Ectomycorrhizal fungal maladaptation and growth reductions associated with assisted migration of Douglas-fir
Summary
- Climatic adaptations are the foundation of conifer genecology, but populations also display variation in traits for nitrogen (N) utilization, along with some heritable specificity for ectomycorrhizal fungi (EMF). We examined soil and EMF influences on assisted migration of Douglas-fir (Pseudotsuga menziesii var. menziesii) by comparing two contrasting maritime populations planted up to 400 km northward in southwestern British Columbia.
- Soil N availability and host N status (via δ15N) were assessed across 12 maritime test sites, whereas EMF on local and introduced hosts were quantified by morphotyping with molecular analysis.
- Climatic transfer effects were only significant with soil N concentrations of test sites as a covariate, and illustrated how height growth was compromised for populations originating from relatively dry or cool maritime environments. We also found evidence for EMF maladaptation, where height declined by up to 15% with the extent of dissimilarity in EMF communities of southern populations relative to local hosts.
- The results demonstrate how geographic structure in belowground environments can contribute to conifer genecology. Differences in the inherent growth potential of conifers may be partly related to nutritional adaptations arising under native soil fertility, and optimization of this growth potential likely requires close affiliation with local EMF communities.
Introduction
Spatial gradients in environmental factors have produced geographically based, differentiated populations of trees that provide a valuable genetic heritage for forest management. Variation in growth and phenology – the adaptations most well established for conifer species – evolved in response to spatially structured gradients in climate (St Clair et al., 2005). The nutritional status of trees, by contrast, is rarely quantified as a geographically based trait (e.g. Martins et al., 2009), despite evidence for inherent differences in population growth rates related to nutrient utilization, especially nitrogen (N) (Miller & Hawkins, 2003; Pritchard & Guy, 2005). Landscapes such as the Pacific coast of North America have complex topography, diverse parent materials, and wide gradients in precipitation and temperature that have influenced soil development and produced disparate N regimes (Littke et al., 2011). The scale of these gradients in N supply, over many hundreds of kilometres, may provide enough spatial separation to allow for differential selection among tree populations in growth traits related to soil fertility. Our understanding of conifer genecology and management of genetic resources may therefore be improved by considering both nutritional and climatic adaptations of tree populations.
Forest genetic research has established that faster growing families of trees typically have more efficient N uptake (higher rates of net NH4+ or NO3− influx and accumulation) and utilization (amount of biomass produced per unit of leaf or plant N concentration) than slower growing families (Mari et al., 2003; Hawkins, 2007; Miller & Hawkins, 2007). Much of this nutritional research has used nonmycorrhizal conifer seedlings, however, and in some cases N utilization levels have been affected when ectomycorrhizal fungi (EMF) were included (Leski et al., 2010; Sousa et al., 2012; Velmala et al., 2013). Host population interactions with EMF deserve closer attention, as conifer feeder roots in natural settings are almost entirely colonized by these symbiotic fungi, providing key services to the host tree in the mobilization of N from soils and enhanced uptake of organic N, NH4+ and NO3− (Chalot & Plassard, 2011). Furthermore, the geographic structure of the mutualism is considerable, with typically diverse communities well aligned with climatic and edaphic conditions (Bahram et al., 2013), along with coevolutionary selection pressures possibly linking local populations of host trees with native EMF species (Piculell et al., 2008; Hoeksema et al., 2012).
As climates change, forests established with locally derived genetic stock may have heritable traits (growth rates, cold hardiness, bud phenology) that increasingly deviate from their adaptive peak (‘maladaptation’; St Clair & Howe, 2007). One strategy under discussion to mitigate productivity losses is to establish plantations with nonlocal seed sources better adapted to future climates (i.e. assisted migration) (Aubin et al., 2011; Pedlar et al., 2012). The productivity of transplanted tree populations over wide climatic gradients has been quantified using individual and pooled transfer functions (Wang et al., 2010), but nonclimatic variance in field trials is sometimes high (O'Neill et al., 2007), perhaps because population response to soils is rarely assessed. Although climatic adaptations are clearly important, the likelihood that assisted migration would ultimately benefit forests might also depend upon optimal uptake of soil nutrients and water, particularly if foreign trees have a less effective symbiosis with local EMF communities (Thompson et al., 2002). Mycorrhizal maladaptation, in this context, is defined as a relative decline in host fitness (survival, growth, nutrition) due to increasingly altered mycorrhizal communities from native settings. Evidence for higher fitness of local host populations with local soils and native mycorrhizal inoculum has been found in reciprocal transplants with arbuscular mycorrhizal plants (Moora et al., 2004; Ji et al., 2010; Johnson et al., 2010). Some spatial complexity was also observed in an EMF mutualism by the extent of divergence in local EMF communities with foreign host populations (Kranabetter et al., 2012). We could therefore envision two possible outcomes with assisted migration, where either inherently fast-growing populations consistently outperform over all others across landscapes, regardless of EMF associations, or a more geographically structured response, where growth and nutritional status of foreign populations diminish with increasing transfer distance from host origins due to interactions with native EMF.
We used a 40-yr-old provenance trial of coastal Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco var. menziesii) to examine the belowground influences of the environment on conifer genecology and assisted population migration. We focus on two maritime (low elevation, mild winter) populations from contrasting precipitation environments that were transplanted up to 400 km northward into 12 maritime test sites in southwestern British Columbia, Canada (we use the term ‘population’ to imply an inter-mating set of individuals, and ‘provenance’ as the geographic origin of those individuals). Test sites and populations were selected to sample a wide range in precipitation but narrow range in temperature to minimize possible confounding effects of cold hardiness and growth phenology (Darychuk et al., 2012; Gould et al., 2012) on nutrient utilization and tree productivity. We assessed the EMF community and N status (via δ15N) of each population, along with the soil properties of test sites, to test how strong the differences in survival and growth traits were, and whether transfer distance affected these outcomes. The objectives of our study were to: test the significance of soil N availability on individual and pooled climate transfer functions for tree height; test how the dissimilarity in EMF communities relative to local populations correlate with foreign host growth rates; and test whether N status of the populations varies spatially in accordance with EMF community and soil N interactions. Our hypothesis was that the relative productivity and N status of southern populations would decline with transfer distance because of the geographic structure underlying host-mutualist communities and their adaptations to local site conditions.
Materials and Methods
Materials and sites
Materials used in this study were obtained from a coastal Douglas-fir provenance trial (EP599.03) established between 1969 and 1971 by the British Columbia Ministry of Forests to identify superior seed sources for use in reforestation (Krakowski & Stoehr, 2009). The trial consists of five open-pollinated populations tested at each of 23 test sites in southwest British Columbia (BC), along with a local control population at each test site. We examined two of the five populations (Hoh and Duncan) and the Local population (which originated an average of 17 km from the test sites) at 12 of the test sites (Fig. 1; Table 1). We selected populations originating from slightly warmer (1–2°C mean annual temperature) environments than the test sites to simulate the magnitude and direction of migration distances being proposed or adopted in forestry-assisted migration (O'Neill et al., 2008). Populations and sites were also selected to evenly span the full gradient in precipitation regimes (Fig. 1; Table 1) comprising the range in natural habitat of coastal Douglas-fir in BC.
| Moisture class | Lat (°N) | Long (°W) | Elev (m) | MAT (°C) | MAP (mm) | AH:M | |
|---|---|---|---|---|---|---|---|
| Provenance | |||||||
| Hoh | Wet | 47.80 | 124.07 | 244 | 10.1 | 3384 | 5.9 |
| Duncan | Dry | 48.75 | 123.75 | 61 | 9.3 | 1080 | 17.9 |
| Test site | |||||||
| Saltspring Island | Dry | 48.75 | 123.48 | 365 | 9.1 | 1072 | 17.8 |
| Texada Island | Dry | 49.63 | 124.42 | 85 | 9.4 | 1144 | 17.0 |
| Dove Creek | Dry | 49.72 | 125.00 | 275 | 8.1 | 1634 | 11.1 |
| Quadra Island | Dry | 50.00 | 125.25 | 180 | 8.7 | 1613 | 11.6 |
| Lois Lake | Moist | 49.87 | 124.25 | 185 | 8.7 | 1559 | 12.0 |
| Harrison Lake | Moist | 49.53 | 121.75 | 245 | 7.3 | 1890 | 9.2 |
| Chilliwack Low | Moist | 49.08 | 121.72 | 215 | 6.9 | 1608 | 10.5 |
| Kelsey Bay | Moist | 50.32 | 125.85 | 395 | 8.2 | 2437 | 7.5 |
| Coal Harbour | Wet | 50.62 | 127.52 | 45 | 8.9 | 2019 | 9.4 |
| Chehalis | Wet | 49.48 | 122.00 | 275 | 7.9 | 2491 | 7.2 |
| Kennedy Lake | Wet | 49.12 | 125.55 | 335 | 8.9 | 4611 | 4.1 |
| Jeune Landing | Wet | 50.33 | 127.32 | 350 | 8.3 | 5320 | 3.4 |
- Lat, latitude; Long, longitude; Elev, elevation; MAT, mean annual temperature; MAP, mean annual precipitation; AH:M, heat:moisture index.
Planting stock was produced by sowing seed into irrigated and fertilized (40 kg N ha−1) field seedbeds in the first year, followed by a second year in irrigated and fertilized transplant beds. Seedlings were planted in spring of their third year in test sites situated in recently harvested operational cutblocks. The ectomycorrhizal status of the Douglas-fir stock before outplanting is unknown, but colonization was likely dominated by a native EMF species common to coastal nurseries, Thelephora terrestris (Kranabetter et al., 2012).
Experimental design and measurements
At each of the 12 sites, seedlings of the three populations (Hoh, Duncan, and the Local population) were planted at 3 m × 3 m spacing in four randomized complete blocks of 35-tree row plots per population. Four of the test sites (Coal Harbour, Harrison Lake, Kennedy Lake and Chilliwack) were thinned systematically at age 15 yr to retain 50% of the original number of trees. Cumulative mortality and tree height (measured by telescoping height poles or Forester Vertex (Haglöf, Sweden)) were assessed at ages 10, 20 and 30 yr. Population fitness was assessed by mortality rates and growth response via height. Although tree height may not be as direct a measure of population fitness as survival or fecundity, it is nevertheless an important attribute (Rehfeldt et al., 1999; Ying & Yanchuk, 2006) and can be measured more accurately and readily on large trees and across extensive field trials than other traits. Height growth reflects adaptive processes such as competition for light and photosynthetic activity (Falster & Westoby, 2003), and is strongly related to adaptation at all stages of ontogeny, except where trees approach their maximum height (King, 1990). Although we lacked the resources to remeasure stands at age 40 yr, when the current study was undertaken, the growth trajectories of Duncan, Hoh and the local seedlot were stable relative to each other over the first 30 yr (repeated measures ANOVA P-values for provenance × year = 0.852). For this reason we expect stand measures at age 30 yr to adequately reflect relative height differences at age 40 yr.
The close spacing of tree rows in the experimental design precluded random sampling of root tips. Therefore, to ensure host population identity, we sampled near the base of the bole where root attachment to the selected tree was more certain. Moss and duff within 30 cm of the tree bole were gently removed to reveal radiating structural roots of the sample tree. Thin feeder root systems that were attached to structural roots were clipped and removed. Ectomycorrhizal root tips were collected in the early summer (June 2012) and autumn (October 2012) to capture any seasonal effects, with the exception of the Kennedy site where all samples were taken in June due to difficulties in access. Root sampling was performed on three randomly selected trees from each of the three populations in each of the four blocks at each of the 12 test sites, for a total of 432 samples.
One hundred root tips were assessed from each root sample and all unknown morphotypes were submitted for DNA extraction and PCR amplification of the fungal ITS region of nuclear rDNA. Detailed molecular methodology can be found in Supporting Information Methods S1. A total of 577 root tip samples were subjected to DNA extraction and sequencing. The common and easily distinguished species Cenococcum geophilum, Piloderma byssinium, Piloderma fallax, Amphinema byssoides, Rhizopogon (tuberculate) vinicolor/vesiculosus and Lactarius rubrilacteus were tested molecularly at least three times, and then counted based on morphotype identification. Sequencing the many observations of these common fungi was impractical, and although we recognize potential lumping of closely related taxa within these morphotypes (Lim & Berbee, 2013), we feel the autecological differences among cryptic species likely to be sufficiently minor for the objectives of the study. Less familiar taxa with distinct morphotypes (e.g. Elaphomyces granulatus, Truncolumnella citrina) were confirmed from multiple collections across different sites, whereas indistinguishable morphotypes (e.g. Cortinarius and Inocybe species) were collected upon each observation of a root cluster. The sequences were BLAST searched (Altschul et al., 1997) against the GenBank and UNITE databases to suggest taxonomic affinities of the samples, and closely aligned sequences (e.g. species within Tomentella, Cortinarius, and Inocybe genera) were classified into separate taxa using a 97% similarity criteria with CD-HIT Suite (Huang et al., 2010). Representative sequences for each species were deposited at GenBank (accessions KM402878–KM403071).
Mineralizable N (min-N) and total N concentrations of the upper mineral soil were assessed as relative metrics of soil productivity based on the consistent utility of these attributes in delineating ecosystem types and assessing site productivity for coastal Douglas-fir (Kabzems & Klinka, 1987; Klinka & Carter, 1990). Min-N, measured through an anaerobic laboratory incubation, is a useful predictor of plant-available N in forests, more so for mineral soils than forest floors (Klinka et al., 1994). Given the utility of mineral soil, the distances between sites, and the effort required to obtain soil bulk density, we chose not to undertake repeated in situ measures of N availability and to exclude forest floors in the ranking of soil fertility to permit statistical analysis with N concentration rather than N content (kg ha−1) of the soil profile.
At each site, soils were sampled at eight randomly selected locations within each block. Mineral soils to a 20 cm depth were extracted with a soil auger for chemical analysis in June 2012. The eight samples in each block were pooled into two bulk samples, for a total of eight samples per site. The soils were air-dried, ground and sieved (2 mm) for chemical analysis. Mineralizable N was determined by saturating the soil for an anaerobic incubation of 2 wk at 30°C, followed by a 1N KCl extraction and colorimetric analysis for ammonium N (Kalra & Maynard, 1991). Total N was measured using combustion elemental analysis with a Fisons/Carl-Erba NA-1500 NCS analyzer (Thermo Fisher Scientific, Waltham, MA, USA).
The N status of the trees was assessed using 15N isotope concentrations of wood (Kranabetter et al., 2013), as accurate foliar sampling from the ground by pruning pole or shotgun was not feasible for these tall trees with overlapping canopies. Three trees per block of each population were cored to a depth of 5 cm with an increment borer for extraction of wood samples in June 2012. Bark was discarded, and the peripheral 0.5 cm of each wood increment core (equivalent to the last 4.0 yr of growth increment, on average) removed for drying (60°C for 24 h) and analysis. The three wood core segments per block from each population were combined (4 blocks × 3 populations = 12 samples per site) and ground to 1 mm. 15N isotope in the ground wood was measured using a Carlo Erba NC2500 elemental analyzer (Thermo Fisher Scientific, Waltham, MA, USA) interfaced with a Thermo Delta V+ isotope ratio mass spectrometer (IRMS). Combusted samples pass through a CO2 trap, a magnesium perchlorate water trap and a GC column (50°C) before entering the IRMS. Two different known standards (USGS 40 and USGS 41 or in-house standards that have been calibrated against the internationally known standards) were run to normalize values for each batch of samples. The long-term standard deviation of quality control standards is 0.3‰ for δ15N. The final isotopic values are expressed relative to international standard air for nitrogen (0.3660‰).
Climate data and analysis
Mean values of mean annual temperature (MAT), mean annual precipitation (MAP) and heat:moisture index (AH:M = (MAT + 10)/(MAP/1000)) for the 30-yr period 1961–1990 were obtained for each test site location and provenance by querying ClimateWNA v4.72 (Wang et al., 2012) with latitude, longitude and elevation. Climate transfer distance was calculated as the difference between test site and provenance (e.g. for Hoh at Saltspring Island, ΔAH:M = 17.8 − 5.9 = 11.9).
The general linear model procedure was used with Type III Sums-of-Squares and provenance as a class variable to examine stepwise multiple regressions of mean tree height or wood δ15N with transfer distance, site properties, and EMF community parameters as continuous variables (SAS Institute, 2011). Soil effects were nearly equivalent when solving with either total N or mineralizable N and for brevity only one variable (total N%) was reported. Species richness and Shannon–Wiener diversity index were determined by PC-ORD 5.0, whereas percentage dissimilarity (1−PS) between Duncan, Hoh and Local provenances was determined by relative Sørensen index using species abundance (% colonization) (McCune & Grace, 2002). Differences in EMF communities among populations were tested by blocked multi-response permutation procedure (MRBP) to isolate inherent site differences in EMF communities using PC-ORD (McCune & Grace, 2002). The MRBP analysis used the relative Sørensen distance measure based on species abundance, and included pairwise comparisons of provenances. A nonmetric multi-dimensional scaling (NMS) using the relative Sørensen distance was performed with PC-ORD on each provenance using Pearson and Kendall correlations to test site variables, with the ordination graphs rotated to soil mineralizable N concentration (McCune & Grace, 2002).
In order to test those factors that influence EMF species abundance, we fitted a completely randomized split-plot design separately to data from each EMF species using PROC MIXED (reml) of SAS (SAS Institute, 2011). Only those sites where the EMF species of interest was observed were used in the analysis, and all observations from the two sample periods were included. The P-values for testing Site (main-plot effect), Provenance and the Site × Provenance interaction were based on permutation tests, via a custom SAS macro wrapper that selected a random subset (10 000) of the total number of possible permutations for each effect (Cassell, 2002). Holm's correction was applied post-hoc to further adjust P-values for multiple testing.
Results
Test site characteristics
The test sites differed by over 4000 mm in precipitation (MAP 1072–5320 mm), by 2.5°C in temperature (MAT 6.9°–9.4°) and by almost 15 units in heat:moisture index (AH:M 3.4–17.8) (Table 1). Both total N and mineralizable N of the upper 20 cm of mineral soil ranged widely over the test sites (0.06–0.34%, and 1.0–82.8 mg kg−1, respectively; Table 2), and were positively correlated with MAP (N% = 0.07 +0.000047(MAP), P = 0.004, r2 = 0.58; and Min-N = 6.4 +0.012(MAP), P = 0.010, r2 = 0.48). Soil textures ranged from loamy sand to loam, and were predominantly acidic, averaging 4.9 in pH (range 4.4–5.9).
| Test site | Soil N (%) | Min-N (mg kg−1) | Local Ht (m) | ∆Ht Hoh (%) | ∆Ht Dunc (%) |
|---|---|---|---|---|---|
| Saltspring Island | 0.09 (0.01) | 5.5 (1.1) | 14.9 (0.3) | 8.4 | 6.6 |
| Texada Island | 0.13 (0.01) | 33.2 (8.7) | 17.4 (0.2) | −9.5 | −12.5 |
| Dove Creek | 0.13 (0.01) | 18.1 (3.9) | 24.9 (0.3) | 7.3 | 0.6 |
| Quadra Island | 0.13 (0.02) | 19.3 (4.3) | 21.7 (0.4) | −3.6 | 0.1 |
| Lois Lake | 0.06 (0.01) | 0.1 (0.01) | 11.4 (0.3) | 7.4 | 0.4 |
| Harrison Lake | 0.20 (0.03) | 45.5 (5.9) | 22.7 (0.5) | 3.0 | −10.6 |
| Chilliwack Low | 0.17 (0.01) | 37.1 (3.1) | 22.3 (0.2) | 3.9 | −13.6 |
| Kelsey Bay | 0.14 (0.01) | 23.9 (5.1) | 15.7 (0.2) | 2.0 | −4.8 |
| Coal Harbour | 0.28 (0.02) | 54.6 (8.2) | 23.0 (0.3) | 14.0 | 3.0 |
| Chehalis | 0.16 (0.01) | 29.0 (1.7) | 21.6 (0.2) | 3.1 | −15.9 |
| Kennedy Lake | 0.34 (0.02) | 82.8 (4.6) | 19.0 (0.5) | 14.6 | 2.0 |
| Jeune Landing | 0.27 (0.02) | 51.6 (6.5) | 19.1 (0.2) | −3.3 | −4.4 |
- Min-N, mineralizable nitrogen.
Population fitness by transfer functions
Mortality averaged 16% over the 12 test sites, with no significant differences in survival between Duncan and Hoh populations by physical (km; P = 0.444) or climatic transfer distance (ΔAH:M; P = 0.773). Relative differences in height for Duncan and Hoh (Δ%Ht) by transfer distance were also insignificant, both by physical (km; P = 0.876) and climate parameters (ΔMAT, P = 0.365; ΔMAP, P = 0.736; ΔAH:M, P = 0.813), although the net gains by Hoh were significantly greater overall than Duncan (+3.9% and −4.1%, respectively, P = 0.002). A better predictor of Δ%Ht was soil N% of the test sites (P = 0.002, r2 = 0.51), but the correlation was inversely curvilinear, indicating that sites with intermediate fertility had some of the largest relative reductions in height growth (Table 2; Fig. 2a). The pooled transfer function (Hoh and Duncan combined) was not correlated with climate distance as a quadratic function (P = 0.143), whereas a multivariate correlation that included soil N% of the test sites was considerably stronger (Fig. 2b; P = 0.003, r2 = 0.56). As a supplemental analysis, the significant contribution of soilN% to climate transfer analysis was also confirmed for the full set of maritime provenances (Jeune and Noeick from northwest Vancouver Island and midcoast BC, respectively, in addition to Hoh and Duncan; see Krakowski & Stoehr, 2009) planted on these test sites (P = 0.201 for ΔMAT and ΔMAP alone; and P < 0.001, r2 = 0.53 for ΔMAT, ΔMAP and soil N%; see Fig. S1). The pooled transfer function illustrated how host populations originating from dry or cold sites were likely to have more significant reductions in relative productivity on warm, wet sites (Figs 2b, S1).
Dissimilarity in ectomycorrhizal fungal communities
Virtually all feeder roots were found to be colonized by EMF. A majority of the morphotype samples were successfully sequenced (88%), leaving only 2.5% of the root systems as unidentified and excluded from statistical analysis. A total of 195 EMF species were identified, with 101 species occurring only once. EMF species richness and Shannon-Wiener diversity index did not differ among populations, or by transfer distance, averaging 21 species per site with an average diversity index of 1.83. Percentage dissimilarity (PD) of EMF communities in relation to Local ranged from 0.69 to 0.38, and was negatively correlated with ΔMAT (P = 0.006, r2 = 0.39), including a slight difference between populations (P = 0.062) (Fig. 3a). The increase in EMF community dissimilarity of Hoh and Duncan in relation to Local was also well correlated with reductions in relative tree height (P = 0.001, r2 = 0.48) (Fig. 3b). The correlation differed between Hoh and Duncan (P = 0.007), but with no PD × population interaction (P = 0.457). The correlation between PD and Δ%Ht was not nullified nor improved by adding any other transfer variables, including physical distance (P = 0.876), differences in elevation (P = 0.586), ΔMAT (P = 0.173), ΔMAP (P = 0.736), or ΔAH:M (P = 0.837) as covariates.
MRBP analysis of EMF communities depicted no difference overall in population alignment (P = 0.108; A = 0.0026), and in pairwise comparisons Duncan was only marginally significantly different from Hoh (P = 0.088; A = 0.0036). In the NMS ordination we found the EMF communities of each population were well aligned with site N status, but the correlation was slightly better for Hoh (r = 0.836 for min-N along axis 1, cumulative r2 = 0.814) than Duncan (r = 0.681 for min-N along axis 1, cumulative r2 = 0.696) or Local (r = 0.645 for min-N along axis 1, cumulative r2 = 0.788) (Fig. 4). The majority of individual EMF species were too infrequent to adequately test the effect of host population on abundance. Of those tested, a handful of species showed at least marginally significant host effects or site interactions, but none of these effects remained significant after post-hoc error correction for multiple testing (Table S1).
Population nitrogen status
Wood δ15N became less depleted with increasing soil N% (P = 0.001), and with finer soil textures (clay %, P < 0.001), but there were no differences in wood δ15N trends by population (P = 0.594) (Fig. 5a). Soil δ15N was also included as a covariate to standardize the extent of isotopic fractionation, but this step did not substantially improve site correlations. There was a significant change in isotope abundance relative to Local values (Δwood δ15N), with Hoh becoming less negative with physical transfer distance (P = 0.027, r2 = 0.40), in some contrast to Duncan (prov P = 0.067) (Fig. 5b). The effect of distance on Δwood δ15N was not negated, nor the correlation improved, by including climate variables or soil N% as covariates. No such correlation, however, was found for Δwood δ15N and PD (P = 0.689).
Discussion
The wide gradient in site fertility across southwestern British Columbia (BC) was ideal for examining the belowground structure of conifer genecology in more detail, particularly how soils might affect population fitness when seed is moved for reforestation or assisted migration. The correlation of soil N availability with increasing mean annual precipitation was likely linked to higher historic N inputs from red alder (Alnus rubra), lower N volatilization losses from wildfire, and greater microbial biomass and enzymatic activity in moist soils (Perakis & Sinkhorn, 2011; Brockett et al., 2012). Other significant influences on soil N availability were related to specific landscape features at each test site, such as mesoslope position, soil texture and drainage (Kranabetter et al., 2013). The upper limit in mineralizable N (83 mg kg−1) of the test sites was consistent with wet maritime forests of BC (Klinka et al., 1999), but the N capital of these rich morainal soils might still be only half of that found with sedimentary-derived soils of Douglas-fir forests further south (Littke et al., 2011; Perakis & Sinkhorn, 2011). The seed sources for Hoh and Duncan, despite being physically separated by only 110 km, encompassed a wide spectrum in precipitation, suggesting that these two populations originated from contrasting soil N environments (by a factor of c. 2.5 times, or 47 and 19 mg kg−1 min-N for Hoh and Duncan, respectively, based on the regression with MAP).
The results of the study indicate that interactions in population response to soil fertility, both as an inherent trait and by association with local EMF communities, had significantly affected climate transfer functions. Hoh often had the greatest differential in height from Duncan on rich sites (e.g. Harrison Lake, Chilliwack, Coal Harbour, Kennedy Lake), where greater supplies of N likely matched the greater host capacity for N uptake and utilization. This growth pattern would be generally consistent with the root physiological traits of fast- and slow-growing populations (Mari et al., 2003; Hawkins, 2007; Miller & Hawkins, 2007). But in addition to this inherent trait, we found the degree of overlap in EMF communities of the foreign populations with that of local hosts had modified the relative gains or losses (up to a net change of 20%) in height growth (Fig. 3b). This is, to the best of our knowledge, the first field evidence for maladaptation of an EMF symbiosis, where an increasing divergence in the EMF community of foreign hosts in relation to local populations was matched by declines in host productivity. This maladaptation effect was most evident for Duncan on sites of intermediate to high soil fertility, and likely contributed to the unexpected inverse curvilinear correlation between soil N% and relative height growth. An alternative explanation may be that photosynthate carbon (C) was reduced on these sites, leading to altered EMF communities, but these relatively small differences in tree heights are unlikely to cause the same degree of C deprivation to roots as defoliation or girdling (Saikkonen et al., 1999; Druebert et al., 2009). Consistent with studies of grassland plants and arbuscular mycorrhiza (Moora et al., 2004; Ji et al., 2010; Johnson et al., 2010), we suggest that this correlation indicates that a host tree's genetic potential for growth is maximized by close affiliation (both by species composition and relative abundance) with local, site-adapted EMF communities.
The considerable range in EMF community dissimilarity among populations (0.7–0.4), with no significant overall trend in provenance alignment, demonstrated the inconsistency in host heritability effects across these test sites (Kranabetter et al., 2012). A portion of the divergence in EMF community composition was related to ∆MAT (Fig. 3a), which suggests that extensive migration distances by southern populations to colder environments could lead to increasingly stronger heritability influences on EMF symbioses. Temperature effects may be related to coevolution selection mosaics (Piculell et al., 2008), based on the biogeography of certain EMF species, and in response to temperature gradients underlying soil fertility, such as the relative availability of amino acids, NH4+ and NO3−. There were, however, no effects of transfer distance on average EMF species richness or diversity (see also Velmala et al., 2014), which might have indicated much more significant incompatibility between these foreign host populations and native mycorrhizal species. Small shifts in EMF species occurrence and abundance were instead more apparent among populations, as indicated by the slight differences in Pearson correlations with soil min-N. In addition to ∆MAT, clinal shifts in species composition might be partly due to preferential allocation of photosynthate to fungi most adept at nutrient uptake (Bever et al., 2009), which could differ between populations because of contrasting nutrient demands (Kranabetter et al., 2012). Clear differences in host compatibility with individual fungal species (Hoeksema et al., 2012) were apparently rare, based on % colonization, but would also be difficult to detect in field studies encompassing such diverse EMF communities.
The general pattern in reduced discrimination against the 15N isotope with increasing soil N supply was consistent with EMF trees from similar productivity gradients (Kranabetter & MacKenzie, 2010; Mayor et al., 2012). The net difference in δ15N over the site N gradient was relatively narrow, however, at 1.4‰ on average, which was much less than might be expected with foliar attributes (> 5‰; Kranabetter & MacKenzie, 2010; Mayor et al., 2012). It is possible that the relative insensitivity of wood δ15N as a measure of tree N status (i.e. Drake et al., 2011) obscured some potential differences in N uptake among populations, despite the differences in growth rates and EMF community dissimilarity. In addition, some important N utilization traits might not be reflected by δ15N values, such as greater allocation of biomass to shoots rather than roots, or more efficient growth at a given plant N status (Hawkins, 2007; Velmala et al., 2014). Nevertheless, 15N abundance was not entirely consistent among populations, and the increase in Δwood δ15N with physical distance suggests that the inherent advantages in N uptake (more so for Hoh than Duncan) had narrowed with a northern migration. The relative extent of N deficiencies would presumably be minor, given no associated declines in tree height, but we would also caution that spatial patterns in host N status might only be fully discernible with foliar attributes.
A potential physiological adaptation of conifers that was not assessed in our study is drought tolerance and water use efficiency (WUE). High WUE is often correlated with greater growth height among conifer populations (Guy & Holowachuk, 2001), and selecting and deploying drought tolerant populations or genotypes during reforestation or afforestation may prove to be an important climate change adaptation strategy (Eilmann et al., 2013). On our test sites there appeared to be no advantages for Duncan or (rarely) Local provenances on the draughtier (high AH:M) sites, however, indicating there had been little consistent differentiation among these populations for any adaptive traits related to moisture stress (Darychuk et al., 2012). It should also be noted that WUE can be an indirect measure of tree N status, rather than water relations, as differences in δ13C values among conifer populations have been attributed to increased foliar N content and enhanced photosynthetic rates (Guy & Holowachuk, 2001; Ripullone et al., 2004).
A second, more unpredictable, future avenue of maladaptation will be whether current site-adapted EMF communities decline in effectiveness with projected changes in climate over the coming decades. Both soil fertility and EMF community composition could undergo sizeable transformations with shifts in temperature or precipitation regimes (Butler et al., 2012; Pickles et al., 2012), and consequently any lags in mycorrhizal fungal response to altered site conditions may reduce both local and foreign host fitness and shift growth trajectories over time. The outcome is difficult to predict, because although we found some geographic structure in host-fungal populations of coastal Douglas-fir, we suspect that these selection pressures are well balanced by the robustness of the symbiosis over wide transfer distances and site conditions that promotes a good degree of adaptability and resilience. Extrapolating recent growth trends of conifers into the future is already difficult for a number of reasons (Coops et al., 2010), and ongoing monitoring of provenance trials may continue to shed light on these many questions regarding implications of climate change and assisted migration on all facets of forest genecology and productivity.
Conclusions
Transfer and response functions are important genecological tools in establishing safe seed procurement and deployment distances, and for assessing potential population response to climate change (Wang et al., 2010). Our study emphasized how adaptations to soil biotic and abiotic factors combine with climate to contribute to population fitness and growth. Soil N availability of test sites explained a significant portion of the variation in relative population growth, possibly in response to nutrient utilization traits selected for under contrasting native soil conditions. Close affiliation of host populations with local EMF communities also maximized fitness, which illustrated an important aspect of geographic structure with the symbiosis. Based on the results from this trial, we suggest that productivity could be best maintained by relatively short directional transfer of seed (wet to dry and warm to cold environments), as these populations generally do as well or better than local populations (Krakowski & Stoehr, 2009), with no inverse disadvantages on low fertility soils (Miller & Hawkins, 2003). At the same time it would be prudent to move multiple populations incrementally northward, or in elevation (Gray & Hamann, 2013), in recognition of the potential EMF maladaptation and declines in N status with transfer distance.
Acknowledgements
We thank Doug Ashbee and Jodie Krakowski of the BC Ministry of Forests, Lands and Natural Resource Operations for maintenance and data management of EP599. Heather Klassen of the BC Ministry of Forests, Lands and Natural Resource Operations provided valuable assistance in the often arduous fieldwork of this study. Molecular analysis was undertaken by Grace Ross, Tyler Dyer and Dave Dunn of the Pacific Forestry Centre, Victoria. Statistical advice and the wrapper SAS program were provided by Peter Ott of the BC Ministry of Forests, Lands and Natural Resource Operations. Soil analysis was undertaken by Clive Dawson and Andre Bindon of the Ministry of Environment Analytical Lab in Victoria, whereas N isotope analysis was completed by the Central Appalachian Isotope Facility at the University of Maryland, USA. Todd Davis of the West Coast Regional office produced the map figure. Funds for the research were provided by the Forest Genetics Council of British Columbia.
