ground. The following isolation and culturing procedures were used for both sampling dates. Twig segments were rinsed with tap water to remove adhering debris, and then surface sterilized by immersion for 5 min in 70% ethanol, followed by 5 min in 50% Clorox bleach, then three successive rinses in sterile reverse osmosis water (Bailey et al., 2005). During each step, twig segments were vigorously agitated. Twig segments were individually placed on Petri dishes of 2% potato dextrose agar (PDA), and left for one (2006 sample) or two (2008 sample) months. Petri dishes were examined weekly for fungal growth. Colonies of fungi were sorted into morphological categories (morphotypes) based on mycelia color, texture, and growth characteristics at the center and edges of the isolate, hyphal height above the medium, hyphal depth into the
469
medium, characteristics of reproductive structures, growth-rate, and color changes in the PDA. Endophyte isolation frequency was expressed as the percentage of twig segments from a single tree that grew fungi. Representatives of each morphotype were photographed, and a subset from the 2006 sample was archived as live vouchers. Molecular methods were used to identify operational taxonomic units (OTUs) that approximate fungal species. A piece of mycelium (~2 mm diameter) was removed from each colony, and genomic DNA was extracted using DNeasy 96-well Plant Kits (Qiagen, Valencia, CA, USA). PCR amplification of the internal transcribed spacer (ITS) region of the ribosomal DNA (including the 5.8S gene), located between the 18S and 28S genes, was accomplished using the primers ITS1F and ITS4 (White et al., 1990; Gardes and Bruns, 1993). Amplified PCR product was visualized on an agarose gel. For the 2006 samples, the PCR product from each isolate was digested with the restriction enzymes MboI and HinfI and unique banding patterns were treated as OTUs. PCR-RFLP profiles were obtained from 154 isolates, including at least 20% of the isolates of each colony morphotype. An Applied Biosystems 3730 Genetics Analyzer (Foster City, California, USA) at the University of Arizona Genetics Core facility was used to obtain sequences of the ITS region from 25 isolates, representing 1 to 5 individuals of all but one OTU (Ogden_32), a singleton with a unique morphology and PCR-RFLP profile that did not provide a good quality sequence. For the 2008 sample, restriction enzyme digestion was not conducted. Instead, the ITS region of a random selection of 10% or more of the fungal colonies representing each unique morphotype (except OTU Ogden_33, a singleton with a unique morphology that would not amplify) was sequenced using an Applied Biosystems 3730 Genetics Analyzer at Northern Arizona University’s Environmental Genetics and Genomics Laboratory (102 sequences total). For both sample dates, the forward and reverse sequences for isolates were edited and combined into a consensus sequence using the program BioEdit (Hall, 1999). From the 2008 sample, 28 of the isolates had good quality sequences in only one direction, and these forward or reverse sequences were also used. The program cd-hit-est (Huang et al., 2010) was used to group sequences at the 95% similarity level, which has been demonstrated as an appropriate similarity cutoff for conservative delimitation of species boundaries of endophytic Ascomycota (U’Ren et al., 2009). All sequences were subject to BLASTn searches in the NCBI DNA sequence database (http://blast.ncbi.nlm.nih.gov) to obtain tentative taxonomic information. Representative sequences from each OTU were deposited in the NCBI DNA sequence database under accession numbers JX978230 to JX978260. Concurrent with endophyte sampling in 2006, twigs were collected to measure CT concentrations. Twigs were selected at a height between 2 and 5.5 m from the ground, the same as for endophyte sampling. From each tree, the first ~20 cm behind the bud scar for the current year was clipped from a single twig and flash-frozen (Fig. 1). This segment encompassed the same age as the tissue
Fig. 1. The locations of the endophyte samples (2006 and 2008) and twig-level plant traits (2006 only) taken from Populus angustifolia twigs. Segments are labeled with the year in which they grew.
470 TABLE 1.
[Vol. 101
AMERICAN JOURNAL OF BOTANY Endophyte variables from each genotypes, for each sample date. Twig segments
RFLPs or sequencesa
Isolates
Tree genotype by year
Trees per genotype
Total
Mean/tree ±1 SD
Total
Mean/tree ±1 SD
2006 10 1005 1008 1012 1017 1023 996 H-10 T-15 WC-5 Total:
4 4 5 4 9 4 9 3 3 4 49
59 60 74 58 135 59 132 45 45 59 726b
14.8 ± 0.5 15.0 ± 0.0 14.0 ± 0.4 14.0 ± 1.0 15.0 ± 0.0 14.8 ± 0.5 14.7 ± 0.5 15.0 ± 0.0 15.0 ± 0.0 14.8 ± 0.5
18 16 25 21 57 30 57 9 20 8 261
4.5 ± 2.6 4.0 ± 2.0 5.0 ± 2.9 5.3 ± 1.0 6.3 ± 2.6 7.5 ± 1.0 6.3 ± 2.9 3.0 ± 1.0 6.7 ± 1.2 2.0 ± 1.8
12 12 15 10 35 15 32 6 13 4 154
2008 1000 1023 11 18 999 Total:
2 4 5 3 3 17
64 128 160 96 96 544
32.0 ± 0.0 32.0 ± 0.0 32.0 ± 0.0 32.0 ± 0.0 32.0 ± 0.0
24 89 125 26 37 301
12.0 ± 5.7 22.3 ± 4.3 25.0 ± 4.9 8.7 ± 3.2 12.3 ± 2.5
14 28 35 11 14 102
Isolation frequency, Mean/tree ±1 SD
OTU richness, Mean/tree ±1 SD
Simpson’s index, Mean/tree ±1 SD
3.0 ± 2.6 3.0 ± 2.2 3.0 ± 2.7 2.5 ± 0.6 3.9 ± 1.5 3.8 ± 1.9 3.6 ± 2.2 2.0 ± 1.0 4.3 ± 2.1 1.0 ± 1.2
30.5 ± 17.5 26.7 ± 13.3 33.9 ± 19.7 36.5 ± 8.4 42.2 ± 17.3 51.0 ± 7.5 43.2 ± 19.7 20.0 ± 6.7 44.4 ± 7.7 13.8 ± 12.9
3.0 ± 1.6 3.0 ± 0.8 3.0 ± 1.2 3.3 ± 0.5 3.1 ± 0.9 3.0 ± 0.8 3.4 ± 1.1 2.0 ± 1.0 3.3 ± 1.2 1.5 ± 12.9
0.51 ± 0.3 0.62 ± 0.1 0.50 ± 0.3 0.64 ± 0.1 0.53 ± 0.2 0.57 ± 0.1 0.60 ± 0.2 0.38 ± 0.3 0.59 ± 0.2 0.51 ± 0.4
7.0 ± 0.0 7.0 ± 2.4 7.0 ± 2.0 3.7 ± 0.6 4.7 ± 0.6
37.5 ± 17.7 69.5 ± 13.6 78.1 ± 15.3 27.1 ± 10.0 38.5 ± 7.9
6.5 ± 0.7 5.8 ± 0.5 6.8 ± 1.3 5.0 ± 0.0 6.0 ± 0.0
0.78 ± 0.0 0.63 ± 0.1 0.73 ± 0.1 0.74 ± 0.1 0.75 ± 0.1
Total Mean/tree ±1 SD
a
The total and average/tree number of RFLPs are reported for 2006 and ITS sequences for 2008. Nine segments were not included in the final 2006 sample because a small number of trees did not yield 15 branches. Note: ITU = internal transcribed spacer, OTU = operational taxonomic units; RFLP = restriction fragment length polymorphism.
b
from which endophytes were isolated (3-yr-old), as well as younger and older tissue. Condensed tannin concentrations in different woody tissues (e.g., twigs and bark from the trunk) of P. angustifolia tend to be similar when compared within an individual tree (L. J. Lamit et al., unpublished data), so the CT concentrations of slightly older and younger tissue should be representative of the twig material that was sampled for endophytes. Samples were freeze-dried and ground in a Wiley mill to pass through a 40-mesh screen. The concentration of soluble CTs was quantified with the acid butanol method of Porter et al. (1986) using purified CTs from P. angustifolia as the standard (Hagerman and Butler, 1994). Two scales of tree growth traits were measured in conjunction with the 2006 endophyte sample to examine potential links between twig endophytes and tree vigor. First, growth traits of individual twigs were measured to represent tree growth in the immediate location of the endophyte community. To assess the possibility that shoot growth during the growing season of community establishment (2003) had a lasting influence on the endophyte community in the 3-yr-old twig segments from which endophytes were isolated, shoot growth from 2003 was measured as the distance between the bud scars created during budbreak in 2003 and 2004 (Fig. 1). These measurements therefore reflect shoot growth during 2003. Additionally, diameter at the middle of the same twig segment was measured (Fig. 1). As twig diameter continues to expand each year, it captures the width expansion from the time of shoot formation in 2003 to the time of diameter measurement (April 2007). Twig width measurements also reflect the differences in sizes of twig segments cultured for endophytes because although the length of the segment was standardized (i.e., 1 cm), the width still varied. On most trees, 15 twigs were measured and averaged. However, fewer twigs were measured on a small number of trees due to lack of available twigs at the standardized sampling height (2–5.5 m). Second, total aboveground tree biomass was measured, which is an indicator of tree size and an integrative metric of a tree’s performance during the course of its life. Total aboveground tree biomass was estimated for each tree using the Populus biomass equation of Fischer et al. (2006), from diameter at breast height (DBH) measured in fall 2007. Because all trees were the same age and grew in the same environment, differences in growth traits among tree genotypes can be attributed to tree genetics. Statistical analyses—The effect of tree genotype on univariate variables was examined using the program R 2.8.1 (R Foundation for Statistical Computing, Vienna, Austria) with the packages lme4, vegan, and RLRsim. Linear models for total isolation frequency, OTU richness, Simpson’s diversity index
and all tree traits (2006 data set only; CTs, twig length, twig diameter, total aboveground biomass) were fit with restricted maximum likelihood (REML; Shaw, 1987), treating genotype as a random effect. After rank transformation, the same methods were used to test for variation among tree genotypes of the two most common endophyte OTUs, whose top BLAST hits closely matched Cytospora chrysosperma (Pers.) Fr. (OTU = Ogden_08) and Phoma Sacc. (OTU = Ogden_22). These OTUs are of special interest because they represented >50% of the isolates recovered in both sampling dates, and species belonging to Cytospora and Phoma are frequently isolated as endophytes from Populus spp. but are pathogenic in some circumstances (Christensen, 1940; Chapela, 1989; de Gruyter and Scheer, 1998; Hutchison, 1999; Santamaria and Diez, 2005; Desprez-Loustau et al., 2006). For each variable, a restricted likelihood ratio test (Crainiceanu and Ruppert, 2004) was used to examine the effect of genotype. Broad-sense heritability (the proportion of total phenotypic variance due to differences among tree genotypes: H2) was calculated from variance components estimated with REML (Shaw, 1987). The effect of tree genotype on community composition (i.e., the individual abundances of all OTUs in an observation treated together as a multivariate variable) was tested using distance-based permutational multivariate analysis of variance (PERMANOVA; Anderson, 2001), coupled with nonmetric multidimensional scaling (NMDS) and indicator species analysis (see McCune and Grace, 2002). A PERMANOVA was performed in the program Primer 6.1.15 with the PERMANOVA+ 1.0.5 add-on (PRIMER-E, Plymouth, UK), while ordinations and indicator species analyses were run in the program PC-ORD 5 (MjM Software, Gleneden Beach, Oregon, USA). Bray–Curtis dissimilarity was used for both NMDS ordinations and PERMANOVA. First, PERMANOVA was run with each data set using the matrix of raw abundances (i.e., the number of isolates) for each OTU. In a second set of PERMANOVA analyses, the abundances of each OTU were relativized by their OTU’s maximum abundance observed in a data set (McCune and Grace, 2002; hereafter referred to as relativized by OTU maximum). This relativization puts the abundances for all OTUs on the same relative scale, thereby reducing the likelihood that compositional differences are driven by a small number of OTUs isolated in high frequencies. PERMANOVA was conducted with genotype as a random effect. Variance components were used to estimate broad-sense community heritability as the variance attributable to genotype divided by the total variance, which is analogous to the method used for univariate variables. To further understand the magnitude of endophyte community variability among tree genotypes, we also report the standard deviation of the effect of genotype,
March 2014]
LAMIT ET AL.—TREE GENETICS AND TWIG ENDOTYPES
calculated as the square root of the variance component due to tree genotype. Variance components from PERMANOVA are expressed in terms of squared units of dissimilarity, Bray–Curtis in this case, which is scaled as a percentage in PRIMER. The square root of the genotype variance component can therefore be interpreted as the average percentage community dissimilarity among genotypes (see details in Anderson et al., 2008). To graphically represent composition, NMDS ordinations were performed with data relativized by OTU maximum values. Unique groupings of genotypes were observed in the NMDS ordinations; therefore, indicator species analysis was applied to the OTU matrices to identify which OTUs were driving the clustering of genotypes. The 2006 endophyte sample was used to test our hypothesis that the four quantitative tree traits (CTs, twig length, twig diameter, aboveground biomass) were linked to variation in the endophyte community. These analyses were focused on two endophyte community traits, total isolation frequency, and OTU composition (using data relativized by OTU maximum values), which showed the most consistent effects of tree genotype. First, bivariate relationships between each tree trait and each endophyte variable were examined using individual tree phenotype values, and mean values for each genotype. Pearson’s correlation was used for tests with total endophyte isolation frequency, and distance-based linear model regression (DISTLM) was used for endophyte composition. To obtain a statistic from DISTLM that was comparable to the r-value from Pearson’s correlations, the R2 value was square-root-transformed (Lamit et al., 2011). To understand how the four tree traits related to each other, we ran Pearson’s correlation analyses between each pair of traits using individual tree phenotype and genotype mean values. Next, the effect of each tree trait on endophytes was examined while controlling for the effects of the other tree traits and tree genotype. With these analyses, we hypothesized that if a tree trait remains significant while holding genotype and the other traits constant, the significant trait may be linked to the underlying mechanism by which tree genotype influences endophytes. Total endophyte isolation frequency was examined with a REML-based linear model with tree genotype as a random effect and all tree traits as covariates. A restricted likelihood ratio test was used to test the effect of tree genotype, while Wald χ2 tests were used to test covariates. Equivalent models were examined for OTU composition using PERMANOVA with Type III sums of squares. These analyses were performed with the packages lme4, RLRsim and car in R 2.8.1 and Primer 6.1.15 with PERMANOVA+ 1.0.5. PERMANOVA and DISTLM were implemented with Bray–Curtis dissimilarity. Total aboveground tree biomass was natural-log-transformed in all analyses to linearize its relationships with response variables. Patterns of endophyte OTU co-occurrences, which may be indicative of species interactions, were examined with the C-score (Stone and Roberts, 1990). The C-score is computed as the average number of checkerboard units (i.e., one OTU occurs where another does not) among all OTU pairs in the community. This metric is robust to Type I error (Gotelli, 2000) and has been used previously to infer patterns of species interactions among endophytes (Pan and May, 2009). To test the significance of the observed C-score, we used a null modeling approach in which presence–absence matrices of OTUs were randomized with an algorithm that maintained observation totals and the relative number of isolates for OTUs (i.e., fixed-fixed algorithm; Gotelli, 2000). This algorithm was chosen because relative and total abundances were expected to differ among OTUs and observations, respectively. A total of 10 000 randomizations were run for each analysis. A single P-value for each analysis was calculated as the number of simulated C-scores equal to or more extreme than the observed C-score for each community. Because our hypothesis about co-occurrences was nondirectional, we used an alpha-level of 0.025 to test each C-score. A standardized effect size (SES), the difference between the observed C-score and the mean of the simulated C-scores, was generated for all analyses to facilitate comparisons among analyses. Higher SES values are indicative of negative relationships, as this results from OTUs tending not to co-occur (i.e., more checkerboard units). SES values less than zero are indicative of positive relationships that result in aggregation of species (i.e., positive co-occurrences). Three co-occurrence analyses were conducted separately for both endophyte sample dates: (1) using all trees pooled together, with the communities from each tree treated as replicates (generating a single C-score and SES value), (2) using the mean values of species abundances for each genotype in a single analysis (also generating a single C-score and SES value), and (3) individual analyses for each genotype with only communities from replicates of a single genotype at a time (generating a C-score and SES value for each genotype). Genotype 1000 (2008 sample) was excluded from analyses because it only had two replicates. Co-occurrence analyses were conducted in R using the commsimulator function in the bipartite package.
471 RESULTS
The endophytic fungal community and its response to tree genotype— A total of 33 fungal OTUs were recovered (Fig. 2; Appendix S2). Fungi emerged from 261 of the 726 twig segments in 2006 (~36%). These fungi grouped into 12 unique PCRRFLP types, seven of which were singletons (Fig. 2A; Appendix S2). Several isolates in the 2006 data set were difficult to separate morphologically or with PCR-RFLP; BLAST results suggested these represented several members of the genus Alternaria, and they were lumped into one OTU group for statistical analyses of the 2006 data set (OTU = AltComp2006 in Fig. 2A). Of the 576 twig segments plated in 2008, 301 grew endophytes (~52%), which represented 27 OTUs including 13 singletons (Fig. 2B; Appendix S2). All top BLAST matches were fungi from the division Ascomycota, and many were taxa sampled from plant tissues, including Populus spp. (Appendix S2). Consistent with our hypothesis that tree genotype influences endophyte community structure, some endophyte variables exhibited significant variation among tree genotypes in the 2006 data set (Table 2; Figs. 2A, 3A, B). Total endophyte isolation frequency differed among tree genotypes, with broad-sense heritability estimates indicating that 22% of the total variation in isolation frequency was due to tree genotype (Table 2; Fig. 3A). In contrast, species richness and Simpson’s diversity showed no evidence of a response to genotype (Table 2). Community composition exhibited a significant nonrandom pattern among tree genotypes, but because total variation of endophyte composition was high (i.e., variance due to genotype plus residual variance), broad-sense heritability estimates suggested that tree genotype accounted for 9% of the total variation in composition (using raw or relativized data; Table 2; Fig. 3B). However, in terms of Bray– Curtis dissimilarity, the compositions of endophyte communities were 13.7–15.2% dissimilar among genotypes, on average (square root of genotype variance component from PERMANOVA for the raw community matrix = 13.7, and for the matrix relativized by OTU maximum values = 15.2). Between the two distinct groupings of communities in tree genotypes observed in the NMDS ordination (Fig. 3B), indicator species analysis revealed that OTU Ogden_06 was an indicator of group 1 (genotypes WC-5, 10, H-10, 1008, 1005; indicator value = 39.5, P = 0.050), while Ogden_12 (indicator value = 51.1, P = 0.002) and Ogden_08 (indicator value = 67.7, P < 0.001) were associated with genotype group 2 (genotypes 1012, 1017, 996, 1023, T-15). Of the two individual OTUs who were examined individually, Ogden_08 exhibited significant variation among tree genotypes (RLR = 2.501, df = 1, P = 0.043, H2 = 0.210) while Ogden_22 did not (RLR = 0.000, df = 1, P = 1.00, H2 = 0.000). Communities isolated for the 2008 data set were also sensitive to tree genotype (Table 2; Figs. 2B, 3C, 3D), lending further support to our hypothesis that tree genotype influences endophyte community structure. Broad-sense heritability estimates indicated that tree genotype explained 72% of the variation in total isolation frequency and had a marginal effect on OTU richness, while Simpson’s diversity showed no evidence of variation among tree genotypes (Table 2; Fig. 3C). Community composition exhibited a significant nonrandom pattern among tree genotypes, and broadsense heritability estimates suggested that tree genotype accounted for 24% and 11% of the total variation in composition (using raw and relativized data, respectively; Table 2; Fig. 3D). In terms of Bray–Curtis dissimilarity, the compositions of endophyte communities were 17.4–22.8% dissimilar among genotypes, on average (square root of genotype variance component from PERMANOVA
472
AMERICAN JOURNAL OF BOTANY
[Vol. 101
Fig. 2. Relative isolation frequencies for endophytes from the (A) 2006 and (B) 2008 data sets, based on mean values of each operational taxonomic unit (OTU) for each genotype. Rare OTUs are present 1–3 times in a year’s entire data set.
for the raw community matrix = 22.8, and for the matrix relativized by OTU maximum values = 17.4). Among the three clusters of genotypes observed in the NMDS ordination (see Fig. 3D), indicator species analysis revealed that OTU Ogden_03 was a marginally significant indicator of group 1 (genotype 18; indicator value = 48.5, P = 0.078), Ogden_12 (indicator value = 56.5, P = 0.032) and Ogden_22 (indicator value = 74.1, P = 0.002) were associated with genotype group 2 (genotypes 999, 11, 1023), while group 3 (genotype 1000) had no significant indicators. Of the two individual OTUs who were examined individually, Ogden_08 exhibited significant variation among tree genotypes (RLR = 6.281, df = 1, P = 0.006, H2 = 0.590) while Ogden_22 showed marginally significant variation among genotypes (RLR = 2.080, df = 1, P = 0.064, H2 = 0.386). Relationships between tree traits and twig endophyte communities— All plant traits varied significantly among tree genotypes (Table 2; condensed tannins: mean = 2.11% dry mass, SD = 0.52% dry mass; aboveground tree biomass: mean = 198
711.1 g, SD = 169 968.5 g; 2003 twig length: mean = 4.83 cm, SD = 3.01 cm; 2003 twig diameter: mean = 0.33 cm, SD = 0.066 cm). Tree genotype explained 21–48% of the variance in the tree traits, demonstrating that these traits have a genetic basis (Table 2). Aboveground tree biomass exhibited significant positive phenotypic and genotypic correlations with twig diameter, while twig diameter exhibited a positive phenotypic correlation with twig length of the year of establishment and a similar (but nonsignificant) trend with twig length using genotype mean values (Appendix S3). Results suggest that variation in aboveground tree biomass and twig diameter are associated with variation in the endophyte community (Tables 3 and 4; Fig. 4). Correlation analyses show that total aboveground biomass and twig diameter were positively associated with total endophyte isolation frequency using tree phenotype data, but only aboveground biomass showed a significant positive relationship with isolation frequency using genotype mean values (Table 3). Condensed tannin concentrations and twig length were not significantly
March 2014]
LAMIT ET AL.—TREE GENETICS AND TWIG ENDOTYPES
TABLE 2.
Effect of tree genotype on endophyte community variables for each sample date and on tree traits for the 2006 data set. RLR or Pseudo-F
Response variable 2006 Endophyte isolation frequency Endophyte OTU richness Endophyte Simpson’s diversity index Endophyte OTU composition Endophyte OTU composition (relativized by OTU maximum) Condensed tannin concentration (% dry mass) Aboveground tree biomass (g) Twig length: year of community establishment, 2003 (cm) Twig diameter: year of community establishment, 2003 (cm) 2008 Endophyte isolation frequency Endophyte OTU richness Endophyte Simpson’s diversity index Endophyte OTU composition Endophyte OTU composition (relativized by OTU maximum)
P
H2
2.747 0.000 0.000 1.453 1.500
0.040 0.435 1.000 0.051 0.024
0.22 0.00 0.00 0.09 0.09
12.487
